EEWeb Pulse - Volume 28

Nathan SeidleSparkFun Electronics

PULSE EEWeb.comIssue 28

January 10, 2012

Electrical Engineering Community

EEWeb

Contact Us For Advertising Opportunities

[email protected]

www.eeweb.com/advertising

Electrical Engineering CommunityEEWeb

Digi-Key is an authorized distributor for all supplier partners. New products added daily. © 2011 Digi-Key Corporation, 701 Brooks Ave. South, Thief River Falls, MN 56701, USADigi-Key is an authorized distributor for all supplier partners. New products added daily.

www.digikey.com/techxchange

It’s all about connections.

The user-to-user forum is for everyone, from design engineers to hobbyists, to discuss technology, products, designs and more. Join the discussions that match your interest or offer your expertise to others.

Join the discussion now at:

discussions

hobbyists

engineers

industry experts

resourceslinks

technical documentswhite papers

reference designs

application notes

community

power

microcontroller

lighting

wireless

sensor

students

http://bit.ly/kDYHm7

EEWeb | Electrical Engineering Community Visit www.eeweb.com 3

TABLE O

F CO

NTEN

TSTABLE OF CONTENTS

Nathan Seidle 4SparkFun Electronics

SparkFun: Why Open Source? 9BY PETE DOKTER WITH SPARKFUN ELECTRONICS

Featured Products 11What Bumblebees and Models of DFE Have in CommonBY MICHAEL STEINBERGER WITH SISOFT

Real-World Range Testing 17 BY CHRISTOPHER HOFMEISTER WITH LS RESEARCH

RTZ - Return to Zero Comic 23

Dokter explains SparkFun’s reasoning for being open source.

Interview with Nathan Seidle - Owner and CEO

A detailed look at the creation and use of IBIS-AMI models of DFE.

An outline of the procedure for a successful RF range test providing data on RF link performance.

12


INTERVIEWFEA

TURED IN

TERVIEW

Owner and CEO of SparkFun Electronics

NathanSeidle

How did you get into electronics/engineering and when did you start?I started in electrical engineering at the University of Colorado in 2000. At the time I really didn’t understand what electrical engineering was,

and was under the impression that I was always going to be some sort of chip designer. This was way back before I had any kind of comprehension of what I was going to be doing. Then, in 2002, I discovered microcontrollers. I

thought, “Wow, you don’t need a full computer; you can program these little things and they can actually do quite a lot!” And at the end of 2002 I blew out my PIC programmer. That’s when I started scouring the Internet for a cheaper source, because as a student I didn’t even have money for the first programmer. So when I blew it up I was really in a bind. After searching, I found out that most of the web sites in 2002 were very difficult to order from. There weren’t very good pictures, and a phone call was often necessary to place the order. At this point I thought, “Maybe I could do better. Maybe I could start a website that had online checkout, as well as some nice, clean pictures of some electronics.” And off I went!

You were still in school when you started SparkFun?Yes, I was a junior in the electrical engineering program and it took me about a year and a half to graduate. Once I graduated, all my professors were asking me what companies I was applying to work for, and I said, “Actually, I’m not really applying anywhere. I’m just going to run this


INTERVIEWFEA

TURED IN

TERVIEW

SparkFun thing I’ve been doing for the last year and a half.” They were a little scared, but they’ve been very receptive, and now I go back and guest lecture for a few of them.

Can you tell us more about what SparkFun is like today?It’s turned into a wonderful monster. We are based in Boulder, Colorado—the same town as the University of Colorado. We are now at, I believe, 134 employees. Like I said, it started with just me, and then in about a year and a half I graduated, and decided that I had just barely enough work to hire a friend of mine. I didn’t have enough money to pay him, but I certainly had enough work for him. Luckily, the work brought in some money and I was able to pay him. Then very quickly it became obvious that we needed a third person, and we’ve just sort of scaled up from there. Our building now is roughly 50,000 square feet, so the company has grown to a pretty decent size since it started.

In the early days, were there any particular products or components you started selling that helped you take off?In the very beginning, while I was an engineering student, I had a fascination with GPS. I really wanted to play with these brand-new, dual-axis accelerometers. But when I talked to the company that sold them, they said it would cost me $500–$600 for an evaluation kit. I wasn’t willing to pay for it, nor did I need the entire evaluation kit. I just needed the breakout board—an accelerometer soldered to a

little board, with a 0.1” spaced header. They said that they didn’t sell anything like that. So from the beginning, that was sort of the secret sauce to SparkFun. We found these cool technologies, and then made them accessible, for example by providing breakout boards or evaluation boards for the GPS module or the accelerometer. Since then, it’s turned into all sorts of fun sensors and technologies, and we’ve really expanded into a bunch of different fields. As of just the recently, we have just over 2,000 products for sale on the SparkFun web site.

How would you characterize your business?It is a mix. When we first started, we were just a regular distributor. We found cool parts from other people and resold them. Once I graduated from college I actually had the time to design stuff. That’s when the SparkFun products started. First came a breakout board for a USB serial chip, or a breakout board for an accelerometer. Now, here in Boulder we have all of our manufacturing, testing, packaging and shipping all in the same building. So out of those roughly 2,000 products, we design and build about 400–450 in-house. And with the other products, we provide the best quality from the best manufacturers.

What is the work culture like at SparkFun?It’s hard to describe until you see it for yourself, but for better or for worse I never had a “real job.” So I had no preconceived notions of what the work environment should be like. I made a place where I felt

like I would want to work. The work culture here is pretty laid back; there’s certainly no dress code. Also, over time people started asking if they could bring their dogs into work. My response was, “Well I’m not really a dog person, but if it would make you happier and you can take care of your dogs, go ahead and bring them in.” Then the skateboards showed up, then the loud music. So now we have this wonderful culture of controlled chaos. It’s crazy, but it makes for a pretty good work environment, and now we just have a bunch of friends working together.

How many of your staff have technical or electronics backgrounds?I believe that our engineering crew is made up of 13 people, so about 10 percent of the company. That obviously means that there are a lot of support staff that do the manufacturing, shipping and testing. But out of those in the engineering group, there is a very interesting variety of technical backgrounds. Some come from physics, some from the expected electrical engineering or electrical and computer engineering. We also have one or two self-made engineers who didn’t take the traditional, formal academic route, but eventually showed that they can build some really amazing stuff, which is why we are fortunate to have them working for us.

Are all of your designs part of the open-source hardware movement, or are they proprietary designs?Since about 2004, I started writing


INTERVIEWFEA

TURED IN

TERVIEW

tutorials, releasing the firmware and posting the schematic saying essentially, “If this product isn’t tailored to what you need, here’s the firmware. So if you need it to do something different, go ahead and make it happen.” We release stuff in the public domain mostly so that our customers can tailor it to their needs. In the past two to three years there has been a big push toward this new open-source hardware movement, which we really like. I see it as a clear explanation of what you’re going to get. So whenever a customer downloads one of our products and says, “Oh okay, this is marked with the open-source hardware logo,” the customer now knows that he or she is going to get the engineering design (EAGLE CAD) files. It’s all very open, and we encourage customers to take what we give them to modify the products for themselves, or even resell them for profit.

Can you tell us about the community and the people around SparkFun?When I started SparkFun, being an engineer myself, I really expected to sell products to my fellow engineering students, and maybe some other people in Colorado. Of course, I (thankfully) grossly underestimated the demand. Our customers come from every walk of life. We see some really exciting orders coming from big institutions such as Carnegie Mellon, Stanford and even NASA. We also see a lot of orders from just regular folks that have some sort of problem, whether it’s their cat door, or a Halloween costume they are trying to build. These folks have really creative

ideas, and come up with some really interesting solutions that I, as an engineer, don’t even come up with. So our community is this mixed bag of really exciting people doing all sorts of really wacky things.

Besides selling products, what do you do to encourage interest?Recently we actually had an open house at SparkFun, which was really fun. We had about 650 people come through the front gate, and we just sort of said, “Here’s what we do, and here are some of the projects that we’re working on.” We gave building tours, had a big tent with some music, and had a bunch

The ideal market that I would love to get into would include young, elementary school-

level kids. So ideally our products would

end up in educational toy stores to spark

the interests of kids.

of tables set up for people to bring their own projects that they wanted to show off. It was a great chance for a lot of people in and around the area to show off some of their projects. We had one of the SparkFun employees show off his balloon satellite, which

was basically a weather balloon attached to some high-definition cameras and tracking devices. He captured some amazing footage from an elevation of about 100,000 feet. Along with that, bunches of other people were showing off their projects as well, like robots and other electronics. It was all really neat.

We often open our doors to the public for somewhat of a classroom environment. We inform people about upcoming events or courses on things like e-textiles or soldering, for example. The idea started a few years ago when I was teaching a surface mount soldering course at the university. It became apparent to me that our customers also really wanted to learn about surface mount soldering. From there it got a little bigger and more complex, and now we have a department of five people that make up the “Department of Education.” They are responsible for taking SparkFun out into the educational world. I would love to change the way that electronics is taught at the pre-university level.

We’re very open about the educational information we provide, and it’s all available on the SparkFun website. We’re working on broadcasting videos of the classes and tutorials as well, but that is still in the works.

What should we expect to see from SparkFun in the near future?As I mentioned before, one of our main goals revolves around education. We’re trying to come


INTERVIEWFEA

TURED IN

TERVIEW

up with low-cost kits that kids as young as eight years old can play with. We have a couple kits already that include some LEDs, a coin cell battery holder, some needles and some conductive thread. This all comes with a piece of fabric that kids can sew it all to so they can create this thing that lights up. It’s a tool that can lead into further education about electronics, voltage and current.

Another kit we’re working on is more for high school students. It’s sort of a logic gate puzzle. Students can put it together and learn about logic gates.

On the other side of the spectrum with regard to company goals, we are really trying to package up more of our products for retail. So whenever customers go into a brick and mortar store, we want to have approachable electronics that are more attractive to customers with less experience.

Are there particular types of businesses you’re targeting for this?The ideal market that I would love

to get into would include young, elementary school-level kids. So ideally our products would end up in educational toy stores to spark the interests of kids.

Other places where we would like to see SparkFun products sold are places like museums or university bookstores. These are more of the types of avenues we are looking at.

Can you tell us about Arduino? How has it impacted your business?I believe it was in 2005 that Tom Igoe of New York University told me that SparkFun should really carry the Arduino board. I looked at it, and at the time it was just a through-hole kit. I explained that, to me, it looked like just another development board. Then about nine months later, Arduino came out with a fully assembled USB version. And Tom Igoe again said to me that we should carry the board. He said that he’d been using it to teach his students and that they really liked it. So I agreed to carry it, but it took me about a year or two to really see the value in it. I realized that whenever

you’re teaching someone or trying to build a project, you don’t want to have to worry about how to initialize the UART or how to get the I2C protocol to work, because that’s a really bad pitfall. Arduino has been so successful, and has been a game changer for people that have a bigger project and just need a little board to control some basic inputs and outputs.

Are you still involved in product development and design, or is the business so big that you are no longer able to put your resources into that?I have to wear my business hat more and more, which kind of pulls at my heart strings. I do still get to play and build electronics about four or five hours per week. But for the most part, engineering is its own department. They don’t really like it when I get in there and start messing with things, but I do still get to play with some of the parts that we build, and I get to have some input in the new products that we’re coming out with. ■

EEWebElectrical Engineering Community

Join Todaywww.eeweb.com/register

http://bit.ly/jd6Wcw

MoonstoneTM Tri-color Power LED light source from Avago Technologies: ASMT-MT00

Avago’s MoonstoneTM Tri-color Power LED light sources are high performance energy efficient devices which can handle high thermal and high driving current. The exposed pad design has excellent heat transfer from the package to the mother board.

The advanced 3x1Watt RGB MoonstoneTM package delivers typical 108 lumens at 350mA.

The robust package design makes it an excellent choice for color mixing applications.

Applications

•ArchitecturalLighting

•Decorative&MoodLighting

•RetailLighting

•Gardenlighting

•ShowLighting

Avago Technologies

MoonstoneTM 3x1 Watt

RGB High Power LED

gives you a reliable,

energy efficient, long

life operation product

for ease of design

in any type of color

mixing application.

www.avagotech.com/led

Avago Technologies LED Lighting Solutions Technology

Any Color. One Device.

For more information or to request a sample please go to:



PROJECTFEA

TURED PRO

JECT

SparkFun: By Pete Dokter

Over the past few years, SparkFun has had some amazing success. And that’s due in no small amount to the idea of open source. How so? Through extremely careful and considered planning… Yeah, that’s a lie. We were carried by a tide. Make no mistake; we were exactly where we wanted to be. We just had no idea how big the tide would be.

Early on, we used to just post source code for some of our products. And we resisted this, not because we wanted to hold on to our IP, but because we thought that anyone who saw our code would be appalled. But our thinking was, if we’ve messed up the code somehow, the least we can do is provide it to somebody else so they can have a chance at fixing it, right? Assume that on the night before a crucial lab you’re trying to get something working and there is a problem with something in our module. We are then essentially responsible. The very least we can do is give you a crowbar to pry yourself out with.

Next came schematics. Anyone who has worked with a large company’s development board knows how vital schematics are, so these really are a no-brainer. You don’t keep the shop manual from the guy that just bought your car.

Then came Gerbers. Wha...? Why would anyone need these? Are they really going to go to the trouble to make their own board? It turns out they might. And what were

Why Open Source?Pete Dokter - Director of Engineering


PROJECTFEA

TURED PRO

JECT

we saving, really? We were still amateurs at this. They’re simple two-layer layouts, and our breakout boards are almost always the application circuit in the part data sheet. We’re already providing everything else, why not these, too?

But honestly, if people are savvy enough to take our designs and make their own boards, they’re likely savvy enough to improve upon our work. We can’t really fault them for that. They did our job better than we did. So now we emulate them as they emulated us, rolling their improvements into our own design and trying to improve further. We share attribution, everybody’s got a better product making the customer happier, and we improve the species just a little bit. Therein lies the essence of open source.

So it made sense to us from both a practical and ethical standpoint; we wanted to be more useful to the customer than not, and being forthcoming with all of our materials seemed a simple and obvious thing. But it turned out to make practical business sense, too. We’ve never had to worry about IP or patents, or, for the most part, lawyers (I hear Qualcomm spends a bit of time chasing down patent infringements). We just want to focus on what we love to do: build stuff with our toys and share our love of electronics with others. Being open source allowed us to be closer to that ideal, and further away from what might be considered “business norm.” And it turns out

that there are a lot of people out there that feel as we do.

However, I recently saw a presentation by a young gentleman (and I have to be careful here because I don’t want to out him) that impressed me. At my behest some weeks earlier, he had created an online tutorial for some work he had done in an effort to attract more people to his particular pursuit. He mentioned this in his presentation, and then lamented having potentially given a leg-up to his competition. But the pursuit happens to be a highly dangerous one, and doing this particular task incorrectly could ultimately result in someone’s death. That’s a lot more responsibility than I referred to earlier, and I’d certainly rather be faulted for giving someone too much information than too little in such a situation.

Fortunately for us, it’s pretty hard to kill yourself with low-voltage electronics. Saving someone’s life isn’t necessarily one of our motives, but the reasoning is more-or-less the same. If you’re going to trust us to buy our products, we’re going to do as much as we can to ensure your success. Being open source is the natural extension of that. It’s not the new logo, nor the protections, nor the notoriety that seems to accompany the title that attracts us to open source. It’s the mentality behind the choice to be so, and that’s something that pervades our group from the top to the bottom. And I think it’s something from which a lot of companies could benefit. ■

SparkFun’s Production Department


FEATURED

PROD

UCTS

FEATURED PRODUCTS

200-MHz Handheld OscilloscopesAgilent Technologies Inc. announced the addition of two new oscilloscopes to its portfolio of handheld instruments. The 100-MHz U1610A and 200-MHz U1620A – the first handheld units to include a color VGA display – will be available for ordering on Feb. 1. With up to three viewing modes (indoor, outdoor and night vision), these instruments

enable engineers to view signal waveforms by zooming in to capture glitches under all lighting conditions. The outdoor viewing mode, coupled with the 5.7-inch transflective display on these oscilloscopes, makes it possible to read waveforms even in bright sunlight. Key measurement capabilities include a sampling rate of 1 or 2 GSa/sec and two safety-isolated input channels. Analysis capabilities include deep memory, 1,000-times zooming, and dual zoom windows for overview and detailed displays. For more information, please click here.

Dual-Core GyroscopeSTMicroelectronics, a global semiconductor leader serving customers across the spectrum of electronics applications and the leading supplier of MEMS (Micro-Electro-Mechanical Systems) for consumer and portable applications 1, has introduced the market’s first dual-core gyroscope capable of handling both user-motion recognition and camera image stabilization. With the innovative system architecture of the ST device, equipment manufacturers only need to use a single gyroscope for the two different functions, reducing the size, system complexity,

and cost in mobile phones, tablets, and other smart consumer devices.The innovative design of ST’s dual-core gyroscope employs separate output paths optimized for the two different functions in a 4×4×1 mm package. The device addresses, at the same time, motion and gesture recognition and optical image stabilization for sharper mobile-camera photos. For more information, please click here.

MEM

S S

ENS

OR

MU

XM

UX

C - converter

V S C �lter ADC Digital �lter s

S PI

BW S election

Actuation Interface

C - converter

V S C �lter ADC Digital �lter s

I2C

BW S election

Actuation Interface

Full s caleS election

FIFO

OI S Readin g Chain

UI Readin g Chain

frequency with no external matching components required. Typical VCAs in this performance category have poor inherent attenuation versus temperature and poor nonlinear attenuation versus control voltage characteristics. For more information, please click here.

Voltage Controlled AttenuatorRFMD’s RFSA3023 is a fully monolithic analog voltage controlled attenuator (VCA) featuring exceptional linearity over a minimum temperature compensated 30dB gain control range. It incorporates a revolutionary new circuit architecture to solve a long standing industry problem: high IP3, high attenuation range, low DC current, broad bandwidth and temperature compensated linear in dB control voltage characteristic. This voltage controlled attenuator is controlled by a single positive control voltage with on-chip DC conditioning circuitry. The slope of the control voltage versus gain is selectable. The RFSA3023 draws a very low 1mA current and is packaged in a small 3mmx 3mm QFN. This attenuator is matched to 75Ω over its rated control range and

4

CONTROLBLOCK

ATTEN

RFIN RFOUT

VC

VD

D

MO

DE

5

3

2

1

6 7 8

9

10

11

12

13141516

GND

NC

NC

GN

D

GN

D

GN

D

GN

D

NC

NC

GND

GN

D

http://www.agilent.com/about/newsroom/presrel/2012/04jan-em12001.html

http://www.st.com/internet/com/press_release/p3260.jsp

https://estore.rfmd.com/RFMD_Onlinestore/Products/RFMD+Parts/PID-P_RFSA3023.aspx?DC=25


Bumblebees

DFE

What

and Models ofHave inCommon

Michael SteinbergerLead Architect, Serial Channel Products

While at first it may seem impossible or impractical to create an Input-output Buffer Information Specification, Algorithmic Modeling Interface (IBIS-AMI) model of Decision Feedback Equalization (DFE), it’s actually quite easy as long as you understand some unavoidable artifacts in the results.

1.0 “It Can’t Be Done”

Legend has it that aerodynamic analysis proves bumblebees can’t fly. Similarly, there is a line of reasoning that proves that IBIS-AMI [1] [2] models of DFE can’t possibly support statistical analysis, and cannot support time domain simulation of some popular DFE architectures.

Just as the (possibly apocryphal)

aerodynamic analysis fails to account for the bumblebee wing’s increased thickness (due to boundary layer thickness at low Reynolds number), and therefore its increased lift coefficient, there exist arguments about DFE models that are more focused on the problem than the solution.

This article will demonstrate how IBIS-AMI models can be written for some complex DFE architectures, and then demonstrate how such models can be written to support statistical analysis as well as time domain simulation.

2.0 Speculative DFE Architecture

The primary role of the IBIS-AMI specification is to define a standard

interface to the model. While having a standard interface enables many solutions, it also imposes some constraints. One of those constraints is that the model has a single input and a single output for the primary data signal, with supplementary inputs and outputs only available for including crosstalk in statistical analysis. There are, however, many receiver architectures that have multiple data paths, with the detected data selected from the outputs of these paths. How can IBIS-AMI modeling represent such architectures?

The speculative DFE architecture has been used for a number of years, and is a good example of the challenges posed by multiple data paths. In the speculative DFE


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

Figure 1: Speculative DFE Operation

architecture, there are two data paths that receive the same incoming data signal but apply equal and opposite low-frequency offsets to that signal. The magnitude of this offset is equal to the value of the first DFE tap. Both data paths make a decision at every bit time, but then the data bit delivered downstream is selected based on the previously detected bit. If the previously detected bit was a one, then the result from the negative offset data path is chosen. And if the previously detected bit was a zero, then the result from the positive offset data path is chosen.

The advantage of the speculative DFE architecture is that it is not necessary to feed the detected data bit back to the decision circuit input in time for the detection of the very next data bit. Instead, the choice of detected bit is deferred for one bit time, therefore providing a lot more timing margin in the circuit, at the expense of some duplication of circuitry.

To avoid complicating the example, Figure 1 illustrates the operation of a speculative DFE receiver with a single tap. In Figure 1, the magenta waveform is the incoming data signal, the blue waveform is the incoming data signal offset in the positive direction by the DFE tap and the yellow waveform is the incoming data signal offset in the negative direction by the DFE tap. This figure also shows markers for the edges of some of the data symbols. The data will be detected halfway between these markers.

Ideally, a model of this architecture would maintain both the positive and negative offset data paths and choose between them just the way

Time (ns)

Volts

(mV)

0

400.0

300.0

200.0

100.0

0.0

-100.0

-200.0

-300.0

-400.0

0.52

0.372n 0.572n 0.772n 0.972n 1.172n 1.372n

1.0 1.52 2.0 2.52 3.0

Waveform Incoming data signal Positive offset data path Negative offset data path

the real receiver does. The IBIS-AMI model interface will not support this. For any given bit, however, all that really matters is the value of the waveform that will actually be used to detect the value that will be delivered downstream. The model is capable of determining, for each bit, which of these two waveforms will be used, and is therefore able

to deliver the pertinent waveform segment for each bit. This is illustrated by the magenta waveform in Figure 2. Each detected data bit determines which waveform segment will be output for the next bit.

Figure 3 is the eye diagram that was produced from the model’s output waveform in Figure 2. Note that the

Time (ns)

XX

X X X X X : Timing & Locationof data detection

Volts

(mV)

0

400.0

300.0

200.0

100.0

0.0

-100.0

-200.0

-300.0

-400.0

0.52

0.372n 0.572n 0.772n 0.972n 1.172n 1.372n

1.0 1.52 2.0 2.52 3.0

Waveform Modeled waveform Positive offset data path Negative offset data path

Figure 2: Speculative DFE Modeling


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

middle of this eye diagram looks like an eye diagram that would be measured in the lab, but that the discontinuities in the model’s output waveform are clearly evident at the edge of the eye. Every model is an illusion to some extent, and in this case the illusion only works in the center of the eye.

Because the values that affect the bit error rate analysis are taken from the center of the eye and not the edge, the discontinuity at the edge of the eye is not relevant. If, however, one were to perform a jitter analysis based on this waveform, then the discontinuities in the model’s output waveform would render the results completely invalid. If a jitter analysis is to be performed, it should be performed on the model’s clock output and not its output data waveform.

3.0 Statistical Analysis

In order to be rigorously valid, statistical analysis must use the response of a linear, time-invariant

system. The input to a DFE tap is the output of the data detector, which is nonlinear. Furthermore, DFE is usually controlled by a continuously operating control loop. Therefore the tap weight is time-varying. So DFE is neither linear nor time-invariant. One would therefore conclude that statistical analysis in not applicable to a link that has DFE.

Consider the following assertions, however:

1. In steady state operation, the taps weights of a DFE are essentially constant.

2. The transmit amplitude of a high-speed serial channel is essentially constant.

3. In normal or expected operation, the receiver bit error rate is very low.

We define the following symbols:

x(t): The data signal driving the channel

Volts

(mV)

Prob

abili

ty

0

400.0

300.0

200.0

100.0

0.0

-100.0

-200.0

-300.0

-400.0

Persistant Eye DiagramOne tap speculative DFE

100.0 200.0 300.0

_ 1.3E-2

_ 3E-3

_ 1.6E-4

_ 2.8E-8

_ 1.9E-10

Time ( s)

Figure 3: Speculative DFE Eye Diagram

y(t): The signal at the receiver’s data decision point

z(t): The output data from the receiver

hc(t): The impulse response of the channel including any transmit equalization and any receiver linear equalization

wk: The k-th DFE tap weight

T: Transmit symbol duration

Tc: Channel delay

Then the signal at the receiver decision point is the sum of the linear channel/equalization response plus the DFE signal:

The third assertion implies

( ) ( )z t x t cx= -

Substituting Equation 2 into Equation 1,

( ) ( ) ( ) ( )y t h t x t w x t n cc k

k

n

1

7 x x= + - -=

/( ) ( ) ( ) ( )y t h t x t w x t n cc k

k

n

1

7 x x= + - -=

/

( ) ( ) ( ) ( )y t h t x t w z t nc k

k

n

1

7 x= + -=

/ (EQ1)

( ) ( ) ( ) ( )y t h t x t w z t nc k

k

n

1

7 x= + -=

/

(EQ2)

(EQ3)

Employing the Dirac delta function δ(t),

( ) ( ) ( ) ( ) ( )y t h t x t w t n c x tc k

k

n

1

7 7d x x= + - -=

/( ) ( ) ( ) ( ) ( )y t h t x t w t n c x tc k

k

n

1

7 7d x x= + - -=

/(EQ4)

( ) ( ) ( ( )y t h t w t n x tc k c

k

n

1

7d x x= + - --

c m/( ) ( ) ( ( )y t h t w t n x tc k c

k

n

1

7d x x= + - --

c m/(EQ5)

Note that Equation 5 appears to be a linear, time-invariant equation in which the DFE taps are represented as Dirac delta functions. The equation is not rigorously linear


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

in that it’s only valid if the transmit and receive data have the same amplitude and the bit error rate is low. Nonetheless, these assertions are true under a wide enough range of conditions that this equation is a useful engineering approximation.

Reference [2] gives more of the details on the practical application of this modeling approach. For the example given in the previous section, Figure 4 shows the impulse response of the end-to-end channel and Figure 5 shows the statistical eye diagram.

4.0 Conclusion

Bumblebees do fly, and IBIS-AMI models of DFE, correlated with measured data, are used routinely to produce accurate performance estimates for high-speed serial channels. The only caveat is that users need to understand the limitations of the model and therefore need to ignore the discontinuity at the edge of the eye diagram.

5.0 References

[1] IBIS (I/O Buffer Information Specification) Version 5.0, August 29, 2008.

[2] Michael Steinberger, Todd Westerhoff, Christopher White, “Demonstration of SerDes Modeling using the Algorithmic Model Interface (AMI) Standard”, DesignCon2008, paper 7-TA3, February 5, 2008.

About the Author

Michael Steinberger, PhD, has over 30 years experience in the design and analysis of very high-speed electronic circuits. Dr.

Time (ns)

Channel

Volts

(GV)

0

0.0

-10.0

-20.0

-30.0

-40.0

-50.0

-60.0

1.0 1.52 2.0 2.52 3.0

Impulse ResponseOne tap speculative DFE

DFE

Figure 4: Impulse response with single tap DFE

Figure 5: Statistical eye diagram with DFE

Steinberger began his career at Hughes Aircraft, designing microwave circuits. He then moved to Bell Labs, where he designed microwave systems that helped AT&T move from analog to digital long-distance transmission. He was instrumental in the development of high-speed digital backplanes used throughout Lucent’s transmission product line. Prior to joining SiSoft, Dr. Steinberger led a group of over

20 design engineers at Cray, Inc. responsible for SerDes design, high-speed channel analysis, PCB design, and custom RAM design. ■

Join Today

www.eeweb.com/register


http://bit.ly/pjX3rr



RANGETESTING

REALWORLD

Christopher HofmeisterFirmware Engineer

This article outlines the procedure for a successful RF range test that provides quantitative data on how the RF link performs in its intended environment. This enables informed decisions about network architecture.

The article also explains subtleties that might be overlooked during a RF range test that can affect performance. Although these principles apply to any radio, we give special attention to features of the 802.15.4 standard that can affect range test results.

The design of a RF range test is broken down, and specific elements are explained in a real-world range test using LS Research’s RF modules.

So Many Variables

One of the most frequently asked questions about a given radio is, “How far will this radio transmit?” Unfortunately, it is also one of the most difficult questions to answer, as so many variables apply. Besides obvious variables like transmit power and receive sensitivity, a host of other factors come into play. Without careful attention to understanding and controlling as many of these

variables as possible, the results of an RF range test will leave more questions than answers.

For example, consider this common scenario. An engineer evaluating a radio puts the radio in a spot, walks some distance from it while monitoring an LED that indicates a received packet, and hopes to find a magical line on the ground at which the radio works on one side and doesn’t work on the other. In real life the link will become marginal at some point and continue to work intermittently, perhaps even improving at certain points, forcing the engineer to make a judgment call on how far the radio worked. A week later the same test performed by the same or a different individual yields drastically different results. What happened?

As we attempt to reconcile the results, questions come up. Even if the test was done outdoors, many factors apply. What antennas were used? How were they oriented? How far off the ground were they? How were the radios powered? If batteries were used, what was their voltage? Was the same version of firmware used? What was the over-the-air data rate? How many bytes were in the RF


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

packet? What was the RF environment like during each test? How was it determined that the link was bad? Were RF acks and retries used? What obstructions were in the path of the radios? The answers to these questions reveal that the trials had little in common, and the results are inconclusive.

Add to this the complexities of an office or industrial environment. The RF environment is constantly changing as laptop computers move about and cordless phones or even microwave ovens are operated, affecting the 2.4GHz band in particular. Doors open and close, office equipment is turned on and off, equipment is moved, and so on. Also, the building’s physical construction certainly has an effect; an addition may make it appear that an outside wall of concrete, stucco or metal siding is an interior wall.

These situations illustrate why it is so hard to answer the question, “How far will this radio transmit?” Clearly, as many variables as possible need to be controlled, and those that cannot be controlled at least need to be understood. When radio manufacturers list a range distance, be sure to ask:

• How were those results obtained?

• Is this the environment in which my radios will operate?

• What was the performance of radios at this distance?

Good Range Test Design

Specifying Range

Most manufacturers specify range in an optimum situation, specifically line-of-sight outdoors. Besides putting their best foot forward, this enables them to repeatedly reproduce the test conditions and obtain similar results. This does not mean, however, that users of the radio can obtain the same results in a typical environment.

Before running the range test, determine the criteria for specifying how far the radio transmits. This is especially important when trying to compare different radios. Some may specify this as the farthest point at which 99 percent of packets are successfully transmitted. Others may plot out the average receive sensitivity over distance to determine range. A combination of both metrics could also be employed.

Write Everything Down

It should go without saying, but it’s easy to take shortcuts under the premise that we will remember. In reality, hours turn into days, and days into weeks, and we can’t recall the details. So take the time to write everything down. You could use the example at the end of this paper to create a form that ensures all important details are noted.

Collect Quantitative Data

A good range test will collect useful information, such as the number of packets transmitted, the number of packets received, RSSI, LQI, the amount of noise on the channel, and other test conditions such as timestamps on the packets, the number of bytes in the packet, and power supply conditions. Such information should be logged into a PC to enable analysis of the data collected.

Understand the Difference Between LQI and RSSI

RSSI is an initialism for Received Signal Strength Indicator. LQI is an initialism for Link Quality Indicator. Some use the terms interchangeably, but they are different. RSSI is an absolute value that can define the magnitude of the received signal, and it can be converted into units of dBm. LQI is a relative number between 0 and 255 that represents the quality of the received signal. Since there are no official rules on how it is calculated, every stack vendor usually calculates LQI differently. It is generally a function of RSSI and possibly other indicators such as the quality of the modulation on the received signal.

When range testing a radio, it is generally good to track both RSSI and LQI and to understand how the LQI is being calculated. A strong RSSI but weak LQI (assuming LQI is factoring in the quality of the modulation) would indicate the presence of RF noise or interference.

Understand Acknowledgment Requests and Retries

Many radio standards today, such as 802.25.4, employ a RF acknowledgment/retry mechanism. When they are mentioned, it is important to understand that they can exist in the MAC layer and the application layer of software. In 802.15.4, acks/retries can be enabled on a packet-by-packet basis in the MAC. If a message is sent with MAC layer ack, the recipient of the message will send a very small RF acknowledgment back to the sender, confirming receipt. If the sender does not receive this ack, it will resend (retry) the message again, up to


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

three times. If you are doing the math, this means that the same message could be transmitted up to four times: the original message and up to three retries.

It is often assumed that if a message is retried, it was not

it may appear that the link is very good when in reality it is on the fringes. A clue that this is happening could be gathered from statics kept on each side of the link. If the originator side shows 100 packets were sent out and the recipient shows 300 messages were received, this is a sure sign that the network relied heavily on the ack/retry mechanism.

Although not the focus of this paper, application layer acks/retries have more practical value in hopping networks or in networks where extremely high reliability is required. The MAC layer acks/retries only guarantee that the message made it to the next hop—not the final recipient. Also, even in single-hop systems, this method does not guarantee that the message made it to the application. For example, a node could receive an RF message, send the MAC layer ack, and then try to allocate memory to send the message to a PC or display but run out of memory. The message originator would have no knowledge of this problem.

ORIGINATOR RECIPIENT

RF Message withMAC Ack Request

Ack

Figure 1: MAC layer acks/retries

heard by the recipient. In practice, it is not uncommon for the acknowledgment to go unheard by the originator, prompting the originator to resend the message. This means that it’s possible for the recipient to hear the same message multiple times.

RECIPIENT

RF Message withMAC Ack Request 3

Ack 3

Ack 1

Ack 2



ORIGINATOR

Figure 2

It is important to consider that using MAC layer acks/retries may create a false sense of security regarding RF-link quality. For example, if the ack/retry mechanism results in each message being sent two or three times,

ORIGINATOR RECIPIENT

RF Alarm Message withMAC Ack Request

MAC Ack

Figure 3: Pitfall of MAC-only acks

LS Research Range Test

LS Research has developed a protocol for evaluating range of its radio modules, and the principles contained therein can be used to conduct a successful range test. One part of the process involves using a simple form, like the one at the end of this paper, to write down important details about the test, including details that are easy to overlook. The second part of the process combines the firmware on the radio module being tested with PC software to display and record the results. The range test is also designed to achieve the following goals:

• Obtain quantitative range test results (Packet Error Rate, RSSI, LQI, Background RF Noise).

• Establish the ability to restart data collection remotely —that is, look at performance in various locations


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

without having to walk back to a PC to restart the test.

• Enable a single user to perform the entire test.

• Provide a means to log the results.

• Record as many variables (such as battery voltage) as possible.

• Use only one PC.

Materials for Range Test

The range test in this paper was performed with two of the ModFLEX series radios, the ProFLEX01 (2.4GHz DSSS) and SiFLEX02 (900MHz DSSS). Development kits for the radios are commercially available. The module firmware and PC software (ModFLEX Test Tool Suite) are available for download from the LS Research web site at http://www.lsr.com.

Over-the-Air Packet Structure

Necessary parts of range test results are transmitted over the air to allow either end of the system to collect data. Each time the master transmits a packet, it increments its packet ID. If the “application ack” is set to 1, the slave device will transmit a packet back to the master, making it a round-trip test. The following example shows what these packets would look like.

Master-to-Slave Packet

• The packet ID indicates how many packets have been transmitted by the master.

• The power supply voltage of the master is sent to the slave.

• If the slave receives the packet, these can be determined: 1 packet was transmitted, 1 packet was received (100% success); and RSSI, LQI, and battery voltages.

• Slave information is 0 as this information needs to be filled in from the slave.

Slave-to-Master Packet

• The slave appends information into the packet: RSSI and LQI of the master-to-slave packet; battery voltage.

• If the master receives the packet, these can be determined: 1 packet was transmitted, 1 packet was received (100% success); RSSI, LQI, and battery voltages of itself and the slave.

Missed Packets

If the slave misses several packets, the statistics will correct themselves as soon as another packet is received. For example, if the slave does not hear packets 90–99 but does hear packet 100, the over-the-air message would look like the one in Figure 5. It would be determined that the master sent 100 messages but the slave heard only 90.

Restarting Test in the Field

Smarts are built into the range test application to start/restart the test if the packet ID of the received message is less than the packet ID of the previous message. This would allow a user to restart the packet-error-rate calculation at distance intervals without having to reset both ends of the test. Table 4 provides an example of how this would look in the over-the-air packet.

RF Acknowledgments and Retries

In the LSR range test, application layer acknowledgments can be turned on or off. Turning them off makes the test one-dimensional; turning them on creates a round-trip test. Application layer acks/retries can be used in conjunction with MAC layer acks/retries.

Host Serial Message Packet Structure

Applicable information from the received range test RF packet and additional information is formatted into a serial message and passed on to the PC. The ModFLEX Test Tool Suite uses this information to display and log the results of the range test. More information on the serial message is available in the respective modules’ Host Protocol Guide at the LS Research website: http://www.lsr.com.

SiFLEX02 Results

Results were graphed for RSSI vs. distance on each antenna type and data-rate. The results from each end of the test were graphed: what the slave saw from the master (MS) and what the master saw from the slave (SM).

At 40kbs, the SiFLEX02 with the +2dBi dipole antenna had adequate link margin to the end of the lake at 1.5 miles. If the data were extrapolated, it is reasonable to assume the link margin would be adequate to at least 2.0 miles.

http://www.lsr.com


TECHN

ICA

L ARTIC

LETECHNICAL ARTICLE

At 40kbs, the SiFLEX02 with the wire antenna had adequate link margin to the end of the lake at 1.5 miles. If the data were extrapolated, it is reasonable to assume the link margin would be adequate to at least 1.75 miles (Figure 4).

ProFLEX01 Results

Figure 17 shows RSSI vs. distance on each antenna type. The graph shows results from each end of the test: What the slave saw from the master (MS) and what the master saw from the slave (SM).

SiFLEX02 (900 MHz)

RSSI

(dBm

)

-65

-70

-75

-80

-85

-90

-95

-100

-1050.25-69-74-69-73

Dipole 40kbps S-->M

0.5-75-81-75-80

0.75-76-87-75-86

1-82-92-82-91

1.25-86-97-83-97

1.5-88-96-89-96

1.75 2 2.25

Wire 40kbps S-->M

Dipole 40kbps M-->S

Wire 40kbps M-->S

The ProFLEX01 unit showed adequate link margin to about 1 mile (Figure 5).

About the Author

Christopher Hofmeister, a senior software engineer at LS Research, has spent more than 15 years in the low-power wireless industry. He has been involved in hardware and firmware design as well as testing of low-power RF transmitters, receivers and transceivers for both engineering validation and production testing. ■

Figure 4: RSSI vs. distance for wire antenna and +2dBi dipole antenna

ProFLEX01 (2.4 GHz)

RSSI

(dBm

)

-75

-80

-85

-90

-95

-100 0.25-78-86-77-85

0.5-89-84-88-83

0.75-94-87-93-86

1-96-95-96-94

1.25-97-97-97-96

1.5

-99

-98

Dipole Antenna S-->M

F-Antenna S-->M

Dipole Antenna M-->S

F-Antenna M-->S

Figure 5: RSSI vs. distance for F-antenna and +2dBi dipole antenna

Dual 800mA Low Quiescent Current 2.25MHz High Efficiency Synchronous Buck RegulatorISL78228The ISL78228 is a high efficiency, dual synchronous step-down DC/DC regulator that can deliver up to 800mA continuous output current per channel. The supply voltage range of 2.75V to 5.5V allows for the use of a 3.3V or 5V input. The current mode control architecture enables very low duty cycle operation at high frequency with fast transient response and excellent loop stability. The ISL78228 operates above the AM radio band as well as the 2.25MHz switching frequency, allowing for the use of small, low cost inductors and capacitors. Each channel is optimized for generating an output voltage as low as 0.6V.

The ISL78228 has a user configurable mode of operation-forced PWM mode and PFM/PWM mode. The forced PWM mode operation reduces noise and RF interference while the PFM mode operation provides high efficiency by reducing switching losses at light loads. In PFM mode of operation, both channels draw a total quiescent current of only 30µA, hence enabling high light load efficiency in order to maximize battery life.

The ISL78228 offers a 1ms Power-Good (PG) to monitor both outputs at power-up. When shutdown, ISL78228 discharges the outputs capacitor. Other features include internal digital soft-start, enable for power sequence, overcurrent protection, and thermal shutdown. The ISL78228 is offered in a 3mmx3mm 10 Ld DFN package with 1mm maximum height. The complete converter occupies less than 1.8cm2 area.

The ISL78228 is both AEC - Q100 rated and fully TS16949 compliant. The ISL78228 is rated for the automotive temperature range (-40°C to +105°C).

Features• Internal Current Mode Compensation

• 100% Maximum Duty Cycle for Lowest Dropout

• Selectable Forced PWM Mode and PFM Mode

• External Synchronization up to 4MHz

• Start-up with Pre-biased Output

• Soft-Stop Output Discharge During Disabled

• Internal Digital Soft-Start - 2ms

• Power-Good (PG) Output with 1ms Delay

• TS16949 Compliant

• AEC - Q100 Tested

• Pb-free (RoHS Compliant)

Applications• DC/DC POL Modules

• µC/µP, FPGA and DSP Power

• Rear Camera Systems

• Navigation Systems

• Infotainment Systems

FIGURE 1. EFFICIENCY CHARACTERISTICS CURVE

OUTPUT LOAD (A)

EFFIC

IEN

CY

(%

)

40

50

60

70

80

90

100

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

2.5VOUT-PFM

1.8VOUT-PFM

2.5VOUT-PWM

1.8VOUT-PWM

VIN = 5V

May 2, 2011FN7849.0

Intersil (and design) is a registered trademark of Intersil Americas Inc. Copyright Intersil Americas Inc. 2011All Rights Reserved. All other trademarks mentioned are the property of their respective owners.

Get the Datasheet and Order Samples

http://www.intersil.com

http://bit.ly/nHxRhB


RETURN TO

ZERORETURN TO ZERO


RETURN TO

ZERORETURN TO ZERO

Contact Us For Advertising Opportunities

[email protected]

www.eeweb.com/advertising


http://bit.ly/l6u6Cx

EEWeb Pulse - Volume 28

Documents

Transcript of EEWeb Pulse - Volume 28