circuit-cellar-363-2020

Prévia do material em texto

 



2 CIRCUIT CELLAR • OCTOBER 2020 1363 
OUR NETWORK 
(• drcuit cellar f i I [. ftHxpress 
VOICE•t COIL 
~ LinuxGizmos Ll SLOUDSPEAKER INOUil•Y SOURCEBOOK 
All Electronics Corp. 77 
CCS, Inc. 77 
congatec, Inc. 25 
EasyOEM 13 
Micro Computer Control Corp. 77 
IAR Systems, Inc. C3 
Newhaven Display International, Inc. 11 
PCBWay 9 
Pico Technology 17 
SlingShot Assembly 43 
Technologic Systems, Inc. C4, 77 
University of Cincinnati Online 67 
NOT A SUPPORTING COMPANY YET? 
Contact Hugh Heinsohn 
(hugh@circuitcellar.com, Phone: 757-525-3677, Fax: 888-980-1303) 
to reserve space in the next issue of Circuit Cellar. 
THE TEAM 
PRESIDENT 
KC Prescott 
CONTROLLER 
Chuck Fellows 
FOUNDER 
Steve Ciarcia 
COLUMNISTS 
EDITOR-IN-CHIEF 
Jeff Child 
SENIOR ASSOCIATE EDITOR 
Shannon Becker 
TECHNICAL COPY EDITOR 
Carol Bower 
GRAPHICS 
Grace Chen 
ADVERTISING COORDINATOR 
Nathaniel Black 
ADVERTISING SALES REP. 
Hugh Heinsohn 
PROJECT EDITORS 
Ken Davidson 
David Tweed 
Jeff Bachiochi (From the Bench), Bob Japenga (Embedded in Thin Slices), 
Robert Lacoste (The Darker Side}, Brian Millier (Picking Up Mixed Signals), 
and Colin O'Fiynn (Embedded Systems Essentials} 
1111Uel63 October 2020 IISS111521-06011 
QRCUIT CELLAR® (ISSN 1528-0608) Is publish~ monthly by: 
KO< Media Corp. 
PO Box 417, <Mse City, VA 23924 
Periodical rates paid at <Mse Oty, VA, end additioiKll offiCI!S. 
One-)'l!llr (12 issue<) sub<criptlon ralr! US and possessions 
$50, Canada $65, Foreign/ P!JW $75. AI SIJbscription orders 
payable in US funds only via Vtsa, MasterCard, inWIKltional 
postal money order, or check drawn on US bank. 
SUBSCRIPTION MANAGEMENT 
OniiM "-llllt .._ ........... ,, circuitcellar.com/ac00101t 
Renew I Change Address/E-mail 1 Check Status 
CUSTOMER SERVICE 
1!-tu•: customerservlce4Pclrcuitcellar.ccm 
Phone: 434.533.0246 
Mlolt Circuit Cellar, PO Box 417, Chase City, VA 23924 
Postmaster: Send addross changos to 
OmJit Cellar, PO Box 417, Chase Oty, VA 23924 
NEW SU!lSCRII'HONS 
drcuitx:ellar.ccm/SIJbscription 
ADVERTISING 
Contact: Hugh Heinsohn 
Phone: 757-525-3677 
Fu: 888-980-1303 
E-mail: hheinsohnOclrcultceller.ccm 
Advertising rates and terms available on request. 
NEW PRODUCTS 
E-md: editor@clrcultcellar.com 
HEAD OFFICE 
KO< Media Corp. 
PO Box417 
Cha~ City, VA 23924 
Phone: 434-533-0246 
CDPYRIGitT NOTICE 
Entire contents copyright 11:) 2020 by KCK Media Corp. 
All rights reserved. Circuit Cellar Is a registered trademark 
of KCK Media Corp. Reproduction of this publication In 
whole or in part without written consent from 
KCK Media Corp. Is prohibited. 
DISCLAIMER 
KCK Media Corp. makes no warranties and assumes no 
responsibility or liability of any kind for errors In these 
programs or schematics or for the consequences of any such 
errors printed in Circuit Cellar®. Furthermore, because of 
possible variation in the quality and condition of materials and 
workmanship at r!!ader-os~mbled projects, KCK Media Corp. 
disdaims any responsibility for the sofe and proper function 
of reader-assembled projects bas~ upon or from plans, 
descriptions, or irlormation published in Circuit Cellar®. 
The irlormation provided in Circuit Celar® by KCK Media 
Corp. is for educational purposes. KO< Media Corp. makes 
no claims or warrants that readers have a right to build 
things based upon these lde1ts under pat:l!nt or other 
relevant intelectual property law In their jurisdiction, or 
that readers have a rl!lht to construct or operate any of 
the devices described herein under the relevant patent or 
other intellectual property law of the reader's jurisdiction. 
The reader assumes any risk of Infringement liabifity fur 
constructing or operating such devices. 
© KCK Media Corp. 2020 Printed In the United States 
E mbedded system design is a broad landscape. For a large segment of applications, microcontrollers are the main computing and control element-a 
key focus for Circuit Cellar obviously. But when you 
move above that realm, into the so-called mid- and 
high-end ranges of embedded computing, single-board 
computers (SBCs) have been the traditional center of 
gravity. By center of gravity I mean that all the elements 
of a system needed to revolve around the SBC. Because 
most systems needed a variety of functions beyond the 
SBCs, an SBC would often need to be in some standard 
form factor, and connect with compatible I/0 boards, 
DSP boards or whatever over some sort of standards-
based parallel bus-or serial buses in more recent 
times. This meant choosing a bus architecture, a card 
cage and an SBC, plus any additional I/0 to fulfill your 
system requirements. 
In some high-end systems-or legacy systems-
such slot-card style systems are still popular. But as 
levels of chip and module integration progressed, such 
systems no longer made sense for many applications. 
Instead, what's emerged are a myriad of box-level 
systems. These complete system boxes provide a 
complete, tested and enclosed computing solution that 
eliminate complex integration for customers. Because 
they are often sold as turn-key "ready-to-run" systems, 
system developers can begin creating their application 
software to run on these systems immediately. 
What's interesting is that box-level systems haven't 
replaced SBCs by any means. SBCs are still vitally 
important in lots of mid- to high-end embedded 
computing applications. That said, I'd argue that box-
level systems have become-alongside SBCs-a second 
set of gravity in the embedded computing world. 
At one time, these box-level systems were mainly 
sold as fully custom offerings, built for specific customer 
needs. But around 10 years ago, give or take, these box-
level embedded computers started to become standard 
catalog products. In fact, SBC vendors were among the 
leaders of this new trend. Fast forward to today, and 
you see that box-level systems as a product category 
has exploded-driven by many traditional SBC vendors 
as well as companies specializing in such systems. 
circuitcellar.com 3 
The Box as a Component 
It's difficult to get one's arms around all the 
variations in today's box-level systems. In this issue's 
Datasheet section, we focus narrowly on the "embedded 
PC" segment, but it truly is a vast arena that's hot with 
design activity. Beyond the generic embedded PC type, 
there are flavors of box-level systems of all kinds-
vehicle computers, vision systems, AI/IoT processing 
systems, systems targeted for railway applications, 
digital signage boxes and more. 
Box-level systems are often being used and upgraded 
in much the same ways that SBCs have been. When 
new levels of processing and memory are available, 
system developers can swap in a new box. Connectivity 
has become simpler too in these days where IP-based 
Ethernet networking is a straightforward way for box-
level subsystems to communicate and interoperate with 
each other. 
Another advantage of box-level systems is that they 
off-load whole phases of engineering from your system 
development process. Instead of integrating hardware 
and OS software yourself, it's already done for you. 
Meanwhile, environmental testing-for temperature, 
vibration and so on-are done by box-level system 
vendors (although not always, and not always to the 
same extent). Again, to be clear, SBCs and box-level 
systems are the stars of what I call the mid- to high-end 
of embedding computing. In the more deeply embedded 
realm, MCUs and embedded processors remain the king 
of the castle. 
The bottom line is that box-level embedded computing 
systems have evolved to the point where embedded 
system developers can think of them as components. Like 
nearly every other advance in electronics technology, 
these solutions let you move up a level in your perspective 
and focus more on broad concerns like your application 
software, user interfaces and so forth. Circuit Cellar will 
continue to follow this exciting area oftechnology. e 
4 CIRCUIT CELLAR • OCTOBER 2020 #363 
DATASHEET 
50 Embedded PCs Integrated Solutions 
By Jeff Child 
56 Embedded in Thin Slices 
LoRa (Part 7) 
Building a Prototype LoRa Device 
By Bob Japenga 
60 From the Bench 
White Hot: Measuring 
Color Temperature 
Using an Arduino Mega 
By Jeff Bachloch/ 
68 The Darker Side 
An Introduction to Stub Filters 
A Wire to Nowhere 
By Robert Lacoste 
TECH THE FUTURE 
7 The Future of Modular Edge Processing 
Three Real-World Intelligent 
Infrastructure Applications 
By Dan Demers 
75 : PRODUCT NEWS 78 : TEST YOUR EQ 
ci rcuitcellar. corn 5 
6 Building a Power Rail Probe An Exercise in Low-Noise Design 
By Andrew Levido 
14 Design a Wireless Tracking Device 
Smart Antenna Choices 
By Geoff Schulteis 
18 A Review: VS Plus Drone Flight Controller 
Hands-On Evaluation 
By Raul Alvarez-Torrico 
26 Code Size vs. Memory Footprint And Why They Matter 
By Steve Graves 
30 Balancing a Ball on a Touchscreen Using a PID Controller 
By Gregory Kaiser and Samuel Feibel 
SPECIAL FEATURE 
35 Smart Cities Leverage loT Building Blocks 
Chips, Systems and Services 
By Jeff Child 
44 
TECHNOLOGY SPOTLIGHT 
Raspberry Pi Roundup 
A Bounty of HATs and Clones 
By Jeff Child 
~ @editor _cc .:31 
' @cirtuitcellar a • cirtuitcellar 
W hy would someone spend several thousand dollars on a probe just for power rails? 
The reason is that standard 
passive oscilloscope probes are not really 
suited to the task. Let's see why. In probing a 
power rail with an oscilloscope, you probably 
want to do one or more of three things: (1) 
look at the level of ripple and noise on the rail; 
(2) chase fast spikes or switching artifacts or 
(3) look at load transients such as over- or 
under-shoots, settling times and so forth. 
These are all usually millivolt-level signals 
riding on top of the DC rails, with periods 
ranging from nanoseconds to seconds. 
You might try to use a standard passive 
probe with your oscilloscope switched to 
DC coupling, and rely on the vertical offset 
controls to center the waveform on screen. 
Unfortunately, you will find that at millivolt-
per-division resolution, you are limited to just 
a volt or two of offset range, so you may not 
even be able to get a 3.3V rail on the screen, 
let alone a 12V rail. 
If you flick the AC-coupling switch, you 
will be able to get the trace on the screen 
with millivolt resolution, but you will have 
just introduced a high-pass filter with a -3dB 
cut-off at a few tens of Hertz. You therefore 
will not be able to make much sense of load 
transients with a time constant of longer than 
a few tens of milliseconds. 
ADC NOISE 
If we are interested in measuring the noise 
on the power rail, the displayed noise will 
include that introduced by the oscilloscope's 
analog-to-digital converter (ADC), your chosen 
attenuation path and the input probe. There 
is not much you can do about the ADC noise 
(except buy a more expensive oscilloscope), 
but the attenuation path and probe noise are 
things we can control. With most oscilloscopes, 
the lowest noise is achieved when you switch 
in the SOn input termination. This is not 
practical with power rail measurements, since 
son would impose a significant load on the 
power supply, and the SOQ input is usually only 
rated to :!::SV to limit the power dissipation in 
the termination resistor. 
A dedicated power rail probe is designed 
to overcome these limitations. It presents a 
relatively high-impedance load to the power 
rail, and provides a means to inject a DC offset 
to the signal, so the DC coupling can be used 
with high vertical sensitivity. It should have a 
very wide bandwidth, from DC to at least the 
oscilloscope's bandwidth, be compatible with 
the son input and should introduce as little 
additional noise as possible. 
The commercial units I could find seemed 
to be specific to each manufacturer's 
proprietary active probe interface, and were 
integrated into the oscilloscope's on-screen 
user interface. I wanted to see what would 
be involved in rolling my own version of a 
power rail probe, so I did a quick survey of 
the available units to come up with a basic 
specification for mine. 
1. It should be able to null out DC offsets of 
±20V, and be safe to connect to ±60V. 
2. A DC input impedance of SOkQ seems to be 
the de-facto standard. 
3. I aimed for an attenuation of 1:1.2 or 
better, and flat to <ldB over the range DC 
to lGHz. 
4. It should accommodate an active signal 
range of ±lV, and be designed to operate 
into a SOQ load. 
5. I set myself an ambitious target of the probe 
contributing no more than 10% additional 
noise beyond a typical oscilloscope's noise 
floor. 
DESIGN 
This design turned out to be an interesting 
exercise in low-noise design, and included a 
little bit of AC circuit analysis into the bargain. I 
think it might be interesting to walk through the 
highlights of the design process step by step. 
I did not think a low-noise active circuit 
with a bandwidth in the gigahertz range was 
a practical proposition. Instead, I looked at 
splitting the probe into two parallel signal 
paths-an active, low-frequency path that 
deals with the DC offset, and a passive, high-
Jl 
In 
V-
Oft>« 
RI 
50k 
.... ~ .,..._...,.--.~vv......._.~ 
FIGURE 1 
V• 
R3 
OUl I 
V• 
V-
Re 
200 
frequency path. I started the design with the 
low-frequency path. This can be an ordinary 
summing amplifier followed by an inverting 
stage to restore the polarity of the output 
signal, as shown in the schematic (Figure 1). 
This circuit pretty much designs itself, 
because we know the input impedance must 
be 50kQ, and the feedback resistor R2 must 
be the same for a gain of -1. If we assume a 
split SV supply (±2.SV), then we can choose 
R2 to be SkQ for a gain of -10, giving an offset 
range of ±25V, which is nicely in line with 
our specification. Since we are connecting 
R2 directly to the wiper of the potentiometer, 
the input impedance we will see on the 
offset summing input will only be SkQ at the 
extremes of the potentiometer. 
In the middle, the impedance will be SkQ 
plus the two "halves" of the potentiometer 
resistance in parallel, reducing the gain from 
-10 to something less than that. For the SkQ 
pot I used, the gain of the offset branch is 
reduced to -8 at the center point. This means 
we will have slightly increased sensitivity at 
lower offset voltages. This is by no means 
a bad thing, and the alternative of adding a 
buffer between the pot and R2 is not desirable, 
because it will add more noise to the circuit. 
Because we are adding a high level of 
offset on purpose in this circuit, for once we 
do not have to be too concerned about the 
input offset voltages or bias currents of the 
op amps we will select. This also means we do 
not need to add resistors in the non-inverting 
inputs of the op amps, and can connect them 
directly to our circuit ground. Similarly, for 
the inverting amplifier, we just select R4 = RS 
C l 
J2 
R6 
SWIA 10 Ou1 l OOn 
~ - 81 
J-82 
u 2 R7 DP.t r-
10, 10 DAV99V : 
I 
I 
I 
I 
L . 
I 
V< J-a• 
LEDl 
~ .... SWl8 
ci rcuitcellar. corn 7 
l PI 
C5:1 c~ • OV• 
47~ ·~ 10V -::::-
GND 
~ 
1P2 
c1: eel I 0 \J .. 
4 7~ lOOn+ lOV -:;-
Full schematic for the power rail probe. The DC path is a simple summing amplifier followed by an inverting stage. The AC path is a direct connection. The two paths are combined 
via Cl and Ll. R6 and R7 ensure a flat frequency response. 
B CIRCUIT CELLAR • OCTOBER 2020 #363 
a} c 
FIGURE2 
The circuit on the left is a simplified 
view of how the low and high 
frequency paths combine. The 
functionally equivalent circuit on the 
right is instantly recognisable as a 
classic LC notch filter. 
FIGURE3 
The RC and RL series combinations 
on the left can be converted to their 
parallel equivalents, as shown on 
the right, via the transformation 
equations. This transformation helps 
us understand the impedance ofthe 
network at resonance. 
b) c 
for a gain of -1. We will choose these values 
a little later, since they will affect the noise 
performance of our circuit. 
We can now look at the high-frequency 
path. Since we don't have to worry about 
the DC offset, this path is just a direct 
connection between the input terminal and 
Cl. Figure 2a shows a simplified view of 
how the low-frequency active path and the 
passive high-frequency path come together. 
The capacitor C and the load impedance RL 
form a high-pass RC filter, and the inductor 
l similarly forms a low-pass RL filter with RL. 
This is functionally equivalent to Figure Zb, 
which you may recognize as a classic LC notch 
filter. A steep notch appears in the frequency 
response at the resonant frequency given by: 
1 
fo = 21t./LC 
The depth of the notch is determined by 
the Q of the filter. Naturally, we do not want 
this notch in our frequency response, so we 
need to lower the Q of the filter by adding 
some series resistors R5 to the circuit, 
to provide some damping (Figure 3). We 
also need to preserve the symmetry of the 
branches, so each branch must have the same 
resistor value. 
These series resistors reduce the gain 
of the circuit in each branch, in that they 
form wideband voltage dividers with the 
load resistance RL. How do we choose these 
resistors for a flat frequency response? The 
answer is not immediately obvious, but we 
can make this easier by exploiting the series-
parallel transformation property of complex 
impedances. This simply says that for every 
series combination of real and imaginary 
Series 
impedances, there is a corresponding parallel 
combination that will have the same overall 
behavior when viewed from the terminals. 
Figure 3 shows how this looks for our circuit. The 
series C5 + R5 and L5 + R5 combinations become 
parallel Cp 11 Rp and Lp 11 Rp combinations. 
At DC, our series circuit will look like an 
impedance of R5, since the capacitor C5 is 
essentially open, and the inductor L5 a short. 
Similarly, at very high frequencies, the series 
circuit will also look like an impedance of R5, 
because the inductor L5 will be essentially 
open and the capacitor C5 essentially a short. 
It is not obvious from the series circuit what 
the impedance will be at resonance. 
This is where we need to look at the 
equivalent parallel circuit. We can see here that 
at resonance, when the lp 11 Cp combination is 
presenting an infinite impedance, the end-to-
end impedance will be the parallel combination 
of the two resistors. If we want the gain to 
be flat across the spectrum, this should also 
be equal to R5• Substituting Rp = 2R5 into the 
series-parallel transformation equation for Rp 
given in Figure 3, and using the expression 
above for the resonant frequency, a little 
algebra delivers the expression: 
Rs= {& vc; 
Now we can choose some values. Let's 
start by choosing C to be lOOnF, which doesn't 
seem like too much of an additional capacitive 
load, given that power rails are generally 
bypassed by many lOOnF capacitors. A value 
of lOOnF will form an RC high-pass filter with 
a corner frequency of around 31.8kHz. 
We want to choose the inductor l such 
that the corner frequency of the Rl low-pass 
filter is considerably higher than this, but not 
so high that we create an unnecessary op amp 
design challenge. Choosing lMHz, for example, 
gives us an L value of 7.91JH. Rounding this up 
to a sensible value like l01JH gives a corner 
frequency of around 796kHz, which should 
work fine. We can now finally calculate Rp, 
which turns out to be lOQ-a nice, round value 
that gives us an overall gain through the probe 
of 1:1.2, exactly our design brief. 
Parallel 
NOISE 101 
Before we can we can select the op amps 
for the circuit, let's look at the noise issues 
involved. This op amp selection will be driven 
largely by the need to add the lowest possible 
amount of noise to the measurement. Low-
noise design is a whole field in itself, but let's 
recap on some "Noise 101" basics. The three 
main sources of noise we need to consider in 
modern electronics are Johnson noise, shot 
noise and 1/f or "flicker" noise. 
Johnson noise is essentially the thermal 
noise caused by the movement of electrons 
with temperature. For example, a 10kQ 
resistor lying unconnected on your bench at 
room temperature will develop a noise voltage 
across its terminals of 1.31JVRMS• if measured 
with a voltmeter with 10kHz bandwidth. 
This brings us to an important point 
about noise. If we want to calculate a noise 
voltage or current, we must always specify 
the bandwidth over which we are concerned. 
Because we don't always know what the band 
of interest is when specifying components, 
we tend to talk about noise densities, which 
is a measure of the noise voltage or current 
per unit of bandwidth. Noise densities are 
expressed in terms of voltage or current 
per square root Hertz. You just multiply the 
voltage or current noise density by the square 
root of the bandwidth to get the root mean 
square (RMS) voltage or current. 
While we are on the subject of noise, 
another important thing to keep in mind is 
how noise in a circuit adds. Since noise is 
random, as long as the noise sources are 
not correlated, the two sources don't add 
arithmetically, but rather, as the square root 
of the sum of squares: 
en= ~e;l +e:2 + ... 
With these two points covered, let's get 
back to Johnson noise. The voltage noise 
density of Johnson noise in a resistor is given 
by: 
where k is Boltzmann's constant and T is the 
temperature in Kelvin. At room temperature, 
this can be reduced to: en = 127 x 10·12 in 
units of "V/...fHz". We will use this equation 
later to calculate the noise produced by the 
resistors in our circuit. 
The two other sources of noise are 
generally specific to semiconductors. Shot 
noise is the noise produced by the discrete 
nature of charge when charges are acting 
ci rcuitcellar. corn 9 
PCBWay -
PCBWay is a lead1ng PCB manufacturer m Shenzhen speciahztng tn PCB protolyptng and 
assembly serviCe, and havmg been loved by over 1M engmeers around the world for tt 
comprehensive advantages 
•m• n------- lnsumt Ouo1e 
Tumaround ltme can be as short as 24 hours - -
CIJ.!.fii 
Pr~ee ts as low as 10 PCBs for only 5 USD 
10 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE4 
Noiseless 
OpAmp 
The noise model of an op amp has a voltage noise source in series with the input, and sources forcing current 
noise out of the input pins. Op amp noise is specified in terms of these noise densities and the f recuency fc, 
at which the flicker noise begins to dominate. 
ABOUT THE AUTHOR 
independently, as they do in semiconductors 
(but not in conductors). 1/f or flicker noise is 
an artifact of the materials used and methods 
of construction of the devices. 
Johnson and shot noise are called "white 
noise," and have equal energy in every unit 
of bandwidth. Flicker noise is "pink" noise 
and has equal energies in every decade 
of bandwidth. If you plotted the energy of 
flicker noise versus frequency, we would see 
the same energy in the 1-10Hz decade as 
the 10-100Hz or 100Hz-1kHz decades. The 
amount of energy therefore falls off with the 
inverse of frequency-hence the 1/f name. 
In a semiconductor device such as an op 
amp, there will be a white-noise component 
that is flat with frequency, and a pink-noise 
component that will fall off at 1/f. The point 
where the 1/f noise has the same value as 
the white noise is designated fc. This figure is 
used if the frequency of interest is low enough 
that 1/f noise becomes important-as we shall 
see later. 
We don't need to concern ourselves with 
the detail of all the noise sources in op amps, 
since manufacturers specify the total noise in 
terms of three parameters-a voltage noise 
Andrew Levido ,(P'a-n""'dr_e_w_.l.-e-vi""'d,....o@.,..---m-a-;,il,.....c-o-m..,......arned a bachelor's degree in 
Electrical Engineering in Sydney, Australia,in 1986. He worked for many 
years in R&D for power elect ronics and telecommunicati on companies, 
before moving into management roles. Andrew has maintained a hands-on 
interest in electronics, particularly embedded systems, power electronics and 
analog design. Over the years, he has written articles for various electronics 
publications, and provides consulting services as t ime allows. 
density and a current noise density referred 
to as the op amp input, and an fc figure as 
described earlier. 
As shown in Figure 4, the noise model of 
an op amp looks like a noise voltage source 
en in series with the op amp input, and noise 
current sources in driving current out of the 
input terminals. These currents will create 
noise voltages across the source resistance 
as seen by the input terminal. There will also 
be a Johnson noise component in the source 
resistance itself. The noise that appears at 
the op amp output terminal will be the sum 
of these three sources multiplied through the 
op amp gain. If there is noise on the op amp 
power rails, this too will affect the overall noise, 
as seen at the op amp output. I will ignore this 
for now, but check whether this is reasonable 
once we have designed the power supply. 
COMPONENT SELECTION 
We now need to choose the op amps. In 
this case our primary requirement is for low 
noise. We also need a bandwidth well above 
the 796kHz corner frequency of the LC filter-
say SMHz or higher. We want to operate from 
low voltage supply (:1::2.5V), and U2 needs to 
be able to drive a volt into a SOQ load. 
One candidate is the Texas Instruments 
(TI) LMP7731. This has a GBW product of 
22MHz, swings almost to the rails with a SV 
supply, and has excellent noise figures with 
en = 2.9nV/..JHz, in = 2.2pA/>/Hz and fc = 3Hz. 
Quiescent current is 4mA, and it costs less 
than $2. The only shortcoming is that its 
output drive capability is not enough. Another 
candidate is TI's TLV2460. This has a GBW 
product of 6.4 MHz, good output swing, SOmA 
drive capability, quiescent current of 1mA, 
costs under $2, but has slightly worse noise 
figures of en= 11nVNHz, in= 0.13pN..JHz and 
fc =150Hz. 
Given that we have control over the 
values of the resistors around the inverting 
amplifier, which will help us reduce the noise 
in that stage, we will use the TLV2460 for 
the inverting amp, and the LMP7731 for the 
summing amp. 
We can calculate the noise in the 
summing amplifier by adding up the various 
noise sources, as shown in Figure 5. At 
the inverting input of U1 we will see three 
sources of noise-the amplifier voltage noise, 
the amplifier current noise reflected across 
the source resistance, and the Johnson noise 
of the source resistance itself. The source 
resistance is simply the parallel combination 
of R1, R2 and R3, which is around 4.2kQ. 
Because the gain of the stage is 1, the output 
voltage noise will be same as the input noise. 
Remembering that we add noise as the 
root sum of squares, the total voltage noise 
density at the summing amplifier output can 
be calculated to be 12.7nVNHz. 
The calculation for the second stage is 
similar, but this time we also have to add the 
noise produced by the first stage. In this case, 
we will choose the resistors R4 and RS to each 
be 1kQ, so the source resistance is only 5000, 
thus reducing the noise in this part of the 
circuit. Again, the gain of the stage is 1, so the 
output noise will equal the input noise. Doing 
the math gives us a voltage noise density of 
17.0nVNHz. The only other contributor to 
the low-frequency path noise will be the lOQ 
resistor, which will contribute just 0.4nVNHz. 
This is negligible and can be ignored. Note 
that we can't make this simplification in the 
high-frequency path, as we shall see later. 
NOISE AT OUTPUT 
Now, to calculate the noise voltage at the 
output of the low-frequency path, we need to 
specify a bandwidth. The lower limit is DC, so 
that is easy. You might think the upper limit 
should be the -3dB bandwidth of the LC filter 
(796kHz), but the reality is that the LC filter 
is far from a brick-wall filter, so there will be 
noise energy at higher frequencies. It turns 
out that we must adjust this upper frequency 
by a factor of n/2 to account for the gradual 
roll-off. This means our upper frequency 
Working together to design 
the ideal display for your 
application. Let's get started 
on your next project. 
~~ .. ~ •T::." .. , 
l!l.. . ~ 
bound is 1.25MHz. Crunching the numbers 
gives a noise voltage of 19.01JVRMS· 
We have not, however, taken the 1/f 
noise into account at this stage. In our case, 
bandwidth of interest straddles fc so we can't 
ignore it. The 1/f noise of the TLV2460 will 
dominate, because its fc is 150Hz, compared 
to 3Hz for the LMP7731. The equation for 
calculating the noise voltage where the 
bandwidth of interest straddles fc is: 
v, = e"~(f;, - J; + fcln(J, I J;) 
Instead of just multiplying the noise 
density by the square root of bandwidth fh 
- f1 we now have an extra logarithmic term 
that accounts for the 1/f noise. For obvious 
reasons, we can't plug OHz into this equation 
for the lower limit. So, we choose a suitably 
ci rcuitcellar. corn 11 
FJGURE 5 
When calculating the noise in an 
inverting amplifier, we add the root 
sum of squares of the amplifier input 
noise voltage, the voltage produced by 
the input noise current in the source 
resistance, and the Johnson noise 
produced by the source resistance 
itself-all multiplied through the op 
amp st:lge gain. 
12. CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE6 
The finished PCB assembly. The input 
is at the top, with the output and 
controls at the bottom. The whole 
assembly slides into an aluminum RF 
screened case (not shown). 
low frequency, say l~Hz, and again use 
1.2SMHz for the upper limit. This gives us 
a revised noise voltage of 21.2~V which just 
tells us that the 1/f noise contribution is small 
but not negligible. 
Finally, we need to calculate the noise 
voltage due to the high-frequency path. In the 
same way as before, we need to extend the 
lower frequency cut-off from the -3dB value 
of 31.8kHz by a factor of n/2 to 20.3kHz. The 
upper bound in frequency is our design limit of 
lGHz. The only contributor to this noise is the 
0.4nV/v'Hz Johnson noise in the 10n resistor. 
Calculating this gives a voltage noise in the 
high-frequency path of 22.5~VRMs-not too 
different from that for the low-frequency path. 
If we add these two together, using the 
root of the sum of squares, of course, we get 
a final noise figure of 30.9~VRMS· The noise 
floor of a good oscilloscope will be in the 
order of lOO~V - 300~V at lmV per division. 
Adding our measurement noise to that of 
the lower end of the range will increase the 
For detailed article references and additional resources go to: 
pww.clrcultcellar.com/artlde-materlals 
RESOURCES 
oscilloscope noise floor to lOS~V, well inside 
the 10% figure we set ourselves. 
POWER SUPPLY 
The op amps we have chosen work over a 
voltage range of 1.8 to S.SV and 2.7 to 6.0V, 
respectively. The TLV2460 must be able to 
drive 20mA into son, at which point it can 
swing to ±400mV of the rails, so we need a 
minimum supply of :!:1.4V and a maximum 
of about ±2.SV. Adding up the quiescent and 
load currents of the two op amps, we reach 
a total of about 25mA. The alkaline AA cells I 
looked at had a capacity of 3,000mA-hours at 
2SmA, for a nominal life of 120 hours, which 
should be enough. 
Two cells in series for each rail would give 
a voltage range of :1:3.0 to :1:2.0V over the 
discharge life of the battery. Adding a diode 
drop to each rail will reduce this range to 
nominally :1:2.4 to :1:1.4V, which is pretty close 
to what we need. The only downside is that 
at low battery voltages, the maximum offset 
we will be able to inject will reduce to :U4V. If 
this is a problem for you, you can adjust the 
value of R2. 
The noise calculations above assumed 
a perfectly noiseless powersupply. If there 
were noise in the supply, it would appear at 
the amplifier outputs, attenuated by the op 
amp's power supply rejection ratio (PSRR). 
This is typically in the order of -100 dB at 
low frequencies, but gets a lot worse at high 
frequencies. The ops amps we chose have a 
PSRR of around -3SdB at lMHz. 
Alkaline AA battery noise data is not easy 
to find, but what I could find suggested a 
noise density in the order of 0.2nV/VHz and 
an fc of around 1kHz. After -3SdB of PSRR 
attenuation, this is less than O.OlnV/v'Hz, 
which confirms we were justified in ignoring 
this noise in our analysis discussed earlier. 
I added a battery-voltage-monitoring circuit 
U3 in Figure 1, using a BU4828G voltage 
detector (Rohm Semiconductor). When the 
supply voltage falls below 2.8V (:1:1.4V), 
LEDl lights up. This embellishment is not 
strictly necessary, but does make for a more 
practical instrument. 
CONSTRUCTION AND RESULTS 
I laid out a simple, two-layer PCB as 
shown in Figure 6. The high-frequency path 
is a straight son controlled-impedance trace, 
running down the center of the board under 
the batteries. The rest of the layout is not 
super critical. It is important, however, to 
enclose the circuitry in a well shielded case to 
prevent stray noise from being coupled into 
the circuit. I used a shielded aluminum case 
from Hammond Manufacturing. It was the 
most expensive component in the project. 
To use the probe, simply connect to your 
oscilloscope via a SOQ BNC-to-BNC cable, 
and switch to SOQ input termination. For 
the best results when using the probe, it 
should be connected to the circuit through a 
shielded cable soldered directly to the circuit 
in question. Terminate the braid close to the 
core (across a bypass capacitor is a good 
place). Don't use a passive probe for critical 
measurements. 
So, how did it work out in practice? 
Real-life components have tolerances and 
parasitics, and the physica l construct ion 
also inf luences their actual performance. 
Measuring the gain over the 8 decades of 
frequency shows a fairly f lat r esponse from 
DC to just under l OOMHz, which is the limit 
of my measu rement capabilities (Figure 7). 
There is a dip centered on 160kHz, which 
tells us that the matching of the damping 
resistors to the capacitive and inductive 
impedances is not perfect. This error is 
less than ldB at its worst point, just on the 
design target. Given that I used standard 
components, this is not a bad result. 
The noise f loor of my 200MHz oscilloscope 
in the lmV-per-division range is around 
1551.JVRMS• and adding the power rail probe 
increases this to around 200~VRMs-noticeably 
worse than my calculations suggest, but 
not catastrophic. If I switch-in the 20MHz 
bandwidth limit, the noise floor drops to 
96~VRMS• and with the probe connected, the 
noise increases to 108~VRMS• which is much 
more in line with what I expected. 
This suggests the noise is mostly in the 
20-200MHz range, and most likely relates to 
the imper fec t shield ing of the enclosure. I 
might try to address this in a future itera tion 
of the design, but to be frank, I am happy 
enough with this result, given the type of 
work I do, and the fact that total cost of par ts 
is well under $100. 
circuitcellar.com 13 
FIGURE 7 
Measured ga 1n o~~er tre frequency 
range I could measure With my setup. 
Tre gain is nat to Within !dB. Tre 
slight d1p at tre resonant frequency of 
around 160kHz IS due to the non-ideal 
nature of real-life components. 
Portable trackers offer a very particular set of 
problems for designers. In this article, Geoff 
takes a quick look at the design requirements of a 
portable tracking device, the choice of embedded 
antennas and some of the points to consider when 
designing a product where the performance of the 
antenna in situ could determine the commercial 
success of the product. 
D es1gmng portable trackers means facing some unique challenges, including selecting 
the right antenna solution. By 
their very nature, it is difficult to design for 
Global Navigation Satellite System (GNSS) 
frequency bands. Particularly in small devices, 
where a whole host of issues can arise during 
development. Here, we consider the factors 
that make a successful design for a tracking 
device. 
Portable trackers need to be small enough 
to be easily carried around and are often 
used in applications where they are attached 
to another item, or they may be used as a 
portable or wearable device for a living being 
such as a pet, a child or an adult person. 
This means that the design needs to be neat 
and discrete, because the whole idea of the 
tracker is to provide an inobtrusive way to 
monitor location. This means that the device 
is likely to be small, and the antenna that 
drives it should also be small (Figure 1). If the 
antenna is to be embedded into the PCB, its 
location needs to be planned right from the 
very earliest stages of product design. 
There are plenty of antenna options 
available, including choices of network: LTE, 
Bluetooth, WLAN and so on, and the satellite 
options of GPS, GNSS which can pinpoint the 
location for navigation, mapping, gee-tracking 
and fitness trackers. For a design example 
using a Bluetooth radio, see the sidebar "Design 
Example: Smart Bike Lock" on p.17. 
In selecting an antenna, you will need to 
consider the ground plane requirements and 
proximity loading of lossy bodies (such as 
the human extremities), to be sure that the 
design allows enough space for the antenna to 
radiate effectively. You may want to consider 
the different forms of antenna as well, which 
could be a flexible printed circuit, ceramic or 
an FR4 surface mount chip antenna. 
ANTENNA DIRECTIONALITY 
The antenna within a portable device needs 
to be as omni-directional as possible, so that 
it can send and receive signals from anywhere 
in its environment (Figure 2). This might pose 
a difficulty because it means that the RF 
design will need to mitigate interference that 
could affect the antenna from any direction. 
Choosing the right antenna and arranging 
the components correctly within the device 
becomes particularly important. We have 
sometimes seen trackers where the RF 
functionality has been compromised by the 
operating environment, so it is important 
to design the RF circuity correctly and carry 
out tests to be sure how the final device will 
perform. 
Portable trackers are generally small, so 
the space on the PCB will be limited and the 
antenna may be placed very close to other 
components. Components such as batteries 
and power regulators are known for creating 
electrical noise, which could interfere with the 
antenna signal. Some antennas are designed 
to support eo-location with noisy components 
and other antennas, but others require strict 
"keep out" areas, which will be specified in the 
datasheet for the antenna. Designers need to 
be very aware of how the layout of the other 
components will affect signal transmission and 
reception. One way to be sure is to submit your 
Gerber file for verification by an RF expert, 
before proceeding too far with the design. 
There are also human factors to consider. 
Some trackers are used to track goods, such 
as shipping containers or pieces of luggage. 
However, if the device is a fitness tracker or 
one that is going to be used to track pets or 
livestock, it will be placed directly on the body. 
The challenge is that the body reflects and 
absorbs radio signals. This could potentially 
prevent the device from operating in the real 
world, even if it performs perfectly in a test 
chamber. 
Designers therefore need to consider 
the environment where the tracking 
application will be used and create their 
designs accordingly. Here are some effective 
solutions: some designs use antennas with 
ceramic housing to isolate interference, or 
some even use flexible antennas that can be 
inserted within textiles and clothing. These 
can be effective waysto ensure that a tracker 
functions correctly when placed against the 
body it will be tracking. 
RANGE AND POWER 
Consider the effective operational range 
for the device you are building. A tracker 
should be able to send and receive signals 
continuously as it moves around. If the device 
can easily be moved out of range, it has failed 
in its mission. It is not easy to guarantee the 
range for the antenna within a small tracker, 
particularly when it is integrated alongside 
other components, as discussed earlier. Real-
life use scenarios can be difficult to predict 
and may reveal oversights in otherwise 
theoretically "correct" solutions. The best way 
to overcome this is to use an extensive range 
of testing scenarios to discover the realistic 
operational range of the tracker. 
For an antenna to maintain a reliable signal 
in large, noisy environments, it will inevitably 
draw more power than devices operating in 
smaller controlled spaces. Also, the volume 
and frequency of the data transmitted has a 
direct bearing on battery life. 
The battery needs to be large enough 
to keep a device operating for a reasonable 
period of time, and there should be a clean 
way to replace or recharge it. If the means of 
providing power are not user-friendly, it can be 
a frustrating user experience. Nobody wants to 
have to remove and charge their dog's collar 
every night when all they were hoping for was 
ci rcuitcellar. corn 15 
Wrist mounted we;moble exoomole ror Bluetooth aoolicatlon 
an easy way to keep track of their new puppy. 
One good way to help keep a tracking device 
charged is to make it compatible with one of 
the wireless induction chargers. 
CERAMIC OR CHIP ANTENNA? 
The latest tiny chip antennas for GNSS allow 
even small devices to receive signals from 
over 20,000km away. In RF terms, they have 
taken strides pastthe traditional ceramic patch 
antenna. A ceramic patch antenna will usually 
provide superior performance in open space, 
with plenty of ground plane, however, they 
require a large antenna footprint. They are 
also relatively cheap, but it is unlikely that your 
design brief will be able to dedicate enough 
space to accommodate one of these, and the 
smaller ceramic patch antennas are limited by 
their efficiency and rarely collect enough RF 
energy to perform as well as the larger ones. 
This means that smaller patches measuring 
less than 17mm x 17mm will typically be 
unable to achieve omni-directional gain. This 
will limit their performance in devices that are 
not operating parallel to the source radiation, 
in this case, the sky. They are also moderately 
difficult to amplify without allowing signal 
1 
5 
FIGURE 1 
Wearable devices with small PCBs 
have limited space to provide 
clearance between the antenna and 
other components. 
----------..... _ ---... 
3 6 
2 
FIGURE2 
A flexible antenna for GNSS showing antenna radiation in all six spatial directions 
16 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE3 
The Sinica SMD (surface mount device) 
antenna fur 1559-1609 MHz measures 
7.0mm x S.Bmm x 0.4mm. It is a low-
prufi le antenna with the performance 
of a ceramic patch. 
FIGURE4 
M20047 is a 1.584GHz general purpose 
integrated trace RF antenna. 
FIGURE 5 
This shows the antenna keep 
out area on the PCB. 
ABOUT THE AUTHOR 
noise to impact the performance of the device. 
In small tracker designs, a surface-mounted 
antenna will have a smaller footprint, and will 
usually outperform a ceramic patch at a lower 
cost (Figure 3). 
Ceramic patch antennas need line of sight 
to the sky, which means they need to be 
pointed upward wherever possible. Most cars 
and drones operate parallel to the horizon so 
this may not be an issue in those use cases, 
Geoff Schulteis is RF Antenna Application Specialist with Antenova, Americas. 
For more informat ion about antennas and their integ rat ion, please see 
www.antenova.com. 
For detailed article references and additional resources go to: 
-w.circuitcellar.com/article-materials 
RESOURCES 
Antenova I www.antenova.com 
but it does limit their usefulness for trackers, 
wearables, handheld and other devices that 
require an antenna with low directivity to 
perform well for a quick time-to-first-fix. 
There are many satellite frequencies 
within GNSS and many devices support 
multiple systems, but it becomes difficult 
to achieve sufficient performance across 
varying wavelengths with small ceramic patch 
antennas. If the device is smaller than 25mm2, 
ceramic patches will only effectively operate 
on narrow frequency bands. In contrast, a 
surface mount antenna can work effectively 
on wide frequency bands. This may not pose 
a design issue. If the device is to operate 
parallel to the horizon, on a single frequency, 
a small ceramic patch antenna may provide 
sufficient performance. 
ANTENNA GROUND PLANE 
Due to the size of the satellite wavelengths, 
an embedded patch antenna will require 
a greater than normal ground plane size, 
typically upward of 60mm x 60mm. Surface-
mount antennas allow more flexibility, and 
some overcome this by operating on the 
corner of the PCB, whereas the ceramic 
patch antenna must be placed in the center 
of the ground plane for omni-directionality. 
It is important to select the most appropriate 
antenna for your design and minimize the 
challenges of RF integration. 
The ground plan functions like a runway 
for transmissions to and from the antenna, 
reflecting signals into and away from the 
antenna to minimize interference and optimize 
the radiation pattern. For fixed antennas, 
the Earth's surface functions as the ground 
plane, reflecting radio waves into and away 
from the antenna. For portable antennas, the 
ground plane must be part of the PCB itself. 
This can become a real issue when designing 
small GNSS tracking devices. Space confines 
and proximity to other components can 
compromise the efficiency of a ground plane. 
Figure 4 shows the M20047, a 1.5B4GHz 
general purpose integrated trace RF antenna. 
In order for the ceramic antenna to be 
effective, a ground plane needs to have a 
radius of at least a quarter wavelength of 
the radio waves the antenna is tuned to. For 
GNSS and GPS, which broadcast at either 
1575.42MHz or 1227.60MHz, this means a 
ground plane with a radius of around Scm. 
This space needs to be planned in your design 
early on. It can be problematic if it is not 
considered early in the design process. 
The antenna uses the host PCB ground to 
effectively radiate. As such, a GND plane must 
be placed on the host PCB on at least two 
layers. In the example shown in Figure 5, the 
only area void of GND is the antenna keep-out 
area. The solder mask is removed to make the 
copper visible. 
If it proves tricky to allow sufficient space 
for the antenna and its ground plane, it 
may help to consider the options of surface-
mounted antennas and flexible printed circuit 
(FPQ antennas. It is crucial that the layout of 
your PCB does not affect the ground plane. 
This is a particular issue in GNSS tracking 
devices, where space is at a premium and 
components need to be close to one another. 
REAL-WORLD TESTING 
The ground plane will allow effective signal 
transmission and minimize interference, but 
the antenna may still receive reflections from 
glass and buildings in the surrounding area, 
which could compromise functionality. 
Portable tracking devices are particularly 
vulnerable to this. They may be carried into 
all sorts of environments, including potentially 
hostile ones, but they still need to be capable 
of transmitting and receiving reliably, so we 
would recommend that designers keep this in 
mind and plan a test program accordingly. G 
ci rcuitcellar. corn 17 
Design Example: Smart Bike Lock 
The antenna choice is crucial in the design of the bike lock because 
the Bluetooth low Energy (BLE) radio is the only way to communicate 
with the smart phone a pp that monitors the lock. In a real-world design 
alongthose lines, the design required the Bluetooth signal to be as 
strong as possible, to maintain the connection between the bike and 
the smart phone so that the users can monitor the security of their bike 
from a distance. 
BLE solutions typically use a chip antenna eo-located on a radio 
module, but this may not be a reliable solution. Depending upon how 
the antenna is integrated, it may work next to a plastic surface but not 
on a metal one. The challenge was to work within the small dimensions 
of the host PCB and accommodate the antenna in close proximity to the 
lock mechanism and the bike itself. The antenna was shielded below the 
electronics, but the integration did not allow a eo-located antenna on 
the BLE radio module. 
The antenna was therefore located away from sources of interference 
on the main board, where it could be protected from knocks and bumps 
during day-to-day use. For the design, the engineers chose Antenova's 
Weii ceramic antenna, which measures just l.Omm x O.Smm x O.Smm, 
with the matching network tuned for the application. This achieves good 
efficiency-good enough to operate at a range of a few hundred feet. 
It uses an antenna topology known as magnetic-dipole, which remains 
very stable and on-frequency in this application. 
The NEW PicoScope® 6000E Series 
Smarter scopes for faster debug r~, 
-,_ 
•' • e: · t • • 
By 
Raul Alvarez-Torrico 
Over the past year, Raul has written several articles about drone 
design. Here, he offers a product review of CUAV Tech's VS Plus drone 
flight controller and its associated NEO V2 GNSS module. Raul dives into 
specifications and functions of these devices in great detail, then shares 
the results of some flight tests using the products. 
T his is a product review article for the VS Plus drone flight controller and the NEO V2 Global Navigation 
Satellite System (GNSS) module-
both manufactured and distributed by CUAV 
Tech. The VS Plus flight controller is promoted 
by its manufacturer as an advanced drone 
flight controller aimed at academic research 
and commercial applications. It is based on 
the Pixhawk FMUvS open hardware design 
standard, and is modular, consisting of a core 
plus a separate carrier board. It is compatible 
with PX4 and ArduPilot firmware, supporting 
a number of vehicle types and configurations. 
The accompanying NEO V2 GNSS module 
can concurrently process signals from GPS, 
Galileo, GLONASS and BeiDou global satellite 
positioning systems, and includes a digital 
compass. 
In this review, I begin by describing 
the flight controller and GNSS module's 
specifications and functionality in detail. Next, 
I some introductory concepts about 
Hardware-in-the-Loop (HITL) simulation, 
to put in context some HITL tests I carried 
out with the VS Plus. I also flight log 
analysis review as a tool to assess aircraft 
performance. Finally, I the results of 
some field flight tests I made with the VS Plus 
and the NEO V2 GNSS module, both installed 
in a test quadrotor. 
I received the VS Plus flight controller and the 
NEO V2 GNSS module from CUAV Tech for free, 
to test and evaluate. This review reflects my 
personal experiences and opinions about these 
products, in the context of the type of drone 
applications I normally do as part of my day 
job: advising undergraduate college students 
with drone autonomous flight projects. 
TECHNICAL SPECIFICATIONS 
CUAV Tech is a Chinese company working 
since 2012 in research and development, 
production and salesofunmanned autonomous 
vehicle (UAV) applications and system modules, 
such as flight controllers, digital radio link 
systems and global positioning devices. One 
of their most recent products is the VS Plus 
flight controller, designed in collaboration 
with the PX4 autopilot team. CUAV Tech is 
a "silver member" of Dronecode-the non-
profit organization governed by the linux 
Foundation that oversees the development 
of four open-source projects in the PX4 
ecosystem: PX4 autopilot, MAVlink, MAVSDK 
and QGroundControl. 
The VS Plus is based on the Pixhawk 
FMUvS (Flight Management Unit vS) open 
hardware design standard, and uses Pixhawk 
standard pinouts to interface with external 
devices. It comes out of the box pre-installed 
with PX4 firmware, but is also compatible 
with ArduPilot firmware. Figure 1 shows the 
VS Plus core module (in silver) attached to its 
carrier board (in black). The detachable core 
contains the flight controller's main hardware 
components (microcontrollers (MCUs) and 
onboard sensors), and the separate carrier 
board contains the interfacing connectors to 
external devices. 
That modularity gives you the choice of 
either using the core with one of the of-the-shelf 
carrier boards from CUAV Tech, or designing a 
custom one. CUAV Tech made available an open 
carrier board reference design that can be used 
as a starting point to design a custom one. 
The main processor in the core module is the 
STMicroelectronics STM32F76S MCU that runs 
at 216MHz and contains 2MB of flash memory 
and 512KB of RAM. This improves computation 
power in comparison with previous flight 
controllers, such as the Pixhawk V3x, also made 
by CUAV Tech. 
The core module implements multi-sensor 
redundancy by combining five sets of sensors. 
It has two TDK InvenSense integrated 
circuits-the ICM-20689 and the ICM-20602-
each of which contain both a 3D accelerometer 
and a 3D gyroscope. Additionally, it has a 
Bosch Sensortec BM1055, which also contains 
a 3D accelerometer and 3D gyroscope. An 
iSentek Technology IST8310 3D compass and 
a MEAS Switzerland MS5611 atmospheric 
pressure sensor (altimeter) complete the set 
of onboard sensors (the MSS611 is sold by 
TE Connectivity). The flight controller reads 
the multi-channel sensor data in real time, 
and switches to a redundant sensor whenever 
there's a failure in the previous one, thus 
improving safety and reliability. 
Figure 2 summarizes the hardware 
specifications for the VS Plus flight controller. 
Additional key characteristics worth 
mentioning are: a secondary input/output 
processor (the STMicroelectronics STM32F100 
MCU); two battery monitoring systems; 
support for high-accuracy, centimeter-level, 
Real-Time-Kinematic (RTK) positioning, 
with the use of corresponding modules and 
support for LTE telemetry and video links-
which provide a much wider coverage area 
than conventional ones. 
The core module has a built-in damping 
system that separates the sensors from the 
main board and provides high-performance 
shock absorption. This reduces signal noise 
in the sensors due to mechanical vibration, 
making unnecessary the use of external anti-
vibration systems (Figure 3). The VS Plus fully 
adheres to the PX4 official standard, to ensure 
compatibility and interoperability with other 
hardware and software products in the PX4 
ecosystem. Available VS Plus carrier boards 
are also fully compatible with the new X7 
flight controller cores announced by CUAV 
Tech in May, 2020. 
The NEO V2 GNSS module has a U-blox 
MBN, 72-channel GNSS receiver and an 
iSentek IST8310 digital compass (Figure 4). 
The MBN sub-module is a dual-mode parallel 
GNSS receiver than can concurrently process 
signals from GPS/Galileo and GLONASS 
FJGURE 1 
ci rcuitcellar. corn 19 
- VS Plus core and standard carrier boand 
VS Plus Core 
FJGURE2 
Main FMU Processor: STM32F765 
32 Bit Arm• Cortex•·M7, 216MHz, 2MB FLASH, 512KB RAM 
10 Processor: STM32Fl00 
32 Bit Arm• Cortex•·M3, 24MHz, 8KB SRAM 
On-board sensors: 
3 x Accei/Gyro: ICM-20689, ICM-20602, BM lOSS 
1 x Magnetometer: IST8310 
1 x Barometer: MS5611 
Interlaces: 
8·14 PWM outputs (6 rrom 10. 8 from FMU) 
3 dedicated PWM/Capture Inputs on FMU 
Dedicated R/C Input for CPPM 
Dedicated R/C input for PPM and S.Bus 
Analog / PWM RSSI input 
S.Bus servo output 
5 general purpose serial ports 
4 12C ports 
4SPI buses 
2 CAN buses with serial ESC 
2 battery monitoring inputs 
Power System: 
Power: 4.3-5.4V 
U5B Input. 4.75-s.25V 
Servo Rail Input: Q-36 V 
Welcht and Dime nsions: 
Weight: 90g 
Dimensions: 85.5"42•33 mm 
Other Characteristics: 
Operating temperature: -20- 80 •c 
VS Plus summary of specifications 
ZO CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE3 
VS Plus built-in damping system 
FIGURE4 
NEO V2 GNSS receiver and compass 
module 
FIGURES 
VS Plus Hffi simulation running 
or GPS/Galileo and BeiDou global satellite 
positioning systems. This feature ensures 
much better positioning accuracy in urban 
canyons or any environment with weak GNSS 
signal reception. It has an update rate of 10Hz 
(maximum), a cold start acquisition time of 
26 seconds and a hot start acquisition time of 
1.5 seconds. 
The NEO V2 module incorporates a safety 
switch, a buzzer and RGB status lights, 
otherwise separately added to the drone. I 
received from CUAV Tech a kit containing the 
VS Plus core and standard carrier board, the 
NEO V2 module, an HV_PM power module, 
a PW-Link Wi-Fi telemetry module, an IlC 
expansion board, a USB to UART converter 
and some additional accessories. 
HITL SIMULATION 
Hardware-in-the-loop (HITL or HIL) is a 
simulation mode in which a real hardware flight 
controller runs regular PX4 firmware to control 
a computer modeled vehicle inside a simulated 
environment. In contrast, Software-in-the-
Loop simulation (SITL) is the mode in which 
the PX4 flight stack is run typically on the same 
computer on which both the modeled vehicle 
and virtual world are running. The benefit of 
HITL over SITL simulation is that HITL allows 
testing most of the actual PX4 autopilot flight 
code on the real hardware. PX4 supports HITL 
simulation for multi-rotor drones with jMAVSim 
and Gazebo simulators. 
SITL and HITL simulation are great tools, 
not just for development, but also for teaching 
and learning. For instance, they are useful for 
autonomous drone application development. 
You ea n write, test and debug your autonomous 
flight code right in front of your development 
computer, without needing to go outside with 
a real drone until it is absolutely necessary. 
This is particularly helpful when you integrate 
other technologies-such as computer vision 
and cloud data connections-to your drone 
application, and you need to do indoor tests 
simultaneously with your autonomous flight 
code. 
I tested the VS Plus in HITL simulation by 
running some MAVSDK-Python autonomous 
flight code examples I had on hand. For 
example, it ran well with the write_my_initials. 
py program example from my previous article 
uwriting MAVSDK/PX4 Drone Applications" 
(Circuit Cellar 361, August 2020) [1]. Figure 5 
shows a screen capture of the HITL simulation 
running. As I sometimes do with new drone 
builds, I also made a couple of HITL tests 
I• 
---
-· --
_____ ,.. 
_....,. ..... ~ .. ---
flying the simulated drone with the real radio 
control system connected to the hardware 
flight controller, and it worked well. 
In my previous article [1], I discussed 
the installation procedure of the PX4 SITL 
simulation environment for developing 
autonomous flight applications with 
the MAVSDK-Python library. The same 
environment works for HITL simulation. 
If you are interested in autonomous flight 
application development, check that article 
for a quick introduction to the subject. Once 
you have the simulation environment correctly 
installed, the procedure for running an HITL 
simulation is as easy as for SITL when using 
the jMAVSim simulator. 
It is somewhat more involved for HITL with 
--· 
.. -. .. _.._ 
-..---·---
--
l x:M 
__ , 
P-lt'""' : ~~ "";""--- l 
1:- ....,. 
_.J ~ 
-·- --
ci rcuitcellar. corn 21 
FIGURE6 
Plot examples from a flight log review 
22 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE 7 
Test quadrotor with the VS Plus and 
NEO V2 module 
the Gazebo simulator. For HITL simulation 
to work, you typically connect the flight 
controller with the USB cable to the computer 
running the simulation, so the flight controller 
can connect to the virtual environment. You 
can use a stand-alone flight controller with 
nothing else connected to it, or one in a finished 
drone build. The file run_hitl_simulation.md 
included in this article's materials details the 
steps for running HITL with both jMAVSim 
and Gazebo. The file is available via a link on 
Circuit Cellar's article materials webpage or 
via down load from Circuit Cellar's article code 
and files page. If you've tried SITL before, 
running HITL is almost the same. 
FLIGHT LOG ANALYSIS 
Pixhawk flight controllers store flight 
logs on a Micro SD card. These flight logs 
can be later analyzed to assess the drone's 
performance and help debug issues. The 
flight logs are stored in Ulog format and 
contain data from sensors, actuators, control 
algorithms, the radio control system, power 
system, processor and memory and the data 
logging system itself. Some examples of data 
stored in log files are: altitude, roll/pitch/ 
yaw angles and angular rate, local position 
(X, Y, Z) and velocity, manual radio control 
inputs, actuator control and outputs, raw 
acceleration, angular speed and magnetic field 
strength, vibration level, GNSS uncertainty, 
noise and jamming, thrust and magnetic field 
(for correlation), power system, health report 
(estimator watchdog), RC quality, CPU and 
RAM, sampling regularity of sensor data and 
logged messages (warnings, errors). 
When a flight goes wrong, log analysis is 
a great tool that can potentially help us find 
issues in the aircraft build quality, as well as 
hardware failures, flight controller wrong 
parameter configuration or even issues with our 
piloting skills. But even if nothing goes wrong 
and your aircraft flies well, log analysis can help 
you assess how well your aircraft is performing. 
Figure 6 shows plot exam pies of data logged 
by a Pixhawk flight controller. Of the many tools 
for log analysis, one of the most commonly 
used is Flight Review. This is a free online tool 
available on the cloud that doesn't need local 
installation and makes it easy to analyze flight 
logs and share the results over the web. You just 
need to go to the cloud application web address 
ps: rev1ew. x .10, pload your flight log, 
analyze the data and, if you want, share the 
link with others. Flight logs are also handy 
when asking for help in the official PX4 support 
channels. Other flight log analysis applications 
include pyulog, pyflightAnalysis, FlightPiot, 
PX4Tools, MAVGCL and Plotluggler [2]. 
Understanding flight log plots isn't 
difficult, if you have basic knowledge of how 
a drone works and what roles are played by 
sensors and actuators in the system, along 
with a basic understanding of the main control 
algorithms involved, such as PID. But even if 
you know how to interpret the data graphs, 
it will take a while until you understand what 
values are acceptable in each case, and which 
plots are better to look at when dealing with 
specific problems, such as frame mechanical 
vibration. The PX4 site has good introductory 
information on how to analyze flight log plots 
and identify common problems. On the Circuit 
Cellar article materials webpage for this 
article, I've provided links to PX4's website 
pages containing more detailed information 
about flight reporting and flight log analysis. 
FIELD FLIGHT TESTS 
Figure 7 shows the quad rotor drone I used 
with the VS Plus for my flight tests in the field. 
It has the following hardware: 
• An SSOO frame 
• Generic 2212/920KV brushless motors 
• DYS 30A 2-SS Electronic Speed Controllers 
(ESCs) 
• DJI 9450 and Gemfan 1045 propellers 
• A generic 5,000mA-hours, 3S, 30C Lithium 
Polymer (LiPo) battery 
• A FlySky FS-I6S radio control transmitter 
paired to a FS-IA10B 10-channel receiver 
• A PW-link Wi-Fi telemetry module 
• The VS Plus flight controller with the NEO V2 
GNSS module 
I also made a couple of flights with another 
quadrotor with the same type of hardware, 
except for a DJI F450 frame. 
For my firstflight tests with the VS Plus, I 
used the HV_PM (High-Voltage Power Module) 
that was included in the kit I received from 
CUAV Tech. This power module supports 
3S-14S batteries and voltages between 10-
60V. However, from the very beginning I had 
an interesting issue with this power module. 
After take-off of the quadrotor, the "low-
battery" warning was being triggered in 
QGroundControl less than a minute into flight, 
even though the batteries were fully charged. 
Looking at QGroundControl's GUI, I noticed 
the "battery voltage" indicator was dropping 
from 12.6V to 10V in just 3-4 seconds, and 
the "remaining-battery" indicator fell from 
90% to 10% in just 8-10 seconds, making 
QGroundControl emit audible "low-battery" 
warnings. I tested the quadrotor with three 
different batteries, which I knew were fine 
because I had used them previously with 
other quadrotors of similar configuration 
without any issues-and got the same results. 
Additionally, I had an external LiPo battery 
alarm connected, which never got triggered 
by the apparent low-voltage condition. I was 
flying with PX4 firmware, but I also changed 
to ArduPilot to check if the issue was specific 
to PX4. I got the same instant "low-battery" 
warnings with ArduPilot as well. 
USER GROUP HELP 
The CUAV Tech user group channel at PX4's 
Slack workspace helped greatly to determine 
the cause of the issue. The first hint I was 
given was that perhaps the power cable 
connecting the battery to the power module 
was too long. And it was-around 15cm in 
length-because of an adapter I was using to 
connect my T-connector batteries to the XT60 
connector in the HV_PM power module. From 
my own flight logs, it also was brought to my 
attention that apparently, the power cable 
length was causing a considerable voltage 
drop, which in turn was causing more current 
to be sourced from the battery. 
Referring to the gray curve in Figure 8, 
you'll notice that the current sourced when 
the quadrotor is hovering is around 22A on 
average-with a peak as high as 52A! And 
the measured battery voltage (the dark-
blue curve) drops from 12V to lOV almost 
Power 
50 -
40 
30 -
instantly. Of course, in reality, the battery 
wasn't getting discharged that fast, otherwise 
my external LiPo alarm would have been 
triggered as well. It was just that, for some 
reason, the flight controller was apparently 
getting erroneous readings from the current 
and voltage sensors in the power module. 
I got rid of the adapter cable and did a flight 
test again. The flight time improved to around 
4-6 minutes, before the warning signal sounded 
again. It then appeared that the cause for the 
early warnings was that the voltage and current 
sensors in the HV_PM power module were not 
working well with the 35 batteries. I was also 
told that, apparently for a 35 LiPo battery, a 
current of around 15A was normally expected 
for the kind of quadrotor configuration I 
had, instead of the 22A I was getting in the 
flight logs. From past logs from other similar 
quadrotors, I found that currents between 13A 
to 15A were typical. 
It was suggested that I try using a LiPo 
battery with more cells, so I did tests with 
45 batteries; the drone flew well but with a 
bit of oscillation. Perhaps it would have been 
fixed by tweaking the default PID calibration, 
but I didn't go that route. I always tend to use 
35 batteries with quadrotors of this size and 
configuration, anyway. 
Instead, I replaced the HV_PM power module 
with a regular 25-85 Pixhawk 1 power module, 
and the flight time increased easily up to 8-10 
minutes with the default battery configuration 
in the flight controller. This default configuration 
causes the flight controller to do conservative 
estimates of the remaining battery power and 
available flight time. So, after tweaking these 
battery parameters a bit, the flight time rose a 
little more. 
In retrospect, perhaps the HV_PM module 
could work well with higher-voltage batteries, 
but I didn't test it with more than 45 ones. 
I had never previously encountered any issue 
like this before, and it took me some days to 
understand what was happening. As you know, 
- Battery voltage [V] 
- Battery voltage filtered [V) 
- Battery current [A] 
- Discharged amount [mAh / 1001 
- Battery remaining (0 =empty, 10 =full) 
SV 
3.3 V 
20 "' .A. }\ ,.....-"'\ /\.... /\ I \ ~ 
V 
l ....... = / ~ ....... V - -
10 
0 
2:10 2:15 2:20 2:25 2:30 2:35 2:40 
ci rcuitcellar. corn 23 
FIGURES 
V5 Plus "power" plot from a flight log 
review. lh is graph is representEd 
here in the same fonmat that it's in 
the PX4 cloud application. But some 
notes about the axis units are needed . 
The Y axis displays both units of 
Voltage M and Current (A) . The X axis 
shows time in Minures: Seconds. 
Z4 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE9 
VS plus drone fiight test 
failure is a master that treats roughly but 
teaches well. Figure 9 shows the quadrotor 
during a flight test. 
CAN PDB CARRIER BOARD 
Figure 10 shows an optional carrier board 
for the VS Plus core, also made by CUAV Tech. 
The CAN PDB carrier board is a combination of 
a baseboard for connecting external devices, 
a Power Distribution Board (PDB) and a Power 
Management Unit (PMU). The PMU part of the 
CAN PDB has a Universal Battery Elimination 
Circuit or UBEC (which is really just a voltage 
regulator) based on an Analog Devices LT86455 
65V, 8A step-down regulator that provides a SV, 
8A output, and a Texas Instruments TP554561 
60V, SA step-down regulator that provides a 
12V, 4A output. Both outputs are convenient 
to power external devices, such as off-board 
sensors and actuators, cameras, wireless 
transceivers and companion computers. 
Because of their specifications, they are even 
more convenient for external devices that 
consume relatively higher power. 
For detailed article references and additional resources go to: 
-w.circuitcellar.com/article-materials 
References [1] and [3] as marked in the article can be found there. 
RESOURCES 
Analog Devices I www.analog.com 
Ardupilot I https:f/ardupilot.org 
Bosch Sensortec I www.bosch-sensortec.com 
CUAV Tech I www.cuav.net 
Dronecode I www.dronecode.org 
iSentek Technology I www. isentek.com 
PX4 Autopilot I https://px4.io 
STMicroelectronics I www.st.com 
TE Connectivity I www.te.com 
Texas Instruments I www.ti.com 
The PMU has its own dedicated MCU, 
a 100MHz 5TMicroelectronics 5TM32F412 
running CUAV Tech's Impedance Temperature 
Tracking {ITT) algorithm. According to the 
manufacturer, it accurately measures voltage 
and current in real time, and uses CAN bus 
communication and the UAV CAN protocol to 
send power consumption data to the core. 
The manufacturer says its ITT algorithm 
corrects measurement errors caused by 
impedance changes at different operating 
temperatures, to maintain consistent high-
precision voltage and current measurements. 
By adopting the UAV CAN dig ita I communication 
protocol, attenuation and interference errors 
are avoided. These can be present in regular 
power modules that feed the flight controller 
with voltage and current measurements in 
analog format. To accurately calculate power 
consumption, the PMU's MCU samples voltage 
and current at a rate of more than 100Hz. 
These data are then passed to the flight 
controller. The PMU can accurately measure 
currents between 0-180A. 
CAN PDB supports input voltages from 
lSV to 62V, which roughly correspond to LiPo 
batteries from 45 up to 1S5, and the power 
distribution board supports up to 180A of 
continuous current [3]. The baseboard has 13 
output channels and several connectors for 
telemetry, radio control input, GNSS module, 
ADC, I2C, CAN, UART, and the SV and 12V 
high-current outputs, among other things. 
The PDB power lines are built with 95cm2 
of copper sheeting, which also acts as a 
heatsink, helping reduce heat caused by the 
lines' internal resistance when high currents 
circulate. The CAN PDB carrierboard is 
compatible with the VS Plus core and the new 
X7 flight controller cores from CUAV Tech. 
Because of their voltage and current specs, 
CAN PDB can be suitable for big drones 
carrying high payloads. CUAV Tech also sent 
me a CAN PDB carrier board to test, but 
I wasn't able to do a flight test because of 
global health emergency restrictions due to 
the COVID-19 pandemic. 
CONCLUSIONS AND FURTHER 
TESTS 
The kit I received came well packed in a 
fine box. From an aesthetic standpoint, the 
VS Plus core, the standard carrier board and 
CAN PDB carrier board, and the NEO V2 GNSS 
module seem well crafted and robust, with a 
nice-looking finish. When powered, both the 
flight controller core and GNSS module emit 
an elegant, halo-like status light surrounding 
their bodies. 
After changing the original power module, 
I had no further "low battery" sudden 
warnings, and the flight controller performed 
well in the next flight tests. The sensor's 
triple-redundancy gives confidence such that, 
in the event of an accelerometer or gyroscope 
failure, there are still two additional redundant 
ones the flight controller can use to maintain 
control of the aircraft. This highly reduces the 
probability of a crash due to multiple-sensor 
failure. This kind of reliability is important, 
for instance when your aircraft carries 
expensive remote sensing equipment, such as 
professional video cameras, high-end LIDARs 
or high-resolution thermal cameras. Likewise, 
it is important when the drone has been given 
a highly sensitive task and can't afford a 
failure that would lead to a subsequent crash. 
Currently I'm planning to test the VS Plus 
with some autonomous flight applications that 
integrate computer vision for object detection 
and tracking. I tested the VS Plus mainly with 
PX4, because it's the platform I generally use 
for autonomous flight development, but I 
flashed it with ArduPilot firmware as well and 
made a couple of flight tests. The configuration 
process is equally easy with both platforms. 
With a price tag of around $3S8 for the 
complete kit (flight controller core plus standard 
carrier board, GNSS module, power module, 
Wifi telemetry module and accessories), the 
VS Plus is not cheap. But whether you work with 
drones professionally, or you are a drone 
enthusiast always eager to try new things, 
perhaps you should take a serious look at the VS 
Plus. It could be just the type of reliable flight 
controller you are looking for. e 
ABOUT THE AUTHOR 
ci rcuitcellar. corn 25 
FIGURE 10 
CAN PDB <:ilrrier board 
Raul Alvarez-Torrico has a BEng in electronics and is the f ou nder of 
TecBolivia, a company offering services in physical comput ing and educational 
robotics in Bolivia. In his spare time, he li kes to experiment with wireless 
sensor networks, robotics and artificial intelligence. He is also committed 
to publishing articles and video tutorials about embedded systems and 
programming in his native language (Spanish), at thei r company's site 
www.TecBolivia.com. You may contact him at raul@tecbolivia.com. 
C onsider these two questions: 1. Are "footprint" and "code size" the same thing? 2. And, is either 
relevant in the current age? 
Answer 1: No. Code size is a subset of 
footprint. In the context of an in-memory 
database system, "footprint" consists of: 
• Code size 
• Stack memory used 
• DRAM used for anything other than raw data 
With respect to code size, we must have 
an all-encompassing view. We're not the only 
embedded database vendor that boasts about 
a "small footprint." That said, eXtremeDB 
from McObject might be the only embedded 
database that has no external dependencies. 
For purposes of this discussion, it means 
that using the eXtremeDB core in-memory 
database does not cause the application to 
link in the C run-time library for things like 
malloc/free, string operations and so 
forth. A claim of "small footprint" is hollow 
when the application is forced to link a 
megabyte of the C run-time library. 
Answer 2: Yes, these terms are relevant. 
Code size matters. We can ascribe two things 
to a small code size: The function call depth 
and the number of CPU instructions for any 
given operation. A core embedded database 
engine with a code size of 150KB is going to 
require fewer function calls, and each function 
call is going to require fewer CPU instructions 
than another embedded database engine 
with a code size of 1.5MB. Whether you're 
executing on a 200MHz Arm processor or a 
3GHz Intel Sky Lake processor, a database 
lockup that requires 100 CPU cycles is lOx 
faster than a database lockup that requires 
1,000 processor cycles. 
STACK SIZE 
Stack size in embedded/real-time 
operating systems is quite limited. For 
example, the default stack size with VxWorks 
is just 20KB. Unlike Windows and linux, the 
stack does not grow dynamically. If the stack 
size is exceeded, the program fails. This ties 
directly back to function call depth, but also 
influences how much, and how, information 
is passed between functions, which is a 
consideration that must be considered when 
the database system is designed. This is a key 
differentiator between an embedded database 
that was written for embedded systems, and 
an embedded database that is merely capable 
of running on embedded systems. Figure 1 
illustrates the function call stack. 
It should be obvious, but for an in-memory 
database, DRAM is storage space. Any 
memory that is used for anything other than 
raw data limits the amount of raw data that 
can be stored in any given amount of DRAM. 
In addition to storing raw data, memory is 
used for meta data (data about the data), 
indexes and database internals. 
Let's use some real numbers to Illustrate. 
Suppose that in-memory database "A11 
imposes 100% overhead on the raw data. That 
means that 20MB of DRAM will be required to 
store 10MB of raw data. In-memory database 
"B" only imposes 30% overhead, so it only 
requires 13MB to store the same 10MB of raw 
data. In an embedded system like a portable 
media player that only has 32MB of memory 
Top of stack 
r 
Local Variables 
Return Address 
Func1 (Arg1, Arg2 .. . ) --< 
ArgN 
... 
Argl 
>-
Local Variables 
Return Address 
Func2 {Arg1 , Arg2 ... ) --< 
ArgN 
... 
Arg t 
'-
Call Stack 
to begin with, that is a meaningful difference! 
eXtremeDB is an example of an in-memory 
database type "B" in this comparison. The 
relationship of code, static data, stack and 
heap is illustrated in Figure 2. 
For its part, eXtremeDB employs several 
techniques to minimize the overhead. The 
first-and biggest-influencer is index types 
and how indexes are organized. eXtremeDB 
offers a wider variety of index types than other 
database systems, but the most commonly 
used are hash and b-tree indexes. By their 
nature, hash indexes use less memory (and 
are faster) than b-tree indexes. By far, the 
most common type of index is the b-tree 
index. To understand the disadvantages of a 
b-tree you need to have at least a high-level 
understanding of their organization. 
A I-TREE INDEX 
As shown in Figure 3, a b-tree is an upside-
down tree, with the "root" node at the top. Each 
node has many slots where the number of slots 
is a function of the size of the node (in bytes) 
and the size of the "payload.u Simplistically, 
number-of-slots = node size I payload. Each slot 
contains (1) a pointer to a node with key values 
that precede the key value in this slot, (2) this 
key value, and (3) a pointer to the actual data 
record that is being indexed. 
As you can see in the illustration, 
b-trees are normally partially empty (up 
to 45% empty), so there's built-in space 
inefficiency. The key value exists in the slot 
so that database systems can implement an 
optimization called a "covered query,n which 
• Call Stack is a stack that contains 
a collection of stack frames. 
• A Stack Frame, also called 
an activation record, contains 
information about a function call.• When the program runs, the call stac k 
contains at least one stack frame, 
which is for the main function. A stack 
frame is created whenever a function 
is called. 
• Arguments are inserted into a stack 
frame from right to left. 
• A Stack Point Register (register 1) is 
used to mark t he current "top" of 
the stack . 
simply means that ifthe column(s) queries are 
only the indexed column(s), then the query 
results can be obtained by only scanning the 
key. There is no need to follow the pointer to 
the data record and the amount of logical I/0 
is reduced by one-half. But this optimization 
only makes sense for persistent databases, 
because of the speed of storage media. 
Following a pointer in memory to the 
data record takes just nanoseconds, whereas 
following a pointer to a track and sector on 
a solid-state disk can be as low as l~s, and 
generally on the order of 10 milliseconds for 
a hard disk drive. The use of precious DRAM 
to store redundant key values is unjustified 
Heap 
In-memory rl database lives 
! 
here 
t 
StackN 
Data 
Code 
ci rcuitcellar. corn 'D 
FIGURE 1 
An illustration of tl1e Call Stack function 
FIGURE Z 
This shows the relationship of the 
code, the static data, the stack and 
the heap. 
15877542
15877542
Servo control is just one area where microcontrollers 
shine. In this article, see how these two Comell 
students designed and built a ball-balancing platform 
based on PIC32 MCU, using a resistive touchscreen 
and servomotors attached to a gimbal mount. A PlO 
control algorithm is used to balance the ball. 
FIGURE2 
W hile it may seem a simple task, balancing on one foot is no small feat. It requires a 
constant cycle of observation 
and adjustment. Getting an electronic device 
to perform a similar task in just a few weeks 
is not easy and takes quite a bit of tuning-
just as it takes us time to learn to balance as 
children. 
mgsin(O) mgsln(O) 5 (B) 5 
8 a - --1- =--2- = ?IJsln .. 70 
m+RT m+~m 
Illustrated here is a free body diagram of an isolated direction of the ball on an angled plate, and an equation 
of predicted motion. 
In this article, we present a ball-balancing 
platform that uses a touchscreen and a 
Microchip PIC32 microcontroller (MCU) to 
keep a steel ball bearing stable on its surface. 
A proportional, integral, derivative (PID) 
control algorithm and two serves are used. 
This project was designed to inspire interest 
in control theory and its applications, and to 
motivate deeper thought about the physics at 
play in this system. Our end result is shown 
in Figure 1. 
PHYSICAL ANALYSIS 
A ball rolling on a perfectly flat plate is 
technically metastable, since, if there are no 
forces applied, the ball will remain static. 
Of course, this assumption is terrible, and a 
real ball sitting on a real plate is going to fall 
off unless we do something about it. A ball 
rolling on a tilted platform roughly follows a 
linear relation between angle of the platform 
and acceleration along the surface (Figure :Z). 
The no-slip condition makes this linearizing 
assumption possible, because we essentially 
assume that each rolling direction is totally 
independent. 
In this analysis, m represents the mass 
of the ball, I is the moment of inertia of a 
rotating sphere, and a is the acceleration 
of the ball parallel to the surface. At small 
angles, sin (8)=8, so we show a linear relation 
at the end between angle and acceleration. 
We tilted the platform over a range of about 
±15 degrees, or about ±0.3 radians, which 
gives a maximum error of approximately 
l.So/o due to the small angle approximation. 
Given a relatively heavy, 1" steel ball, we 
knew we needed servos that could hold their 
position with the mass of the ball (0.07kg) at 
the servo arm's maximum distance from the 
pivot (Bcm). This meant we needed serves 
that have about 0.56 kg-cm torque, which 
translates to around Boz-in. The serves we 
chose had a 4.8V stall torque of 9.4kg-cm 
or over 125oz-in. Although we planned to 
run on SV power and thought we might be 
underestimating the maximum torque on 
the motor, this was plenty strong. The main 
trade-off is between strength and precision 
of the motors. As it turns out, our serves 
were extremely strong for this task, causing 
a precision issue in the tuning stage. 
MECHANICAL DESIGN 
The main pieces of the platform were the 
base plate, pedestal, top plate, serve and pivot 
hinges, and the servo mounting brackets. By 
securing the serves to the base plate, we 
hoped that the structure was rigid enough to 
handle rapid movement. When starting this 
project, the biggest decision to make was 
to first choose a resistive touchscreen. The 
dimensions of that piece of hardware largely 
determine the rest of the design, and it is 
the most unique item on the supply list. Our 
project used a medium-sized, rectangular, 7" 
diagonal resistive touchscreen overlay from 
Adafruit [1], and we cut our top plate to match 
its dimensions (165mm x 105mm). 
From here it makes sense to do a bit of 
math to answer the following questions: How 
far off the table does the plate need to be 
to accommodate two serves and the table's 
full range of motion? How powerful does each 
serve have to be to accelerate the ball when 
it is moving quickly toward the edge? A good 
estimate can go a long way-even if it means 
the design is tweaked at the end. 
The resistive touchscreen sits on an acrylic 
plate, which is mounted to a base using a 
partially 3D-printed Cardan joint (Figure 3). 
This allows the plate to pivot around its hinge 
in two directions independently. Each direction 
is controlled by a serve motor, which is 
attached to the underside of the plate with a 
ball socket connection. 
We used 1/8" acrylic for all pieces, and 
attached them with acrylic cement. In 
our implementation, the platform had an 
approximately 32-degree range of motion in 
the short direction, and 36 degrees in the long 
direction. Placing the pivot mechanism at the 
center of the plate, while easy to build, did not 
allow for rotations outside of this range. This 
mechanism has the only 3D-printed part: a 
cross-shaped connector that joins the top plate 
and pedestal. 
We first laser cut the acrylic and glued 
together all the supports, noting the order of 
operations for attaching the pivot mechanism. 
The pedestal and cross piece were assembled 
first, and the servo joint attachment points 
(shown in red in Figure 3) were glued to 
the top plate separately. Then the center 
supports were attached to the top plate, while 
simultaneously joining them to the cross piece. 
The ball joint attachments to the serve motors 
were then easy to attach to the top plate and 
were snapped onto the servos mounted to the 
bottom plate at their minimum or maximum 
extent. To do this, we first zeroed the serves 
and then set the angle of the platform to our 
desired zero angle. 
The success of this project was heavily 
dependent on the care with which the 
FIGURE3 
CAD drawings of the acrylic pieces, fully assembled 
ci rcuitcellar. corn 31 
32. CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE4 
Full schematic, including our attilched 
hardware, of parts of the development 
board [4] created by Sean Carrell for 
ECE 4760 [5] 
mechanical integration was performed. Tightly 
fitting each joint to the servo motors, and 
making sure the pivot point didn't wobble, both 
yielded better performance in the final design. 
ELECTRICAL DESIGN 
The main electrical components include a 
Microchip PIC32 MCU [2], two serve motors, the 
resistive touchscreen and a serial connection to 
a separate computer monitor for the tuning 
step (see the schematic in Figure 4). 
The touchscreen generates a voltage 
proportional to the X (long dimension) and Y 
(short dimension) position of an input touch. 
Using Adafruit's touchscreen control chip to 
receive data proved time consuming, but 
there were several options for making the job 
easier. The touchscreen uses four pinsto set 
the voltage of strips of conductive material 
that line the outer perimeter of the screen. 
By holding a SV potential across the screen 
in one direction, the voltage read by one of 
the perpendicular pins makes the screen 
330 
essentially a potentiometer. The value read by 
that pin changes based on where you press 
on the screen. The fourth pin is left to float, 
since it must be used for the other direction's 
touch. By performing this process for one 
direction and then the other, the position of a 
point of pressure on the screen is located [3]. 
The serve control used a standard pulse-
width-modulated (PWM) signal. The Tower 
Pro MG995 took a 50Hz pulse with an on 
time between lms and 2.5ms. The serial 
connection to a computer and monitor was 
purely a convenience for tuning parameters. 
This is highly recommended to make tuning 
less time consuming. 
By watching the acceleration of the ball at 
maximum tilt, we estimated that we needed 
at least a 1kHz sampling rate for an accurate 
reading of the ball's position. Because we can 
only measure the ball's position in one direction 
at a time, this means that each direction must 
be sampled at that 1kHz frequency. We then 
update the servo position as fast as possible. 
330 330 
"' •33V a: 
U4 9 
VOD 
GPBO GPAO 
CPBl GPAl 
GPB2 GPA2 
GPBJ GPA3 
GPB4 GPM 
GPB5 GPAS 
CP!l6 GPA6 
GPB7 GPA7 
SCK 
l~ SI so 
•cs 
20 AO INTA 
19 
Al INTB 
A2 10 
' RESET vss 
MCP23Sl7 
l.JO 
~ 
~ 
_L ~~n ••0 
~ LED 
ON::. CS 
~ 
~ 
TFTCS 
J l O 
~ 
PECS 
+JJV 
PGD= RBO 
PGC• RBl 
1\da~ulr TFT LCD DosfJIIIY 
PROGRAM STRUCTURE 
The program structure runs on two threads 
and one interrupt service routine (ISR), with 
a few other helper functions. The meat of 
the functionality is in the ISR, which runs at 
2kHz and toggles between reading the X and 
Y positions. After each read, the PWM signal 
to the corresponding servo is also set. The 
two scheduled threads are for writing output 
to a screen or exchanging serial information 
with our command structure. Both of them 
can be implemented in many different ways, 
depending on what is convenient. We used 
code and explanations from Bruce Land's 
MCU design class [5] to use Adam Dunkels' 
Protothreads library [6] for the high-level 
design of our code structure. 
Our main () function sets up the analog-to-
digital converter (ADC) ports, output compare 
units (for generating PWM signals), and sets 
up our first read from the touchscreen, before 
jumping into the infinite cycle of writing to/ 
reading from the screen and the serial port. 
The ISR then handles cycling through the 
actual control loop. 
Two helper functions called rea dXp rep () 
and readYprep() set the ports that control 
the touchscreen to their appropriate values 
for a read in one direction versus another. 
Following the touchscreen specifications, one 
pin must be set to logic high, one to low and 
one is set as an input, so that its value nfloats" 
and does not affect the reading. The fourth 
pin is then read by an ADC on the PIC32. 
The PID controller is a feedback loop that 
tries to minimize the error between the input 
and a desired input. The "proportional" term 
applies a correction that is linearly related to 
the error. The "derivative" term is essentially 
a damping term and serves to slow down 
the signal when it is moving too quickly. This 
means that oscillations around the zero-error 
point caused by the proportional term will 
be smoothed out over time. The "integral" 
term simply corrects for steady-state error. 
We did not need an integral term, since the 
system was never settled enough to measure 
a steady-state error for which to correct. 
During one control cycle, the ISR first reads 
the position of a touch, shifts by an offset value 
we determined to be at the "edge" of our screen's 
sensitive area, and then scales to convert to a 
position in units of O.lmm. This step allowed 
us to check that the positions were being read 
correctly. The error is calculated by subtracting 
the read position from our pre-set desired 
position, and is low-pass filtered to make these 
position reads less jittery. We used an exponential 
filter to low pass the output, and found that even 
after heavy filtering, the error did not lag enough 
to matter, because the ball did not move fast 
enough to make it inaccurate (Figure 5). 
ci rcuitcellar. corn 33 
Setpoint + 
p, 
p, = a p, _, + (1 - a)p , 
FIGURES 
The P!D control loop with filtered raw position data from the touchscreen 
ABOUT THE AUTHOR 
Samuel Feibel is a graduate student pu rsuing a M.Eng. in Mechanical and 
Aerospace Engineering at Cornell University. Email: o 
With the right tools 
designing a microprocessor can DP. ~~~ 
Okay, maybe not easy, but certainly less complicated . Monte 
Dalrymple has taken his yea rs of experience designing embedded 
architecture and microprocessors and compiled his knowledge into 
one comprehensive guide to 
processor design in the real 
world. 
Monte demonstrates how Verllog 
hardware description language 
(HDL) enables you to depict, 
simulate, and synthesize an 
electronic design so you can 
reduce your workload and 
increase productlviry. 
34 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE6 
A QR link to a demonstration video of our device's capabilities. The video is also posted on Circuit Cellar's, 
article materials webpage. 
This filtering was done to desensitize 
the system, given our very coarse servo 
positioning {discussed in the Results section} 
and slow update speed. The servo signal 
was running at 50Hz, but the update in the 
ISR took place at 2kHz. Because of that, the 
position data needed to be slowed down so 
that the servos updated with the most recent 
accurate data over the last update period. 
To calculate the derivative term, we stored 
five values of the error-the four previous 
values of the error, and the current error. Of 
the different ways to calculate the derivative, 
we opted for subtracting the most recent error 
from the oldest, multiplying by derivative gain 
and adding that to the output signal. 
After uploading this version of the 
program, dragging a finger across the screen 
causes a reaction in the motors, which should 
not go outside the ranges set. If this is not 
the case, fixing the safety parameters before 
putting the ball on the platform is a good 
idea. Watching your delicate device-or more 
harrowingly, HEARING your device break itself 
due to an overzealous servo-is not a required 
step and should be avoided. 
For detailed article references and additional resources go to: 
w.circuitcellar.com/artide-material 
References [1] through [7) as marked in the article can be found there. 
RESOURCES 
RESULTS 
Tuning the parameters takes a while, so 
once you get to this step go ahead and settle 
in. We achieved some starting values for each 
direction independently by attaching a thin 
railing to the platform, which constrained the 
ball to a single direction of linear motion. The 
derivative coefficient and the proportional 
coefficient, along with a parameter that 
changes the strength of the low-pass filter 
on our error calculations, was performed 
through a serial connection to a PC. After 
getting the X direction tuned to a reasonable 
level, we tuned Y, and then combined them by 
removing the railing constraints. 
Our video demonstrating the device in 
action can be accessed by scanning the QR 
code in Figure 6. With our best tuning on our 
best run, we were able to balance the ball 
for about 44 seconds. In about 24 seconds, 
each servo performed around 46 direction 
changes, matching the observed orbit around 
the setpoint. Visually, it was balancing better 
than a human could if constrained to a pivot 
point. 
During those runs where stability was not 
ideal, a few issues could account for the lack 
of steadiness: 
• Minor looseness in the joints could couple 
the X and Y directions of motion, throwing 
off the algorithm. 
• The servos mapto rotations, and don't 
linearly map to the angle of the platform, 
especially at high angles. 
• TheY-dimension of the board was about 2/3 
as large as the X-dimension, and therefore 
had a harder time stabilizing the ball. 
• The servos would only react to a change in 
PWM on time of around 0.06ms to O.OBms. 
Since servos like ours have unpredictable 
behavior outside the range of about 1ms to 
2.Sms of on time, this means we effectively 
had about 25 or so positions over 180 
degrees. This was further clamped to 
about 10 distinct positions with which the 
PID algorithm could work. Buying precise 
servos that can react to smaller changes 
in PWM signal would make this platform 
much more effective. 
CONCLUSION 
This project was fun to make, impressed 
our peers, and taught us a lot. It has an 
exciting amount of movement, involves 
interesting dynamics, and has great potential 
for improvement. While the PID control setup 
was adequate for getting the results we 
expected, the dynamics of a ball on a plate 
has been explored in more detail by similar 
groups of students employing more complex 
control algorithms [7]. G 
Huge opportunities exist today for implementing systems that build Smart 
Cities. From controlling energy waste to Smart Lighting and traffic control, 
loT technologies including sensors, AI platforms, cloud services analog ICs 
and wireless connectivity products all have roles to play in Smart Cities. 
I n some ways, Smart Cities can be thought of as the ultimate form of the lnternet-of-Things (loT) at work. While loT implementations in factories and agriculture 
can be fairly large scale, Smart Cities are inherently 
large-scale deployments often involving diverse public and 
private networks and stakeholders. Sub-segments of the 
Smart City include traffic control, public security, waste 
management, Smart lighting and more. 
Serving the needs of Smart City system developers, a 
wide range of technologies have emerged over the past 
12 months, all useful for various large and small elements 
of Smart City networked systems. These range from box-
level AI solutions and sophisticated loT cloud services, to 
chip-level solutions for implementing LoRa wireless edge 
nodes and battery-powered loT devices. Meanwhile, new 
concerns related to COVID-19 have increased the urgency 
for implementing many kinds of Smart City applications. 
SMART ENTRANCE COUNTER 
There's no doubt that the COVID-19 pandemic has created 
needs for special kinds of Smart City solutions. Governmental 
regulations all over the world have driven an urgent need 
for solutions to secure social distancing in public buildings in 
order to support slowing down the spread of COVID-19. With 
that in mind, lnfineon Technologies developed a system that 
counts people while entering and leaving buildings or rooms 
and ensures social distancing at the same time. The Smart 
Entrance Counter solution is a miniaturized, discrete radar 
board (20mm x 15mm) that accurately and anonymously 
counts people with one single 60GHz radar sensor and 
integrated software (Figure 1). A traffic light system informs 
whether an entry is allowed or not. 
lnfineon Technologies says its Smart Entrance Counter 
solution is a closed system. On one hand, it prevents 
overcrowding, on the other hand it enables businesses to 
keep their operations running. And most importantly, due 
to the use of radar technology, personal data are 100% 
protected. The system counts a person, but does not know 
who it is, says the company. 
Infineon expects volumes for this kind of solution 
to amount to 90 million units globally. Infineon's Smart 
Entrance Counter solution with XENSIV 60GHz radar sensor 
is contactless and can easily be installed on the side or 
ceiling of an entrance or exit. It can be implemented in all 
kinds of building types, such as public buildings, retail and 
grocery stores, restaurants, schools or corporate spaces-
for example, cafes, offices and the like. 
SECURE CONTACTLESS TRANSACTIONS 
Large-scale contactless payment capability was already 
a key technology for Smart City applications, in mass 
transportation, for example. But new requirements for social 
36 CIRCUIT CELLAR • OCTOBER 2020 #363 
Radar 1/0 board 
Radar FF board 
•) b) (Reference Hardware Platform} 
FIGURE 1 
Infineon's Smart Entrance Counter solution (a) is a miniatl.Jrized, discrete radar board (b) (20 mm x 15 mm) that accurately and anonymously counts people with one single 60GHz 
radar sensor and integrated software. 
FIGURE2 
distancing have amplified these demands. 
And the worldwide impact of COVID-19 makes 
contact-free transactions a necessity. Serving 
such needs, in June NXP Semiconductors 
announced its new MIFARE DESFire EV3 IC, 
a chip that provides advanced security and 
seamless integration of mobile services for a 
new era of security and connectivity in smart 
city services (Figure 2). 
As the third evolution of NXP's proven 
contactless MIFARE DESFire portfolio, the 
latest IC is backward compatible and offers 
enhanced performance with a greater 
operating distance and improved transaction 
speed. In combination with its advanced 
security features, the new IC delivers faster, 
more secure transactions that are truly 
The MIFARE DESFire EV3 IC chip enables secure transactions that are truly contactless, such as paying fur 
parking, accessing offices or campuses and using other essential city services-all touch-free. 
contactless, such as paying for parking, 
accessing offices or campuses and using 
other essential city services-all touch-free. 
NXP's MIFARE product portfolio is used 
in Smart City installations worldwide. The 
new MIFARE DESFire EV3 IC builds upon 
the portfolio's heritage of powering mass 
transit ticketing. An extensive set of security 
features in the MIFARE DESFire EV3 IC provide 
more ways to protect data and help ensure 
privacy. The IC hardware and software are 
certified to Common Criteria EAL 5+, and the 
IC supports a broad choice of open crypto 
algorithms. A card-generated MAC helps to 
securely authenticate transactions, and a 
new Transaction Timer feature helps mitigate 
man-in-the-middle attacks so it's harder for 
an attacker to interfere with the transaction. 
Meanwhile, the MIFARE DESFire EV3's new 
Secure Unique NFC (SUN) messaging feature 
offers a more secure method for maintaining 
data confidentiality and integrity. Each time a 
card, phone or ticket is tapped with the SUN 
featured enabled, a tap-unique authentication 
message and crypto-secure URL are generated 
that can be sent to a server for verification, 
which makes taps unclonable. 
Support for MIFARE DESFire EV3 will be 
integrated into NXP's MIFARE 2GO cloud 
service, which manages digitized MIFARE 
product-based credentials and helps 
streamline mobile integration via NXP's 
ecosystem. With this, Smart City services 
can be seamlessly deployed to NFC-enabled 
smartphones, wearables and other mobile 
devices. 
For simplified deployment, each MIFARE 
DESFire EV3 IC is pre-configured with keys 
to enable delegated application management, 
which supports seamless, over-the-air 
updates to already deployed smart cards using 
NFC-enabled smartphones. In addition, NXP's 
upcoming MIFARE Plus EV2 IC will provide a 
drop-in replacement for upgrading existing 
MIFARE Plus and MIFARE Classic product-
based installations for higher security. 
loT PASSENGER COUNTING 
Mass transit passenger counting had 
already been a popular emerging technology. 
But, as with so many other things, the COVID-19 
era brings the importance of this capability to 
the forefront. Eurotech's passenger counters, 
already installed on urban and intercity bus 
lines, allow accurate real-time passenger 
monitoring, remotely and onsite. 
Eurotech's technology was selected by 
Thales as a supplier of embedded hardware and 
loT software for the CAVE project »Automatic 
Counting of Passengers" in some of the Grand 
Paris Express' metro lines. In additionto 
using the power-over-Ethernet (PoE) version 
of Eurotech's DynaPCN passenger counter, the 
solution is based on ESF (Everyware Software 
Framework) for the onboard software for data 
collection and remote configuration, and EC 
(Everyware Cloud) for the ground software to 
enable Societe du Grand Paris to securely access 
passenger counting information and to use this 
data internally or by sharing it externally. 
The DynaPCN 10-20 is a compact, low 
power, autonomous device based on non-
contact stereoscopic v1s1on technology 
(Figure 3). It has been specifically designed 
for passenger counting above the doorways in 
buses and trains. It can also be used to count 
people as they enter or leave buildings or 
any area with restricted access. Stereoscopic 
cameras capture images of the area below 
the device. Thanks to the integrated high 
luminosity infrared LED indicators, it can 
operate in any type of lighting condition. 
The extended temperature range capabilities 
allow integrators to use the device in a wide 
range of climatic conditions. 
The DynaPCN 10-20 analyzes the height, 
shape and direction of any object passing the 
field of view. If the object is recognized as 
a person entering or leaving, the incoming 
and outgoing counters are incremented 
accordingly, along with time and date 
information. Data transfer is made via an 
Ethernet interface. The onboard insulated 
® 
digital I/0 interfaces can be used to directly 
communicate with intelligent doors or flow 
control systems, guaranteeing optimal 
functionality at all times. For example, it can 
stop counting when the doors are closed. 
LoRa FOR CONNECTIVITY 
Because of the inherent spread-out nature 
of Smart City network implementations, 
LoRa has emerged as a popular wireless 
connectivity technology for these applications. 
Over the past couple years, the release of LoRa 
(long range) chip solutions have accelerated 
as technology providers race to keep pace 
with demands. loT networks based on the 
LoRaWAN specification have been deployed in 
100 countries. for more on loRa as a solution 
for Smart Cities, see the sidebar "LoRoWAN: 
A foundational Network for Municipal loT 
Solutions" on p.41. 
In an example of a recent LoRa device, 
in August STMicroelectronics (ST) added a 
QFN48 package to its STM32WLE5 wireless 
system-on-Chip (SoC) portfolio. The 
STM32WLES combines ST's STM32l4 ultra-
low-power microcontoller (MCU) technology 
and Semtech SX126x sub-GHz radio IP 
optimized to meet local radio-equipment 
regulations worldwide (Figure 4). The chip's 
single-silicon-die integration helps save 
bill-of-materials (BOM) costs and simplifies 
system designs. ST targets the device for use 
in connected, smart devices for applications 
in metering, city management, agriculture, 
retail, logistics, smart buildings and 
environmental management. 
The STM32WLES's new 7mm x 7mm QFN48 
package option makes it suited to a simplified 
two-layer board design that further eases 
manufacturing and reduces BOM costs. The 
STM32WLES supports multiple RF-modulation 
ci rcuitcellar. corn D 
FIGURE 3 
The DynaPCN 10-20 is a compact, 
low power, autonomous device based 
on non-contact stereoscopic vision 
technology. It has been specifically 
designed for passenger counting 
above the doorways in buses and 
trains. 
38 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE4 
schemes including LoRa spread-spectrum 
modulation as well as the (G)FSK, (G)MSK 
and BPSK used by various sub-gigahertz 
long-range protocols including proprietary 
protocols. Users have flexibility to apply their 
desired protocol stack, whether created in-
house, sourced externally or chosen from 
off-the-shelf software such as the LoRaWAN 
and wM Bus (wireless M-Bus) stacks that are 
available from ST and authorized partners. 
ST has engineered the integrated RF stage 
to address global markets while improving 
performance and easing manufacture. Features 
include dual low-power (14dBm) and high-power 
(22dBm) transmitter modes with highly linear 
performance from 150MHz to 960MHz, spanning 
the sub-lGHz unlicensed frequency range, 
which ensures technical compatibility with RF 
regulations in all world markets. Sensitivity 
down to -148dBm helps to maximize RF range. 
Only a single crystal is needed to synchronize 
the high-speed external (HSE) clock and the 
radio, delivering a further savings in BOM. 
The new QFN48 package extends the 
STM32WLE5 portfolio, which also includes 
devices in a Smm x Smm BGA73. Three different 
flash-memory density options are available per 
package, offering a choice of 64KB, 128KB or 
256KB, and all devices feature a high proportion 
of user-assignable GPIOs. All feature ST's ultra-
low-power MCU technologies, including dynamic 
voltage scaling and proprietary adaptive real-
time ART Accelerator that allows zero-wait 
execution from flash. 
LoRa INFRASTRUCTURE 
In order for loRa wireless connectivity to 
be deployed on a large scale, cities need to 
The STM32WLES l.oRa SoC combines ST's STM32L4 ultra-low-power MQJ technology and Semtech SX126x 
sub-gigahertz radio IP optimized to meet local radio-equipment regulations worldwide. 
put LoRa infrastructure technology into place. 
Feeding such needs, Semtech announced 
a collaboration with Nesten, a developer of 
specialized blockchain and IoT-based solutions 
for optimized communications networks. The 
effort supports the continued roll out of the 
Nesten's LoRa-based node platform. 
Following the successful 2019 deployment 
of its first-generation G1 wireless node based 
on Semtech's LoRa devices and the LoRaWAN 
protocol, Nesten has increased the integration 
of its LoRa-based wireless infrastructure in 
several key markets throughout the US and 
abroad, and has begun mass production of 
its node platform to empower individual loT 
management. 
Nesten's G1 node is a wireless gateway 
enabling loRa, Wi-Fi, GNSS and Bluetooth-
based data communication in real time. 
Nesten recently announced the deployment 
of more than 270 Gl nodes throughout the 
Dallas metropolitan area, with additional 
loRa-based infrastructure deployed across 
major markets in the US, including Houston, 
Georgia, California, the New York, New Jersey 
and Washington, D.C. suburbs, as well as select 
markets in Canada, Mexico and South Korea. 
Nesten's consumer-based deployment 
model enables end users to deploy their own 
flexible LoRa-based applications built on the 
Gl node, and operate and maintain their own 
IoT network without the associated costs 
traditionally found with telecom carriers. 
leveraging the end user and blockchain as the 
cornerstone of each LoRa-based deployment, 
Nesten is also able to provide extensive data 
protection and privacy. Several loT use cases 
currently utilize Nesten's open-source LoRa-
based infrastructure. These applications exist 
within a wide variety of vertical markets, 
including smart parking, personal emergency 
detection and pet tracking and geolocation, 
home security and more. 
LoRa EXPRESSWAY APPLICATION 
In an example of LoRa technology in 
action, in August Semtech announced that 
Korea Expressway Corp. (KEC) has built a 
network based on Semtech's LoRa devices 
and the LoRaWAN protocol for its expressways 
as the first stage in its ongoing four-year IoT 
deployment plan in the country (Figure 5). 
Beginning in 2019, KEC deployed 78 loRa-
based gateways over 306km of expressway. 
Woojoo Telecom, a systems integrator and SK 
Telesys, a loRa-based gateway manufacturer 
and network server provider, were selected 
as operators for this project, conducting 
communication tests on roads and in tunnels 
to establish the high-quality loT network. 
Currently, the network offers many strategic 
functions including parking space and trash 
bin fill monitoring at rest areas, and real-
time condition tracking for shock-absorbing 
barriers and guardrails. 
In the coming years, KEC plans to expand 
the LoRa-based network's applications to 
include landslip detectors,road-freezing 
sensors, electronic displays, fire extinguishers 
in tunnel and expressway lights. By continuing 
to migrate applications from legacy 3G/LTE 
connectivity to LoRaWAN, KEC expects to reduce 
operating cost by up to $2 million per year. LoRa 
devices and LoRaWAN networks cover hundreds 
of uses cases essential to city and often 
countrywide infrastructure, says Semtech. 
These deployments increase efficiency and cost 
savings for those government entities providing 
public infrastructure such as nationwide 
expressways, and LoRa devices represent an 
ideal technology and connectivity backbone. 
CELLULAR loT PLATFORM 
Implementation loT over existing cellular 
infrastructure is a popular option for Smart 
City system developers. To smooth the 
way, in July Cloud of Things announced 
a collaboration with Monogoto, a cellular 
provider implementing cellular connectivity 
alongside secure private LTE networks. The 
two firms rolled out a global offering to 
smart product manufacturers that reduces 
complexity and risk, and speeds time to 
market with advanced services supporting 
secure edge-cloud deployments. 
The Cloud of Things' DeviceTone loT Suite, 
which includes its patented CloudSwitch 
technology, has been integrated with 
Monogoto's global cellular connectivity 
platform using the Monogoto loT SIM 
(Figure 6). This makes it simple to spin up 
and scale connected systems, including 
automatically and securely registering end-
points to the network, whether those end-
points are fixed or mobile, on land or at sea, 
are indoors or outdoors and include single or 
multiple sensors. 
This means that Smart City system 
developers can take full advantage of a 
real-time operating system using a single 
network with security embedded merged 
with a single platform based on Microsoft 
Azure's worldwide infrastructure that can 
connect to any cloud. The combination of the 
two platforms is enhanced with APis, which 
means smart product/utility companies and 
solution providers can benefit from a unified, 
customized provisioning, management and 
maintenance environment. 
Monogoto offers global coverage in all 
major markets, including their innovative 
ucellular suitcase" and the ability to roam 
into private networks like CBRS in the US for 
service in remote places. With its multiple 
cellular profile container, the Monogoto 
Secured loT SIM has a globally redundant 
footprint in more than 170 countries and 2G, 
3G, 4G and LTE CAT-Ml connectivity federating 
more than 550 cellular networks. Supporting 
standard UlCC (triple form factor), eUICC and 
iUICC, the Monogoto loT SIM card guarantees 
compatibility to any loT device/use case. 
The first joint offering from the two 
companies is an end-to-end prototyping 
package, which includes the DeviceTone 
Genie, a ready-to-run device launched in 
June, and made available on the Microsoft 
Azure loT marketplace, and Monogoto's kit, 
which includes zero code service creation for 
rapid flow formation and global loT cellular 
connectivity. Designed specifically for the 
connected field service market, the DeviceTone 
Genie Starter Kit that comes with a full featured 
loT environment and with a Monogoto SIM and 
complementary support services. 
GPS: A CRITICAL ASSET 
Because Smart City networks are deployed 
on large scales, many of their applications 
depend on GPS location and timing to ensure 
ongoing operations. This can include power 
utilities, financial services, mobile networks, 
ci rcuitcellar. corn 39 
FIGURE 5 
Korea Expressway Corp. built a 
network based on Semtech's LoRa 
devices and the LoRaWAN protocol for 
its expressways as the first stage in 
its ongoing four-year loT deployment 
plan in the country. 
40 CIRCUIT CELLAR • OCTOBER 2020 #363 
Assets 
FIGURE6 
• oevtc.lone 
cen 
~Iona 
Gate .. ay 
Designed specifical~ for the connected 
field service market, the DeviceTone 
Genie Starter Kit comes with a full 
featured loT environment. 
FIGURE 7 
• Device Tone M~~ag~ BJs nwss Appl cat•ons 
transportation and more. As a result, GPS 
has become a critical asset in Smart City 
implementations. To protect that asset, in 
June Microchip Technology announced the 
release of a major software update for its 
BlueSky GNSS Firewall product, providing 
a higher level of resiliency against GPS 
vulnerabilities for systems dependent on GPS 
signal reception. 
Microchip's BlueSky GNSS Firewall 
Software Release 2.0 performs real-time 
analysis to detect jamming and spoofing 
for protecting reception of the GPS signal 
and hardening response and recovery to 
avoid signal disruption (Figure 7). BlueSky 
GNSS Firewall Software Release 2.0 includes 
charting and advanced threshold settings of 
Global Navigation Satellite System (GNSS} 
observables such as satellites-in-view, 
carrier-to-noise, position dispersion, phase 
time deviation and radio frequency (RF) 
power level to simplify system turn-up and 
deployment. 
Microchip's BlueSky GNSS Firewall Software Release 2.0 performs real-time analysis to detect jamming and 
• spoofing far protecting neception of the GPS signal and hardening response and necovery to avoid signal 
disruption. 
BlueSky GNSS Firewall Software Release 2.0 
includes improvements developed by Microchip 
as a result of participation in an industry live-
sky testing event hosted by the US Department 
of Homeland Security (OHS} Science and 
Technology Directorate and open to all providers. 
Microchip's participation in the OHS-hosted 
GPS Testing for Critical Infrastructure (GET-
0} events, with scenarios including spoofed 
signals, has helped the company to identify new 
solutions to prevent signal disruptions. The Blue 
Sky GNSS Firewall Software Release 2.0 release 
improvements are based on the result of the 
20191ive-sky testing and other input. 
Microchip's BlueSky GNSS Firewall 
Software Release 2.0 is provided as part of 
a subscription service and is installed on the 
BlueSky GNSS Firewall device. When connected 
to Microchip's TimePictra management 
software, critical infrastructure operators 
also can monitor and analyze GNSS signals in 
cities, across geographic regions, throughout 
,a country and even globally. 
BUCK-BOOST DC-DC CONVERTER 
Smart metering is key subsegment of 
Smart Cities. An ability run battery-operated 
smart metering devices is vital in today's 
applications where data-collection via IoT 
is the only practical option. That means 
devices that can operate for at least 10 years 
without changing batteries. Along these 
lines, in September Texas Instruments (TI} 
rolled out what it claims is the industry's 
first DC-DC buck-boost converter to combine 
programmable input current limit and 
integrated dynamic voltage scaling to extend 
battery life by at least 50%. The TPS63900 
maintains the industry's lowest quiescent 
current (IQ}, 75 nA, with 92% efficiency at 
l01JA and delivers up to three times more 
output current than competing devices, says 
TI (Figure 8). 
It can be challenging for engineers to 
design for low IQ while providing enough 
output current to send signals between 
connected smart grid applications and a 
network via commonly used radio-frequency 
standards, such as narrowband IoT (NB-IoT}, 
Bluetooth Low Energy (BLE} and long-range 
and wM-Bus. The TPS63900 helps engineers 
conserve energy in wirelessly connected 
applications that run on batteries. 
The TPS63900 integrates dynamic voltage 
scaling to deliver power while keeping the 
system at the minimum voltage required 
to operate efficiently, maximizing battery 
ci rcuitcellar. corn 41 
life and reducing required maintenance 
for industrial applications. This feature 
enables design engineers to optimize power 
architectures for ultra-low-power sensors 
and wireless connectivity ICs, supporting 
applications that can operate for at least 10 
years using the primary battery. For example, 
the buck-boost converter can be paired with 
TI's MSP430FR2155 in security sensors or 
wireless loTsensors to monitor the vibration 
of water pumps for predictive maintenance 
and help drive down costs. The TPS63900 
is now available in a 2.5mm x 2.5mm wafer 
small outline no-lead (WSON) package. 
Extend battery life by 50% with the 
industry's lowest 10 buck-boost converter 
FIGURES 
AI SYSTEM FOR SECURITY 
Security is another area of Smart City 
development where technology can make 
a difference, and specifically technology 
combining AI and loT. Exemplifying this trend, 
in July IEI Integration introduced the FLEX 
AloT Dev Kit equipped with 9th Generation LGA 
1151 Intel Core/Xeon processor, Intel Q370/ 
C246 chipset and DDR4 memory (Figure 9). 
The FLEX AloT Dev. Kit is specifically designed 
for edge learning inference computation and 
features modularized, rich interconnectivity. 
Various input/output interfaces are provided 
for engineers to integrate cameras, sensors 
and motion control equipment to fast respond 
to accidental event. 
The FLEX AloT Dev. Kit features rich 1/0 
and two PCie slots (xB) to support add-ons 
like accelerator cards (Mustang-F100-A10 
and Mustang-V100-MX8) or PoE cards (IPCIE-
4POE) to enhance performance. The Mustang-
F100-A10 is a PCie-based accelerator card 
using the programmable Intel Arria 10 FPGA 
that provides the performance and versatility 
of FPGA acceleration. The Mustang-V100-MX8 
is a PCie-based accelerator card using an 
Intel Movidius VPU that drives the demanding 
workloads of modern computer vision and AI 
applications. 
The TPS63900 maintains the industry's 
lowest quiescent current (IQ), 75nA, 
with 92% efficiency at lO>JA. It 
helps engineers conserve energy in 
wi relessly connected applications that 
run on batteries. 
The FLEX AloT Dev Kit's Mustang 
accelerator card can be used for analyzing 
multiple streams of video data and handling 
LoRoWAN: A Foundational Network for Municipal loT Solutions 
Like many industries and organizations, municipalities 
have the opportunity to transform their operations and 
service delivery by making smart technology investments 
and building successful partnerships. The process of 
becoming a usmart City" is supported by the evolution of 
Internet of Things (loT) products and services which has 
become a global market with significant activity in most 
countries and regions. 
As municipalities explore opportunities to deliver city-
wide loT services, they are in a unique position to partner 
with first-movers in their region to deploy and manage 
smart city solutions on a common and open network 
architecture. Partnering with water utilities, for example, 
provides municipalities with a unique opportunity to create 
a shared network infrastructure capable of supporting a 
wide range of commercial and citizen-facing applications. 
LoRaWAN networks designed for water metering are 
among the most capable. They are high-density, built 
with abundant capacity and secure. With a carrier-grade 
LoRaWAN network in place to support Advanced Metering 
Infrastructure (AMI) deployments, municipalities and 
utilities can partner to rapidly deploy and expand their loT 
initiatives. Easy LoRaWAN gateway deployment and cloud-
based network management allow for multi-service utilities 
(water, gas, electric) to deploy service and safety applications 
at a low cost and in a collaborative fashion. Beyond utility 
metering and service monitoring, municipalities can rapidly 
deploy LoRaWAN solutions for waste management, smart 
parking, environmental monitoring, asset tracking, pest 
management, public safety and more. 
Additionally, municipal leaders should consider that 
short-term measures taken in response to global crises 
often lead to changes that last for decades, making today's 
technology decisions as important as ever. The COVID-19 
crisis has made access to reliable, real-time environmental 
data an absolute necessity. Contactless data collection 
not only supports a wide variety of business reentry 
initiatives, but it will be a key component of driving long 
term operational efficiencies, improving service delivery, 
implementing sustainable practices and ultimately 
achieving a higher quality of life for citizens. ,....---....,..--...... 
- Contributed by Senet, www.senetco.com 
42 CIRCUIT CELLAR • OCTOBER 2020 #363 
b) 
FIGURE9 
FLEX AioT Dev Kit (a) equipped with 9th Generation LGA 1151 Intel Core/)(eon 
pnocessor, Intel Q370/C246 chipset and DDR4 memory. The FLEX AioT Dev. Kit is 
specifically designed for edge learning infenence computation (b). 
RESOURCES 
Texas Instruments I 
real-time tasks such as age, gender and face 
identification. The data collected can be used 
in precision marketing, VIP customer service 
and customer flow analysis to further provide 
relevant product information through digital 
signage display to improve product sales and 
inventory control. 
In addition to face identification, the 
solution speeds up the characteristic 
recognitions of vehicles such as license plates, 
vehicle models and colors. It can be applied 
in services like traffic steering analysis, illegal 
parking/helmet detection and trajectory 
tracking. Furthermore, vehicle license plate 
analysis can be deployed on highways for 
electronic toll collection, and can provide 
enhanced security by establishing a database 
of suspicious vehicles. 
MORE AI AND SMART CITIES 
late last year, Integrated Device 
Technology (IDT), a subsidiary of Renesas 
AI Inference Sy$lem 
Electronics, announced a partnership with 
industrial AI specialist ONE Tech to provide an 
end-to-end, integrated loT and edge analytics 
solution aimed at optimizing the performance 
and use of city assets. 
The joint loT with Edge AI platform 
combines lOT's BLE and 6loWPAN wireless 
interconnectivity devices, sensors and MRAM 
non-volatile memory data storage with 
ONE Tech's loT Platform and Micro Al. ONE 
Tech's platform and AI deliver performance 
optimization, security and analytics for 
connected devices and assets (Figure 10). 
The combination results in an end-to-end loT 
ecosystem-from node-to-edge-to-cloud-
that is well suited for a variety of Smart City 
applications, including connected community, 
water and waste management, and energy 
and asset management. 
According to IDT, the loT with Edge 
AI platform is a good alternative for loT 
developers who do not want to develop 
their own cloud platform, nor want to use 
other major cloud platform solutions. Such 
development requires expertise in configuring 
wire less modules, sensors and other connected 
devices. The platform is designed as a ready-
to-go solution that configures and provisions 
sensor and wireless products, allowing loT 
developers to easily scale and customize the 
platform to the unique requirements of their 
applications and devices. 
Securing connected assets and optimizing 
their performance are the most critical aspects 
of managing Smart City environments, says 
ONE Tech. By avoiding unexpected asset 
downtime, increasing the productivity of 
field teams and protecting all data, cities can 
maximize their return on investment for all 
... 
J-
their public service projects while redirecting 
any savings to other critical initiatives. 
Using the loT with Edge AI service, users 
can activate and manage any IDT/Renesas 
device. It lets you define workflows for 
automating business processes and create 
real-time event notifications. The platform 
V sllngS1ot 
....... L Y 
NOT YOUR 
AVERAGE SLINGSHOT 
c 20 (J 7 0 
• Sin fti"CJkO 
enables you to store big data and apply AI/ 
machine learning (ML) to perform advanced 
analytics and gain deeper insights. You can 
monitor real-time location and performance 
for all your loT connected assets and 
orchestrate an end-to-end digital experience 
for your Smart City loT deployment. G 
ci rcuitcellar. corn 43 
FIGURE 10 
The joint loT with Edge fJJ. platform 
delivers performance optimization, 
security and analytics for connected 
devices and assets. The combination 
results in anend-to-end loT 
ecosystem-from node-to-edge-to-
cloud-that is well suited for a variety 
of Smart City applications . 
44 ORQIIT CELLAR • OCTOBER 2020 #363 
Raspberry Pi Roundup 
A Bounty of HATs and Clones 
The MCC 1n is 1 bio-dllnnel, high-speed DAQ HAT for making SOJnd and vibralian 
meiiSUI'8Tienls fhm lEPE sensors like aa:eleramet!rs and microphones. Up ID ~ MCC HATs 
a111 be stadced onto one Raspberry PI. 
T he popularity of the Raspberry Pi product line of open-spec SBCs is truly astounding. In surveys 
of LinuxGizmos's readers, we've 
found that year after year Raspberry Pi 
SBCs are the most popular open-spec SBCs. 
(LinuxGizmos.com is Circuit Cellar's sister 
website). Raspberry Pi's targeted audience 
has traditionally been DIY builders, makers, 
electronics hobbyists and the like. But the 
open-spec nature of Raspberry Pi has led to 
huge interest among professional engineers 
for two reasons. 
First, board vendors have been able to 
develop clones of Raspberry Pi SBCs, that 
either match their functionality or simply 
offer the same connector scheme. Second, the 
emergence of Raspberry Pi add-on boards, 
called HATs, has resulted in the development 
of a wide variety of add-on functions that 
professional engineers need for all kinds 
of applications. HAT stands for "hardware 
attached on top," a hardware specification 
for add-on modules for the Raspberry Pi 
model B+ SBC. HATs have several advantages 
compared to older add-on modules for the 
Raspberry Pi, such as no soldering required. 
In other words, HATs can be plugged right 
onto the Raspberry Pi's connector. 
With all that in mind, the real story about 
Raspberry Pi can arguably said to revolve 
around HAT products, rather than the SBCs 
themselves. Every month, new HAT products 
have emerged targeting an impressively wide 
set of functions and application needs. In fact, 
such new products are so frequent, that you'll 
notice that many of the products covered in 
this article were released as recently as July 
and August. 
MACHINE CONDITIONING HAT 
Exemplifying these trends, the latest 
Raspberry Pi offering from Measurement 
Computing is its MCC 172 IEPE Measurement 
HAT for Raspberry Pi released in May. Well 
suited for machine condition monitoring and 
edge computing applications, the MCC 172 is a 
two-channel, high-speed DAQ HAT for making 
sound and vibration measurements from IEPE 
sensors like accelerometers and microphones 
(Figure 1). It features two, simultaneous, 
24 bit analog inputs, with sample rates up to 
51.2KS/s per channel. 
Up to eight MCC HATs can be stacked 
onto one Raspberry Pi. MCC offers a variety 
of DAQ HATs that enable users to configure 
multifunction, Pi-based solutions with analog 
input, output and digital I/0. The open-
source MCC DAQ HAT Library of commands 
in C/C++ and Python allows users to develop 
applications on Linux. The library is available 
to download from GitHub. Comprehensive 
API and hardware documentation are also 
provided. 
The MCC 172 header plugs into the 40-pin 
general purpose I/0 (GPIO) connector on a user-
supplied Raspberry Pi SBC. The MCC 172 was 
tested for use with all Raspberry Pi models with 
the 40-pin GPIO connector. Multiple MCC 172 
HATs can be synchronized to a single sampling 
clock. The clock is programmable for sampling 
rates between 51.2KS/s to 200 samples/s. 
HAT configuration parameters are stored 
in an on-board EEPROM that allows the 
Raspberry Pi to automatically set up the GPIO 
pins when the HAT is connected. The MCC 172 
is powered with SV provided by the Raspberry 
Pi through the GPIO header connector. The 
trigger input (terminal TRIG) is used to delay 
an input scan until a specified condition is met 
at the trigger input. 
AUDIO AMPLIFIER HAT 
While Raspberry Pi is aimed at hobbyists 
and nmakers,n the rich variety of HATs 
available has attracted professional 
embedded system developers that want to 
take advantage of what's available. Some 
such HATs include products that are aimed at 
users and makers. In an example along those 
lines, in May Infineon Technologies released 
what it claims is the world's first fully self-
contained Raspberry Pi audio amplifier HAT 
board (Figure Z). 
The board (KIT_ 40W_AMP_HAT_ZW) offers 
high definition audio at boom box power levels 
in a small form factor. The Infineon proprietary 
multilevel technology ensures minimum size 
and consumption, state of the art power 
efficiency and HD audio quality for Raspberry 
Pi users and makers. Infineon markets the 
product for systems such as active speakers 
with wireless music streaming. 
The board is compatible with Raspberry 
Pi Zero W and Raspberry Pi 3 and 4 SBCs. 
It leverages the MERUS multilevel class D 
amplification enabled by the MERUS MA12070P 
amplifier, which allows for filter-free amplifier 
design that does not need to use a filter-coil at 
the output filter. This significantly reduces the 
BOM cost and enables PCB area reduction. The 
board furthermore scores high output power 
in small form factor up to 40W instantaneous 
peak power at 4Q. The solution offers best-
in-class efficiency and playback time up to 20 
hours with a 6700mA-hour power bank. There 
is no need for additional power supplies, only 
a single 5V/2.5A USB power supply is required 
for both the Raspberry Pi and the HAT. 
For a quick and easy audio system setup, 
the board is compatible with the main Linux 
distributions such as Raspbian, Volumio, 
moOde Audio or JustboomPiayer. Dual-channel 
bridge-tied load (BTL) or single-channel 
parallel bridge-tied load (PBTL) configurations 
for Multi room, TWS, or subwoofer applications 
are also possible. 
FIGURE 2 
ci rcuitcellar. corn 45 
The KIT_40W_AMP_HAT_ZW board is a fully self-contained Raspberry Pi audio amplifier HAT board. The 
board is compatible with Raspberry Pi Zero Wand Raspberry Pi 3 and 4. 
FIGURE 3 
The Indoor Air Quality HAT for Raspberry Pi features an on-board calibrated Renesas ZMOD4410 sensor that 
measures the concentrations of total volatile organic compounds (TVOCs) from parts-per-billion to parts-
per-million and measures estimate carbon dioxide (eC02) levels. 
46 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE4 
pHAT FOR GNSS 
Once again illustrating the diverse 
ecosystem of Raspberry Pi add-on boards 
available, in July SparkFun launched ZED-F9R 
GPS pHAT, a high precision, sensor fusion 
GPS board with impressive configuration 
options. It embeds U-blox's Automotive 
Dead Reckoning (ADR) technology. SparkFun 
says the ZED-F9R module provides a highly 
accurate and continuous position by fusing 
a 3D IMU sensor, wheel ticks, a vehicle 
dynamics model, correction data and GNSS 
measurements. Sparkfun follows its pHAT (for 
partial HAT) approach, which doesn't support 
all the specs of the official Raspberry Pi HAT. 
The ZED-F9R module is a 184-channel 
U-blox F9 engine GNSS receiver, meaning it can 
receive signals from the GPS, GLONASS, Galileo 
and BeiDou constellations with approximately 
0.2-meter accuracy. Such accuracy can be 
achieved with an RTK navigation solution 
when used with a correction source. Note that 
the ZED-F9R can only operate as a rover, so 
you will need to connect to a base station. 
The module supports concurrent reception of 
four GNSS systems. The combination of GNSS 
and integrated 3D sensor measurements 
on the ZED-F9R provide accurate, real-time 
positioning rates of up to 30Hz. 
Compared to other GPS modules, this pHAT 
maximizes position accuracy in dense cities or 
covered areas, says the company. Even under 
poor signal conditions, continuous positioning 
is provided in urban environments and is also 
available during complete signal loss (such 
as short tunnels and parking garages). The 
ZED-F9R is a solution for autonomous robotic 
The Cl'i-A150WR is a Raspberry Pi-based industrial panel PC with a 15" touch-panel display. Leveraging the 
compact Raspberry Pi 3 SBC, the ComfilePi Cl'i-AlSOWR inherits the Raspberry Pi 3's 1.2GHz 64 bit quad-core Arm Cortex-A53 processor, a Broadcom VideoCore IV GPU and 1GB of RAM. 
applications that require accurate positioning 
under challenging conditions. 
The U-blox receiver supports a few serial 
protocols. By default, SparkFun chose to use 
the Raspberry Pi's serial UART to communicate 
with the module. With pre-soldered headers, 
no soldering is required to stack the pHAT 
on a Raspberry Pi, NVIDIA Jetson Nano, 
Google Coral or any SBC with the 2x20 pin 
connector header. The board breaks out a few 
0.1"-spaced pins from the U-blox receiver. 
A Qwiic connector is also added in case you 
need to connect a Qwiic-enabled device. 
AIR QUALITY SENSOR HAT 
Some Raspberry Pi HAT manufacturers 
have been keen to leverage the technologies 
of sensor vendors. In an example along those 
lines, in August, Avnet launched a Renesas-
sensor-based Indoor Air Quality HAT for 
Raspberry Pi called the AES-RHSEN-ZM44-G 
(Figure 3). The HAT features an on-board 
calibrated Renesas ZMOD4410 sensor that 
measures the concentrations of total volatile 
organic compounds (TVOCs) from parts-per-
billion to parts-per-million and measures 
estimate carbon dioxide (eC02) levels. These 
are important indicators for monitoring indoor 
air quality. All sensors are electrically and 
chemically (gas) tested with lab calibration 
data stored in each ZMOD4410 sensor's built-
in nonvolatile memory. 
In addition to the ZMOD4410 sensor, the 
HAT incorporates a Renesas HS3001 high-
precision relative humidity and temperature 
sensor, along with software-controlled status 
LEDs. Connection points are also available 
to measure the ZMOD4410 active current 
consumption, which is useful when integrating 
the sensor and its software into products 
powered by extended life batteries. 
Potential applications for the board 
range from air/health monitoring systems, 
smart home appliances, smart thermostats, 
smart speakers and smart fans to smoke 
alarms, vacuum cleaners, garage openers, 
security systems, HVAC controls, air purifiers 
and building automation. The ZMOD4410 
indoor air quality platform is available in 
four software configurations providing 
various sensor behaviors targeting unique 
applications indoors. 
Avnet provides a pre-compiled test 
application that runs on the Raspberry Pi 
operation system and can measure TVOC 
and eC02 out of the box. The application 
is built with Renesas' proprietary licensed 
algorithms. For production purposes, 
customers may obtain the algorithms directly 
from Renesas under terms of their software 
license agreement. 
PANEL PC FOR RASPBERRY Pi 
In July, Comfile introduced the ComfilePi 
CPi-AlSOWR, a Raspberry Pi-based industrial 
panel PC with a 15" touch-panel display. 
Leveraging the compact Raspberry Pi3 
SBC, the ComfilePi CPi-A150WR inherits the 
Raspberry Pi 3's 1.2GHz 64 bit quad-core Arm 
Cortex-A53 processor, a Broadcom VideoCore 
IV GPU and 1GB of RAM (Figure 4). The new, 
larger version has the same capabilities 
as its predecessors with smaller screen 
sizes: the 7" CPi-A070WR and the 10.2" CPi-
A102WR. All models in the series now support 
AdvancedHMI software, based on the .NET 
framework, which uses the popular Visual 
Studio design environment. AdvancedHMI 
enables the creation of HMis that are not 
possible with other off-the-shelf packages. It 
creates a true, fast executable for the Linux-
based ComfilePi. 
The CPi-A150WR's display is a 24-bit calor 
LCD with a resistive touchscreen, and the 
controller itself provides 22x ESD-protected 
GPIO lines, 3x USB 2.0 host ports, lx RJ-45 
Ethernet port, lx 12C port, lx RS-485 port, a 
1x RS-232C port and a battery-backed RTC. 
It also features stereo audio output as well 
as a piezo buzzer. Wi-Fi is also possible using 
an external USB dangle. This 15" ComfilePi 
requires a 12-24VDC 21W power input. The 
CPi-AlSOWR is housed in a flame-retardant 
ABS enclosure with an IP65 water resistant 
front panel, and is specified for an operating 
temperature range of oac to 70°C. 
While Raspberry Pi SBCs typically run the 
Linux operating system, the ComfilePi can 
be programmed in almost any language, 
including C, C++, Java, JavaFX, Javascript, 
python, C#, vb.net and even emerging 
languages such as Rust, D and Nim, as well 
as .Net Core 2.0 applications. (Android and 
Windows 10 loT are not supported.) Due to 
the built-in Broadcomm VideoCore IV GPU, 
the ComfilePi can even render real-time 
3D graphics. CODESYS can be used on the 
ComfilePi by installing one or more of the 
CODESYS Raspberry Pi Control Modules. 
CODESYS is a development environment 
for programming controller applications 
according to the international industrial 
standard IEC 61131-3. Made in South Korea 
by Comfile Technology, ComfilePi products are 
available now from the distributor Saelig. 
1/0 CARDS FOR RPi 
For its part, Sequent Microsystems makes 
a variety of Raspberry Pi-compatible I/0 
boards serving diverse applications. In July, 
the company Kickstarted a stackable HAT 
called the Mega-RTD. Resistance Temperature 
Detector (RTD) sensors, which are used for 
highly precise temperature measurements 
in industrial and laboratory systems, can 
be found on embedded systems. Sequent 
Microsystems won Kickstarter funding for a 
Mega-RTD HAT that can be stacked to provide 
the Raspberry Pi with up to 64 RTD channels. 
Shipments are due in October. 
Sequent's next Raspberry Pi offering, 
listed on its website now as "coming soon," 
is a Raspberry Pi stackable card for industrial 
automation under the product code MEGA-IND 
(Figure 5). The board will be the company's 
second-generation industrial automation 
card, and is designed to bring to Raspberry 
Pi all the necessary inputs and outputs to 
implement complex industrial automation 
projects, says Sequent. 
The MEGA-IND features four optically 
isolated digital inputs with status LEDs, four 
OV - 10V or 0 to :!:lOV analog inputs and four 
optically isolated 4-20mA inputs. You also 
get four optically isolated open drain 24V/4A 
outputs, four 0-lOV analog outputs and four 
4-20mA analog outputs. The card provides 
TVS protection on all inputs. 
Other features of the MEGA-IND include 
RS485 and CAN ports, 24VDC power supply, 
on-board hardware watchdog and on-board 
resettable fuse. The board provides a real 
time clock with battery backup, pluggable 
connectors, support for eight-level stacking 
with all mounting hardware included. Command 
line and Python driver support is provided. 
PICO-ITX WITH RPi EXPANSION 
Some embedded board vendors have 
taken advantage of Raspberry Pi's open 
specifications to create their own SBCs that 
FIGURE 5 
ci rcuitcellar. corn 47 
This Raspberry Pi st.ackable card for industTial automation, called the MEGA-IND, is designed to bring to 
Raspberry Pi all the necessary inputs and outputs to implement complex industrial automation projects. 
48 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE 6 
The Uno-220 system is an loT gateway designed to be powered by a bring-your-own Raspberry Pi 4 Model B 
SBC. The Uno-220 exposes all the ports on the RPi 4 and adds a custom HAT add-on with serial and GPIO 
connectivity. 
RESOURCES 
Advantech I www.advantech.com 
Avnet I www.avnet. com 
Hilscher I www.hilscher.com 
include Raspberry Pi expansion support. That 
opens the door to, for example, more rugged 
SBC offerings that can work with the same 
HATs as Raspberry Pi SBCs. Along exactly 
those lines, in August ICP (Industrial Computer 
Products) Germany released an i.MX8M-
based Pico-ITX SBC that it's positioning as an 
industrial replacement for a Raspberry Pi SBC. 
The 100mm x 72mm board serves up a 
l.SGHz NXP i.MXBM in either quad- or dual-
core versions. Memory options include 1GB, 
2GB or 4GB LPDDR4 SDRAM with 1600MHz 
clock rate. There's a standard Mini-PCie 
with PCie and USB signal. A Raspberry Pi-
compatible 40-pin header with GPIO signal is 
available as an expansion slot along with an 
API library. 
In the basic version, ND108T comes with 
HDMI 2.0with 4K resolution, GbE, 2x USB 3.0, 
lx USB 2.0, lx RS-232/422/485 and audio line-
out and mic-in. A USB 2.0 OTG and a UART are 
available for programming and debugging. 
ICP Germany I www. icp-deutschland.de 
Infineon Technologies I www. infineon.com 
Measurement Computing I www.mccdaq.com 
Renesas Electronics I www.renesas.com 
Saelig I www.saelig.com 
Sequent Microsystems I www.sequentmicrosyst ems.com 
Sparkfun I www.sparkfun.com 
Trenz Electronic I www. trenz-electronic.de 
For mass storage you get an eMMC with 8GB, 
16GB or 32GB storage and a microSD slot. 
The basic version can be upgraded if 
required, says ICP. Dual GbE or other display 
connectors like LVDS or Display Port are 
optional. The ND108T operates from a 12V 
power supply in a temperature range of 0 to 
60°C and complies with the CE/FCC Class B 
standard. Operating system support includes 
Board Support Packages (BSPs) for Yocto 
Sumo and Android 9 Pie, both with Linux 
Kernel 4.14.98. 
BOX-LEVEL RPi SYSTEM 
Support for Raspberry Pi appears to 
have even infiltrated the box-level embedded 
system realm. As reported by Circuit 
Cellar's sister website LinuxGizmos, in July, 
Advantech launched what appears to be its 
first Raspberry Pi-compatible system. The 
$90, lOOmm x 70mm x 33mm Uno-220 loT 
gateway is designed to be powered by a bring-
your-own Raspberry Pi 4 Model 8 (Figure 6). 
The Uno-220 exposes all the ports on the RPi 4 
and adds a custom HAT add-on with serial 
and GPIO connectivity. The system ships with 
an 8GB microSD card that is preconfigured 
with a Raspberry Pi OS based AdvRaspbian 
stack. AdvRaspbian is said to offer plug-and-
play installation with the Raspberry Pi 4. The 
source code is also available on GitHub. 
External access is provided for the new 
RS-232/485 port with auto direction control 
and a terminal block interface. There are 
also 4x GPIOs and inside the box is a real-
time clock with battery installed on the HAT. 
TPM 2.0 is available on request. Standard 
options are wall- and stand mounts, while 
DIN-rail mounting is optional. The system 
ships with all necessary screws to integrate 
the Raspberry Pi. 
The aluminum-built loT gateway is 
protected per IP40 specs, supports 0 to soac 
temperatures, and provides 10-95% RH at 
40°(, non-condensing humidity resistance. 
The enclosure's improved heat dissipation and 
electrostatic discharge protection provides 
the optimum protection for Raspberry Pi 
SBCs, according to Advantech. 
FPGA-BASED RPi ZERO CLONE 
As discussed earlier, the open-spec nature 
of Raspberry Pi makes it ripe for board 
developers to design clones of the SBCs. Along 
just those lines, in August Trenz Electronic 
launched its open-spec ZynqBerryZero, a 
pseudo-clone of the Raspberry PI Zero SBC 
with a Zynq-7010 Arm/FPGA SaC, a microSD 
slot, dual micro-USB, mini-HDMI, CSI-2 and 
40-pin GPIO (Figure 7). 
In 2017, Trenz Electronic rolled out its 
ZynqBerry SBC from, which combines a Xilinx 
Zynq-7010 FPGA SaC with a Raspberry Pi. Now, 
this new ZynqBerryZero (TE0727) SBC takes a 
similar approach, but this time imitating the 
Raspberry Pi Zero. The 65mm x 30mm SBC 
offers a 40-pin GPIO header that can load 
Raspberry Pi HATs that support the Zero. 
The ZynqBerryZero trades out the Zero's 
lGHz, Armll-based Broadcom BCM2836 for 
the Zynq-7010, which combines dual 667MHz 
Cortex-A9 cores with an FPGA with 28K logic 
cells, compared to 85K for the Zynq-7020. The 
SBC comes with 512MB DDR3L DRAM, 16MB 
flash and a microSD slot. The SBC's coastline 
ports are limited to a mini-HDMI Type-C port, 
dual micro-USB 2.0 ports, an OTG port and a 
port for JTAG/UART debugging and SV power 
input. The 0 to 70°( tolerant board is also 
equipped with a 40-pin header with 26x GPIOs 
and a MIPI-CSI-2 camera connector. 
loT GATEWAY FOR RPi 
loT gateways are yet another area that 
has embraced the Raspberry Pi phenomenon. 
With that in mind, in July Hilscher released its 
Raspberry Pi based netPI Real-Time Ethernet 
gateway. The netPI series is based on an 
industrial-ready, customized Raspberry Pi 3 
design made to run any custom applications in 
edge automation (Figure 8). The model RTE 3 
comes with Hilscher's multiprotocol netX SoC 
additionally implementing all popular Industrial 
Ethernet networks. With their powerful RPi 3 
1.2GHz quad-core Arm CPU, the platforms are 
ready-made for any demanding Cloud, Industry 
4.0 and Industrial loT (lion application. 
The model netPI RTE 3 includes two extra 
Industrial Ethernet ports extra to connect to 
systems such as PROFINET, EtherNet/IP and 
others, as supported by netX. An expansion 
slot at the unit's bottom accepts extension 
modules made for sensor/actuator level 
communications such as RFID, digital 1/0s, 
others or your own custom boards. 
To meet EMC standards EN 55011 and 
IEC 61000, netPI is housed in a robust metal 
chassis and uses two additional PCB layers 
compared to a standard Raspberry Pi 3 SBC. 
The radio antenna extends beyond the chassis 
for improved wireless coverage. A hardware 
Real-Time Clock with a supercapacitor 
as backup power source is supported. A 
nonvolatile auxiliary ferroelectric memory 
(FeRAM) guarantees high endurance for data 
to be rewritten billions of times (Model RTE 3). 
The netPI hosts an AppArmor-secured Yocto 
Linux build. By design, the system software 
complies with the IEC 62443 cybersecurity 
standard for automation and control systems. 
User access is granted via web browsers over 
https-secured connections. The device boots 
secure and allows system updates with Hilscher 
integrity-checked software only. G 
ci rcuitcellar. corn 49 
FIGURE 7 
The ZynqBerryZero is a pseudo-clone of the Raspberry PI Zero SBC with a Zynq-7010 Arm/FPGA SoC, a 
microSD slot, dual micro-USB, mini-HDMI, CSI 2 and 40-pin GPIO. 
iiii 
.. 11 
FJGURE8 
--.. --•••• iiii 
D 
a ... ~ 
docker 
The netPI Real-Time Ethernet gateway features an industrial-ready, customized Raspberry Pi 3 design made 
to run any custom applications in edge automation. The model RTE 3 comes with Hilscher's multiprotocol 
netX SoC additionally implementing all popular Industrial Ethemet networks. 
By Jeff Child, 
Editor-in-Chief 
Gone are the days when box-level embedded computers, also llntNIII 
as Embedded PCs, were stridly custom-built offerings. Today, several 
vendors offer extensive produd lines of embedded PC of all 
and configurations, and for diverse applications including i 
system monitoring, energy system control and ······~-
For its system that enables bats and wind turbines to coexist, Fleximaus chose the KBox A-150-APL 
embedded PC from Kontron. The KBox A-150-APL embeds an Intel Pentium Quad Core N4200 processor. 
F or many years, the mid- to high-end of embedded computing revolved around single-board computers 
{SBCs). But a decade or so ago, a 
new class of computing solution emerged-the 
notion of an enclosed, integrated computing 
system to which system integrators could 
link their interfaces. Many such systems 
used to be common as custom-built units, 
but vendors shifted to offering them as off-
the-shelf products. These systems are called 
a number of names including industrial 
PCs, rugged stand-alone box systems and 
embedded PCs. These systems are attractive 
to system developers that want a high-level of 
integration to start with-so that don't have 
to deal with making different SBCs and 1/0 
boards work together, for example. 
Attempting to keep this Datasheet 
roundup focused was a challenge because 
box-level embedded computing has grown 
into a broad category of products. With that 
in mind, the representative products included 
in this article's product gallery is restricted to 
embedded PCs in the literal sense of systems 
with PC-architecture computing (lntel or 
AMD processing) and Windows OS support. 
The gallery also excludes box-level systems 
that are marketed as application-specific 
systems-aimed at vehicles, digital signage, 
railway systems, vision system and soforth. 
It gets murky because many of the systems in 
this article can and are used in transportation, 
for example, so there is that kind of overlap. 
An example of an embedded PC application 
is a system developed by Fleximaus that 
allows bats and wind turbines to coexist 
{Figure 1). The Fleximaus solution is based on 
intelligent software that is able to evaluate 
the environmental conditions detected by rain 
and temperature sensors and use this data to 
effectively control the wind farms in order to 
drastically reduce the dangers for hunting 
bats. An essential technical component of the 
solution is an industrial PC with integrated 
soft PLC. For the embedded PC, Fleximaus 
chose the KBox A-150-APL from Kontron. 
According to Fleximaus, the KBox A-150-APL 
was selected because of its fanless, rugged 
design, which guarantees an extended service 
life and high system availability. Other factors 
in the selection included KBox's compact 
design, its wide range of interfaces and its 
expansion options. The KBox A-150-APL 
embeds an Intel Pentium Quad Core N4200 
processor. G 
Fanless System Embeds 8th 
Gen Coffee Lake 
The BOXER-6641 from Aaeon 
offers a range of processors from the 
Bth Generation Intel Core and Xeon 
processor family (formerly Coffee 
Lake), from Celeron G4900T and 
Pentium G5400T up to Xeon E-2124G 
(up to 71W). The system provides up to 
32GB of ECC or non-ECC ODR4 2666MHz 
RAM and support for two hard drives. 
• i3/i5/i7/i9 processor for A1 model 
(max. TOP up to 65W) 
• ... or i3/i5/i7/i9/Xeon processor for 
A2 model (max. TOP up to 71W) 
• DOR4 ECC or Non-ECC SODIMM, Up 
to 64GB 
• Intel GbE LAN x 4 (i211 x3 + i219 x1) 
• Intel AMT (A2model) 
• HDMI x 2 dual display 
• USB 3.2 gen 1 x 8 (A2 model) 
• BIOS selectable RS-232/422/485 x6 
• Wide range DC 10V to 35V input 
DATASHEET URLS: 
AAEON 
www.aaeon.com 
System Blends CPU and GPU 
Computing 
AOLINK Technology's Matrix series 
MVP-5100-MXM is an integrated fanless 
embedded computer featuring Intel 
Core i7/i5/i3 processors and embedded 
GPU for industrial AI applications. The 
system includes a comprehensive I/0 
set to collect sensor measurements and 
allow data-driven actions to be taken in 
a timely manner. 
• 9th Gen Intel Core i7/15/i3 LGA 
processor 
• Dual SOOIMMs sockets for up to 
32GB DOR4 
• Up to 4x additional OP 1.4 from 
MXM, 2x OP++, DVI, VGA, 3x GbE, 
3x COM, TPM2.0 
• 3x USB 3.1 Gen 1, 3x USB 2.0 
• 2x 2.5" SATA, M.2 2280 
• Front accessible I/0 and adaptive 
Function Module 2.0 options 
• Slots for Mini PCie, M.2 3042, 
2x USIM 
• Embedded GP/GPU computing 
options built-in 
ADLINK Technology 
www.adlinktech.com 
AAEON http://data-us.aaeon.com/DOWNLOAD/2014%20datasheet/Systems/BOXER-6641.pdf 
ci rcuitcellar. corn 51 
Embedded Computer Features 
lU Height 
EPC-T4286 from Advantech is an 
embedded computer designed for self-
service applications. As the newest 
member of Advantech's EPC-T series, 
this embedded computer features a 
highly-functional small lU-high form-
factor design. Powered by a 9th Gen. 
Intel CPU platform, EPC-T4286 combines 
high computing performance with 
diverse 1/0 ports, making it well suited 
for applications in self-service kiosk 
terminals. 
• Intel Core i CPU 
• Thin compact design with 44.2mm 
in height 
• Thin barebones system with multi-10 
ports 
• Suited for environments with OC-in 
design 
• Easy and installation for additional 
peripherals 
• 6x COMports, 6x USB ports, 
3x LAN ports, 
• 1x 2.5" shock-resistant drive bay 
• Waii/VESA/Rack mounting kits 
Advantech 
www.advantech.com 
ADllNK Technology www.adlinktech.com/Products/Download.ashx?type=MDownload&isDatasheet=yes&file=l815%5cMVP-5100-MXM_Series-20200304.pdf 
Advantech https:/ /advdownload.blob.core.windows.net/productfi le/PIS/EPC-T4286/file/EPC-T4286-Series-System_DS(OB1020)20200903164229.pdf 
52. CIRCUIT CELLAR • OCTOBER 2020 #363 
Embedded PCs 
Apollo Lake Fanless Box 
Targets Industrial loT 
LYNX-6110 from American Portwell 
Technology is a compact, fanless and 
cable-less PC adapting a low power 
Intel Apollo Lake processor. Developed 
to meet heavy industrial standards, 
the rugged design enables it to work 
in harsh environment. Providing 
basic I/0 interfaces in a palm-
sized box, the system is suitable for 
industrial automation, smart factory, 
edge computing and IIoT gateway 
applications. 
• Intel Celeron N3350, 2 cores, 
1.1GHz/2.4GHz, 6W TOP 
• 4GB onboard LPDDR4, 2400 MT/s, 
non-ECC; up to 8GB (optional) 
• 32GB onboard eMMC 5.0 flash; 
up to 256GB (optional) 
• Expansion 1x M.2 Key E 2230 for 
wireless module 
• Display 1x DP v1.4 4096 x 2160 at 
60Hz 
• USB 2x USB 3.0 
• Ethernet 2x 10/100/1000 GbE 
• Power input 12-30VDC (+/- 20%) 
• Operation temperature 0°( to 50°( 
American Portwell Technology 
www.portwell.com 
DATASHEET URlS: 
Coffee Lake System Offers 
Flexible 1/0 Design 
Axiomtek's eBOX640-521-FL is 
powered by the 9th/8th generation Intel 
Core i7/i5/i3 and Celeron processor with 
the Intel H310 chipset (Coffee Lake). 
This embedded system features a wide 
choice of front-accessible I/0 interfaces 
for easy cabling and maintenance. It 
also has a flexible I/0 window for ease of 
customization for versatile applications 
and customer requests. 
• LGA1151 socket 9th/8th gen Intel 
Core i7/i5/i3 
• ... and Celeron processor (Coffee 
Lake) 
• Supports 2x COM, 6x USB and 8 
channels of digital I/0 
• Two 2.5" SATA drive bay 
• M.2 Key E 2230 for Wi-Fi 
• Front I/0 connectivity design 
• 12/19 to 24 VDC power input via 
DC-Jack 
• Flexible I/0 window supported 
Axiomtek 
https://us.axlomtek.com 
American Portwell Technology ....ww.portwell.com/pdf/iot/LYNX-6110.pdf 
Axiomtek https://us.axiomtek.com/Download/Spec/en-US/ebox640-521-fl.pdf 
!base Technology ....ww.ibase.com.tw/englishfProductDetaii/EmbeddedComputingj/lSB200-918 
Fanless Box Sports AMD 
Ryzen ¥1000/RlOOOs 
The ASB200-918 series from lbase 
Technology is based on the IB918 
3.5" form factor SBC supporting AMD 
Ryzen SoCs (V1605B/V1202B/R1606G/ 
R1505G). The fanless ASB200-918 runs 
two simultaneous HDMI 2.0a displays 
at Ultra HD 4K resolution (3840 x 2160) 
with a frame rate of 60 fps. Measuring 
240mm x 162mm x 42mm, it is 
equipped with rugged components. 
• AMD Ryzen Embedded V1000/R1000 
on board 
• 2x DDR4 50-DIMM, Max. 32GB, 
supports ECC 
• Supports 2x M.2 sockets (E-Key/ 
M-Key, NVMe) 
• 2x HDMI (2.0a), 2x Intel Gbit LAN 
• 4x COMports, external GPIO 
• 4x USB3.1, 1x USB2.0 
• 12V (-10%) to 24V (+10%) DC-in 
• Operating temperature 0°( to 45°( 
Ibase Technology 
www.ibase.com.tw 
Embedded PC Serves Up 
9th Gen Core Processors 
Kontron's KBox B-202-CFl offers 
high performance with maximum 
expandability and low noise level 
(maximum 34 dB(A)). The embedded 
Box-PC with Intel Core 9th generation 
processors, with up to eight processor 
cores, provides a performance boost 
with constant power dissipation. This 
enables the KBox B-202-CFl to handled 
computationally complex processes and 
large amounts of data. 
• Up to 9th Gen Intel Core i7/i5/i3 
• Raid functionality with up to 
4 storage devices 
• 2x PCie expansion slots for graphics 
or network cards 
• low noise design 
• Removable 2.5" SSD 
• EN55032 Class B certified 
• Operating temperature 0°C to 45°C 
• Size: 190mm x 120mm x 190 mm 
DATASHEET URLS: 
Kontron 
www.kontron.com 
Apollo Lake Industrial PC is 
LYE Certified 
lanner's lEC-7242 is an industrial 
environment gateway IPC driven by Intel 
Apollo lake CPU, is CE, FCC and PTCRB 
certified for LTE or Wi-Fi. It is designed 
for IoT/IIoT edge, wireless security 
and multi-site management solution 
deployment in commercial, enterprise 
and public sector applications. Power 
consumption ranges between 5.4W to 
13.5W. 
• lntel Apollo lake CPU 
• FCC/CE/PTCRB certification for LTE 
or Wi-Fi 
• 2x GbE RJ45, 1x RS232/422/485 
COMport 
• 2x USB 3.0, 1x DisplayPort, 1x HDMI 
• 1x M.2 3042 B+M Key socket with 
duaiSIM 
• 1x Mini-PCiesocket with dual SIM 
• Onboard lPDDR4, max 8GB and 
onboard 64GB eMMC 
• TPM support 
Lanner 
www.lannerlnc.com 
Kontmn www.kontron.com/products/systems/embedded-box-pcjkbox-b-series{kbox-b-202-cfl.html 
l.anner www.lannerinc.com/products{i ntelligent-edge-appliances/embedded-platform/ lec-7242 
Onlogic https://'iJ:iltic.onlogic.com/resources/manuals/Onlogic-Helix-Pmduct-Manuai-Vl.pdf 
ci rcuitcellar. corn 53 
Comet Lake System Targets 
Industrial Edge 
The Helix 500 from Onlogic brings 
desktop processing performance into 
a fanless, silent mini-PC. Powered by 
Intel Comet Lake Core processing, 
Helix 500 can accommodate up to 
64GB of RAM, has room for dual SSDs 
(including NVMe options) and features 
available VPU accelerators, including 
Intel Movidius and Google Coral. 
• Intel Comet lake Core i3/i5/i7/i9 
processors 
• Intel Q470 chipset; Intel UHD 
Graphics 630 
• Up to 64GB DDR4 50-DIMM 
(non-ECC) 
• 6x USB 3.1 Gen 2 ports, 
2x USB 3.1 Gen 1 ports 
• 2x Gbit lAN ports, 3x DisplayPorts 
• lx 12V to 24V, 4-pin DC jack 
• 2x M.2 2280 slots (NVME PCie x1/ 
SATA) 
• Operating temperature: 0 to 50°C 
• Size: 154mm x SO.Bmm x 210 mm 
On Logic 
www.onlogic.com 
54 CIRCUIT CELLAR • OCTOBER 2020 #363 
Embedded PCs 
... ·- .. 
-· "• ", · o ~HHH• 
I 1 • . 
Fanless Computers Blend GPU 
and lntel Gen 10 CPU 
Sintrones' ABOX-5210G Edge AI GPU 
Computing is powered by Intel Gen 10th 
10 Cores Xeon, i9 CPU with 10x GbE 
with Nvidia GeForce GTX GPU. It can 
provide high computing performance 
on graphic processing by harnessing 
the power of its GPU. The fanless box 
computers feature dual hot swappable 
SATA storage RAID 0,1,5. 
• Intel Gen 10 Cores Xeon W-1290TE/ 
i9-10900TE 
• Nvidia Geforce up to GTX 1660 GPU 
• Supports 1408 CUDA Cores 
• Bx GPI, 4x GPO and 
3x RS-232/422/485 
• 1x M.2 B key, 1x M.2 A-E key 
• 2x Mini PCie expansion slots 
• Dual hot swappable SATA storage 
RAID 0,1,5 
• Operating temperature.: 
-40°C to 70°C 
• 10x Gbit Ethernet RJ45 
DATASHEET URLS: 
Sintrones 
Industrial PC Boasts 
Six: Gbit Ethernet LANs 
The Pl-50030 from WIN Enterprises 
is a fanless, high-performance 
networking system based on the Intel 
7th Gen Core Processor. With six Gbit 
Ethernet LANs and 4x USB 3.0 and 2x 
USB 2.0 ports, it is well suited for small 
call centers that demand quiet, but 
high performance. Network expansion 
is through PCie interface modules, M.2 
and Mini PCie technology. 
• Intel 7th gen Core processor 
• DDR4 SODIMM up to 32GB 
• Support for three independent 
displays 
• Rugged and fanless embedded 
system 
• 4x PoE + 2x LAN, Sx COM, 4x USB 
3.0, 2x USB 2.0, 1x M.2x NVMe slots, 
16-bit DIO 
• lx PCie X16 or lx PCI, 2x Mini PCie 
• Quiet fanless operation 
• Wi-Fi antenna 
Sintrones IMNW .si ,:rones. cor"' f datasheels/ ABOX-52 lOG. pdf 
WIN Enterprises IMNW. w i n-ent. com/i mag es/sto •i es/down load/catasheets/M B-50030_ v2_DS. pdf 
WIN SYSTEMS https ://resources. wi nsystems. eo m/ da:asr eets/sys-i t:x-n-3900-ds. pdf 
Embedded Box: PC Offers 
Wide Temp Range 
WINSYSTEMS' SYS-ITX-N-3900 is 
an industrial embedded computer 
based on the Intel Atom E3900 Apollo 
Lake processor family. It uses less 
than 12W for fanless applications 
and performs reliably in industrial 
operating temperatures ranging from 
-200 to +600C. It includes a SODIMM 
socket supporting up to 8GB of DDR3 
system memory and internal mounting 
support for one 2.5" SATA HDD/SSD. 
• Intel Apollo Lake E3950/E3940 
processor 
• 2x 10/100/1000 Mbps Ethernet ports 
• DisplayPort and VGA outputs for 
dual displays 
• Up to 8GB DDR3-LV RAM 
• 2x USB 3.0, 1x serial port, 
audio output 
• lx MiniPCie Socket with 
mSATA support 
• lx M.2 Socket, with E key, type 2230 
• 1x SATA 3.0 ports and lx Micro SD 
for storage 
• Operating temperature: 
-20°C to +60°C 
WIN SYSTEMS 
ms.co 
56 CIRCUIT CELLAR • OCTOBER 2020 #363 
Embedded in Thin Slices 
By 
BobJapenga 
FIGURE 1 
R ecently I was in a meeting with 12 young professionals. We met over Zoom, and some of us gathered 30 
minutes before the meeting began 
in order to share what's happening in our 
lives. One of the young men was excited about 
working on a Raspberry Pi loRa concentrator 
{gateway). His company makes sensors and I 
was excited to see that his company is now 
getting into loRa sensors. 
That's what we've been discussing in the 
past six articles. This month we wrap up our 
not-so-deep-dive into loRa. We are going to 
look at building a prototype loRa device for 
The MQSparl<. development kit comes with four Grove sensors for measuring temperature, sound, distance/ 
speed and orientation. 
the MachineQ network using a development 
kit and the example software. For testing, 
we will be using the off-the-shelf MachineQ 
loRaWAN gateway. As I did with building a 
device and gateway for The Things Network 
{TTN), I want to emphasize the prototype 
nature of what I did. That's because I don't 
want to mislead you into thinking: "Wow! 
That part is all done and all I need to do is 
concentrate on my specific application." 
Au contraire. Using this hardware and 
software will get you going for a proof-
of-concept prototype, but it is a long way 
from a certified, ready-to-build-and-deploy 
product-especially in terms of the software. 
We all have used example software that was 
little more than 10% of the finished product-
all-the-while hoping that we could use that 
example software and be done with it. 
In addition, I am a little concerned about 
the robustness of the off-the-shelf gateway 
I was using from MachineQ. Two days ago, 
I disconnected the Ethernet cable from the 
gateway at my router and put the cable 
into a different port. The gateway stopped 
connecting for 36 hours until I powered it 
down. I was already told there were problems 
with Wi-Fi in this area. This does not bode 
well for 24/7 unattended operation. 
REMOTE RESET WOES 
As a side note, we experienced this several 
years ago with an off-the-shelf cellular modem. 
We were in the midst of a feeding frenzy in 
Ontario when solar energy suppliers were 
knocking on doors asking homeowners if they 
could rent their roofs to install solar panels. 
Canada was recognizing that it needed to supply 
more electrical energy than its infrastructure 
could provide, but did not want to invest in 
the fossil fuel infrastructure [1]. The provincial 
government was paying an outrageous amount 
for homeowners to put electricity onto the grid 
using photovoltaics {$0.802/k.W-hour! [2]). I 
pay just over $0.09/k.W-hour for fossil fuel 
generated electricity. 
Many companies were tripping over each 
other to get solar panels on every possible 
roof as fast as possible. We rushed out a 
modification of a design of a solar power 
energy monitor for one of our clients that 
used an off-the-shelf cellular modem powered 
by USB. Once a few were in the field, we found 
that some modems would stop communicating 
and require a reset. We quickly installed a 
fix to allow us to remotely power-cycle the 
modem. Ever since then something equivalent 
to this has been part of virtually all of our 
designs that incorporate complex (software) 
off-the-shelf systems. Basically, we don't trust 
suppliers to create robust devices. You just 
cannot send technicians to the field to reset 
thousands of devices. So, if you are using 
anyone's off-the-shelf gateway in a place that 
cannot be easily reset, consider providing the 
ability to remotely reset it or power cycle it. 
Getting back on track, this month we will 
look at the hardware and the software 1 used to 
put together one LoRa node on an off-the-shelf 
LoRaWAN gateway on the MachineQ network. 
THE HARDWARE 
The LoRa device: For the LoRa device, 
MachineQ offers two development kits: The 
MQSpark is the simplest and comes with four 
Grove sensors for measuring temperature, 
sound, distance/speed and orientation 
(Figure 1). There are literally hundreds of 
Grove sensors [3] for you to experiment 
with. Robotics,motion, environmental and 
proximity sensors are available to rapidly 
prototype your new design. The device plugs 
into a USB port for power and communications. 
It uses the Arduino IDE for development. 
All that said, I chose to use the other 
offering: the STMicroelectronics (ST) LoRa 
Development kit, which is for those who 
want to design their own sensor and require 
more advanced software and hardware. 
This kit consists of two boards. The main 
board (B-L072Z-LRWAN1) and the X-NUCLE0-
1KS01A2 daughterboard (Figure Z). 
Main board: Also known as the Discovery 
kit for LoRaWAN, Sigfox and LPWAN 
protocols, this main board contains a Murata 
CMWX1ZZABZ-091 module (Figure 3). This 
module contains an ST Arm Cortex MO+ Core 
with 192KB of flash, 20KB of RAM and 20KB of 
Serial EEPROM; a Semtech SX1276 LoRa radio; 
and variety of I/0 (ADCs, DACs, USB, serial 
ports, discrete I/0, 7 LEDs, two pushbuttons 
and more). There is a battery holder on board 
for 3 AAA batteries. However, as shipped, the 
battery is not connected. You have to cut a 
jumper to connect the battery. Unfortunately, 
you then forfeit the ability to use the USB 
for loading, debugging and monitoring the 
software. Ouch! Very unfriendly for real 
development. I decided to buy a USB battery 
charger for a phone and use that for my range 
testing. I was impressed by the supported 
development toolchains: Kell, IAR Embedded 
Workbench for Arm, GCC-based IDE's like 
Eclipse, and Arm embed. I used the Keil IDE, 
ci rcuitcellar. corn S1 
FIGURE2 
The ST LoRa Development kit that 
I uS€d consists of two boards. The 
main board (B-L072Z-LRWAN1) 
(bottom) and the X-NUQEO-IKS01A2 
daughterboard (top). 
U'l.WII' 
IIAA1' 
"" ac: 
STM32L''' U'·Tiol 
use 
CIPIOoi!IW 
FIGURE 3 
Murata CMWXlZZABZ-091 module 
T-s-o OliUoNI 
MCO 
I!IOOTII 
15877542
15877542
Centuries before the science of color temperature was understood, blacksmiths 
already knew how to read the temperature of a heated piece of metal by its 
color emissions. In this article, leff es the math and science behind 
color temperature. He then details his project that uses a color-lighHo-digital 
sensor to measure the color temperature of a light source. 
11 w hite assnow,U 0 white chocolate," 0 White 
wedding dress,u "white lightning" 
and "white diamondsu are common 
descriptions of items in our world that 
include the descriptor "white." You can probably think of many 
others. Some of these describe a whitish hue, while others 
don't connote a color at all. The most truthful might be 0 White 
hot." While today we might use this to describe someone that 
is udrop-dead gorgeous, u according to Merriam Webster, this 
was first used as meaning "being at or radiating white heat.n 
If you've ever sat by a campfire or gazed at a candle, you have 
experienced heat through radiation and colors associated with 
the flame. 
Before we can discuss temperature and color, we need to 
begin with the "black-body radiato~ [1]. This is a theoretical 
object that can absorb all radiation that falls upon it. Radiation 
is the emission or transmission of energy in the form of 
waves (or particles) through some medium. This includes: 
electromagnetic radiation, such as radio waves, microwaves, 
infrared, visible light, ultraviolet, x-rays and gamma radiation 
(y). At absolute zero-the coldest temperature possible-no 
heat is generated in a black body. Absolute zero is -273.15°C 
(Celsius) or -459.67°F (Fahrenheit). Science often uses the K 
(Kelvin) temperature scale to make more sense. 
K = °C + 273.27 
so absolute zero = OK 
... and freezing = 273.16K 
As a black-body radiator's temperature rises, it gains 
energy and will emit radiation beginning at the low end 
of the electromagnetic spectrum. Figure 1 is a good 
illustration comparing radiation type and temperature for the 
electromagnetic spectrum [2]. We experience this emission as 
color once its wavelength enters the visible spectrum: 400nm-
700nm. A blacksmith heats a piece of steel in his forge to 
affect its malleability. While it's not a perfect black body, the 
physical properties of the metal change with temperature, 
and he can estimate its temperature by its color emission 
alone (Figure :Z) [3]. 
We can therefore define the color temperature of a light 
source in terms of the absolute temperature of a black-
body radiator or "chromaticity." Chromaticity is the quality 
of color, independent its brightness or luminescence. 
Chromaticity consists of two independent parameters, often 
specified as hue (h) and as colorfulness (s). Colorfulness is 
alternatively called saturation or chroma. A color space 
Penetrates EarUfs 
Aunosphere? 
y --.-. ·. --
Radrabon Jypc Radio MicrOONave Infrared Visible u ltraviolet x -ray Gamma ray 
1tr 10, 10 .. 0.5• 10 ... 10 8 10 10 10 17 V\lavelenylh (m, 
A pproXJmate Scale 
~· 11~ -P AC ~ of wavelength }) ~I Buildrngs Hun-.ans Bu ne rftre s Needle Pomt Prorozoans Mol ecules AlOrns Aromlc Nuclei 
Frequency (Hz) 
~mperaiUre of 
ObJeCts at whrch 
thrs radrauo n rs the 
most rmense 
wavelengdr emnted 
chromaticity diagram (Figure 3) defines all 
visible calor based on these parameters [4]. 
Note: Temperature calor has been indicated 
within the calor space. Green and violet fall 
outside the black body path or Planckian 
locus (see physicist Max Planck), with green 
(secondary) being a combination of primary 
yellow and blue, and violet (secondary) being 
a combination of primary red and blue. 
ADDI11VE I. SUBTRACTIVE COLOR 
This has to do with the source of the 
calor. If the source is projected, the three 
wavelengths used are RGB (red, green, blue) 
and are additive. If the source is reflected, 
then the pigments used are CMY {cyan, 
magenta, yellow) and are subtractive. Note: 
Many of us were taught early on that the three 
primary colors were RYB (red, yellow, blue). 
Figure 4 shows the computational effects of 
each system. 
These are the hues of chromaticity. The 
other aspect of chromaticity-colorfulness, 
saturation or chroma-has to do with its 
intensity. If our brains receive equal signal 
strengths from all three types of cones (calor 
photoreceptors) in our eyes, then we would 
perceive white light. If these cone outputs 
were reduced from maximum to minimum, 
the resultant perception would be a fade to 
black. Likewise, if some other mix of cone 
outputs were perceived as, say, green, and 
these outputs were reduced to minimum 
while remaining in the same proportions, the 
perception would be a fade to gray. 
10000 ,000K 
-1o.ooo,ooo •c 
the expected colors. However, if the source 
were lacking some frequency, the resultant 
perception would be skewed. The missing 
frequency cannot be reflected, and so the 
result will be lacking. We have two options for 
a light source-the sun or an artificial source. 
You've most likely noticed the temperature 
calor is the opposite of the way we think of 
temperature and the visible calor spectrum 
(as red = hot and blue = cold). With calor 
temperature, red is of a lower temperature 
and therefore cooler than blue. The sun's 
effective temperature, defined by the total 
radiative power per square unit, is about 
5,800K. During sunrise, the tangential angle 
of light passes through more atmosphere, 
scattering the shorter-wavelength blue light 
from our view. As the path becomes a more 
FIGURE2 
ci rcuitcellar. corn 61 
FIGURE 1 
Shown here is the frequency/ 
wavelength /color temperature of the 
electromagnetic spectrum of energy 
[2]. Note those areas of protection by 
our atmosphere. 
We assume that a (white) source 
illuminating a subject is perfect. A perfect 
white source would give the expected result. 
The subject under observation would absorb/ 
reflect the wavelengths, and we would perceive 
A blacksmith uses tlle col or emissions given off by a heated metal to estimate its temperature [3]. This is 
key to being able to rorm and temper it. 
62 CIRCUIT CELLAR • OCTOBER 2020 #363 
0.9 
0.8 
0.70.6 
500 
0.5 
y 
0.4 
0.3 
0.2 
0.1 
FIGURE 3 
direct angle, this scattering can be seen off-
axis as blue skies. 
Figure 5 shows how the source calor 
temperature can vary between the sun 
and artificial lights. The calor temperature 
value of a light source refers only to the 
visual appearance of the source. It does not 
necessarily describe the effect this source 
0.3 0.4 
X 
0.5 0.6 0.7 0.8 
The OE (International Commission on Illumination) 1931 color chromaticity diagram was the first to 
demonstrate the link between wavelengths in the electromagnetic visible spectrum, and physiologically 
perceived colors of human col or vision. Note the plot of chromaticities of black-body light sources of various 
temperatures [4]. 
FIGURE 4 
will have on photographs or digital images. 
The calor temperature does not take into 
consideration the spectral distribution of a 
visible light source, and is not a reliable means 
of selecting suitable filters or creating look-
up tables for calor balance corrections. While 
different light sources may be described as 
having the same calor temperature, without 
proper white balance, spectral output cannot 
determine the correct adjustments for calor 
temperature balance. So, let's look at one 
possible sensor that will allow us to get a feel 
for the spectral content of light exposed to it. 
TCS3472 
The TCS3472 is one of many sensors 
manufactured by ams, a global leader in the 
design and manufacture of advanced sensor 
solutions. It consists of three sets of four 
filtered sensors protected by an IR filter. 
Three of the four filtered sensors have calor 
filter RGB. Sensor values and control registers 
are available via an FC interface. 
The amount of light falling on each set 
of silicon photodiodes is measured by an 
integrating analog-to-digital converter (ADC), 
one for each calor. All ADCs operate in parallel, 
so all ADCs use the same integration time. The 
device covers a wide signal range. For low-
level signals, higher programmable gains and 
longer integration times can be used. At the 
other extreme, higher-level signals will require 
lower gains and shorter integration times. 
The integration time period is chosen as 
a multiple of 2.4ms. A saturated device will 
present a count of 1,024 for each period. So, 
you want to make sure the total count (for any 
sensor) I number of periods is less than 1,024. 
This is an analog saturation check. A sensor's 
total value is held in a byte register pair, and 
therefore is limited to a maximum of 65,535. 
It is possible to have no analog saturation, 
but choose a high number of integration 
periods, where the total count is greater than 
65,535. This is a digital saturation check. You 
would need to adjust gain and/or number of 
integration periods to keep the results high 
On the right is the RYB (red, yellow, 
blue) plot of colors we learned to mix 
with crayons in kindergarten. The 
pigments limit/filter any light reflected 
to our eyes. On the left is the RGB 
(red, green, blue) plot of a monitor 
reproducing calor through the additive 
process of I ight sources. In the center 
the CMYK (cyan, magenta, yellow, 
black) plot shows how a printer mixes 
inks in a process similar to crayons. 
While equal parts of CMY equal K 
(black), it's a waste of colored inks to 
make black, so when black is needed, 
a separate black ink is used. 
RGB (Additive) CMYK (Subtractive) RYB (Subtractive) 
enough to provide good resolution, without 
going into analog or digital saturation. 
The register set for the TCS3472 is 
presented in Figure 6. You'll note that the 
COMMAND register is not really a physical 
register, but rather a pointer used to access 
registers OxOO-OxlB and alternately clear an 
interrupt. Besides pointer data, the COMMAND 
byte also enables/disables the auto increment 
of the pointer register function. 
The device has an interrupt output 
that can be triggered by a clear channel 
conversion count that is less than or greater 
than minimum and maximum values you 
program into registers Ox04/5and Ox06/7. This 
interrupt can be further limited by requiring 
a number of consecutive out-of-range counts 
from 1 to 60, programmed into register OxOC. 
The status of a conversion is found in register 
Ox13. Channel counts follow in registers Ox14-
0x1B. The only interrupt that can be generated 
is the out-of-bounds interrupt, and it is not 
used in this project. Note that the external 
LED is controlled via this interrupt output, and 
it can be enabled and disabled via the AIEN bit 
in the ENABLE register (OxOO). 
You must set a number of registers to control 
the operation of the TCS3472. There are timing 
and amplification areas of configuration. First 
is the general timing of the device. Assuming 
the device has been enabled (AEN, WEN, POW), 
it will continuously idle and sample the sensors. 
The idle or wait time is set by the wCycl es or 
WTIME. Each cycle adds 2.4ms to the operation. 
The cycle values are two's complement values 
(inverts all bits and adds 1). Therefore, to set 
the minimum wCycl es equal to 1 {2.4ms), 
OxFF must be written to the WTIME register 
(Ox03). The CONFIGURATION register (OxOD) 
holds the WLONG bit that, when enabled, will 
multiply the WTIME by 12. The maximum 
WTIME will be approximately 7 seconds. This 
allows the device to remain in idle condition 
for a large percentage of time, lowering the 
average current of an active device by a factor 
of 5. However, device current is only 0.3mA to 
begin with. 
The integration time has a similar register 
(ATIME Ox02). You set the integration time 
in periods of 2.4ms. If you wanted to set the 
integration time to 50ms, you would use an 
aCy cl e count of 50ms /2.4ms = 20.8 (or 21 
periods). The two's complement of 21 (Ox15) 
is 235 (OxEB). There is no multiplier for the 
integration time. However, you can adjust the 
gain of the sensor amplifiers, which is similar 
to a multiplier, without actually adding any 
more time to the conversion. 
You can see that if all the timing registers 
are set to maximum, one sample period would 
be WTIME = OxOO (614ms), WLONG = 1 (times 
12) 614ms x 12 = 7.4 seconds, plus ATIME = 
8,000 K 
6,500 K 
5,500 K 
4,500 K 
3,000 K 
2,000 K 
FIGURE 5 
Outdoor 
source 
Snow, water 
b lue sky 
Large shadows 
blue sky 
Average day 
light, central 
latitudes 
Noon sunlight 
Average day light, 
Northern Hemisphere 
Early morning 
late af te rnoon, 
and evening 
sunlight 
Indoor 
source 
Xenon 
flash 
ci rcuitcellar. corn 63 
Blue bulb 
f lash cube 
Fluorescent 
"Warm w hite" 
tubes 
dear flash 
bulbs 
Photofloods 
Photo lamps 
Household 
150/200W 
60/40W 
25W 
Candle light 
This chart shows how the sun (outdoor source of light) and various types of artificial lights 0ndoors) produce 
calor temperatures from a reddish (relaxing) tint to an bluish (intense) tint. In their art, photographers use 
these subtle tints of white light to produce an emotion. 
AllORESS RoalVII<AME 11/W MGIITEA fUNCllOfl RU ff VALUE 
~ w .....,_........,_ c.oo 
IWXI £NAIIlE 1\W 
.,_ ___ 
c.oo 
a.Dt ... ..._ - AGIIC- ,.,. IIIG3 W'llloiE 1\W '~'~"'- IW'f - ALn MY a..._ ..... _ .... byto IWXI 111111 Al. Ill MY a...-to.r-1'111•""" c.oo 
OIIDII ~ A.W Clool -. ...,_ ..... 01100 
011111 Qffi4 A.W cao.. ............. -N11..,.. IWXI 
o.oc: PEAS MY 
~--·· 
0.00 
Qd) oor.m MY ~ 111100 
01111' CONTROl MY OllfWU 01100 
01112 tO R _., tO 
01113 STATW R o.--.. c.oo 
Oat• COAT N.. R a---- 111100 111115 COATAH R aoor .... ...,_ ClliOO 
01111 IIDIITN.. R -- ..... -~ IWXI 
01117 -TAH A --¥1¥0 IWXI 
!loll OQAW_ " a.--~t.,w Odl Olllt OOATAH R a.-- ...... 0.00 
0111A IIOATN.. R 
__ .,....,.,.. 
IWXI 
01118 BOATAH A 
__ l'oil;ll..,.. 
c.oo 
FIGURE 6 
These are the PC registers of the TCS3472x. Note: the first data byte of a write wi 11 be a COMMAND byte, 
which sets the pointer to a register of interest. Additional data bytes are stored beginning at the pointer. A 
read will produce data,also beginning from that pointer. 
64 CIRCUIT CELLAR • OCTOBER 2020 #363 
OxOO (614ms) 7.4 seconds + 0.614 seconds = 8.014 seconds. 
Reading the resultant registers prior to this time would 
provide the data for the previous sample only. 
USING THE SENSOR 
Since the TSC3472 is an I2C device, it can be easily 
interfaced using a 3.3V Arduino (or a SV Arduino with some 
level shifting on the I2C pins). I'll begin this project with an 
Arduino MEGA. A library for the TCS3472 is available, but I 
won't use it for this project. Instead I'll break it all down into 
visible function calls. 
Let's begin with reading all of the registers. To do this 
we'll need to write a command to the device. As we discussed 
previously, the COMMAND byte sets up the pointer to the 
register we want to communicate with (either to read from 
or to write to). It also instructs the device whether or not 
11 This example code is in the public domain. 
String SignOn = "TCS3472 Probe 5/20/2020"; 
#include <Wire.h> 
byte I2CADDRESS = Ox29; 
#define RB OxOO 
11 repeat same address 
#define AI Ox20 
11 auto increment address 
//define SF Ox60 //special function 
byte TYPE = AI; 
byte REGISTER= 0; 
int REGISTERVALUE = 0; 
const i nt REGISTERCOUNT = OxlC; 
i nt CHARACTER; 
unsigned int BYTEARRAY[ 255 ]; 
const unsigned long I 2CCLOCK = 400000; 
11-----------------------------------------
// setup() 
1/-----------------------------------------
void setup() 
{ 
Wi re.be gin(); 
11 join i2c bus (addres s opt ional for 
ma ster) 
Wi re.set Clock(I2CCLOCK); 
Serial.begin(ll52 00 ) ; 
I I start serial f or output 
Se r i a l . p r i n t l n (S i g nOn ) ; 
} 
//-----------------------------------------
// loop() 
/1-----------------------------------------
voi d loop ( ) 
{ 
} 
r eadi2CCl: 
di splayAll Regi sters( ); 
/ /-----------------------------------------
// readi2C () 
/1-----------------------------------------
void r eadi 2C( ) 
{ 
AODR = 0; 
ifCwrit ei 2C ( 0)) 
{ 
I..ISTlNG 1 
(continued} 
The basic Arduino code needed to read all the registers in the TSC3472x 
to increase this pointer automatically after each register 
access. Using a COMMAND value of OxAO will set the pointer 
to register OxOO and enable auto increment. Since the device 
uses registers Ox00-0x1B, we will need to read Ox1C (28) 
registers values. Listing 1 shows the basic Arduino code 
needed to read all of the registers in the TSC3472x. 
In Listing 1 we've collected 28 registers, and they are in 
8 YTE ARRAY[ J. Displaying these requires the majority of 
programming, and will therefore not be listed here. However, 
Ustlng 2 shows a typical display of the output from this function. 
There are actually two phases for each register. The first gives 
raw register info, register name and hex value found there. The 
second breaks this down-where necessary-indicating any 
special bit significance. 
A check of the value in the ID register (Ox12) tells me the 
values are good. This device is actually a TSC34725 (a version 
(Listing 1 continued) 
int o = 0; 
if (Wire . request From( I 2CADD RESS , 
REGISTERCOUNT)) // re quest bytes fr om 
slave device 
{ 
while ( Wi r e.available()) 
11 slave may send less t han req uested 
{ 
int c = Wi re.rea d() ; 
11 receive a by te as ch ar acter 
BYTEARRAY [ o] = c; 
} 
else 
{ 
o++ ; 
Se r i a l • p r i n t l n C "Nobody There ! " ) ; 
} 
// -----------------------------------------
// writei 2C(int c ) 
1/ -----------------------------------------
bool ean writei 2C(by t e c) 
{ 
byte COMMAN D = OxBO I TYPE I ADDR; 
bool ean q = f al se ; 
Wi re.beg i nTransmiss ionC I 2CADD RESSl ; 
11 t ransmit to dev i ce 
Wire.write(C OMMAN D) ; // sends one byte 
wh i le( c) 
{ 
c-- ; 
I/ 
Wi re.write CBYTEARRAY[REGI ST ER + c ] ); 
send s one byte 
} 
ifCWire.endTr ansmi ssion() 
11 st op tran smitti ng 
{ 
q = t rue ; 
} 
return q; 
== 0) 
that uses an I2CAddress of Ox29 and runs 
off 3.3V). Note that some registers are not 
used (or at least have not been defined). 
With the displayAllRegisters() function 
working, we can expand this program by 
adding a menu to allow us to choose alternate 
functions. 
The next function that will be useful is to 
change a register's contents. Only the first 
registers OxOO-OxOD are R/W (read/write} 
registers. Six of these are byte oriented, and 
four are word oriented. A sub menu will allow 
you choose which of these you want to change. 
----Change Register----
0 - OxOO Enable 
1 - Ox01 Integration Time 
2 - Ox03 Wait Time 
3 - Ox04/5 Low Threshold 
4 - Ox06/7 High Threshold 
5 - OxOC Persistence 
6 - OxOD Configuration 
7 - OxOF Control 
Menu selection? 
When a choice is made here, you get to 
enter a value for the byte or word register 
pairs. REGISTER and REGISTERVALUE 
variables are used to update a single register 
in the TSC3472x. 
The Main menu below gives you total 
control ofthe device at your fingertips. We can 
stop, start and compare conversions to find 
the optimum settings for any light condition. 
----TC53472 Probe----
1 - Read and display all I2C registers 
2 - Change an I2C register 
3 - Get a raw sample 
4 - Toggle external LED 
5- +Gain 
6-- Gain 
7 - + Integrate Time 
8 - - Integrate Time 
9 - Sample and auto adjust 
Menu selection? 
Note that while complete control of all 
registers is available, the main menu has 
shortcuts that allow adjustments of the 
two main parameters that you will want to 
change. The Gain sensor and its Integration 
Time can be altered directly via single-keys 
commands. I added a command to toggle the 
on-board LED light source that can be used 
for reflective measurements. Command 3 
retrieves a sensor sample using the present 
configuration. The final Command, 9, makes 
fine tuning and sampling automatic. 
AUTOMAGICAL 
To make the magic happen, we need to 
make changes continuously to the Gain and 
Integration Time, and look for that point 
just before the system reaches saturation. 
Previous discussion showed both analog and 
digital saturation. It was identified that if the 
Integration Time exceeded a count of 63 (63 x 
2.4ms = 151.2ms), unsaturated sample could 
occur up to the maximum counts of 65,535. 
For Integration Times less than 64 counts, the 
maximum counts allowed were substantially 
less. Using these rules, we can determine if 
saturation occurs for a conversion. 
OxO - Enable States value= Oxlb 
Interrupt is Enabled 
Wait is Enabled 
Sensor is Enabled 
Power is Enabled 
Oxl - Integration Time value Oxec 
Integration Time = 48.00ms 
Ox2 - not implemented = Ox34 
Ox3 - Wait Time value= Ox9 
Wait Time = 592.80ms 
ci rcuitcellar. corn 65 
Ox4 - Clear Interrupt low t hres hol d value CLSB) = OxO 
Ox5 - Clear Interrupt low t hres hol d value (MSB ) = Oxd e 
Clear channel low threshold= 56 832 
Ox6 - Clear Interrupt high t hr eshold value CLSB) OxO 
Ox7 - Cl ear Interrupt high t hreshold va lue CMSB) = OxO 
Clear channel high threshold = 0 
Ox8 - not i mpl emented OxO 
Ox9 - not impl emented = OxO 
Oxa - not i mplemented = OxO 
Oxb - not impl emented = OxO 
Oxc - Interrupt Persist ence Fil ter value = OxO 
Consecuti ve additional Max/Mi n th r esholds = 0 time s 
Oxd - CON FIGU RATION valu e = Ox2 
Wait multipli er= 12 
Oxe - not i mplement ed = OxO 
Oxf - CONTROL value = Ox3 
Gain = 60 
OxlO - not i mplemen t ed OxO 
Ox11 - not impl emented = Ox1 
Ox1 2 - I D value= Ox 44 
TCS347 2115 
Oxl3 - STATUS value = Ox11 
Interrupt is Set 
Sampl e i s Valid 
Ox14 - Cl ear Data value CLSB) = Ox69 
Ox15 - Clear Data value CMS B) = Oxc 
Cl ear channel raw dat a = 3177 
Ox16 - Red Data value CLSB) = Oxee 
Ox17 - Red Data value CMSB ) = Ox 5 
Red c hannel raw da t a = 1518 
Ox1 8 - Green Data value CLSB) = Ox2 
Oxl9 - Green Data value CMS B) = Ox5 
Green channel raw dat a = 1282 
Ox la - Blue Data value Cl SB) = Ox2 a 
Ox lb - Blue Dat a value CMSB ) = Ox4 
Blue c hannel raw data = 1066 
I..IS'TlNG 2 
The output of the display.AI/Registers() function, which includes a bneakdownof the register bit status. Note: 
All numbers are in hexadecimal and pneceded by Ox. 
66 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE7 
Shown here is tne resultant display 
(SSD1306) for a sample oftne mom ing 
sky prior to the sun breaking the 
horizon. There is very little light 0-ux), 
with a reddish color temperature of 
approximately 4,3821<. 
Since we want to determine the point 
just before saturation occurs, my algorithm 
begins with minimum Integration Time and 
maximum Gain. If saturation occurs, then we 
must turn the Gain down a notch, since the 
Integration Time is already at its minimum. 
Gain has four steps-60, 16, 4 and 1. Once 
we have a conversion with no saturation, 
the Gain is now valid, and we can ramp up 
the Integration Time until saturation again 
occurs. 
The Integration Time has 256 steps 1-256 
cycles (where 0=256). It could take a few 
seconds to try all possible steps. As each 
step increases the Integration Time, the 
conversion time also increases. While you 
could shorten the potential number of steps 
by using a different algorithm, I will keep this 
as simple as possible and just increase by a 
single step until saturation occurs, and then 
back off 1 step (count). We should now be 
at the maximum Gain and Integration Time 
without going into saturation. 
LUX a. COLOR TEMPERATURE 
lux and calor temperature calculations can 
be performed on any unsaturated conversion. 
ABOUT THE AUTHOR 
Jeff Bachiochi (pronounced BAH-key-AH-key) 
has been writing for Circuit Cellar since 1988. 
His background includes product design and 
manufacturing. You can reach him at: 
j eff.bachiochi @imaginethatnow.com or at : 
www.imaginethatnow.com. 
Additional materials from the author are available at: 
www.ci rcu itcellar.com/a rticle-materials 
References [1] and [3] as marked in the article can be found there. 
RESOURCES 
Adafrui t I www.adafruit.com 
ams I www.ams.com 
Espressif Systems I www.espressif.com 
Heltec Au tomation I https://heltec.org 
PacTec Enclosures I www.pactecenclosures.com 
The previous configurations changes have been 
done to optimize the conversion values for the 
best possible accuracy when calculating the 
lux (intensity) and calor temperature (surface 
temperature in Kelvin). The intensity of light 
hitting the sensors (RGBC) can be translated 
into a value that closely matches the human 
eye, by totaling the adjusted red, green and 
blue conversions and applying a device factor. 
You can see that as long as the color outputs 
stay in proportion to one another, the lux has 
virtually no effect on the Color Temperature, 
which is essentially a ratio of blue to red 
values. 
So far, we've needed a PC connected to the 
TSC3472x to get a conversion. Heltec makes 
a DIP-style Arduino with an integrated 4x20 
graphic display for around $10 that is based on 
the Espressif Systems ESP8266. The ESP8266 
has Wi-Fi, which we won't be using here, but 
otherwise it is perfect for this project, as you 
will quickly see. I did use Adafruit's support 
for its SSD1306 display, because Heltec's 
library doesn't have good text support. No 
special changes to the Arduino program are 
necessary, beyond that of the display support. 
However, I want to make this simple and the 
menus will have to go. Well ... not really. 
I want to use two functions in conjunction 
with the display. I want the user to be able 
to start an automatic conversion, and also 
have control over the LED. To do this, I need a 
couple of push buttons and a way to convert the 
button push into a simulated menu selection. 
This was done in the getUserlnput() 
routine that collects keyboard input. 
Normally, execution is waiting here for you to 
enter characters (menu choice). I added two 
d i g it a l Re ad () tests-one for each button. 
A button press (grounded input) forces exit 
of the get User Input() routine, returning 
the character "4" (Toggle external LED) or "9" 
(Sample and auto adjust). Other than how 
these buttons produce virtual keyboard input, 
nothing else changes. 
Although the display is graphic in nature, 
you have control over every pixel on the 
screen. I am using it as a four-line-by-20 
character display. Sign-on messages are 
produced on the first two lines, and the second 
two lines are used to show the progress of 
the conversion process. This should end up 
with a display as shown in Figure 7 when 
the conversion is complete. To produce text 
characters, the library routine is responsible 
for providing the character glyph (picture) of 
each character, as well as where to place the 
character within the display's screen. 
The SSD1306 uses separate display 
and buffer spaces. This allows the user to 
completely draw a screen in the buffer space 
before transferring it to the display space. 
Updates therefore look instantaneous. The buffer remains 
intact, and you may continue to add to it or clear it in 
preparation for new text (or graphics). 
PORTABLE 
Up to this point we've received power from the USB port, 
while attached to the PC. We added buttons to make it a 
stand-alone device, but we must do something about power! 
The Heltec device not only has a battery input, but also-
because everything works on 3.3V-a single cellli-lon battery 
is a perfect mate. The Heltec Wi-Fi Kit 8 also has a Li-lon 
battery charger built in to truly make all of this work with 11no 
strings (or wires) attached." 
I mounted the Wi-Fi Kit 8 in a PacTec enclosure (Figure 8). 
I wanted to be able to aim the sensor independently from 
the display, and found an itty-bitty book light that looked 
like it would work well. The four-wire sensor cable had to be 
replaced with a couple of pieces of flat cable that I zipped 
down to two conductors wide. These could be fed through 
the sensor arm quite nicely, without any pinching to the 
wires. 
To test this rig out, I got up before sunrise to make an 
early morning reading before the sun broke the horizon. Look 
back at Figure 7 for this sample. You'll note the Lux value 
is less than 1 and the calor temperature is 4,382K. This is 
just what I would suspect for sunrise. A clear, mid-day sky 
produced a high calor temperature of 5,732K. You get a good 
sense of the many shades of white while driving at night. The 
range in calor temperature for various styles of headlights is 
3,000K to more than 12,000K! 
University of l(~ 
CINCINNATI IONLINE 
ci rcuitcellar. corn 67 
FIGURES 
I used one of the enclosure boxes from PacTec that I already had. Although it was 
a little large, once I added buttons and the Li-Ion battery (which fit nicely in the gv 
battery box), it was fine. 
Picking out a perfect shade of white is a nightmare come 
to life. Pure white? Off white? Simply white? Benjamin Moore 
has more than 35 shades of white! The shade of white (or 
other calor) you apply to your walls will greatly influence the 
way you feel in that room. 
Remember to wash your hands! Too much to do, too 
little time. e 
W elcome back to "The Darker Side." As electronics designers, we all know what a wire is 
and how it works, right? If we 
connect a wire from A to B, a current can flow 
between these two points. If we then disconnect 
the wire from B, no current can flow. The wire 
becomes useless because one of its ends is 
open. 
True? Well, not always. If you don't know 
why, then keep reading because this will 
be my topic this month. More precisely, my 
goal is to explain what a stub filter is, and 
how to build one. I will also show you that 
you may inadvertently make such a filter, for 
example with an improper printed-circuit 
design. This may jeopardize the performance 
of your project, at least if you are using high 
frequencies or fast digital circuits. 
Finally, and as usual, I will not use math 
to convey these principles, but rather words 
and small experiments. So, the good news 
is that this article should be easy to read. 
The downside is that my explanations will 
approximate. Anyway, I hope that you will 
catch theimportant points. I will provide 
some resources for further reading, if you 
want to dig more deeply into the subject. 
SINE+SINE = 0 
Let's start with a very simple mental 
experiment (Figure 1). Assume that you have 
a sine-shaped signal propagating on a wire. 
Such a signal has a given wavelength, usually 
noted as A. (the Greek letter, "lambda"). This 
wavelength is the propagation speed of the 
signal along the wire, divided by the frequency 
of the signal. On a wire, or more exactly on 
a so-called transmission line comprising the 
wire and the accompanying current return 
path, the propagation speed will be close to 
the speed of light in free space, c, which is 
300,000,000m/s. 
Now, imagine that we delay this signal by 
half the wavelength. We get the same sine 
signal, but with an opposite phase. What 
happens if we then sum the original signal 
and its copy delayed by A./2? Naturally, they 
will cancel out, and the result will be zero-a 
null signal. 
I'm sure you've followed me up to this point, 
right? So, you will also easily understand that, 
rather than delaying by 'A./2, we would get the 
same result if we delayed the signal twice-
each time by 'A/4. Now look at Figure 2 for an 
easy way to do it. The signal is coming from 
the left, and is virtually split into two parts. 
Half of it continues to the output on the right, 
whereas the other half goes into a wire that 
introduces a A./4 delay. Assume that there is 
a reflector at the end of this wire. The signal 
will then be reflected back on the same wire, 
will be shifted again by A./4 on its way back, 
and will be merged with the original signal. As 
in my example above, this adds the signal and 
a copy shifted by half a wavelength, and we 
should get zero. 
You may wonder how to actually make 
such a signal reflector, but long-time readers 
will remember an article I wrote several years 
ago called nTime Domain Reflectometryn 
(Circuit Cellar 225, April 2009). When a 
signal propagates on a transmission line and 
encounters an impedance mismatch, part of 
it is reflected back to the source because it 
can't go anywhere else. And, ifthe impedance 
mismatch is 100%, then the full signal is 
reflected back. There are two easy ways to 
make a perfect impedance mismatch: An 
open circuit (infinite impedance) or a short 
circuit (OQ impedance). The difference is the 
phase of the reflected signal-an open circuit 
reflects without any phase change, whereas a 
short circuit inverts the phase. 
Let's summarize. Figure 2 seems to show 
that if we have a transmission line and add a 
wire connected to nowhere (open circuited), 
we may cancel the signal if its length is 
equivalent to a quarter of the wavelength. Is 
it for real? You bet it is! Such a short wire to 
nowhere-or more precisely, a transmission 
line with open end-even has a name. It is 
called a nstub.u 
QUCS TO THE RESCUE 
Why not start to check if it is working 
using simulation software? Here, I propose to 
use QUCS (Quite Universal Circuit Simulator), 
a very good, free and open-source circuit 
simulation software. Don't hesitate to 
down load it and reproduce the example shown 
in Figure 3. It will take you no more than half 
an hour. QUCS can do plenty of things, but 
in particular, it can model transmission lines 
and run so-called S-parameter simulations. 
I selected a signal source, P1, from the 
tool box, connected it to two SOQ transmission 
lines in series, and plugged the other end 
to another source, P2 (Figure 3). This may 
seem awkward, but for an S-parameter 
simulation, all ports should be connected to 
signal sources. The system then simulates 
the propagation paths from one source to the 
other. For example, the propagation from P1 
to P2 is noted 521, or 5[2,1] in QUCS' syntax. 
Finally, I added a 37.5mm, open-ended 
transmission line as illustrated, making a 
T-shaped circuit. I then gently asked QUCS to 
calculate 521 with a signal frequency ranging 
from 100MHz to 8GHz, and to plot it with a 
decibel scale, thanks to a short equation that 
converts volt ratios to dBs. 
The resulting graph is depicted on the 
bottom of Figure 3. What do we see? For 
low frequencies, the transmission is OdB, 
which means no loss. Not a surprise, since 
even a wire with an open stub does transmit 
electricity. However, what happens at a 
frequency of 2GHz? The transmission is close 
to 0 (-40dB, which means that only 10-40/10 = 
1/10,000 of the power is transmitted). Why is 
this happening at 2GHz? Simply because the 
ci rcuitcellar. corn 69 
SUM = O 
FJGURE 1 
The sum of two sine signals cancel out if one is shifted by half a wavelength. 
A/4 
FJGURE 2 
Reflector 
(in phase) 
A/4 
---.. } suM ~ o 
A shift of half a wavelengtn can be made with a quarter wavelength stub, ended by a perfect reflector. 
• • - --• - --I ~ .1--- -
+ • --.... --
FJGURE 3 
""""" P1 IJnol 
- Z~OIIM 
Zo600Mt L•IOO,., 
---11011 . ......., 
P2 -z~a.. 
........ 
Here is a QUCS simulation of an open stub (line 3, 37.5mm long). The response shows a null in the 
transmission close to 2GHz. 
70 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE 4 
open stub is 37.5mm long, and that 37.5mm 
is a quarter of the wavelength for 2GHz1 !..= 
c/f = 300,000,000 m/s I 2,000,000,000Hz = 
0.15m, divided by 4, or 37.5mm. A word of 
caution here: QUCS assumes by default that 
the propagation speed is c. On a real cable it 
will be a little lower, maybe 80% or 90% of c, 
but you get the idea. 
So, such an open stub in fact makes a 
band-reject filter. It passes the signals, except 
close to 2GHz where the attenuation is very 
high. If you look again at Figure 3, you will 
see that there is something interesting when 
Reflector 
(phase inverted) 'A. 
)\/2 
(\j\_;
. . . . . 
. . . . 
CXX)
. . . . . . 
A null can also be achieved with a phase-inverting reflector like a short circuit, but this time with a half-
wavelength stub. 
~·: @ .,_ -• • I - --• - -I «; ;... - --• • ._ __ .... 
il --
FIGURE 5 
l!9!!!!!!'n ':' 
~~ 
dBS21 odiiC Sfl.lll 
d8S 11001115{1, IJ) 
""""""" P1 U.1 
,._, z·~a. 
Z•~Oin l•IClO ... 
A QUCS simulation of a short-circuited stub. See the zero transmission at OHz as expected. 
•• 
the frequency continues to increase. There 
are other frequencies where the attenuation 
is very high, precisely at 6GHz, 10GHz and 
every 4GHz thereafter (although not shown in 
the Figure 3 graph). I'm sure you will easily 
discover why. 
Here's the trick: 
3 X 2 X f../4 = f.. + 2 X 'A./4 ... 
Up to now, I have used open-circuited 
stubs, which are reflecting the signal in 
phase. In contrast, what would be the effect 
of a short-circuited stub? Well, nearly the 
same, but with a twist. A short-circuited stub 
introduces a 180-degree phase shift to the 
reflected signal. A cancellation of the signal 
is still possible, but this time with a 'A./2 long 
stub rather than 'A./4 (Figure 4). Two times 
'A./2 is 'A. so the propagation back and forth 
is in phase, and the short-circuited reflector 
inverts the phase. Therefore, the sum is zero 
again . 
I modified the QUCS simulation to show 
you what happens with a short-circuited stub. 
Figure 5 is exactly the same design as the 
one we previously discussed, but now with 
a short to ground at the end of the 37.5mm 
stub. The first null is no longer at 2GHz but at 
4GHz, because half the wavelength is 37.5mm 
at this frequency, as expected. The signal is 
now canceled at 4GHz, 8GHz and so forth. The 
simulation also shows you an important point. 
With a short-circuited stub, there is also a zero 
transmission at a frequency of OHz, meaning 
for DC signals. That's clear if you look again 
at the design. The stub is a short circuit, so 
no DC current can flow because it is short-
circuited to ground. 
TO THE LAB 
Simulations are nice, but wouldn't it be 
preferable to see an actual experiment? With 
that in mind, I went to my company's labs 
and switched on our Keysight Technologies 
N5230A vector network analyzer. This nice test 
instrumentis, in fact, doing exactly the same 
as the QUCS simulation described earlier. It 
injects a signal of varying frequency into a 
device under test, and plots its transmission 
and reflections on a graph. For a quick 
demonstration, I simply interconnected two of 
its ports with a short coaxial cable, and added 
a SMA T adapter, ended with a 3cm SMA/SMA 
adapter, to simulate a 37mm open transmission 
line (Figure 6). I pressed some buttons, and 
voila! The result is shown in Figure 7. Very 
dose to the simulation isn't it? Cool, so that's 
actually real! Here the first null was measured 
at a frequency of 2.0GHz, and the second at 
6.16GHz. This is not exactly 6GHz as expected, 
probably due to some parasitics somewhere. 
OK, I must admit that I am a bit lucky to 
have such expensive test equipment on hand. 
So, I wanted to find a way for you to reproduce 
the same experiment with equipment you will 
have on your bench, or that you could buy 
for reasonable prices on eBay. And this is 
indeed easy! The key is to get far down in 
frequency. Going to lower frequency means 
longer wavelengths, and therefore longer 
cables as stubs. 
What wou Id you need to do the experiment? 
First, you will need a long length of coaxial 
cable to make a long open-ended stub. The 
longer the better. I found a lOm long cable 
on my bench, so it should cut frequencies of 
wavelength close to 40m ('M4 = lOm), which 
means f = cf'A. =7.5MHz, or, more exactly, 
around 6MHz. That's because the propagation 
speed in the cable is not the speed of light, 
but rather 80% of it. Next, you need a signal 
generator able to go up to 6MHz. If you don't 
have one, building one is a nice exercise! 
Finally, you will need a way to measure the 
transmitted signal, and the easiest way is 
to use an oscilloscope. It could be a very 
cheap one, because lOMHz of bandwidth is 
enough. On my side, I used a Keysight 33521A 
generator and a Keysight DSO-X 3024A 
oscilloscope, but they are truly both overkill 
for this job. 
Now, switch on the equipment, start with 
a frequency of OHz and slowly increase the 
frequency. Take a piece of paper and plot the 
transmitted power versus frequency, and you 
will find a null at 6MHz! The plot can be done 
manually or, if you are lucky, by switching 
the frequency generator to sweep mode as I 
did. This allows you to get the plot directly 
on the oscilloscope, if you trigger it carefully. 
Looking again at Figure 8, the scope shows 
the envelope of the transmitted signal from 0 
to 20MHz. Two nulls are clearly visible, close 
to 6MHz and 3X6 = 18MHz, as expected. 
PRINTED CIRCUIT BOARDS 
I talked about the effects of open-
circuited wires or coaxial cables, but the same 
phenomenon exists for any kind oftransmission 
line. In particular, it will happen if you design 
a PCB and leave a short copper trace open-
ended. This might or might not be an issue, as 
we will see in a minute, because it all depends 
on the signal frequency and trace length. 
Another simulation may be welcome here. 
Rather than using QUCS again, I switched to 
another and a bit more complex tool: Sonnet. 
This software is neither open source nor 
free, but a free, limited Sonnet Lite version 
is available from Sonnet Software's website. 
Unlike QUCS, Sonnet is not a circuit simulator, 
but rather an actual electromagnetic 
simulator. It uses a so-called planar or 2.50 
ci rcuitcellar. corn 71 
FIGURE 6 
This open-drcuited SMA T adapter implements a short open stub. 
FIGURE 7 
Here is the actJ.Jal measurement of an open stub on a vector network analyzer. 
FIGURE8 
The same experiment can be done using far simpler equipment, here at 6MHz using a lOm-long coaxial 
cable as a stub. 
72 CIRCUIT CELLAR • OCTOBER 2020 #363 
0 
... _ 
\(\( 
.. 0 _.. <>-
·5 -- .. - I g · 10 
I ~ · 15 - • I ID I • ·20 ~ .. 
I I 
I dB I 
·25 
-30 
0 2 3 4 s e • • 10 
~r-(OIU) ........... ___ 
FIGURE 9 
Sonnet Lite can simulate PCB 
structl.lres like this open-circuit. 
20mm stub, showing a first 
transmission null at 1.82SGHz. 
TOPB .. 
il 
fR~ 
GND 
Cl u or .lrdg mou~• to roadoot dolavalue; 
FIGURE 10 
model. The design to be simulated must be 
modeled as parallel dielectric layers, with 
metal structures between each layer. This is 
more limited than a full 3D simulator, but far 
faster in terms of calculations. This is good 
enough for PCB simulations. 
The example I made with Sonnet Lite is 
shown in Figure 9. On the far left you can see 
the layer structure- a 1.6mm FR4 substrate 
and air on top of it, with a full ground plane on 
the bottom layer. I then drew two rectangular 
copper traces on the top side of the PCB-one 
from port 1 to port 2, and one open-ended 
stub. The traces are 3mm wide, which is a 
good width for a SO.Q transmission line on a 
[l .m 
" 
An analysis of the current through each copper trace allows us tn visualize the cancellation of the waves at 
the junction point at a precise frequency. 
1.6mm thick FR4 PCB. (See my previous article 
on the subject: "Microstrip Techniques," 
Circuit Cellar 223, February 2009.) The length 
of the open-ended trace is 22mm. 
I then ran the simulation, and a couple 
of minutes later, I got the nice plot shown in 
Figure B. Very close to the theory again! Here 
the first null is at 1.825GHz. Since the open 
trace is 22mm long, we would expect a null 
for a wavelength of BBmm, so a frequency of 
3.4GHz. However, the propagation speed on 
a FR4 substrate is far lower than the speed 
of light. The ratio is, in fact, the square root 
of the so-called effective dielectric constant 
of the trace. With such a trace on FR4, this 
ratio is close to 1.8, and this explains why a 
22mm trace is equivalent to a quarter wave 
at 1.825GHz. 
The Sonnet simulation also shows that the 
other nulls, higher in frequency, are lower 
and lower in terms of attenuation. This effect 
is dearly visible in the graph in Figure 9, and 
is due to the losses in the substrate, well 
simulated by the tool. Finally, such a simulator 
can represent the currents on the traces at 
any frequency, using a calor grading scheme. 
What's happening at the null frequency of 
1.825GHz is shown in Figure 10. This figure is 
helpful for understanding that the incoming 
and reflected signals are canceling out at the 
junction point, with a quarter wavelength 
exactly equal to the trace length. 
Once again, don't hesitate to download 
Sonnet Lite, and try to reproduce this example 
for yourself. You will learn a lot, even though 
you will probably need to spend time learning 
the software through the available tutorials. 
MURPHY, UNFORTUNATELY 
Stubs, intentionally designed using open-
or short-circuited transmission lines, are 
actually used to build radio-frequency filters, 
and now you know how. Several stubs, well 
designed and with proper separation between 
them, can be used together to implement 
far more complex responses than the basic 
examples I've illustrated. This may bring us a 
little too far for this article. 
That said, I would like to share with you that 
you may also build stub filters unintentionally. 
How? Imagine that you design a kind of 
2.4GHz transmitter-maybe Bluetooth-using 
an IC and some passive components around 
it. You must design two variants of your 
product, using two different kind of antennas. 
Say one is ended with a SMA connector, and 
the other is a dipole ended with a differential 
flex cable. Maybe you would design a PCB like 
the one illustrated in Figure 11a. Warning-
this is not an actual design, just an illustration 
for this article. Here the transmitter circuit 
is supposed to be on the lower left, and the 
antenna trace goes simultaneously to the 
SMA connector, Jl, and to some components 
feeding the flex cable connector, 12. That's 
clever, because, depending on the need, you 
will solder either Jl or 12 and its passive 
components, so you will have only one PCB 
and two versions. Right? 
Wrong! This would be a very, very bad 
idea, and I'm sure you nowknow why. 
Imagine that you populate only the connector, 
ll, and leave J2 and the small accompanying 
passive unpopulated. OK, but the trace going 
from Jl to the pads planned for R7 and R9 
FIGURE 11 
will be an open-ended trace. It will therefore 
act as a stub, and will cancel out signals with 
frequencies for which its length is a quarter 
ofthe wavelength. And, do you know what will 
happen? Murphy's Law states that everything 
that can go wrong will go wrong. So, for sure, 
the frequency that will be canceled will be 
exactly the frequency you want to use. And 
your design will simply not work-all because 
of a short open-circuited trace. 
What would be a good solution? Simply 
avoiding any open-circuited stub on any high 
frequency line. Figure 11b shows a possible 
way. Here I added two OQ resistors, RlO and 
Rll. These two components are positioned 
exactly at the Y junction between the two 
intended paths. Of course, only one of them 
will be populated. Therefore, there will be 
no open-circuit line at all, and no Murphy-
related risk. For the lowest possible risk, it is 
indeed a good practice to share a single pad 
between the two components, RlO and Rll, 
as illustrated, thus avoiding even a lmm stub. 
WRAPPING UP 
Here we are. As always, I have only covered 
the basics, and with simplified explanations. 
Stubs can be used to make any kind of filters or 
impedance-matching networks, especially at 
high frequencies. Stubs can also be the source 
of serious headaches for the incautious PCB 
designer. So, knowing what they are about is 
crucial. This is true not only for high-frequency 
(Left) A wrongly designed 2.4GHz transmitter PCB. (Right) A greatly improved version, avoiding stubs on the RF trace 
ci rcuitcellar. corn 73 
74 CIRCUIT CELLAR • OCTOBER 2020 #363 
FIGURE 12 
This HP532A frequency meter is actually an adjustable short-circuit stub. 
ABOUT THE AUTHOR 
Robert Lacoste lives in France, between Paris and 
Versailles. He has 30 years of experience in RF 
systems, analog designs, and high speed 
electronics. Robert has won prizes in more than 15 
international design contests. In 2003 he started a 
consulting company, ALOOM, to share his passion 
for innovative mixed-signal designs. Robert's 
bimonthly Darker Side column has been published 
in Circuit Cellar since 2007. You can reach him at 
rlacoste@alciom.com. 
For detailed article references and additional resources go to: 
www.circuitcellar.com/article-materials 
RESOURCES 
Keysight Technologies I www.keysight.com 
Sonnet Software I www.sonnetsoftware.com 
designs, but also for high-speed electronics. 
A fast, digital signal with nanoseconds-
long transitions should be treated with 
the same care as a radio-frequency signal 
in the gigahertz range. Put simply, a PC 
motherboard with its multi-Gbps PCI express 
links will never work with uncontrolled stubs 
or bad PCB routing! And even if you are not 
designing PC motherboards, I'm sure that you 
encounter fast digital transmissions often. 
For example, USB is replacing RS-232 links in 
many systems-and the frequency range is 
not exactly the same. 
To close, I can't resistthe pleasure of showing 
you an old test instrument that is actually 
based on a stub-like structure (Figure 1Z). It 
is a Hewlett Packard HP 532A frequency meter 
from the 1960s, borrowed from my colleague 
YannickAvelino. (Hewlett Packard later became 
Agilent and then Keysight). How does such a 
frequency meter work? At the bottom of the 
device, there is a transmission line. Here it is 
made with a waveguide, but this is the same 
as a wire transmission. The transmission line 
is ended by a diode detector, clearly visible on 
the right side. 
On top of this transmission line, there is 
a vertical tube, which is an adjustable cavity. 
This cavity is the waveguide equivalent 
of a short-circuited stub. Its length can 
be adjusted, thanks to a precision helix 
associated with a calibrated dial. The use 
is then straightforward. The user injects a 
signal of unknown frequency and turns the 
knob, while looking at the power measured 
by the diode detector. At a given position, 
the signal on the detector will be drastically 
attenuated. At that precise point, the cavity 
length corresponds to half a wavelength, and 
the dial directly indicates the corresponding 
frequency-up to 26GHz! Who said that 
microwaves have to be complicated? 
I hope that you liked this small introduction 
to stubs. I encourage you to read the articles 
on Circuit Cellar's article materials webpage. 
More importantly, I urge you to play with 
stubs, using either simulators or actual 
experiments. Don't be afraid. It will be fun, 
and you will learn a lot! e 
Module Speeds Medical Ventilator Designs 
Trinamic Motion Control, now part of Maxim Integrated, has announced the 
TMC4671+TMC6100-TOSV-REF design board, a BLOC serve driver capable of 12V to 
36V with up to 6A RMS to accommodate medical ventilator and respiratory system 
design. The reference design shows engineers how to build medical ventilators 
using readily available components, bypassing the longer lead times of traditional 
components. To shorten design cycles, the open-source module features a session 
border controller connector in Raspberry Pi style and space for a pressure sensor 
add-on board. 
The reference design board includes a Hall sensor interface and connectors 
for an optional pressure sensor add-on board. As a result, important pressure 
data can be directly visualized on a Raspberry Pi with touch screen using the free 
firmware developed by Trinamic for the reference design board. 
Thanks to the embedded TMC4671 serve controller IC, the module generates a 
pulse-width modulation (PWM} frequency and a current controller clock of 100kHz 
independent of the microcontroller. The complete TMC4671+TMC6100-TOSV-REF 
reference board is available now through Trinamic's distribution channels. All 
hardware, firmware and software are open source under the MIT license. 
Trinamic Motion Control 1 www.trinamic.com 
Automotive Grade Step-Down DC-DC Converters 
c·rrcuitcellar.com 75 
PRODUO NEWS 
Dialog Semiconductor has announced the DA913X-A product family, a new line of highly efficient, high-current, automotive-
grade, step-down DC-DC (Buck} converters. The highly integrated DA913X-A devices require fewer external components than 
competing solutions, enabling a low system BoM cost and a reduced 
solution footprint. The devices operate at efficiency levels above 90%, 
reducing the thermal design challenges for powering high-current 
rails in a wide range of automotive systems, including infotainment, 
navigation, telemetry and Advanced Driver Assistance Systems 
(ADAS). 
The DA913X-A family is comprised of three devices configured 
as single- or dual-output buck converters. The DA9130-A operates 
as a single-channel, dual-phase buck converter, delivering up to a 
lOA output current. The DA9131-A integrates two single-phase buck 
converters, each delivering up to SA output current. The DA9132-A 
also integrates two single-phase buck converters, each delivering up 
to 3A output current. The DA9130X-A devices are AEC-QlOO Grade 2 
qualified and are available in a 3.3mm x 4.8mm a 24-pin FC-QFN 
wettable flank package. Industrial/commercial grade versions are 
also available. 
Dialog Semiconductor 1 www.dialog-semiconductor.com 
Buck Boost DC-DC Converters Deliver 300W 
TDK has announced the addition of 12.5A 300W rated models to the i7C 
non-isolated DC-DC converter series. With an input range of 9 to 53Vdc, the 
output can be adjusted from 5 to 28V. The i7C topology enables a seamless 
transition from buck (voltage reduction} to boost (voltage increase} 
operation. The series is well suited for generating additional high-power 
outputs, from existing 12, 24, 36 or 48V system voltages. 
Efficiencies of up to 96% minimize power losses, allowing the product 
to operate and deliver high useable power in demanding thermal 
environments,says TDK. Under light load conditions, the i7C's control 
techniques significantly reduce power dissipation. All models are certified 
IEC/UL/EN 62368-1, 60950-1 and carry the CE mark for the Low Voltage and 
RoHS Directives. 
TDK-Lambda I www.us.lambda.tdk.com 
76 CIRCUIT CELLAR • OCTOBER 2020 #363 
PRODUO NEWS 
Software for STM32 MCU Condition Monitoring 
STMicroelectronics (ST) has released a free STM32 
software function pack that lets users quickly build, train 
and deploy intelligent edge devices for industrial condition 
monitoring using a microcontroller Discovery kit. Developed in 
conjunction with machine-learning expert and ST Authorized 
Partner Cartesiam, the FP-AI-NANOEDG1 software pack 
contains all the necessary drivers, middleware, documentation 
and sample code to capture sensor data, integrate and run 
Cartesiam's NanoEdge libraries. 
Users without specialist AI skills can quickly create 
and export custom machine-learning libraries for their 
applications using Cartesiam's NanoEdge AI Studio tool 
running on a Windows 10 or Ubuntu PC. The function pack 
simplifies complete prototyping and validation free of charge 
on STM32 development boards, before deploying on customer 
hardware where standard Cartesiam fees apply. 
The FP-AI-NANOEOG1 function pack is available now at 
www.st.com. The STM32l562E-DK Discovery kit contains 
an STM32L562QEI6QU ultra-low-power microcontroller, an 
iNEMO 30 accelerometer and 30 gyroscope, as well as two 
MEMS microphones, a 240x240 calor TFT-LCO module and 
on-board STLINK-V3E debugger/programmer. The budgetary 
price for the Discovery kit is $76.00, and it is available from 
www.st.com or distributors. 
STMicroelectronlcs I www.st.com 
• l::, .1./::•m~~ • .. ,., 
... -· 
<:> " : .. . . . • • . .. r:~ e :o~ .. ~- . _:.•: ·!- -. • 
FP-AI-NANOEDG 1 
to learn and monitor equipment health 
MCU Solution Targets Micromobility Designs 
Renesas Electronics has introduced what it calls its 48V 
Mobility Winning Combination Solution that helps engineers 
accelerate the development of e-scooters, e-bikes, UPS and 
Renesas Reference Design for Electric 
Mobility Solutions 
energy storage systems. This reference design uses a modular 
approach in both hardware and software to showcase core 
and optional functional blocks that can be adopted for many 
24V-48V micromobility applications such as lawn mowers, 
electric carts, robot cleaners, power tools, power banks and 
more. 
It also uses 15 Renesas ICs, including three key devices: 
the ISL94216 16-cell battery front end (BFE), robust HIP2211 
100V MOSFET drivers, and RX23T 32-bit microcontroller (MCU) 
for motor control. The 48V mobility winning combination 
solution is powered from a 25A-hour Li-ion battery that drives 
a 1600W inverter to attain speeds up to 5,000rpm. 
The ISL9421616-cell BFE is available now in a 64-lead QFN 
package from Renesas' worldwide distributors, and is priced 
at $2.99 USD in 1,000-unit quantities. 
Renesas Electronics 1 www.renesas.com 
Low Power Dual-Core MCU Serves Up BLE 5.2 
Maxim Integrated Products has announced the MAX32666, 
an ultra-low-power dual Arm Cortex-M4 MCU with floating-
point unit (FPU) and Bluetooth Low Energy 5.2 (BLE 5.2). 
Its battery life is extended by combining robust memory, 
security, communications, power management and processing 
functions traditionally performed by multiple MCUs into a 
single device, says Maxim. 
This dual Cortex-M4F MCU provides efficient computation 
of complex functions, operating at up to 96MHz. To replace 
the need for a separate PMIC, the MAX32666 features an 
integrated single-inductor, multiple-output (SIMO) regulator, 
which extends the life for small-sized battery applications. The 
MCU offers BLE 5.2, supports up to 2Mbps of data throughput 
and long range (125kbps and 500kbps) and offers transceiver 
output power of +4.5dBm programmable down to -95dBm. 
The MAX32666GWPBT+ is available at Maxim Integrated's 
website for $5.22 (1000-up, FOB USA); also available from 
authorized distributors The MAX32666EVKIT# is available at 
Maxim Integrated's website for $157.50. 
Maxim Integrated I www.maximintegrated.com 
The Directory of 
PRO & SERVICES 
Advertisers must furnish digital files that meet our specifications (circuitcellar.com/mediakit). 
All text and other elements MUST fit within a 2" x 3" format. 
E-mail adcopy@circuitcellar.com with your file. 
circui tcellar.com 17 
For current rates, deadlines, and more information contact 
Hugh Heinsohn at 757-525-3677 or Hugh@circuitcellar.com. 
,. 
~~ Technofogic 
" Systems 
ter 
TS-7250-V2 
1 GHz ARM Computer with 
Customizable FPGA-Driven 
PC/1 04 Connector 
and Several Interfaces 
at Industrial Temp 
ADuC841 
M icrocontroller 
Design Manual 
0 
12c Bus 
Adapters 
CE 
USB I Ethe rne t I RS-232 
Free 
Visual Studio, Lab VIEW & ASCII AP!s 
Win, OSX & Linux Drivers 
Msg & EEPROM Program Apps 
Sw itch Selectab le Options 
Voltage Sense (O.Sv to Sv) 
Power Source l OOmA (3.3v or Sv) 
Pull-Ups (Enable/Disable) 
0 /R V lt & ESD P t et' 
shop-mcc-us.com 
78 CIRCUIT CELlAR • OCTOBER 2020 #363 
Answer 1- The total power dissipation is specified as 
lmW, and the supply voltage is 1.5V. This means that 
the total current consumption of both paths (base and 
collector) can only be 666.7~-JA. 
The "forced W' is specified as 10, which means that 
10x the base current flows through the collector. This 
means that Ib must be 60.61JA and le must be 606.11-JA. 
Since Vc:E(satJ is 0.2V, the collector current must 
produce a voltage drop of 1.5V- 0.2V = 1.3V across Re-
Therefore, Re: must be 1.3V I 606.11-JA = 2,145Q. 
Similarly, since V8E is 0.2V, the base current must 
produce a voltage drop of 1.5V - 0.7V = O.BV across 
Re+ R8• Therefore, the total resistance must be O.BV I 
60.61JA = 13.2 kQ, and the value of RB must be 13200 
- 2145 = 11055 n. 
The closest standard E96 values are 2.15kQ and 
11.0kQ, respectively. Using E12 values of 2.2kQ and 
12kQ yields currents of 5911JA and 56.31-JA, for a forced 
~of 10.5. 
Answer 3- The simplest way to have an arbitrary 
number of inputs is to add both a transistor and its base 
resistor for each additional input. All of the transistors 
would share the same load (collector) resistor. 
Answer 2- Yes, you could build a 2-input NOR gate 
with just one more resistor. 
Clearly, there would be no problem when both inputs 
are high or both inputs are low. However, the situation 
where one input is high and one input is low presents 
a problem. 
Just like with the inverter, the high input drives about 
611JA into the base of the transistor, which sits at about 
0.7V. However, with the other input low (at 0.2V), the 
other base resistor steals about {0.7V - 0.2V) I 11 kQ = 
451JA away, leaving only about 161JA for the transistor. 
Now the transistor must have a minimum P of 6061JA I 
161JA = 38 in order to remain saturated. 
However, you could not add a third input this way. 
A single high input would not be enough to turn the 
transistor on. 
Answer 4- Each additional input connected to the 
output of the gate draws more current through Re, but 
this current must be shared N ways, so the current 
available to each input drops. 
Table 1 shows that even with up to 8 loads, the input 
current to each gate only drops by about half, requiring 
a minimum P of about 20. ..... ,..... .... ..... .... Input 
1 6lj.IA 6lj.IA 
2 105j.IA 52j.IA 
4 163j.IA 41j.IA 
ci rcuitcellar. corn 79 
The Future of Modular Edge Processing 
Three Real-World Intelligent 
Infrastructure Applications 
ystems designed for edge 
applications need to be highly 
flexible. This requires scalable 
processing power, which can be 
provided most efficiently with computer-
on-modules (COMs). Here, we three 
real-world applications-ranging from LoRa 
gateways for smart cities to energy grid 
servers and video surveillance servers, all 
with optional Al-demonstrate the benefits 
of this modular approach. 
Low-power wide-areanetwork 
(LPWAN) technology LoRa is deployed in 
many different application areas. Smart 
cities build networks on the basis of this 
technology to connect smart distributed 
devices-for example distributed parking 
spaces as well as central multi-story car 
parks. Such different parking scenarios are 
one example why LoRa gateways need to be 
highly adaptable. 
LoRa GATEWAYS 
Flexible loRa gateway configurations are 
also required because they can connect any 
LoRa sensor as this wireless protocol needs 
no physical connector. This means that loRa 
gateways can be used anywhere to cater for 
any additional application based on various 
mixes of loRa devices. Smart cities can 
therefore also use loRa parking gateways 
to track their free-floating mobility offers 
and to integrate charging infrastructures 
into parking guidance systems and so on. 
However, extending the tasks of smart city 
gateways leads to the need to integrate 
scalability for future demands to improve 
investment security. 
The LoRa FlexGate FGW264 Gateway from 
the French company Expemb for example 
offers this scalability with COMs (Figure 1). 
By using COM modules for the edge processor 
demands, the new gateway enables a flexible 
hardware and software setup to meet 
any LoRa based loT technical challenge-
including artificial intelligence (AI). 
The platform supports up to eight LoRa 
channels simultaneously to communicate 
with several thousand connected objects. 
It can further host not only one but two 
generations of high-performance and low-
power Intel Atom processors. Such high 
scalability beyond dedicated processor 
sockets has become possible by utilizing 
COMs on the basis of the vendor 
independent Qseven specification. 
Towards the cloud, the LoRa gateway 
offers connectivity with 1 Gbit Ethernet 
and 3G/4G. 
SMART CHARGING 
INFRASTRUCTURE 
Edge systems requiring more hybrid 
local connectivity will need more 1/0 
flexibility. But even in such scenarios, the 
modular approach is a great fit for edge 
computers as we can see in China's recent 
edge layer deployments. China is currently 
deploying an edge layer for a distributed 
management system for smart grids in 
cooperation with Tencent's Energy loT 
Partner (Figure Z). Such an edge server 
design in a smart city infrastructure context 
is quite interesting, as e-mobility has a 
bright future. E-mobility requires charging 
infrastructures that must be fueled by 
regenerative energies to improve the 
ecological footprint. 
Regenerative energy is generated in 
local micro grids or distributed virtual grids; 
and those grids need smart edge devices, 
capable to manage the decentralized 
processes in real-time. Once the Chinese 
edge layer service is fully available, it 
will support thousands of power devices, 
• 2!!:3! 
'": 11 rr: 1 J ~- .. -. . . .. 
)·:.;~~'-· . . , f • I ·u 
.... ( .; '. . . j ' ..... . ,.,. ., ·- -· -
• •·.. l .... . r p 
j , • : • 
Dan Demers, 
congatec 
FIGURE 1 
The LoRaWAN FlexGate FG264 
Gateway from Expemb is designed 
with Qseven ffiMs. 
80 CIRCUIT CELlAR • OCTOBER 2020 #363 
, ~ --
FIGURE2 
The first loT edge router designs in China's edge layer for smart grids are based 
on congatec Server-on-Modules with Intel Xeon DlSxx processors. Alternative 
configurations are based on the Intel AIDm Oxxx processors. 
FIGURE3 
The Christmann microserver family for rack systems is scalable with both COM 
Express Type 6 and COM Express Type 7 Server-on-Modules as well as standardized 
GPGPU cards and FPGA modules. 
hundreds of power protocols and online editing of common 
standard protocols as well as dynamic completion of common 
standard device types. Such a comprehensive ecosystem could 
be highly interesting for many worldwide deployments. 
For the adaptive hardware of the new class of edge devices, 
the edge server designers have chosen a flexible system design 
based on COM Express Type 7 modules. These modules offer 
server-grade performance and functionality on a standardized 
form factor. First systems are based on modules with Intel 
Xeon D1Sxx processors with up to 16 cores and 32 threads. 
Alternative configurations are based on the Intel Atom C3xxx 
processors. Featuring up to 16 cores, those processors are the 
ideal fit for all installations that have to handle smaller package 
sizes in parallel while consuming as little energy as 25W TOP. 
SITUATIONAL AWARENESS 
In rack server designs, standardized COMs can drastically 
cut the long-term investment costs for infrastructure servers 
used in connected smart city applications such as video 
surveillance systems that are being equipped with AI to 
improve situational awareness and public safety. For such 
microserver installations at the infrastructure edges, vendors 
such as Christmann also rely on Server-on-Modules based on 
the vendor-independent COM Express Type 7 as well as the new 
COM-HPC standard (Figure 3). 
What speaks in their favor is not just the flexibility to adapt to 
recent AI needs but also the long term perspective: The second 
generation-required after three to five years as technology 
enhancements evolve rapidly-is expected to cost only about 
50% of the initial investment, as a simple replacement of the 
processor modules will be all that's required in the majority 
of cases. 
To open up the potentia Is of significantly reduced TCO, such 
rackmount servers can be flexibly equipped with up to 27 CPU 
microservers with x86 or Arm architecture. In addition, also 
standardized modular options for parallel processing in the 
form of GPGPU cards and FPGA modules are possible, which 
are physically implemented on a COM Express Basic footprint. 
The modular concept makes the Christmann RECS Box 4.0 
server family extremely scalable. 
MODULARITY AT THE SG EDGE 
These three examples illustrate the tremendous benefits of 
designing edge gateways and edge servers on the basis of 
Computer-on-Modules and Server-on-Modules. For large-scale 
applications, even the fusion of modules with dedicated carrier 
boards is a common strategy. So why not follow a similar 
approach for large-scale applications such as SG edge servers 
or for consumer-grade e-mobility vehicles, implementing first 
a COM &. carrier design for R&D and later a full custom design 
by COM and carrier fusion? Embedded computer vendors such 
as congatec are ready to support such a paradigm change 
from proprietary designs towards closed-loop engineering 
with open modular platforms. e 
RESOURCES 
congatec I www.congatec.com 
Dan Demers holds a B.B.S degree in International Business from Grand 
Valley State University, Grand Rapids, Michigan and an M.B.A. from 
Ashford University, Clinton, Iowa. Mr. Demers has over 22 years of 
experience in embedded computing having worked with Fortune SOD 
companies in the Industrial, Medical, and Communications markets.