Servo magazine 01 2008

Tạp chí Servo

SERVO Magazine (ISSN 1546-0592/CDN Pub Agree

#40702530) is published monthly for $24.95 per year by

T & L Publications, Inc., 430 Princeland Court, Corona, CA

92879 PERIODICALS POSTAGE PAID AT CORONA, CA AND

AT ADDITIONAL ENTRY MAILING OFFICES POSTMASTER:

Send address changes to SERVO Magazine, P.O Box

15277, North Hollywood, CA 91615 or Station A, P.O.

Box 54,Windsor ON N9A 6J5; cpcreturns@servomagazine.com

08 Robytes by Jeff Eckert

Stimulating Robot Tidbits

10 GeerHeadby David Geer

Zeno — The First Complete Character Robot

14 Ask Mr Roboto by Pete Miles

Your Problems Solved Here

61 Twin Tweaks

by Bryce and Evan Woolley

More Than Meets the Eye —

the Mighty Morphing V-Bot

67 Different Bits

by Heather Dewey-Hagborg

Neural Networks for the PIC Microcontroller

Part 4: Self-Organizing Maps

71 Robotics Resources

by Gordon McComb

Small Brains for Your Bot

75 Appetizer by Gerard Fonte

Terms of Endearment

79 Then and Now by Tom Carroll

Robots Take to the Air

PAGE 24

VOL 6 NO 1

ENTER WITH CAUTION!

MEET THE MIGHTY MORPHING V-BOT/pg 61

by Michael Simpson

Part 4: A closer look at the interface needed for each of the GPS modules covered in this series.

by Fulvio Mastrogiovanni

This article focuses on the entire control loop that’s used to allow a rover to safely navigate on Mars using information from Earth.

Microcontrollers Made Easy

Give ARTI ONE a new lease on life.

Features & Projects

Published Monthly By

T & L Publications, Inc.

430 Princeland Court Corona, CA 92879-1300

(951) 371-8497 FAX (951) 371-3052 Product Order Line 1-800-783-4624 www.servomagazine.com

Subscriptions

Inside US 1-877-525-2539 Outside US 1-818-487-4545

P.O Box 15277 North Hollywood, CA 91615

PUBLISHER

Larry Lemieux

publisher@servomagazine.com ASSOCIATE PUBLISHER/

Heather Dewey-Hagborg Fulvio Mastrogiovanni

CIRCULATION DIRECTOR

Tracy Kerley

subscribe@servomagazine.com MARKETING COORDINATOR

WEBSTORE

Brian Kirkpatrick

sales@servomagazine.com WEB CONTENT

Michael Kaudze

website@servomagazine.com PRODUCTION/GRAPHICS

Shannon Lemieux Joe Keungmanivong

ADMINISTRATIVE ASSISTANT

Debbie Stauffacher

Copyright 2008 by

T & L Publications, Inc.

All Rights Reserved

All advertising is subject to publisher’s approval.

We are not responsible for mistakes, misprints,

or typographical errors SERVO Magazine

assumes no responsibility for the availability or condition of advertised items or for the honesty

of the advertiser.The publisher makes no claims

for the legality of any item advertised in SERVO.

This is the sole responsibility of the advertiser Advertisers and their agencies agree to indemnify and protect the publisher from any and all claims, action, or expense arising from

advertising placed in SERVO Please send all

editorial correspondence, UPS, overnight mail,

and artwork to: 430 Princeland Court,

INFRASTRUCTURE

A fundamental aspect of robotics is

that the application domain can range

from ocean beds and table tops to the

nooks and crannies of extraterrestrial

dunes It’s no coincidence that the

techniques and technologies described in

the article featured in this issue of SERVO

can be applied to virtually any application

areas However, if you have a particular

interest in space exploration, then you’ll

find Fulvio Mastrogiovanni’s article,

“Space Robotics,” of particular note.

Fulvio, a PhD candidate from the Mobile

Robotics and Artificial Intelligence at the

University of Genova, Italy, offers a

focused consideration of control theory

applied to the practical challenges

presented by the NASA Mars Explorer

Rover mission The article also hints at an

often ignored and poorly understood

aspect of robotics — that of infrastructure

Space exploration and transport,

together with development of more

traditional military gear, are responsible for

many of the innovations in sensors,

software, and platform designs that trickle

down to civilian robot developers.

However, even if you have access to

the components and algorithms used by

NASA engineers, you’d probably find it

impossible to develop a robot that even

approximates the abilities of the impressive

Martian rovers The missing ingredient is

infrastructure — the robots, conventional

tools, and processes used to develop the

vehicles and robots destined for space.

Consider the array of advanced

robotics used in the construction and

testing of the shuttle’s external tank The

shuttle’s external tank, which has a length

of 55 feet and diameter of 28 feet, is

fabricated at the Michoud Assembly

Facility, near New Orleans When I toured

the facility, I was amazed at the robots

and other automated equipment required

to achieve the tolerances necessary for

space flight In particular, I witnessed a

fully assembled fuel tank mounted

horizontally on a motorized spindle in a

way that enabled robots to apply

insulating foam to the aluminum-lithium tank I tried to imagine the control systems and motors necessary to rotate the 58,000 lb tank and maintain a precise coating depth Although I haven’t seen it first hand, I assume that the robotic equipment used to create the multi-stage Boeing Delta II that transported the Martian rovers is just as impressive.

Sam Christy’s article on programming for FIRST controllers and Tim Paterson’s article on square roots illustrate a key

‘invisible’ component of the infrastructure required to develop robots If you’re like many roboticists, you probably have a drill press, soldering iron, a few boxes of spare parts, perhaps a library of reference texts, and a PC To the uninitiated, the PC may

be simply another artifact in your workshop However, if you’re involved with developing control loops, machine vision, wireless communications, or other computational tasks, then your PC may be your primary development tool In this regard, traditional tools and components often fail to capture the enormity of effort, planning, and time that goes into developing a robot.

If you’re new to robotics, then you’re just beginning to appreciate the depth and breadth of your development infrastructure (or lack of it) Whether you’re primarily involved in developing pneumatic weapons for battle bots or vision recognition algorithms for a commercial robotic platform, you’ll soon discover that a development infrastructure is a prerequisite for efficient, unencumbered robot development.

At the start of my robotics career, I invested almost a year of effort developing

a flexible infrastructure The major components include:

• Drill Press and Bit Assortment

• Vise

• Work Bench

• Dremel and Accessories

• Dual-Trace Digital Oscilloscope

• Air Compressor

• Hand Tools

• Glues and Adhesives

• Nut and Bolt Library

• Clamps

• Cables, Wire, and Shrink Tubing

• Connectors and Pins

• Soldering/Desoldering Station

• R/C Unit

• Nibbler Tool

• PC

• Software (Schematic generation, simulation, compilers)

• Sensors (US, IR, motion, etc.)

• Microcontrollers (STAMP, PICs, ATMEL)

• Storage Bins

• Breadboard System

• Multi-drawer Tool Chest

I’m still working on the infrastructure, but at a much lower

level So, how do you go about building an infrastructure? If

you’re fortunate enough to be financially well positioned, then

the anticipatory approach is a viable option Assuming that you

can accurately anticipate your upcoming needs, then you can

assemble an infrastructure within a few weeks Even with

equipment in hand, you’ll need several weeks to learn how to

operate and apply your new hardware and software.

At the other extreme is the as-needed approach, which

entails purchasing tools and test equipment on an as-needed

basis While easy on the pocket, this approach often results in a

loss of momentum Stopping a project midway to await delivery

of a drill or torque wrench and then learning how to use the

device can derail an otherwise focused project There is also the

defocusing associated with taking time to identify the best

oscilloscope, drill press, or other item.

The two approaches aren’t mutually exclusive For example,

I use a hybrid approach in which a few major purchases — drill

press, multimeter, hand tools, and oscilloscope — are added to as

needed I’ve learned the hard way that when you build an

infrastructure, buy the best that you can afford Don’t be lured by

an inexpensive hand tool or soldering iron that will satisfy your

current project Instead, try to anticipate what you’ll need over

the next five years It costs more to buy a cheap tool and then a

more expensive tool a few months later, than to buy the right tool

to begin with I’m a big fan of eBay where — if you’re patient

enough — you can find good deals on equipment that might

otherwise be out of your reach.

Another option is to outsource your infrastructure by joining

a well-stocked robotics club For modest dues, you can have

access to a supportive infrastructure that may prove invaluable

to your success in robotics A related approach is

to extend your infrastructure with external services, such as

laser cutting I use Pololu Robotics and Electronics

(www.Pololu.com) for laser cutting on large projects.

Outsourcing can be expensive, but it allows you to focus on what

you do best.

As a word of warning, as you construct your infrastructure,

remember that it’s tempting to use robotics as an excuse to

amass a huge collection of fantastic tools and equipment This is

fine if your intent is to collect tools However, if your goal is to

produce functional robots, then do your best to avoid the

seduction of gear SV

Advanced Course3 Days$1995

Program and Control:

GPS Receivers RS232, 485 Serial Communications Wireless Communications IEEE Zig Bee, Bluetooth Distance Sensors RFID Readers and Tags Motor Control W/Feedback

Minneapolis, MN, Jul 21-23 Los Angeles, CA, Aug 18-20 Newark, NJ, Sep 22-24 San Diego, CA, Oct 6-8

Check the web for more dates

Intro Course3 Days$1995

Program and Control:

LED and LCD Displays

DC, Servo, and Stepper Motors Temperature and Light sensors RS232 Serial Communications ADCs and EEPROMs Timers, Interrupts, & Clocks Visual Basic Interfaces

Columbus, OH, Jan 21-23 Los Angeles, CA, Feb 4-6 Huntsville, AL, Feb 18-20 Newark, NJ, Mar 3-5

Check the web for more dates

WoW! Finally, Hands-on Training for

RCG Research

Innovative Ideas in Electronics Design

Monster power!

14-50V 3.15 x4.5”x1.5”

fwd-only channel 5V - 18V

1.6“ x 1.6” x 0.5”

2.5A 12A

6 6 6

5V - 24V 2.7“ x 1.6” x 0.5”

13A 45A Peak!

6 6 6 6 6 6 6

14V - 50V - Dual H-bridges - 150 Adjustable current limiting Temperature limiting Three R/C inputs - serial option Many mixing options - Flipped Bot Input Rugged extruded Aluminum case 4.25" x 3.23" x 1.1”

Closed-loop control of two motors Full PID position/velocity loop Trapezoidal path generator PIC18F6722 CPU

C source for routines provided See

Giant Servo Mode!

www.embeddedelectronics.net

$250

MAD E IN THE USA

H-bridges: Use with Dalf or with your Micro/Stamp

Simple-H

6 6 6 6

6-28V 25

2.25 x2.5 x0.5”

3 wire interface current & temp protection

A!

$79

In re-reading my December 2007 Robotics Resources column, I spotted

a small error regarding the light wavelength of the laser diodes used in

DVD players Most commercial DVD players use 650 nanometer (deep

red) laser diodes, rather than the 780 nanometer (infrared) laser diodes

common in CD players.

Gordon McComb Robotics Resources

Goodbye E-Harmony, Hello

Bot-Harmony

You may have noticed (or tried not

to notice) that some robots are

becom-ing a lot more lifelike and even allurbecom-ing

One example is Dion, a Chinese babe

who is said to mimic all sorts of human

features, including facial expressions,

skin temperature and elasticity, breath,

and heartbeat According to the

manu-facturer, she can even be built to

resem-ble the specific person of your choice

Another deliberately seductive

mechanism is Actroid DER2, developed

at the University of Osaka and

manufac-tured by Kokoro Co This one is

designed “to play an active part for

many occasions as a chairperson with

fluent narrations and booth bunny.” You

can even rent her for a five-day outing

for $400,000 yen (about $3,500 US)

I didn’t locate any specific

infor-mation about how intimate these

yum-bots are designed to be, but if you are

presently settling for a partner who

has to be inflated for each romantic

encounter, you might be interested in

a recent book by AI expert David Levy

The good news is that it someday will

be common for people to marry and

have sex with robots The bad news is

that Levy doesn’t expect it to be legal

and acceptable until about 2050 He

also predicted that Massachusetts will

be the first to legalize it

If you want to follow that line oflogic, pick up a copy of Love and Sex with Robots: The Evolution of Human-Robot Relationships, published

late last year by HarperCollins (www.

harpercollins.com) The hardcover

edition will cost you $24.95, or youcan get the e-book for $19.95

Arm Inspired by the

Real Thing

Ever since the Barrett Hand —

a three-fingered manipulator used extensively by NASA — was introduced in

1997, mechanical arms have grown moredexterous and complex, but still haven’tlooked or operated that much like thehuman counterpart But an eerily lifelikeone, even without any skin covering, hasbeen brought to us by Germany’s Festo

Corp (www.festo.com).

The company is known primarily forits pneumatic and electromechanical systems, components, and industrial controls But perhaps its most interestingaccomplishment is the creation of Airic’sArm Based on the real thing, it even hasmetal versions of the radius and ulna, themetacarpals, and the shoulder and shoulder blade One big difference, how-ever, is that it is powered by 30 of Festo’s

“Fluidic Muscles,” which are tubes ofelastomer reinforced with aramide fibers

When you fill one with compressed air, its diameter increasesand its length shortens, thus creatingmovement This actually gives Airic

an advantage over the rest of us: When the arm contracts, it doesn’trequire any additional power to staycontracted; that is, it can lift some-thing and hold it in place indefinitely.Festo intends to further developthe system by adding vision and tacticsensors and maybe even giving it aneck, a back, and a hip A short video

of the thing in action is presently at

www.youtube.com/watch?v=tVg6 xKHJKY4.

Feeding Robot for the Disabled

On a more specialized level is the

My Spoon feeding robot manufactured

in Tokyo by SECOM Co., Ltd Designed

to allow victims of spinal cord injuries,muscular dystrophy, and other disabili-ties to feed themselves without a care-giver’s assistance, it offers interchange-able controllers and utensils, and aselection of operating modes to accom-modate different disability levels It’s notquite up to the flexibility of a humanoidhand in that foods must be served fromspecific divisions of a meal tray, and theymust be in bite-sized pieces

Both users and therapists seem to

be praising its effectiveness, and it waseven nominated as a finisher in the Top

10 Robots listed by Japan’s Ministry of

Dion, from Beijing Yuanda Super

Robot Technology Co., Ltd., and

Actroid, from Kokoro Co., Ltd.

Airic’s Arm, perhaps the most realistic robotic arm yet.

Photo courtesy of Festo Corp.

The My Spoon™ feeding apparatus Photo courtesy of SECOM Co.

by Jeff Eckert

Economy, Trade, and Industry No price

is given on the company’s website

(www.secom.co.jp/english/index.ht

ml), which probably indicates that it will

run you a pretty nice chunk of change

But the main idea is to give the user

a greater sense of independence, which

is not a bad thing And it could probably

be fitted with a nice demitasse spoon,

making it useful for a fair number of

Hollywood actors and pop stars

Creeper Checks Structural

Integrity

Quite often, even creepy little

robots on tracks can cost an amazing

amount of money For example, it was

reported that the US Army recently

ordered 40 PackBots from iRobot

(www.irobot.com) at a cost of $8.8

million, or $220,000 each It is therefore

interesting that Sanyo appears ready to

introduce a model that is rumored to be

priced at a paltry one million yen (about

$8,900) Sure, the military machines are

more heavily equipped and hardened

for rough duty and, sure, we’re

comparing horses to ponies But still, at

about 1/25th of the cost

In any event, details are spotty, but

Sanyo's new machine — so far, rather

generically named the Underfloor

Robot — is designed to scamper around

underneath potentially crumbling office

buildings, apartments, and other places

that may be suffering from structural

damage It runs about two hours on a

charge, performing visual inspections

and beaming back video of whatever it

finds It is equipped with a zoom lens

for getting down to details and will

automatically create reports and

measure the distance between objects

Other specs include dimensions

(L, W, H) of 420 x 260 x 200 mm (16.5

x 10.2 x 7.9 in), base weight of 9.6 kg

(21 lb), and the ability to traverse

bumps up to 85 mm (3.4 in) in height

Power comes from a lithium-ion

battery pack, as usual More detailed

data should be appearing at www.

sanyo.com by the time you read this.

Watch for Floating Objects

Okay, they’re not the most complexself-operating devices you’ve ever seen,and all they basically do is bob up anddown; measure tempera-

ture, salinity, and velocity;

and relay the informationback home But theamazing thing is that theInternational Argo Project

(www.argo.net), with

support from more than

40 countries, haslaunched more than3,000 of the things since

2000, and they are tered all over the globe

scat-In the operationalcycle, each float spends

10 hours at the surface,descends to its driftingdepth of 1,000 m (3,281ft), and stays there foreight to 10 days, drops

to the profiling depth of2,000 m (6,562 ft), thenmakes its measurementsduring a 10 hour ascent

The objective is toallow the world, for thefirst time, to take thepulse of the oceans on a

continuous basis and relay the information to whoever needs it within hours of its collection And

“whoever” includes you

For the beginner’s guide to

accessing Argo data, log onto www argo.ucsd.edu/Frbeginnersguide html You’ll be giving speeches on

climatology in no time SV

R o b y t e s

Diagram of an Argo Active Float.

Photo courtesy of the Argo Information Centre iRobot’s PackBot Explorer and Sanyo’s upcoming underfloor robot.

Photos courtesy of iRobot and Sanyo Electric Co Ltd.

Zeno — due on the market as a toy

in 2009 — is the closest thing to

human that a robot has become

Its facial expressions are a story all

to themselves, enabling the most

complete robot personality and

human-to-robot emotive interactivity to

date Want some proof? Read on!

Zeno is a 16-inch, six-pound,

interactive robot boy developed by

Hanson Robotics with help from a

number of vendors including

RoboGarage and roboticist Tomotaka

Takahashi, of Japan, who is responsible

for creating Zeno’s body

The body prototype uses 18

servos, enabling the robot to walk, run,

balance on one foot, lie down and

get up, and gesture, as in non-verbal

communication, according to Dr David

Hanson, PhD, who worked with

Takahashi to produce the combined

robot, Zeno

Zeno is Born

Only recently completed (September2007), Zeno, the boy robot, offers everyhuman facial response Zeno (animationsoftware from Massive Software andMaya software) is the world’s first

“complete character robot,” according to

Dr Hanson

Complete character means Zenohas facial expressions, walking and ges-turing expressions, and conversationalcapabilities like a complete humanbeing would have

Zeno smiles, frowns, and getsangry He looks sad, surprised, andafraid He can present confusion or concentration as well, according to Dr

Hanson The mouth moves when Zeno isspeaking, the eyelids each work inde-pendent of the other, and the eyeballsturn side-to-side, and look up and down

“In the toy, we intend for the ears

to wriggle,” says Dr Hanson Zeno cannod, turn, and tilt its head “Thesemotions can be used expressively, aswell as to affect eye contact with people and otherwise look around therobot’s environment,” Dr Hanson says.Zeno — like other Hanson characterrobots — interacts with people with a fullrange of robotically orchestrated facialexpressions and conversation enabled byHanson’s Artificial Intelligence softwareand carried out by Hanson’s patentedrobot mechanics and materials

Thanks to a company calledSensory, Zeno recognizes and understands human speech, according

to Dr Hanson Zeno can respond withsubstantial verbal exchanges

Not short on words, Zeno’s lary is what scientists like Dr Hanson call

vocabu-“arbitrarily large” in size, meaning it ismore than big enough for the job athand “As people talk with the robot, therobot listens especially for words andphrases most pertinent to the recentconversation,” says Dr Hanson

Zeno analyzes human speech toget the intent of the conversation.While the robot makes errors inspeech recognition 10-20 percent ofthe time, it can still respond tohuman commentary, staying ontopic, according to Dr Hanson Thismirrors human exchanges, in whichpeople generally stay on topic butallow conversation to flow and take

Contact the author at geercom@alltel.net

by David Geer

Zeno — The First Complete

Character Robot

Human interaction is its main attraction

Photos are courtesy of Thomas Riccio unless otherwise noted.

Even from the start, Zeno’s head was more

than a hat rack Here is a shell of Zeno’s

head with notes and numbers; a part of

the design process.

A clay mock-up of Zeno from the design phase.

on new directions, Dr Hanson explains

Frubber — It’s All

in the Skin!

Hanson’s patented Frubber

syn-thetic materials technology makes it

possible to give Zeno and other robots

a wide range of life-like, emotive facial

responses Frubber enables Hanson

robots to form near-real human

expres-sions with what is the closest to real

skin that robotics materials have come

so far, while requiring a fraction of the

power needed by other solutions

Hanson’s “structured porosity

elas-tomer” (SPEM) uses a hierarchy of pores

that can pack into other pores down to

smaller and smaller scales — even to the

molecular level — resulting in very

skin-like foam The robot skin stretches skin-like

human skin and packs like skin to enable

more realistic human expressions

Frubber stretches to greater than

eight times its original size or

compress-es to fractions of its original form while

using less than 045 percent of the

power required for other robot face

expression solutions The power

reduc-tion makes it possible to use replaceable

batteries for the facial operations which,

in turn, helps make it possible to create

human-like bipedal robots that run

with-out being tethered to power supplies

Electronics

Hanson’s electrical and computing

technologies include the

patent-pending Character Engine AI

frame-work, which works in conjunction with

animation software

The AI framework and software

enable robotic movement including

human-like qualities that were not

previously transferable from computer

animation to robotic actuation The AI

and machine intelligence can run on a

single, standard PC, which can control

the robot wirelessly, explains Dr Hanson

“In our first prototype, we are

using Bluetooth wireless for the motion

control and the sensor data, and

ana-log composite RF for video

transmis-sion In the final product, we intend to

use Wi-Fi (802.11g),” says Dr Hanson

According to Dr Hanson, theCharacter Engine framework is a modu-lar and extensible software architecturefor developing extremely intelligentcharacters This generally means thesoftware is well designed so it’s easy toexpand on the programming that exists,

to code other things to go with it

The modular approach to codingthe software makes it easy to interfacethe Character Engine with other AI androbotics software like MS RoboticsStudio, Intel Open CV, and a variety ofcognitive architectures

The Character Engine provides agood framework (programming structure) for writing personalities intothe individual robot’s characters TheCharacter Engine enables code/

personality authors to quickly createnew personalities with software toolsjust for that purpose

The Character Engine uses Mayasoftware, as well as Massive software,which will be familiar to games and animations programmers, to

control the animated ior of the robot, so it canexpress its personality So,what games and feature filmanimation language writersand program developers havebeen doing is not simplybringing characters to life onscreen, but paving the way tobring them to life in 4D!

behav-The Massive Softwareprogramming serves as themotor cortex — the interac-tive animation brain forHanson’s robots “Massive is

the right tool for the job, deliveringextremely lifelike character behavior inresponse to sensory stimuli, a capacitywell demonstrated in numerous films,including the Academy Award winningcrowd effects in the “Lord of theRings” movies,” comments Dr Hanson.The Character Engine softwarealso combines state-of-the-art comput-

er vision, face tracking, and motiontracking with other sensory input, soZeno’s personality knows who andwhat it is interacting with and respond-ing to Zeno uses motion detection andcomputer vision to recognize facial

Zeno at WIRED’s Nextfest.

Photo courtesy of Kevin Carpenter.

Full-length Zeno His body is mobile, too Zeno complete with depth of character.

Just look into those eyes!

expressions, human gestures, and

common objects, explains Dr Hanson

Modular

Programming Proves

its Worth Again

The flexibility that modular

pro-gramming offers is imperative to

enable researchers and developers to

combine the complex elements of AI

that make it possible to simulate the

equally complex brains in human

beings, according to Dr Hanson

“We hope this focus on intelligent

characters will foster rapid progress in

sociable AI, helping to give rise to

friendly AI — the sort of AI that feels

compassion towards us, and can help

to prevent the nightmarish future ofunfriendly AI so common in sci-fi.”

“We have authors who constructthe personalities for our robots both asworks of literary art and state-of-the-artAI,” says Hanson These authors definethe robot’s motives in doing things,their intentions, and their will

Hanson’s Character Engine pursues theobjectives of the robot’s will while itinteracts with the world

Zeno is a social robot So, it willrespond positively to socially positivepeople and back away from rude people “If a person is kind to therobot, for example, Zeno will seek tospend more time with them If a person is rude, on the other hand,Zeno will be cold and distant, and seekother people to interact with or will

play alone, ignoring that person.”

In order to tell the kind peoplefrom those who are rude, Zeno uses aunique data structure to rememberpeople; a data structure Hanson callsPeople Objects The People Objectscontain photos of people, their names,memories of interactions with thosepeople, facts about them, topics of discussion and how the robot “felt”when interacting with each person

“This way, the robot builds emotionalmemories which guide his interactionsand motives,” says Hanson

Zeno’s Unveiling

Hanson Robotics unveiled Zeno tothe public in September As a result, an

Associated Press article covering the

fantastic new robot appeared in morethan 200 newspapers with circulations

in the millions

Zeno has since appeared on GoodMorning America, which has five million viewers Zeno has also been onCNN and NBC and in the pages of PC Magazine and any number of other

publications Tens of thousands sawZeno at WIRED’s Nextfest — the WIRED

magazine technology fair Zeno’s estimated exposure as of this writing ismore than 45 million Americans,according to Hanson

Zeno has his own storyboard withother characters of Hanson’s design sohe’s ready to star in his own sci-fi filmproduction Plans for Zeno includemass production at a retail price under

$300 per unit SV

GEERHEAD

Zeno is a 100 percent artificial

telligence robot running from a PC “The

PC sends motor control signals wirelessly

to Zeno, and Zeno sends sensor data

including streaming video wirelessly to

the PC,” says Dr David Hanson The PC

processes the video data comprising

he faces of people around Zeno and

the audio data from people talking

around him.

Zeno can do everything that past

Hanson robots can do and more.

Previous efforts include the Philip K Dick

android — a technological portrait of the

famous author — which won the

American Association for Artificial

Intelligence first place award in 2005.

“Our current Character Engine is

much more powerful than the Philip K.

Dick android software or the software running our Jules, Joey, and Einstein robots,” says Dr Hanson.

Hanson’s robots are fully autonomous, interactive robots that relate in a human-like manner They make eye contact, recognize people, and hold conversations.

“When the $250 Zeno is available in

2009, it will perform most of these functions even without a PC, using an embedded version of our Character Engine With the wireless connection to the PC, though, it will be even smarter, and with the extended AI on our server bank, across the network, little Zeno will

be one of the smartest robots in the world, almost frighteningly smart,” Dr.

Q. It is my understanding that

the HSR-8498HB servos that

are used in the RoboNova

humanoid robot have position

feed-back capabilities, so I bought a couple

of them from Tower Hobbies I have

been trying for several days now to

figure out how to get position data

from these servos From what I have

seen on the Internet, all I have to do is

send the servo a 50 μs pulse, and it will

return a position signal that is similar to

the regular pulse width to move the

servo I am missing something here,

can you help me?

— Pete Senganni

how position feedback for the

HSR-8498HB servo (see Figure 1)

can be measured by sending a 50

μs pulse and then measuring thepulse width of the signal that isreturned from the servo on the samesignal line Since this was on the samedata line, a pullup resistor was needed

on the signal line

At that time, I was planning

on talking about the HMI SerialCommunication Protocol and how

to interface it with a microcontroller

I had all sorts of problems trying toget this to work, which is probablywhy there is almost no information

to be found on the Internet on how

to do this So this month, I will talkabout the problems I had trying toget this to work, and to get all of the

HSR servos from Hitec (www.hitec rcd.com) to work with a computer and

be daisy-changed together; up to 128different servos can be controlled with

a single data line

Programming features includevelocity, start/stop, min/max positionlimits, center position, dead band, proportional and derivative gains, andforward/reverse direction control All

of these features are available throughthe Hitec Multi-protocol Interface(HMI) The HMI protocol allows the servos to be controlled with traditionalposition control using the 1 to 2 mspulse width, position feedback usingthe pulse width method, and via serialdata communications which will be

Tap into the sum of all human knowledge and get your questions answered here!

From software algorithms to material selection, Mr Roboto strives to meet you where you are — and what more would you expect from a complex service droid?

by Pete Miles

Our resident expert on all things robotic is merely an Email away

roboto@servomagazine.com

Figure 1 HSR-8498HB demonstrating

two reconfigurable housings.

Figure 2 HMI Servo Programmer user interface.

discussed here.

These four servos

currently cannot be

pro-grammed with the HFP-10

digital servo programmer

that is used to program the

other digital servos from

Hitec These four servos

can only be programmed

via a PC computer or

a microcontroller The

software for programming

these servos is available for

downloading at the Hitec

Robotics website (www.

robonova.com).

Go to the downloads

section, and the software

subsection Download and

install the program

“HMI_Servo_Programer-V1.0.2.zip.” The current

version of the software at

the time of this writing was

V1.0.2 The version

num-ber may be different by the

time you download the

software, so the program title might be

a little different Figure 2 shows an

image of the software user interface

To connect the PC to the servo, a

special cable is needed Hitec is

planning on selling this cable within the

next few months, but you can build

your own Pressing the “Help” menu

item on the HMI Servo Programmer

window will show you a schematic

drawing for the interface cable I have

redrawn this schematic in Figure 3 and

added some details Note pins 7 and 8

are wired together

Here is where my problems began

Before writing code to get a

microcon-troller to control the servo, I needed to

know if I could control the servo with

the programmer The reason for this is

so that I could monitor the serial data

back and forth between the servo and

the computer to help me understand

how to control the servo

With this cable, I was never able to

get the software to control the servo I

then tried using a RS-232 to TTL

converter that was on the Parallax

Professional Development Board

(www.parallax.com), and I still

couldn’t get it to work For a long time,

I thought it was the servo firmware

version on my HSR-8498 servos Theones I had were version 1.08, and theprogrammer manual stated that itrequired firmware version 1.10

After a while, it occurred to methat these servos were designed to

receive true RS-232 logic levels (i.e., alogic level 1 occurs with a zero or negative input voltage, and a logic level

0 occurs with a high input positive voltage) The RS-232 to TTL voltageconverter that is used on the Parallax

+4.8 - 7.4 V (SERVO POWER)

1

SIGNAL POWER GROUND SERVO CONNECTOR

DATA CARRIER DETECT (DCD)

2 RECEIVE DATA (RD)

3 TRANSMIT DATA (TD) 5

4 SIGNAL GROUND (SG) DATA TERMINAL READY (DTR)

RING INDICATOR (RI) 9

8 7

6 REQUEST TO SEND (RTS) CLEAR TO SEND (CTS) DATA SET READY (DSR) PIN RS-232 FUNCTIONS

Figure 3 Serial interface cable for the HMI Servo Programmer.

HSR-8498HB HSR-5498SG HSR-5980SG HSR-5990TG Operating Voltage 4.8-7.4V 6.0-7.4V 6.0-7.4V 6.0-7.4V

Idle Current Draw (6.0 & 7.4V) 8 & 8 mA 8 & 10 mA 8 & 10 mA 8 & 10 mANo-load Current

Draw, running (6.0 & 7.4V) 200 & 240 mA 200 & 240 mA 300 & 380 mA 300 & 380 mAStall Current Draw

(6.0 & 7.4V) 1.20 & 1.48 A 1.20 & 1.48 A 4.20 & 5.20 A 4.20 & 5.20 APulse Range

(~180°) 600-2350 μs 600-2400 μs 1100-1900 μs 1100-1900 μsCentering Pulse 1500 μs 1500 μs 1500 μs 1500 μs

Dimensions 1.57” x 78” x 1.45”(40 x 20 x 37 mm) 1.57” x 78” x 1.45”(40 x 20 x 37 mm) 1.57” x 78” x 1.45”(40 x 20 x 37 mm) 1.57” x 78” x 1.45”(40 x 20 x 37 mm)Weight 1.92 oz (54.7 g) 2.10 oz (59.8 g) 2.46 oz (70.0 g) 2.39 oz (68.0 g)

Table 1 Hitec robotic servo specifications.

Professional Development board uses

the MAX232E chip The MAX232E

actually inverts the signal, so that a

logic level 1 on the output will have a

+5V signal and logic level 0 will be

at 0V In essence, the servos were

receiving an inverted signal

When I inverted the signal with a

4011 NAND gate, the Servo

Programmer was now able to controland program the servo Figure 4 showsthe circuit that I was able to use tocommunicate with the servos using theHMI Servo Programmer

It took a long time to eventuallyfigure out why the original circuitshown in Figure 3 would not workwith my computer It turns out that it

was my USB to RS-232 adapter Thetechnical gurus at Hitec couldn’tunderstand why this wasn’t working

on my computer since they weresuccessfully using USB to RS-232adapters with their setups For somereason, my adapter would not work.With my adapter, the 3.3K resistorcaused the signal line to always be

DATA CARRIER DETECT (DCD)

5 4 SIGNAL GROUND (SG) DATA TERMINAL READY (DTR)

RING INDICATOR (RI) 9

8 7 6 REQUEST TO SEND (RTS) CLEAR TO SEND (CTS) DATA SET READY (DSR) PIN RS-232 FUNCTIONS

1 2 3 4 5 6 7

11

8 9 10 12 13 14

4011

+5 V

T1OUT R2OUT

R2IN T2OUT R1IN

T1IN

R1OUT T2IN

C2 V GND

SIGNAL GND 4.8-7.4V +4.8 - 7.4V SERVO POWER

HSR CLASS SERVO AND CABLE

SERVO

Figure 5 Wiring an HSR class servo to a microcontroller.

at zero volts But if I replaced the

3.3K resistor with a simple wire (i.e.,

pins 2 and 3 on the DB-9 connector

are connected/shorted together), I

would get the proper signal and the

servo responded to the HMI Servo

Programmer

This little trick happens to work

with my computer and my old

generation USB adapter I would not

recommend that anyone do this

with a regular RS-232 signal since the

purpose of the resistor is to cut the

regular ±15V signals from a proper

RS-232 line down to a safe level

that will not damage the servo (i.e.,

current limiting) If the circuit in Figure

3 does not work with your computer,

then I would recommend you use

a circuit similar to the one shown

in Figure 4

At this point, you can use the HMI

Servo Programmer to test and program

the servo The first thing you do is

select the COM port your serial cable is

connected to, and press the Connect

button If the program connects to the

servo, you will get a message at the

bottom of the user interface “Connect

OK! - HMI Servo Motor Ver- 110.” The

110 means the servo has the 1.10

firmware version

It turns out that the programmer

will connect to servos that have the

older versions of the firmware, just not

all of the features will work with the

older servos

Once connected, select the servo

from the menu list, then press the

“Read Servo” button This will read all

of the parameters that are currently

stored in the servo Now you can

change the various parameters on

the servo based on what is available

on the HMI Servo Programmer

software, such as setting a servo ID

Now, this program does not read

in the current position, voltage, or

current draw from the servo, and

cannot set specific user selected

proportional/derivative gains, or set

the speed of the software

In the ZIP file that the HMI Servo

Programmer came from is a Visual

Basic example program that will show

you how to change velocity and read

position, voltage, and current The rest

of this discussion will be based on

using a microcontroller to control theservos via serial communications

Figure 5 shows a simple schematicfor connecting one of the HSR class ofservos as identified in Table 1, to a highspeed microcontroller The serial I/Oline is bi-directional, and the 4.7K resistor is required for this to work

The serial communication protocol isshown below Note that the communi-cation protocol uses two stop bits

19200 BaudTwo stop bits

No ParityThe serial communication proto-col transmits seven bytes of data Fivebytes are transmitted to the servo,and the servo transmits two bytes ofdata back to the microcontroller

Figure 6 illustrates the data stream

The header is one byte and always0x80 (128 decimal) The checksum isthe 256 minus the sum of the first four bytes of data transmitted tothe servo

The key to getting a troller to work with these servos isthe timing between the CHECKSUMbyte transmitted to the servo and thefirst RETURN1 byte received from theservo The servo will start transmittingthe data almost immediately afterthe last stop bit from the CHECKSUMbyte Thus, the microcontrollerneeds to be able to change the datasignal line state from an output to

microcon-an input microcon-and start looking for thestart bit of the RETURN1 byte Ifthe microcontroller takes too long tomake this change, the servo will eithernot return the two bytes of data ortiming will be off and the data will

be corrupted

When it is time for the servo totransmit the data to the microcon-troller, it may not transmit any data ifthe microcontroller hasn’t changed thesignal line to an input This is why ahigh speed microcontroller is needed

My testing has indicated that thefive bytes of data sent to the servocan be successfully transmitted usingonly one stop bit instead of the specified two stop bits with no timingproblem And receiving the two databytes from the servo using one stopbit also works since the microcon-troller will wait until it gets the firststart bit for each byte of data Theadvantage of sending and receivingthe data to and from the servo usingone stop bit instead of two stop bitswill give the microcontroller a smallamount of time to change the state ofthe signal line from an output to aninput And it gives the microcontrollersome time to start the serial inputprocess to receive the data At 19200baud, each data bit takes 52 ms totransmit Thus, the microcontrollerhas only 52 ms to transition fromtransmitting data to receiving data.Also, since the data line is connected

to a pullup resistor, the troller must pull the data line lowbefore the first byte of data is trans-mitted to the servo, and immediatelyafter the CHECKSUM byte of data istransmitted to the servo This needs

microcon-to be done so the start bits can

be properly distinguished from the normally pulled high signal line

I was not able to get any of the

BASIC Stamps (www.parallax.com)

or the Basic Atom28 (www.basic micro.com) to bi-directionally commu-

nicate with the servos I was able toget the SX28 microcontroller to workwith the servos when I used a 20 MHzresonator The 52 ms time limit toswitch from serial output to serial inputmay be the limiting factor for thechoice of microcontroller used withthese servos

Table 2 shows a list of the differentcommands that works with the HSRservos with firmware version 1.10 Forthe most part, the commands 0x(#ID),0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA,0xEB, 0xEF, and 0xF6 are the ones used

COMMAND PARAM1 PARAM2 CHECKSUM RETURN1 RETURN2 HEADER

CHECKSUM = (256 - (0x80 + COMMAND + PARAM1 + PARAM2) MOD 256)

Figure 6 Serial communication data package.

to control the servos and obtain

feedback information

One of the first things that you

should do to test to see if you have

bi-directional serial communications

established with the servo is to read

the servo firmware and ID number

using command 0xE7 To do this, send

the following five bytes of data to the

servo: 0x80, 0xE7, 0x00, 0x00, 0x99

You should receive the following two

bytes of data back: 0x6E and 0x00 The

decimal equivalent of 0x6E is 110

which is firmware version 1.10 0x00 is

the default servo ID number Unless it

has been previously changed, it should

be zero

To command a servo to move to a

specific target position, the servo ID

number serves as the move command

The target position is in milliseconds,

much like the regular pulse width

commands Since the position range is

typically between 600 and 2,400 ms, it

is always greater than 256 ms, thus the

target position becomes a 16 bit word

This needs to be broken down into

two separate bytes of data to be

transmitted to the servo The following

two equations show how to break thetarget position down into the twobytes of data:

High_Byte = INT(Target_Position/256)Low_Byte = (Target_Position) MOD 256For example, to tell the servo tomove to the 1,500 ms neutral position,the following set of data is sent to theservo: 0x80, 0x00, 0x05, 0xDC, 0x9F

The 0x80 is the header, 0x00 is theservo ID number, 0x05 and 0xDC arethe high and low bytes of the 1,500 mstarget position, and the 0x9F is theChecksum Once this command is sent,the servo will immediately begin tomove to that position

To change the speed of the servo,the 0xE9 command is used The firstparameter sent to the servo is the servo

ID number, and the second parameter

is the speed The speed commandappears to be more like a percentagebetween 0 and 100% The servo’sspeed doesn’t get any faster for num-bers greater than 100 This commandwill also return the current position of

the servo Again, the position is brokendown into two bytes of data that can

be recombined into a single value withthe following formula:

Position = (High_Byte * 256) +

Low_ByteThe HSR servos can be daisy-chained together so that one serialdata communication line can be used

to control all of the servos connected

to that line If the servos are wired thisway, then the above two commands(setting the position, velocity, and position feedback) are the only commands that will address a specificservo, by its ID number The rest of thecommands address all of the servos atthe same time

Thus, if you want to address all ofthe features of the servos, they willneed to be individually wired to themicrocontroller This is true for servofirmware version 1.10 New firmwareversions may change this

Another command for setting theservo’s target position is 0xE6 If theservo’s motion has been stopped using

Command Param1 Param2 Return1 Return2 Description

ID (Servo ID number) Valid Range: 0x00 to 0x7F, Default: 0x00

Table 2 Servo command set.

the 0xEB command, the target position

can be changed, and when the servo

has been re-enabled it will move to the

new target position Command 0xF6

will also do this

Command 0xE5 will read the

current position of the servo in a

similar manner as command 0xE9 The

difference here is that any servo’s

position can be read with the 0xE5

command and the 0xE9 command

reads the position of a specific servo

ID number

Command 0xE8 will read the

current draw and the voltage draw of a

servo The value ranges from 0 to 255

When the servo is unloaded and not

moving, the current draw value is

typically 0 and the voltage value is 127

When I try to force the servo horn to

move by hand, the current draw value

increases and the voltage value

decreases I don’t have any information

about correlating these values to

actual current draw and voltage draw

At the very least, this information can

be used to help determine optimal

reaction load performance and to

determine if the servos are being

overloaded

Command 0xEA can be used to

change the proportional and derivative

gains for the servo to make their

motion “softer” or “stiffer.” I have not

personally experimented with these

numbers, but I have heard that

changing them could make the servos

respond faster to sudden direction

changes, and it could help reduce

servo jitter under certain loaded

conditions

An interesting command is 0xEB,

the servo Stop/Start command When

the servo is commanded to stop

(Param2 = 0), the servo will stop

moving where it is The rotor will be

locked like it was commanded to move

to this position, and will not let you

forcibly change it What is interesting is

that once the servo is stopped, its

current position can be read, the

velocity can be changed, the current

draw can be read, and other

parameters can be changed

When the servo is restarted

(Param2 = 1), the servo will move

from where it is currently located to

its new commanded position This

command can be easily

demonstrat-ed when the servo velocity is set tonear zero, and commanding theservo to move from one extreme toanother Shortly after the servo hasstarted to move, you command theservo to stop, then change the endposition or servo speed, then restartthe servo, and you will notice that itwill move to the new position at thenew velocity

The last motion command is 0xEF

This command will stop all motion control of the servo Unlike the 0xEBcommand which locks the servo position, the 0xEF command will allowthe servo to be moved by hand Servopositions can still be read after the execution of this command This feature enables the ability to manuallymove servos to specific positions, andthen records these locations so thatthey can be played back later This is

a feature that will simplify the programming efforts for animatronicand walking robot routines

In addition to the basic motioncommands described previously, theservos can be reprogrammed with amicrocontroller The key is to knowwhich memory locations contain whichparameter This is where the 0xE1,0xE2, 0xE3, and 0xE4 commands come

in handy These commands can beused to read and write to differentmemory and EEPROM locations

Unfortunately, the memory locations/addresses and their functions are not published But hasthis ever stopped anyone from gettingthe information they wanted? With

the help of a serial port monitor (I like

to use the Free Serial Port Monitor

program from monitor.com) and the HMI Servo

www.serial-port-Programmer, I was able to determineseveral of the EEPROM addresses forseveral different parameters Withthese two software packages, I canmonitor the data transmitted to andfrom the servo, determine whichmemory locations/addresses arechanged, and how they are changedwhen I change a parameter in the HMIServo Programmer software package.For example, I was able to determine that the servo ID is stored atthe EEPROM address 0x29 This can beread using the 0xE1 command, and anew value can be used to change the

ID number using the 0xE2 command

By making a set of systematic changes

in the HMI Servo Programmer, I wasable to determine that every singlechange to the EEPROM also changedthe EEPROM address of 0x2C

The value at 0x2C always seems tochange by the same numerical value as

I changed the other values This gives

an indication that this memory location

is some sort of a Checksum The best Ican tell from monitoring the data usingthe servo reset function (covered next)

is that the checksum is computed usingthe same method for individual com-mand packets Thus, the checksum isequal to 256 minus the mod 256 of thesum of all of the bytes written to EEPROM addresses from 0x00 to 0x2B (Param2 using the 0xE2 command or read by Return1 using the0xE1 command)

Header Command Param1 Param2 Checksum Return1 Return2

tmp = 0x0A - 0x00 = 0x0A (tmp = 10 - 0 = 10) checksum = 0xB3 - tmp = 0xB3 - 0x0A = 0xA3 (checksum = 179 - 10 = 169)

To make a change to an individual

parameter, read both of the values

stored at the checksum and the

parameter EEPROM locations

Calculate the difference between the

new parameter value minus the old

parameter value, then subtract this

value from the checksum value Write

these two new values to the EEPROM

Remember, if the checksum value

becomes negative, it needs to roll over

to 255 again

For example, if I wanted to change

the Servo ID from its default value of

0x00 to 0x0A (10 decimal), then I can

write a program to read the two

EEPROM locations as shown here using

the 0xE1 command (Note the values

shown are from the HSR-5980SG

servo) Then I make the overall

checksum change calculation, also

shown in Table 3

These two new values are then

programmed back to the same

EEPROM addresses using the 0xE2

command as shown Remember,

the checksum value calculated above

is the overall EEPROM checksumvalue, which is sent as Param2, and isnot the same checksum value that iscomputed for each command packetsent to the servo as described earlier(see Table 4)

In order for these parameterchanges to take effect, the servo must

be powered down and restarted Ifthere was an error in calculating thechecksum, the servo will still work butthe motor will be disabled until theproblem is corrected

If for some reason the servo stopsfunctioning properly due to program-ming problems, the HMI ServoProgrammer software has a servo resetfeature that will reset the servo back tothe factory defaults All you need to do

is start the program, select the servotype, and select “Servo Reset” underthe File menu item Then press the Yesbutton, and the servo will be reset back

to its original condition

At this point, you should have allthe information needed to controlthese servos using serial communica-

tions By monitoring the data beingtransmitted between the servo and thecomputer (using a program like theFree-Serial-Port-Monitor), the specificparameters and their functions can bedetermined This information can then

be modified to change your servo performance characteristics

At some point, someone will takethe time and, using the steps present-

ed here, compile a complete list of all

of the parameters and their functionsfor the different HSR servos and publish them to make our lives easier

At 19200 baud, it will take about 4

ms of time to send the seven byte datapackage This takes about twice aslong to command a servo than usingthe traditional pulse width method Butthe command only needs to be sentonce, the 20 ms servo update is nolonger needed, velocity is now a control parameter, and with the feedback information from the servos,true closed loop positional and velocitycontrol of these servos can easily

3DLQH6WUHHW_3RZD\_&DOLIRUQLD___ZZZKLWHFUFGFRP

>_j[Y Ej^[h8hWdZ

.EW

+656*

6SHHGVHF 7RUTXHR]LQ 6WHHO*HDUV

+656*

6SHHGVHF 7RUTXHR]LQ 6WHHO*HDUV

+657*

6SHHGVHF 7RUTXHR]LQ 7LWDQLXP

$//63(&,),&$7,216$792/76

Know of any robot competitions I’ve missed? Is your

local school or robot group planning a contest? Send an

email to steve@ncc.com and tell me about it Be sure to

include the date and location of your contest If you have a

website with contest info, send along the URL as well, so we

can tell everyone else about it

For last-minute updates and changes, you can always

find the most recent version of the Robot Competition FAQ

at Robots.net: http://robots.net/rcfaq.html

— R Steven Rainwater

8-10 Singapore Robotic Games

Singapore Science Center, Republic of Singapore

A plethora of events including Micromouse,autonomous Sumo, RC Sumo, robot soccer, wallclimbing, pole balancing, underwater robots,legged robot marathon, legged robot obstaclerace, and a robot colony contest

http://guppy.mpe.nus.edu/sg/srg25-27 TechFest

Indian Institute of Technology, Bombay, India

Lots of events for autonomous and remote controlled robots including standard Micromouseand several events unique to TechFest: Pixel, acontest for vision-equipped bipeds; Full Throttle:

Grand Prix, remote-controlled, internal combustionpowered cars race on a concrete track; Vertigo, aremote-controlled robot and an autonomousrobot must work together to move blocks around;

Prison Break, remote-controlled robot mustclimb out of a pit and survive a fall to escaperobot-jail; U-571, an obstacle avoidance contestfor underwater robots

http://techfest.org/competitions/department

24-28 APEC Micromouse Contest

Austin Convention Center, Austin, TX

Amazingly fast little autonomous robot crittersrace to solve a maze in this competition If

you’ve never seen one of these events, go seethis one You won’t believe how fast thesethings are

www.apec-conf.org

15-16 Manitoba Robot Games

Tec Voc High School, Winnipeg, Manitoba, Canada

Included in this competition are a mix of eventsfor autonomous and remote-controlled robotsincluding Japanese style mini-Sumo, Western styleSumo, a robot Mini-Tractor Pull, Super Scramble,line following, and the Robo-Critters contestfor kids

www.scmb.mb.ca

Send updates, new listings, corrections, complaints, and suggestions to: steve@ncc.com or FAX 972-404-0269

EZPZ23-HR1 Piezo Motor Servo

The fully intelligent controllers,

measure just 2.25 “ x 2.25“ The Nanomotion HR1 motor,

nine pin DSUB connector plugs directly into the controller

A single four wire bus, containing two power wires and

two communications wires, links up to 16 such piezo

motor controllers in a daisy chain Commands can be

issued from any serial terminal program (such as

HyperTerminal) or from the EZ Servo/Stepper Windows

application

The commands are intuitive and simple For example,

the command A10000 will move the servo motor to the

absolute position 10000 (This communications protocol

is compatible with devices that use the Cavro DT or

OEM protocol.)

The EZ Servo is also capable of stand-alone operation

with no connection to a PC It can be set to execute a

preset string of commands upon power-up (i.e., only

power is required in this mode) The commands include

nested loops and execution halt pending a switch closure,

which is useful in stand-alone applications

Using AllMotion Starter Kits, a first-time user is able to

make the HR1 piezo motor servo in less than half an hour

• RS-232, RS-485, or USB based communications

• Four quadrant operation

• On-board EEPROM for user program storage

• Stand-alone operation with no connection to a PC

• Execution Halt/Branch pending switch closure

• Pre-wired for OptoSwitch inputs

• Position velocity and torque modes

• Industry standard communications protocol

• Homes to an opto or encoder index with a single command

• Fully programmable ramps and speeds

• Quadrature encoder based feedback for position mode

• Velocity mode possible with only Hall sensors forfeedback

• Switch selectable device address

• Software selectable max currents

• 4 MHz max encoder frequency

• Two one amp on/off drivers included

• Optional step and direction mode; 4 MHz step frequency

• Four ADC inputs; Halt/Branch on ADC value

For further information, please contact:

TReX Jr Dual-Motor Controller

Pololu introducesits new TReX JrDual-Motor Controller, aversatile DC motor controllersuited for mixed autonomous andradio control of small- and medium-sized robots Three independent interfaces are offered:radio control (RC) servo pulse interface, analog voltage,and asynchronous serial The serial interface can switchinstantly with one of the other two interfaces, allowingmixed autonomous and remote control For example, arobot could be configured to run autonomously most ofthe time, but a human operator could override theautonomous function if the robot gets stuck or into a dangerous situation If the serial mode is selected as theprimary interface, high-resolution measurements of all five

RC inputs are made available to the autonomous robotcontroller, allowing for complex and unlimited mixing ofoperator control and sensor input

The TReX Jr motor controller operates from 5 to 24 V,and the two primary outputs provide bi-directional controlwith peak currents of 5 A and continuous currents of 2.5

A while a unidirectional auxiliary output delivers over 10 A(continuous) A fourth control channel for invertible robotsallows improved control if the robot does get turnedupside-down, and the fifth control channel determineswhich interface controls the motors The unit measuresapproximately 1.75” x 1.75” x 0.5”, and it is available fullyassembled or in kit form

New Products

MOTOR CONTROLLER

510•471•4000 Fax: 510•471•4003 Email: info@allmotion.com Website: www.allmotion.com

AllMotion

For further information, please

contact:

Advanced Features

for Robot Remote

Control

InRob Ltd a leader in advanced

wireless control systems for

unmanned ground vehicles (UGV),

has announced advanced and

innovative features for its remote

control units for military robots

and UGVs

These features include:

• Secure, clear, and uninterrupted

wireless transmissions in most

operating environments

• Intuitive touch screen for more

efficient operations in the field

• Graphic screen display showing

critical data of the robot, including

fuel level, engine RPM, status of

tools in action, such as cameras, etc

• Networking technology with

option for a single operator to

simultaneously control and

coordi-nate multiple robots

• Advanced safety features for

weapon platforms This feature

is particularly important for

law enforcement and Homeland

Security use, where the robots may

be used near crowds

For further information, please

contact:

REMOTE CONTROL

6000 S Eastern Ave.

Suite 12-D Las Vegas, NV 89119 Tel: 877•7•POLOLU

or 702•262•6648 Fax: 702•262•6894 Email:

www@pololu.com Website:

PP:KLWH ZDWHUFOHDU

/('9

PFG $% 



7KH9HOOHPDQ3HUVRQDO6FRSH

LVQRWDJUDSKLFDOPXOWLPHWHU EXWDFRPSOHWHSRUWDEOH RVFLOORVFRSHDWWKHVL]HDQG FRVWRIDJRRGPXOWLPHWHU +36 

7KHURERWIURJPRYHVIRUZDUG ZKHQLWGHWHFWVVRXQGDQG UHSHDWV VWDUWPRYHIRUZDUG ! VWRS!OHIWWXUQ!VWRS!ULJKW WXUQ!VWRS 65 

6ROGHULQJ

UHTXLUHG

7KRXVDQGV PRUHLWHPVRQ RXUZHEVLWH

$146.

Ask for our FREE 96 page catalog

VISIT OUR ONLINE STORE AT

www.allelectronics.com

WALL TRANSFORMERS, ALARMS, FUSES, CABLE TIES, RELAYS, OPTO ELECTRONICS, KNOBS, VIDEO ACCESSORIES, SIRENS, SOLDER ACCESSORIES, MOTORS, DIODES, HEAT SINKS, CAPACITORS, CHOKES, TOOLS, FASTENERS, TERMINAL STRIPS, CRIMP CONNECTORS, L.E.D.S., DISPLAYS, FANS, BREAD- BOARDS, RESISTORS, SOLAR CELLS, BUZZERS, BATTERIES, MAGNETS, CAMERAS, DC-DC CONVERTERS, HEADPHONES, LAMPS, PANEL METERS, SWITCHES, SPEAKERS, PELTIER DEVICES, and much more

Featured This Month:

Technical Knowledge

by James Baker

Feature

for a Sponsor? by Chad New

Product Review

Power V2 E-Logger and

Power Panel LCD Display

by Don Hebert

ROBOT PROFILE – Top

Ranked Robot This Month:

29 G.I.R. by Kevin Berry

Events

Many people think that theopposite of a robot with akinetic energy spinner is anarmored push-bot I disagree The

KE spinner puts energy into arotating mass as it spins up over aperiod of time Once the energy isthere, delivering it is a very crudeaffair Our friend Isaac Newtontells us that when this spinningweapon hits the opponent, theopponent hits back with equaland opposite force The classic

“immovable object” push-botmeeting our “unstoppable force”

spinner, thus delivers equal energy

to both machines

Once impact occurs, it isout of the driver’s control as toexactly where on the opponentthe energy goes, and how muchenergy is used up In my opinion,this makes spinners and armoredbots the same Neither can besure of where impact occurs,neither decides how much energy

is imparted, and both suffer thesame violent, chaotic energytransfer

An opposite to this would be

a robot that can choose where onthe opponent it attacks, and cancontrol the amount of energy ittransfers to that chosen area over a period of time selected bythe driver This perfectly describesthe crusher

TECHNICAL

KN WLEDGE

● by James Baker

Crushers

usable range of motion

within the weight limit; 100

tons over one inch is easily

possible, but that is of no

use in combat In all weight

classes, it is the robot

chassis that sets the limits

for crushers

Return of the

Crusher

The UK has a long love affair

with crushers We have had many

heavyweight machines successfully

use crushing weapons The

feather-weight class however, rarely saw

the type, as effective crushing

weapons are heavy Circumstances

in the UK at the moment are such

that high kinetic energy spinners

have very few opportunities to

fight Combined with the extra 1.6

kg now allowed in the featherweight

class, it has opened the floodgates

for robots that could never afford

weight for crushing weapons to

make their entrance There are

several crushers being built right

now, and Dragons Claw is the first

of many

Dragons Claw

Tony Booth, a good friend of

mine, decided to build a high power

crusher after winning an eBay

auction for a Linak LA34 linear

actuator This particular item can

deliver a reliable 10,000 newtons

of force at 24 volts, so it’s perfect

for featherweight combat robots

Tony did the right thing

when designing an unfamiliar type

of robot He asked

everyone he could for

advice, listened to the

advice, and used that

advice when designing

the robot It’s amazing

how many people

refuse to do this

Based on a welded

aluminium chassis,

Tony never

underesti-mated the forces

involved in crushing your opponent

He conducted many tests with atemporary chassis and claw,before deciding on the currentconfiguration

Supplying the actuator withthe (quite modest) current itneeds are two packs of 12V NiMHbatteries (in series giving 24V),through a simple servo-operatedDPDT switch and failsafe Thesebatteries also supply the RobotPower Scorpion XL speed controllerwhich, in turn, runs the 18V drillmotor based drive system

This mix of components isproving to be very effective Therobot is fast, turns quickly, and(more importantly) rotates aroundits weapon due to having the driveright at the front The electron-ics are small, light, and overspecification for their currentuse Dragons Claw has beentested and proven itself to be avery effective combat robot Ithas successfully pierced 10

mm plastics, 2 mm titanium,and 6 mm aluminium, to name but a few examples

In fun events, it has provenitself capable of catching its

opponent and bringing the weapondown to pin or crush (Catching theopponent was initially a concern asthe actuator has a speed of just 5

mm per second.)Dragons Claw is able to controlits opponent once it captures them

A light squeeze holds the opponent

in place so it can be taken to arenahazards or for other robots to attack.Increasing the power to crush theopponent puts a full 1,000 kg intothe Hardox claw, and onto the targetrobot through an interchangeabletitanium tip In testing this process,Tony bent the 10 mm steel pivotpins, so they are now 0.5 inchtitanium

So, the robot is ready then? Not quite

CAD of Dragons Claw.

CAD of claw mechanism.

Dragons Claw’s frame.

Prototype claw.

Making weight equals drilling holes.

So you think that you are

ready to get some sponsors,

do you? Have you typed out

your letter and arranged some

nice pictures thinking that you

are ready to rake in the sponsors?

Think again; getting a sponsor,

keeping a sponsor, and mostimportantly making that sponsorhappy with you is much moreinvolved than an email and a fewnice shots!

In this article, I will giveyou some basic guidelines which

may help increase your chances forsuccessfully getting yourself asponsor This will hopefully helpoffset the cost of this sport andalso give you equipment that may

be able to take your robots to thenext level

The cheapest option is to drill holes, which is what Tony did A fullkilogram (2.2 lb.) was lost by takingaluminium away, but you have to becareful not to weaken the chassis somuch that it cannot handle those bignumbers when you hit the “squashhim” button

Thankfully at the same time Tony was drilling holes, the UK robot

combat governing body (theFighting Robot Association)was changing the rules Theextra 1.6 kg now allowed,combined with the manyholes drilled, meant DragonsClaw was now underweight,and ready for its titaniumarmor This was bolted inplace to nylon pillars, allowingflex and deformation A beautiful paint job later andDragons Claw was ready to enter itsfirst competition I have foughtagainst this robot with my fun-bots, and as you can see in thephotos, Dragons Claw is a seriouslycompetitive machine

Robots are never finished Theywill evolve and change to suit theenvironment and their predators orprey This assisted evolution meansthat Dragons Claw may well havechanged again by the time you readthis article, so you can get updatesand further information about it on

Tony Booth’s website at www.team dragon.org.uk SV

SO YOU THINK YOU’RE READY FOR A SP NSOR?

● by Chad New

Weapon control.

Lifting works, too.

Paint matters!

Leverage also matters.

Finished Dragons Claw Crushing the opponent.

The When

As a newbie robot builder, you

have a very low chance at landing

himself a sponsor You are unproven,

your name is not known, and chances

are that your robots are not yet up to

par with your competition It is not an

impossible task, however, your

chances are much greater once you

gain some experience The

appropri-ate time to seek sponsorship is when

you can prove that you know what

you are doing Can you build a robot?

Talk about it; tell a person what all the

parts are, what they do, and how

they are going to work better next

time Have you been to several events

and come home with a robot that you

can still drive, or better yet, a win?

If you can answer “yes” to these

questions, then you might be able to

realistically consider getting yourself a

sponsor You have built up some

cred-ibility that you must now project to

the companies which you would like

to attempt to gain sponsorship from

The Why

Why would anybody sponsor

you? Companies will sponsor who is

willing to work for them They are

not going to shower you with parts

and money for free, no matter how

good your emails are Companies

need testers, feedback, good

com-munication, pictures, and marketing

tools What they need is a person

committed to help sell their product

When a new product comes out,

they need people that know whatthey are doing to test it and givethem feedback, telling them how itcan be made better and the weakpoints of the item When setting up

a booth, they need robots to displayand pictures for their website

Combat robots are great for ing! If Product X can take the abusefrom a combat robot, then inside anR/C car it should last forever! If youmake it clear that you are willing towork for your sponsorship and arenot just looking to be handed some-thing, then you have a chance

market-The How

It all starts with one email or aphone call First impressions are themost important, so you better make it

a good one If you are going to call,have a list of facts ready to talk about;

list the points that you want to makeand most importantly how you canhelp them If you choose to send anemail, make it short and to the point

You do not want the reader to loseinterest and shoot your mail to thetrash bin Again, list the points thatyou really want to emphasize andwhat you have accomplished Youmay also include some pictures ofyour robots However, do not attach avideo because that may take a longtime to download Also, don’t forget

to list what you can do for them!

Ready?

Take all of these tips and

combine them into one great email

or phone call If you are able to scoresome free parts, make sure you putthem to good use! Do not take themand run Give your sponsors updates,feedback, and check to see if theyneed your help with anything With alittle effort and some luck, you toocan have parts for the cost of a littlecomputer work! SV

Iwas always very thorough about

monitoring current draw and amp

hours in my robots Current and

voltage tests can determine the

compatibility of a motor/batterycombination Studying the results led

me to my favorite voltage/amphour/motor combination

At first, an AstroFlight wattmeter would go for a ride, secured

on top of the robot Reading the display while the robot was moving

Quick Tips

1) DO use spell check! You don’twant to sound incompetent

2) Don’t say: “Hi, I would like to build

a robot Please give me parts.”

3) DO follow-up on your emails andcalls Just a simple “hey, just checking

to see if you got my message” works fine

4) Don’t send long random emailsthat have no point

5) DO be polite, even if the answer is

no Next time they need a tester, theymight email you!

6) Don’t be a pest! Do not send people dozens of emails That is anassured no!

7) DO be patient Just because youdon’t get an immediate responsedoesn’t mean the answer is no

8) Don’t give up Work harder at it! Build better robots, take better pictures, and refine your sponsor-ship pitch

PR DUCT REVIEW

Eagle Tree Systems Micro Power V2 E-Logger and

Power Panel LCD Display

● by Don Hebert

was difficult It would move too fast,

so most of my measurements were

done with the wheels up or against a

wall, wheels spinning Wheels-up

testing detected gearbox problems

effectively Bad mechanical

conditions showed higher that

normal amps

This worked okay for most of myneeds But as my lifter got more

powerful, I found the need for faster

measurement Some events, like the

1/2 second of the lift, were not

visible in the display of the watt

meter A data collection device with

sufficient samples per second was

needed The Micro data logger from

Eagle Tree Systems fit the bill

The Micro Power E-Logger from

Eagle Tree Systems (www.eagle

treesystems.com) can measure

±100 amps and 4.5–45 volts It

calculates amp hours and watts All

this is recorded in memory forlater retrieval Sampling rate is

10 per second Data retrieval isvia a supplied USB cable The

PC software is also provided(see Figures 1 and 2)

Real-time amps, volts, watts, andamp hours can be displayed with theoptional Power Panel LCD Think of it

as a dashboard for the robot (seeFigure 3)

The entire combination is halfthe size of the Astro watt meter Figure 4 shows installation in

a 12 lb robot

Preliminary setup prior to firstuse is required The data logger hasmany add-on devices What isinstalled — like the LCD display —must be declared for them to work

For my tests, the LCD was the onlyextra device The current and voltagedetectors on the data logger arealways available

The instructions for what data tocollect is also required I displayedvolts, current, amp hours, andwattage Volts and amps are directfrom the on-board sensors Wattage

is calculated from volts and amps

Amp hours are calculatedfrom amps and time

With the LCD panel,you may conduct realtime monitoring such as awheels-up test You donot have to retrieve thecollected data If thebuffer gets full data, collection simply stops

The real-time display ofdisplayed volts, current,amp hours, and wattagewill continue (Just this is

pretty cool.) The unit is small and willfit anywhere I usually use maskingtape to hold everything in place

A new recording session is

start-ed each time the logger is turnstart-ed on.Plug it in to the battery circuit afterthe power switch I find it helpful toscript the sequence of events ahead

of time Once the data is extracted,current draw will be your only indica-tion to tie the section of the trace tothe test event A wheels-up test atfull power may be indistinguishablefrom full speed reversals on theground or from wheel spin againstthe wall So noting which test wasdone first allows you to link the firstpeak in current draw with that test.After the tests are done, turn offthe power Remove the data loggerfrom the robot and bring it to your computer Start the software.Connect to your PC with the providedUSB cable The connection will power-

up the data logger and connectionwill be established We now have toexplicitly request data from the device.Until the memory is full, you mayplug the logger again and again tocollect more data Each power-on iseasily distinguished as a new session

in the graph For example, I couldsave five separate test sessions in thedata logger before I had to clear dataand start over I saved the data to thesame file name and each time I didanother test, all of the data from theprevious test was still in there Noproblem! The saved file just got larg-

er each time I saved to the hard drive(see the sample graph in Figure 5)

I encountered one problem afterclearing data memory and running afew more data collections I retrieved

FIGURE 1 Micro Power Data Logger.

FIGURE 3 Power panel.

FIGURE 4 Installation

in a robot This is small!

FIGURE 2 Main screen.

the data and saved it to the

previous-ly named file I expected a

concate-nation of the data already saved and

the new data just retrieved But the

new data completely overwrote all of

the previous data; the original data

was gone This was a mistake on my

part, I now know better But there

are no warnings that you are about

to overwrite data

There is a live mode that uses

the main screen of the PC as a

real-time display There are also

additional sensor modules to monitor

RPM and temperature

In conclusion, I heartily endorse

this product The MicroPower V2

E-Logger retails for $69.95, and the

Power Panel LCD Display for $39.95

It records data accurately and meets

my needs Even if you do not use the

data collection, it is still a handy little display SV

ROBOT PR FILE

● by Kevin Berry

FIGURE 5 Graph screen.

Contact the Author

Don can be reached via email at

dhebert562@charter.net

TOP RANKED ROBOT THIS MONTH

Historical Ranking: #1 Weight Class: 6 lb "Mantis"

Team: Chaos Robotics Builder: Dirk Stonehouse Location: Saskatoon, Saskatchewan, Canada

BotRank Data Total Fights Wins Losses

G.I.R has competed in WBX II,WBX III, WBX IV, Kilobots IX,

Kilobots X, and Kilobots XI More

details are listed below:

• Frame: 1-1/2” x 3/4” aluminum

channel, milled to reduce weight

and welded

• Base plate: 1/16” Garolight

• Drive train: Four B62 gearhead

motors

• Voltage: 14.8

• Wheels: 2 inch 1/10 scale pan car

foam tires

• Configuration: Invertible box with

thresher and hinged titanium wedge

• Drive ESC: Two Traxxas EVX

• Drive batteries: Two 2.1 Ah, 7.4V Thunder Power lithium packs

• Weapon: 1-3/4” inch drum thresherwith six two-tooth blades driven atthe center with a dual “O” ring beltconfiguration

• Weapon battery: One 2.1 Ah, 7.4V Thunder Power lithium pack

• Weapon motor: 27 turn stock 540

• Weapon ESC: Tekin Titan

• Armor: Titanium in the front

• Future plans: Upgrade from the 540 weapon motor to a Hacker A30-10XL

Design Philosophy

“I designed G.I.R to be a powerful pusher with the ability to withstand shell and bar spinners Also, to inflict some damage with the drum and be invertible So far, it has proven to be a fairly good design

to be ranked #1.” SV

Photo and information are courtesy of Dirk Stonehouse All fight statistics are

courtesy of BotRank (www.botrank.com)

as of November 13, 2007 Event attendance data is courtesy of The Builder’s

Database (www.buildersdb.com) as of

November 13, 2007.

Results from Oct 14

through Nov 11, 2007

2007 Halloween Robot Terror was

presented by California Insect

Bots in Gilroy, CA on 10/27/2007

Go to www.calbugs.com for more

information Results are as follows:

● 150 Gram Flea Weights — 1st:

“Crisp,” Team Misfit, driven by Orion

Beach; 2nd: “Atom Bomb,” Team

Misfit, driven by Daniel Chatterton

● 1 Pound Ant Weights — 1st:

“Baron Underbheit,” Team Misfit,

driven by Kevin French; 2nd:

“Pooky,” Team ICE, driven by David

Liaw; 3rd: “Mash Potato,” Team ICE,

driven by Pui Shan Wong

● 3 Pound Beetle Weights — 1st:

“Unknown Avenger,” Team ICE, driven by David Liaw; 2nd: “Impact,”

Team Target Practice, driven by DonHebert; 3rd: “Queen of Kings,” TeamMisfit, driven by Shannon Muha

Franklin Institute Robot Conflictwas presented by North EastRobotics Club, Inc., in Philadelphia,

PA on 10/20/2007 Go to www.

nerc.us for more information

Results are as follows:

Bob,” Robotic Hobbies; 2nd:

“TriPolar,” Team Brain Damage; 3rd:

“The Interloper,” Team Interloper

● Sportsmans (30 lb) — 1st: “Bounty

Hunter,” Team Hammertime; 2nd:

“Upheaval,” Team Mad Scientist;3rd: “Mangi,” Team HFA

HORD Fall 2007 was presented bythe Ohio

Robot Club inBrecksville, OH

on 11/3/2007

Results are asfollows:

● Flea Weights — 1st: “Little Buzz,”

Richard Kelley, Team Kelley; 2nd:

“0–2,” Even Gandola, Team Probotics

● Ants — 1st: “Dusty, the Evil Dust

EVENTS

Results and Upcoming Events

Pan,” Jeff Gier, Team Mechanical

Advantage; 2nd: “Heman I,” Evan

Gandola, Team Probotics; 3rd:

“Rumble,” Heman I, Evan Gandola,

Team Probotics

● Beetles — 1st: “D2,” David Timothy,

Team D2; 2nd: “The Box,” Richard

Kelley, Team Kelley; 3rd: “Rumble,”

Sweaver, Greg Shay, Team Fishneck

Robots Live presented a show in

London for the MCM Expo,

10/20-21/2007 Go to www.robots

live.co.uk for more information.

presented by Chicago Robotic

Combat Association in Rosemont, IL

on 10/19-21/2007 Go to www.the

crca.org for more information.

Robowars Metro was presented

by RoboWars in Oakleigh, South

Victoria, Australia on 11/06/07 Go

to http://robowars.org for more

information

Upcoming Events for

January and February

2008

2008 will be presented by

Western AlliedRobotics inSeattle, WA on2/9/2008 Go to

www.western alliedrobotics.

com for more information.

The event will be held at theMonroe County Fairgrounds (near

expo.com/Directions.html) The

event will be held from 6:00pm and safety inspection from9:00am-11:00am If a lot of robotsregister, they may start safety andfights earlier Class: 1, 3, 12, and 30*

11:30am-pound robots (30 lb weapons not allowed) Format:

spinning-Double Elimination or Round Robin(RFL Rules), no ICE or open flames

Entry Fee: $40 for first 30 lb or 12 lbrobot $25 for first 3 lb or 1 lb robot

Additional robots are half price

Special entry fee considerations forbuilders who are under 18 Arena:

12 x 12 with 18” x 18” pit in one corner surrounded by a 2” high wall making it very difficult to accidentally drive into it

Motorama 2008 will be

present-ed by North East RoboticsClub, Inc., in Harrisburg, PA, from

2/15-17/2008 Go to www.nerc.us

for more information

This will be a 150 g to 30 lb combat event Fairies and Ants fight

in 8’ box on Friday; Beetles throughFeatherweights fight in 16 x 16 box

on Saturday and Sunday All

complet-ed forms and entry fees must bereceived by 1/2/07 This is going to

be another awesome event at theFarm Show Complex!

Robots Live will present an event

at the Hermitage Leisure Centre,Whitwick, Leicestershire on 2/2-

3/2008 Go to www.robotslive co.uk for more information.

For the first event of the year,Robots Live heads back to their hometown of Whitwick in Leicestershire.Not only is this their hometown but also their birthday! Come andcelebrate two years of Robots Live!

Robowars has tentatively uled their national title event inOakleigh, South Victoria, Australia

sched-on 1/12-13/2008 Go to http://rob owars.org for more information SV

Tormach PCNC 1100 Features:

Q Table size 34" x 9.5"

Q R8 Spindle 1.5 hp variable speed to 4500 RPM Q Computer controlled spindle speed and direction Q Precision ground ballscrews

QDigitizing and tool sensing support

Q4th axis and high speed spindle options

3 Axis Mill

$6800

plus shipping

When you’re serious about hardware, you need serious tools

Whether milling 0.020” traces on prototype PCBs or cutting ½”

steel battle armor, this CNC mill can do it all Weighing in at more than 1100 lbs, the PCNC can deliver the hardware end of your combined hardware & software projects.

Precision CNC Machining

Mill includes Control, CAD and CAM software Optional stand, coolant system, computer and accessories are extra

Product information and online ordering at www.tormach.com

Because I get to put together all

kinds of neat stuff And, when

I’m done playing with my new

garage-brewed toys, I get to show

them to you and tell you all about

them So, get your soldering irons

hot and stuff your face into that

illumi-nated desktop magnifier We are going

to build an intelligent stepper motor

controller from scratch that is based on

the Allegro MicroSystems A3979

So What??

YIPEE Another stepper controller

NOT! This stepper controller is compact

and when properly heatsinked, handles

two ampere motor loads The Allegro

A3979 is an upgraded version of the

venerable A3977 All of the easy to use

motor control knobs and dials youwere used to on the A3977 are alsopresent on the A3979 DMOS driver

We can even use the same externalcomponents in an A3979 design that

we would use in an A3977 design Thebig difference in the two DMOS drivers

is the addition of 16th stepping capability to the A3979 If you thinkyou’re about to read about a rehashedversion of the A3977 circuits you canget from various Internet points, sitdown, shut up, and read on

Getting Small

If you have ever worked with theA3977, you probably used the 44-pinPLCC version of the part The PLCC is bigand relatively easy to put down on a

printed circuit board (PCB) You can’tget the A3979 in a 44-pin PLCC package; it only comes in a 28-pinTSSOP package Instead of dissipatingheat though pins as the PLCC-packagedA3977 does, the A3979 has an exposedheatsink pad on its belly A woodpeckerwould see that pad on the underside ofthe A3979 in the perspective of Photo 1.The A3979 can be crammed into

a very tiny space if you’re not driving stepper motors the size ofwatermelons The only design pointthat increases the size of the A3979

by Fred Eady

The Super

STEPPER DRIVER

PHOTO 1 Don’t think that you can’t access this

heatsink pad without special soldering equipment.

A few holes in the printed circuit board under the

belly of the A3979 provides access.

PHOTO 2 This was a challenging build as I had to be extra careful not to short the SMT 0805 components on the ground plane extensions under some of the SMT parts A fine tipped soldering iron and an illuminated magnifier are must-have tools for this project.

PCB is heatsinking Thus, as you can see in Photo 2,

I’ve gone all out on the heatsinking of the A3979 by

laying out a ground/heatsink plane onto the entire

top side of the A3979 stepper driver PCB

To keep things simple, I also spread out all of the

A3979’s supporting components in my design If you

trash the serial port and squeeze the PIC18F2620 in

towards the A3979, you can push this stepper

driver/controller into a very small form factor

Let’s Get Started

The very first thing we must do is design and lay

out a suitable PCB to support the A3979 and its

subor-dinate components I want you to be able to build this

stepper controller in your shop I used ExpressPCB and

their tool suite for the PCB fabrication To make

parts-gathering easier for you, I’ve sourced all of the A3979

motor driver board’s components from Mouser and

Digi-Key The A3979 motor driver board ExpressPCB layout

sans ground plane can be seen in Screenshot 1

In that the A3979 has a built-in translator, the computing

support requirements are low Let’s stop here and define

“translator.” Without a built-in translator, we would have to

write every little bit of code required to interpret command

inputs and convert them to motor movements That includes

writing code to correctly phase and drive the A3979’s

H-bridges We would also find ourselves constructing phase

tables for use in the microstepping process The phase tableswould be used to modulate a PWM signal that is fed into aDAC (Digital-to-Analog Converter) The DAC would generatevoltage levels relative to the modulated PWM signal (Trust

me We want to use a translator if at all possible.) Thus, atranslator is — in this context — a combined package of special-ized hardware and firmware that resides in the A3979 silicon.The purpose of the translator is to convert incoming motorcommands into motor actions or A3979 control actions A

REF PFD

MS2 MS1

SLEEP STEP RESET

RXIN

TXOUT RX

RXIN

TXOUT TX

VBB

+5VDC

+5VDC +5VDC

.1uF

C14 1uF

C10 22uF

R5 2

C17 1uF

C8 1uF

11 13

OUT1A

PFD REF

OUT2A MS1

C15

.1uF

R1 50K

R9 10K

U3

ST3232

13 8 10 1

3 4

5

2 6

12 9 14 7 15

16 R1IN T2IN C1+

C2+

C1- V+ V-

C2-R1OUT T1OUT VSS

VDD

C18 1uF R2

50K

OPTIONAL ACTIVITY LED

OPTIONAL 330

ICSP CONNECTOR

1 3 4

6

1 3 4 6

R7 100

C13 1uF

21 23 25 27

28

11 13 15 17

10 9

RB0 RB2 RB4 RB6/PGC RB7/PGD

RC0 RC1/CCP2 RC3 RC5 RC6/TX RC7/RX

OSC2/RA6

MCLR RA4/T0CKI RA5

GND VDD

C19 047uF

BIPOLAR STEPPER MOTOR C12 22uF

C3 001uF

C22 33uF R8 1K

C20 33uF

C21 33uF

R3 30K

C11 22uF C5

.1uF

C9 1uF

1 3 5 7 9

R4 30K

C1 1uF

R6 2

SCREENSHOT 1 This is art However, my wife won’t let me put anything like this up in a frame in the living room It looks better on a computer display, anyhow.

SCHEMATIC 1 The A3979’s translator

is fully serviced by the PIC18F2620 and there are still plenty of spare I/O lines for your use I added an RS-232 port to allow you to control motor movement from a laptop or desktop PC.

motor action is the instigation of motor movement such as

clockwise rotation, counter-clockwise rotation, and halting the

motor The A3979 command inputs STEP and DIRECTION are

used to invoke motor action A3979 control actions include

putting the A3979 silicon to sleep, resetting the A3979’s

inter-nal systems, or enabling or disabling the A3979’s interinter-nal

H-bridges Control actions are also spawned by A3979 command

inputs, which are logic levels applied to the translator command

input subsystem The A3979 translator also provides some

use-ful output in the form of a HOME signal HOME is defined in

A3979 terms as the initial state of the translator The translator

will put the step sequence in the HOME position at power-up

Picking up on the idea that the A3979 doesn’t need an

IBM mainframe to operate efficiently, all we really need to

control the stepper motor by way of the A3979 is a small

microcontroller, such as the PIC18F2620 The PIC18F2620

hookup details can be seen in Schematic 1 As you can

see, it has more than enough I/O to support the A3979 The

ability of the PIC to clock itself internally leaves two more I/O

lines (RA6-RA7) at our disposal

There are only eight A3979 translator I/O lines we need

to deal with The A3979 ENABLE line is optional here and is

put into a permanent enabled state That leaves us with only

seven translator lines to tie to the PIC Since the plan is to

cover the top side of the A3979 PCB with a ground plane, all

of the communication lines between the PIC and the A3979

are routed on the bottom side of the A3979 PCB Note also

that the translator I/O lines are routed as far away from the

power circuitry as possible

The largest current carrying trace needs to be 0.025

inches wide, which will transfer two amperes As you can see

in Screenshot 1, that is a tall order in terms of the A3979 pins

So, we do the best we can and attach the widest trace

possi-ble as close as possipossi-ble to the A3979 motor output pins It is

important to keep the thin traces as short as possible to keep

the trace resistance to a minimum Once we have attached a

trace to the A3979 motor driver pins, we can increase the

copper area to accommodate the higher currents we may

encounter at the A3979 motor output terminations I used

large copper planes instead of traces on the bottom side of the

PCB to connect the A3979’s motor output pins to the four-pin

motor terminal block Copper planes are one of my favorite

fea-tures of the ExpressPCB printed circuit board layout program

The final critical design point is to place the current sense

resistors as close to the A3979’s current sense pins as possible

We also need to consider routing the current sense resistors’

ground return paths The ground return paths to the A3979

from the current sense resistors need to be electrically

unhin-dered I placed the current sense resistors as close as possible

to their respective A3979 current sense pins and used the

vastness of the ground plane as the sense resistors’ ground

return path The A3979 datasheet says to provide a separate

ground path for each sense resistor However, from experience

I’ve found the ground plane method to have no adverse

effects on the operation of the A3979 H-bridge circuitry

Again, I could have saved some board space by

installing fixed resistor voltage dividers for the PFD and REF

potentiometers Having the pots here allows you to adjustthe symmetry of the stepper motor current waveform (PFD)and select the amount of current you want to supply to thestepper motor (REF) with the twist of a screwdriver

No project of mine would be complete without an RS-232port and a standard issue Microchip ICSP programming/debugging portal Note that I’ve used the ST3232 in a five-voltconfiguration in this project The ST3232 can be used in 3.0volt and 5.0 volt environments by simply changing the chargecapacitor values Normally, 0.1 μF charge capacitors would besurrounding the ST3232 in a 3.3 volt project As you can see

in Schematic 1, the 0.1 μF charge capacitors are replaced with0.33 μF charge capacitors and a 0.047 μF charge capacitorbetween the ST3232’s pins 1 and 3

I used a reflow oven to reflow-solder the A3979 motordriver board’s SMT components If you don’t have access to

a reflow oven, you can assemble the A3979 board with a finetipped soldering iron If the reflow process intrigues you, youmay want to investigate the Stencils Unlimited site

(www.stencilsunlimited.com) There you will find all kinds

of SMT soldering aids The A3979’s pins are fine and require

a stencil setup for reflow soldering Because I want you to beable to build the A3979 motor driver board without having

to procure specialized tools, I didn’t go the stencil route thistime and used my Metcal soldering system to manually connect the A3979 pins to the PCB

The Metcal (www.metcal.com) soldering system is a

quick heating precision solder station If you don’t haveaccess to hot air soldering equipment, you’ll need to addsome holes to your PCB layout under the belly of your A3979.The additional holes will allow you to flow solder throughfrom the bottom of the board onto the A3979’s exposedheatsink pad, which must be thermally connected to theground plane I’ve included the A3979 motor driver boardExpressPCB file in the SERVO A3979 project download

package at www.servomagazine.com so you can use it

as a base for your custom A3979 project The A3979ExpressPCB file will reveal the presence of the ground planepassing underneath the A3979 providing a soldering pointfor the A3979’s exposed heatsink pad

Coding the A3979 Motor Driver Firmware

Everything depends on the PIC18F2620 clock At

power-up, the PIC will default to a 1 MHz internal RC clock as wehave specified that it use its internal clocking subsystem.Ultimately, we want to run the PIC at its maximum clockspeed of 32 MHz To do this, we must first load the OSCCONregister with 0x70 This will change the PIC’s internal clockspeed to 8 MHz Then, we enable the 4x PLL to boost theclock speed to 32 MHz This is done by writing a 1 to thePLLEN bit Once the clocking has been taken care of, we canassign the PIC’s port I/O to input or output according to theport pin’s required usage For the PIC18F2620, a 1 identifies

an I/O pin as an input while a 0 is used to define an I/O pin

as an output Here’s the clock and port I/O TRIS code:

For now, we won’t be employing the services of the PIC

‘s analog-to-digital converter (ADC) subsystem So, we’ll just

turn it off If you want to use the ADC in your version of this

project, you can do so as portions of the PIC18F2620’s

PORTA and PORTB I/O subsystem can become ADC inputs

Since we’re not using the ADC, here’s the ADC “OFF” code:

//*******************************************************

//* CONFIGURE A2D AND COMPARATORS

//*******************************************************

ADCON1 = 0b00001111; //all port I/O is digital

ADON = 0; //ADC off

CMCON = 0x07; //comparators off

The ADC OFF code also turns off the PIC18F2620

comparators with a write to the CMCON register

The PIC is endowed with a multitude of timers/counters

So, why not use them? I’ve coded up timer routines for

TIMER1 and TIMER3 Each of the aforementioned timers will

trigger an interrupt every millisecond I’ve set up TIMER1 as

a general-purpose timer that can resolve milliseconds,

seconds, and minutes TIMER3 leans towards being a

real-time clock and if you add an LED to RA7, TIMER3 will

drive the LED at one blink per second The timer setup code

is pretty simple and looks like this for TIMER1:

//************************************************

//* CONFIGURE AND START TIMER1

//* SET TO OVERFLOW EVERY 1mS

//* CONFIGURE AND START TIMER3

//* SET TO OVERFLOW EVERY 1mS

Screenshot 3 is a CleverScope capture of the

TIMER3 1 ms clock driving I/O pin RA7 In the TIMER3

interrupt code, I count 1,000 of these to mark seconds,

which clocks the RA7 LED if it is present in your design

I’ve provided the full code package for the A3979 motordriver board via the SERVO website If you’re interested in the

way the timers interact during an interrupt, you can perusethe download package source code to study the timer interrupt service routines in detail

The final steps of the A3979 motor driver board initialization process include activating the interrupts and setting up the stepper motor:

//******************************************************* //* CONFIGURE EXTERNAL INTERRUPTS

//*******************************************************

quarter_step; //quarter-step mode step_HALT; //stop the motor rst_step = 0; //reset the A3979 mdelay1(100); //delay 100mS rst_step = 1; //bring A3979 out of reset slp_step = 1; //put A3979 to sleep step_HALT; //make sure motor is stopped

The millisecond delay routine (mdelay1()) is actually a Cmacro and its source code can also be found in the downloadcode package The same can be said of the interrupt enablemacros There are a couple of additional macros (quarter_stepand step_HALT) in the stepper motor initialization code thatneed to be defined for you at this time Check this set ofA3979 translator input definitions against Schematic 1:

#define rst_step LATC5

#define slp_step LATC4

#define dir_step LATC0

#define ms1 LATA3

#define ms2 LATA5

Let’s see, there’s a RESET translator input, a SLEEP lator input, a DIRection translator input, and the step resolu-tion selection (ms1-ms2) translator inputs Now you can go

trans-SCREENSHOT 2 To get this shot, I simply toggled the RA7 line every time

a TIMER3 interrupt occurred Using the 1 mS interrupt clock as a time base, I can count the 1 mS pulses to create timings for seconds, minutes, hours, and days.

back to the stepper motor initialization code and resolve the

usage of rst_step and slp_step However, you still don’t have

enough information to decipher quarter_step In fact, you

can’t decrypt any of the ms1 or ms2 logic combinations until

you fix your eyes on the following truth table macro code:

The step resolution definitions you’ve just seen are actually

C macros, which assign logic levels to the ms1-ms2 translator

inputs The macro quarter_step should make sense to you now

If you’re still a bit foggy on the concepts, I suggest getting a copy

of the A3979 datasheet and referencing it as you study the

source code I’m presenting The A3979 logic is very simple and I

guarantee that if you relate what you’re reading here to what

you see in the A3979 datasheet, it will all come together for you

We could write some bit-bang code to feed the A3979

translator’s STEP input However, why do that when we can

conjure up a set of never-ending step pulses from the

PIC18F2620’s PWM generator? And, using the PIC18F2620’s

PWM is easier than writing step code The first step in getting

a string of step pulses out to the A3979 involves setting up

yet another one of the PIC18F2620’s timers for PWM duty:

The values I have loaded into the TIMER2 and PWM

registers worked fine to spin my LIN Engineering (www.lin engineering.com) 5718 series stepper motor You may have

to juggle the PR2 and CCPR1L values for your particular motor.The A3979 will accept step pulses as narrow as 1 μs Iobtained a 1.6 μs step pulse like the one you see in Screenshot

3 by loading CCP1CON with 0x1F and CCPR1L with 0x00.CCPR1L and the two least significant bits of CCP1CON’s uppernibble hold the 10-bit PWM duty cycle value I obtained the1.6 μs step pulse with the PWM duty cycle binary value of0b0000000001 The PWM duty value you see in the codeexample translates to binary 0b0101000001 The lower nibble

of CCP1CON is used to enable or disable the PIC18F2620’sPWM engine as shown in the code snippet that follows:

#define pwm_ON CCP1CON = 0x3F

#define pwm_OFF CCP1CON = 0x00

#define step_CW slp_step = 1; \

dir_step = CW; \ mdelay1(10); \ pwm_ON;

#define step_CCW slp_step = 1; \

dir_step = CCW; \ mdelay1(10); \ pwm_ON;

#define step_HALT slp_step = 0; \

pwm_OFF

If you need to use all 10 bits of PWM resolution, you’ll need

to alter the two least significant bits of the upper nibble of thedefault 0x3F value I’ve posted in my example code I’ve alsogiven the A3979 ten times (10 ms) more wakeup-from-sleeptime (mdelay1(10)) than the 1 ms the datasheet demands.The A3979 download code package also contains an inter-rupt driven EUSART driver for the PIC18F2620 A menu system

is incorporated that will allow you to connect the A3979 motordriver board to a laptop or desktop PC and control the move-ment of the stepper motor We don’t have the page space here

to talk about the EUSART driver in detail However, I describethe EUSART driver down to the bit level in the RS-232 chapter

in my book Networking and Internetworking with Microcontrollers The neat thing about the EUSART

driver is that you can easily drop it into any PIC cation that needs robust RS-232 communications

appli-Firing Up the A3979 Motor Driver Board

It is important that you don’t adjust the REF pot

to get a voltage above four volts on the REF pin So,before you fire up your A3979 motor driver board,

be sure to adjust the REF pot’s wiper to be nearer tothe grounded pot pin It’s a good idea to simply center both of the pots as they will produce around2.5 volts on the PFD and REF pins upon power-up.Once you’ve powered up successfully, adjustthe REF pot to put 1.6 volts on the REF pin This willlimit the stepper motor current to one ampere The

SCREENSHOT 3 This shot was taken with my CleverScope This is as tight a

pulse as I could get with the circuitry we are using.

stepper motor current formula to determine the REF voltage

for a desired stepper motor current limit value is:

REF voltage = 1.6 * Maximum Current Limit Value

The A3979 is rated to handle a maximum motor current

of 2.5 amperes Plugging that into our formula gives us:

REF voltage = 1.6 * 2.5 = 4 volts

Now you know why you don’t want to exceed four volts

at the A3979’s REF pin

You can experiment with the PFD pot setting as the

voltage on the PFD pin determines the symmetry of the motor

current waveform Ideally, you want your motor current

waveform to look like the sinusoidal motor current waveforms

in the A3979 datasheet A trick I’ve picked up is to listen to the

motor as you adjust the PFD pot If the motor sounds like it’s

trying to kill itself, your motor current waveform is not right

Adjust the PFD pot to get the smoothest and quietest sound

from your motor at the desired REF voltage Plug this code into

your main service loop to adjust the PFD and REF voltages:

Spinning Out

That’s all there is to it I’ll include a BOM (Bill ofMaterials) file with the A3979 firmware download soyou can scrounge up the parts you’ll need to buildyour own A3979 motor driver board Who knows, theA3979 just may be the answer you’ve been looking for

to complete those small robotic motor control jobs Asalways, if you have any questions about the A3979 orthe A3979 motor driver board, fire off an email to me.Have fun! SV

For further reading, check out this book, available on the SERVO website:

www.stmicro.com

ST3232 Saelig Corporation

www.saelig.com

CleverScope The A3979 motor driver board

code was written with the

HI-TECH PICC-18 C compiler.

Resources

Fred Eady can be reached via email at fred@edtp.com

Contact the Author

My particular application — as

we discuss here — is in filtering

input data used to calibrate

the odometer in a road rally computer

After computing a best-fit line using a

linear regression on a set of data

points, square roots are needed to

compute the distance of each data

point from the line If a data point is

too far off the line, it is discarded on

the assumption of human error and the

line is recalculated

Square Root on a PIC in the

November ‘06 issue of Nuts & Volts

(www.nutsvolts.com) demonstrated

a simple algorithm for computing

square roots While the

implementa-tion was very compact, the algorithm

has a significant performance issue — it

gets extremely slow as the argument to

the square root function gets large

This is because the algorithm

sequen-tially computes each square starting

at 1 until it finds the one closest to

the argument

Let’s look at an alternative method

of computing square roots It has the

advantage of taking nearly constant

time, regardless of argument — about

150 microseconds for a 32-bit input on

a 20 MHz PIC18 This makes it practical

to compute square roots in a robot

control loop running 100 times per

second, for example Code size is

about 35% larger, but still well under

Figure 1 shows asample long divisionproblem The basicsteps for each digitare:

• Make a guess of the digit

• Multiply the trial digit by the divisor

• Try to subtract that product from the current remainder

– If the product is larger than theremainder, make a new guess with

a smaller digit

– If the subtraction produces aresult larger than the divisor, make

a new guess with a larger digit

In the case of Figure 1, the guessfor the first digit (estimated from 62÷8)initially turned out to be too large The

first trial digit and product are crossedout and a new set was tried

The description for binary long division is the same, but the only twochoices for the trial digit are 0 and 1.This makes it very easy to make the rightguess, and also very easy to computethe trial * divisor

product that is to

be subtracted fromthe remainder

Figure 2 shows

a sample binarydivision problem

By definition

of square root, ifthe divisor equalsthe quotient, then

we have thesquare root of thedividend The algorithm for square roottakes the form of binary division wherethe divisor is continually changed tomatch the quotient

The square root of an N-bit number has N/2 bits; in our example,we’ll be taking the four-bit square root

of an eight-bit number From our binary division viewpoint, we’ll be looking for the square root of the dividend We start with a divisor at themidpoint of the possible result range: a

1 followed by N/2 – one zero; in ourcase, a 1 followed by three zeros

Square roots have a number of possible

applications in microcontroller systems

Square roots have a number of possible

applications in microcontroller systems

√ by Tim Paterson

6 79 897)62518 6279 -5382 8698 -8073 625

1101 1001)1111001 -1001 1100 -1001 0110 -0000 1101 -1001 100

Figure 1.

Decimal long division example.

Figure 2 Binary long division example.

The first half

of Figure 3 shows

an example withthe first bit calcu-lated Notice inthe figure how

we don’t knowthe divisor yet, so I’ve substituted “xxx.”

The binary division procedure requires

multiplying the guess by the divisor, so

“0” is used for each “x” for now

The second half of Figure 3 shows

the partially-completed calculation of

the second bit We make a guess of 1,

multiply the trial bit by the divisor

(which is 11xx now to stay equal to our

quotient), and subtract from the

remainder We must also account for

what we should have subtracted when

we computed the previous bit Back

then, we only subtracted 1000 Now

we have a divisor of 11xx, so we should

have subtracted 1100 before I’ve put

this make-up value to be subtracted on

an additional line Because we’re

shift-ed right one bit, it’s written as 1000,

but you can follow the “1” bit up the

column and see it’s in the right place

So at this point, we have two values

to subtract — 1100 and 1000 — whose

binary total is 10100 This total amount

to subtract is larger than our current

remainder of 1110, so we must

con-clude the trial was too big — instead of

trial bit and the

divisor, and also

The remainder is big enough, so thetrial bit is correct

The second half of Figure 4 showsthe calculation of the final bit Notehow that extra value we subtract (tocorrect for previously subtracting thewrong value) has more “1” bits in it

Each of the 1 bits corresponds to a subtraction for a previous quotient bitwhere we didn’t subtract the bit we’reworking on now So, for each 1 bit inthe quotient (not counting the trial bit),there will be a corresponding 1 bit inthe extra value to subtract to accountfor not subtracting it originally

Once more: For each 1 bit in thequotient, there will be a 1 bit in theextra value to subtract This is the same

as saying the extra value to subtractequals the quotient (and divisor) before

setting our trial bit So, we subtract thedivisor with the trial bit, and we subtractthe divisor without the trial bit Addingthat up, we subtract 2 * divisor + trial.That’s it Since the divisor and quotient are always equal, I’m going tostart calling them the root To calculateeach bit of the root, try to subtract twotimes the root plus the trial bit from thecurrent remainder If it fits, the trial bit

is part of the root; if not, that bit is zero

Square Roots

Listing 1 represents this algorithm

as a function in C You don’t need toknow C to understand it if you look inthe sidebar for the explanation of theoperators This algorithm also includesthe extra step of rounding the resultbased on whether the next bit would

be 0 or 1 This function has been compiled and tested for both Windowsand the AVR microcontroller

101 101x)01111001 -1000 1110 -0000 11100

- 1010

- 1000 1010 1011 1011)01111001 -1000 1110 -0000 11100

- 1010

- 1000 10101

- 1011

- 1010

Figure 4 Third

and fourth bits of

square root being

calculated.

1 1xxx)01111001

-1000

111 11 11xx)01111001

-1000

1110 -1100 -1000

Figure 3 First and second bits of square root are being calculated.

typedef unsigned long UINT;

#define INT_BITS ((sizeof(UINT)) * 8) UINT SquareRoot(UINT arg)

{ UINT trial;

UINT root;

trial = 1 << (INT_BITS - 1); // set up 100000 binary root = 0; // Inside loop, really root * 2 do

{ trial >>= 1; // move trial to next position in root root |= trial; // combine trial bit into root

if (arg < root) // does trial root bit fit?

root ^= trial; // no, remove trial bit else

{ arg -= root; // root fits, remove it from arg root += trial; // double the trial bit

} root >>= 1; // move both root & trial to next trial >>= 1; // position within arg

} while (trial != 0);

// Compute rounding // See if next bit computed would be 1; round up if so.

if (arg > root) root++;

return root;

}

LISTING 1 Fast square root algorithm in C.

The function starts by initializingthe variable trial which contains thecurrent trial bit The function isdesigned to be easily targeted for 16-bit or for 32-bit integers (or anyother size), which explains the some-what complicated-looking expressionthat initializes trial I’ll assume we’reusing 32-bit integers, in which case trialgets initialized to 0x80000000 — a 1 inthe MSB and the rest 0 Note that thefirst thing that happens inside the loop

is that trial is shifted right one bit, sothe first trial bit value is 0x40000000

In the main loop, the variablenamed root actually keeps the value oftwo times the current root Combiningtrial into root gives us the amount tosubtract from the current remainder(called arg) if it fits If it doesn’t fit, theexclusive-OR operation is used to clearthe trial bit in root If it does fit, we need

to double the trial bit so that, as part ofthe new root, it will be root times two

At the bottom of the loop, rootand trial are each shifted right one bit.This moves them to the next bit position for which we’ll be computingthe root Looking back at Figure 4, youcan see how the subtraction for eachbit is shifted right compared to the previous bit Conveniently, when it’stime to end the loop, this right shift of

I have never seen any description of

this algorithm published before I don’t

mean that it’s original with me — in fact, I

believe this algorithm is widely used in

floating-point hardware I became aware

of the algorithm many years ago when

reading about a particular computer

architecture that included hardware

square root.

I have an old Pentium manual that

lists the time to perform a floating-point

divide as 39 clock cycles, and the time to

perform square root as 70 clock cycles –

less than twice as long The only way that

could be happening is if the

floating-point unit implements this algorithm in

hardware.

For a processor that has hardware

divide but no hardware square root, the

most common way to compute square

root is with Newton-Raphson iteration.

This algorithm is very simple:

1) Make a guess at the root.

2) Divide the guess into the argument.

3) Average the quotient with the guess to make the next guess.

This method converges very fast;

every iteration doubles the number of digits (or bits) of accuracy The trick is

to come up with a good initial guess

This is difficult for integer or point arithmetic, where it may take one iteration to get one bit correct Then the subsequent iterations give you 2, 4, 8, and 16 bits of accuracy — typically five total iterations for a 16-bit root of a 32-bit number.

fixed-When using floating point arithmetic, it’s easy to get an accurate initial guess because of the “normalized” format of the numbers With an eight-bit multiply and addition, you can come up with a guess accurate to almost five bits Then only three iterations are needed to achieve single-precision accuracy.

I used this technique when I wrote

the x87 emulator for Microsoft in 1991 This program provided floating-point arithmetic back in the days before it was built into every processor; it hasn’t been needed since the 486SX It is still widely distributed, included as WIN87EM.DLL with every 32-bit operating system Microsoft has shipped.

While researching for this article, I found another square root algorithm

on the Microchip website posted in Application Note TB040 for PICs that have hardware multiply The general idea is to set a trial bit in the root, square the root, and see if it was too big The algorithm calculates the result bit by bit, with one square (multiply) operation per bit According to the app note, with a 32-bit input this method would take 200 microseconds

at 20 MHz, and the code size appeared

to be over 200 bytes So it is both bigger and slower than the algorithm I present.

Square Root Algorithms

SquareRoot:

clrf32 root ; root = 0 clrf16 trial ; trial = 0x80000000 clrf trial+2 ; "

; Didn't fit, remove trial bit

xor32 root,trial ; root ^= trial bra ShiftRight

RootFits:

sub32 arg,root; arg -= root add32 root,trial ; root += trial ShiftRight:

bcf STATUS,C; Clear C flag for following rotate rrf32 root ; root >>= 1 - Always leaves C flag clear rrf32 trial ; trial >>= 1

bnc RootBit ; if trial bit shifted out, we're done

#if ROUND

; Compute rounding Note that root always fits in 16 bits here,

; but it could round up to 17 bits.

movf arg+2,f ; see if bits 17-23 are non-zero bnz IncResult ; if non-zero, arg is big, always > root cmp16 arg,root; if (arg > root) - C flag set if root < arg bnc Done ; "

101 101x )011 11 001 -1000 1110 -0000 11100

- 1010

- 1000 1010 1011 1011 )011 11 001 -1000 1110 -0000 11100

- 1010

- 1000 1010 1

-... Paterson

6 79 897)62518 6279 -5382 8698 -8073 625

1 101 1 001) 1111 001 -1 001 1100 -1 001 0110 -0000 1 101 -1 001 100

Figure 1.

Decimal... pulse with the PWM duty cycle binary value of0b0000000 001 The PWM duty value you see in the codeexample translates to binary 0b0 1010 00 001 The lower nibble

of CCP1CON is used to enable