QRZ! Ham Radio 8/QRZ Ham Radio Callsign Database - Volume 8.iso/pc/files/p_tapr/tapr92.txt

Text File | 1996-06-25 | 72KB | 1,440 lines

A Blow-By-Blow Account of the 1992 TAPR Annual Meeting
======================================================

The following is based on the notes I took during the TAPR annual
meeting.  Any mistakes are mine.  On no account should you assume
that this account represents the official position of TAPR or anybody
else.  But I hope you find it interesting.  You may publish parts or
all of this document if you wish, but if you do please credit the
SANDPAC Newsletter.  Sorry it took so long for me to get these notes
published this time.
        73  -Paul Williamson, KB5MU

The 1992 TAPR Annual Meeting was called to order by TAPR President
Bob Nielsen, W6SWE, on March 7 at 9:41AM at the Inn At the Airport in
scenic Tucson. He introduced Master of Ceremonies "Packet" Pete
Eaton, WB9FLW, who introduced TAPR Office Manager Heather Johnson,
N7DZU.

Heather welcomed everybody to Tucson.  TAPR office hours are Tuesday
to Friday, 10 AM to 3 PM.  Thanks to Bob Nielsen, W6SWE, for his work
as President of TAPR.  Thanks to Pete Eaton, WB9FLW, for taking care
of TAPR's presence at the Dayton Hamvention.  Thanks to Ron Bates,
AG7H, for fielding technical questions.  Thanks to Dave Medley,
KI6QE, for fielding questions on the PK-232 DCD mod and the PSK
modem.  Everybody hates not to have the latest version of software,
so thanks to the new librarian, Lou Nigro, KW7H, for updating all the
software in the TAPR library.  Anybody who has software that should
be in the TAPR library, contact Lou.  And of course, thanks to Lyle
Johnson, WA7GXD, for fielding the daily barrage of miscellaneous
questions. The new TAPR 9600 baud modem is available now at $70, and
the Trakbox is available now at $185.  Thanks to everybody associated
with TAPR for being fun to work with.

Bob Nielsen, W6SWE, took the microphone again, and asked everybody to
introduce themselves.  112 of the usual suspects from all over the
place were present.  Nielsen then announced that the TAPR Board of
Directors had been resized to 9 positions.  The new officers are:

    President:         Bob Nielsen, W6SWE
    Vice President:    Dave Toth, VE3GYQ
    Secretary:         (open, volunteers?)
    Directors:       Tom Clark, W3IWI
                       Jerry Crawford, K7UPJ
                       Jack Davis, WA4EJR
                       Pete Eaton, WB9FLW
                       Greg Jones, WD5IVD
                       Dan Morrison, KV7B
                       Harold Price, NK6K

Bob Hansen, N2GDE, will continue to serve as editor of the Packet
Status Register.  Lou Nigro, KW7H, will take over as software
librarian.

Pete Eaton, WB9FLW, took the microphone at 10:00 AM.  For the first
time, this year's TAPR meeting is recorded by a printed Proceedings.
Thanks to the several speakers who managed to submit papers for the
Proceedings.  He then introduced the first speaker:

==================
Lyle Johnson, WA7GXD
TAPR's New 9600 Baud Modem - What It Is, What It Isn't

TAPR's new 9600 baud modem is compatible with existing K9NG and G3RUH
9600 baud modems.  It's an inexpensive kit, capable of full duplex
operation (like the G3RUH but unlike the K9NG), with improved DCD
(data carrier detect) performance and clock recovery.  It looks up
the transmit waveform in ROM, like the G3RUH, and has a frequency
response compensation adjustment on the receive side.  The board is
designed to mount internally in a TNC-2 (where it's a tight squeeze)
or a PK-232.  The board includes provisions for bit regeneration
(parts optional) for use in a full duplex digipeater.

There are about 5000 G3RUH modems in service overall.  Heathkit is
mostly out of the kit business, so one of the goals of the 9600 baud
modem project was to leverage TAPR's expertise at packaging kits to
make available a 9600 baud modem with better performance than the
K9NG at a reasonable price.  The design address problems with the
K9NG modem (like its half duplex design) and adds features for
network builders (like the bit regenerator).  The state machine in
the K9NG could mistake silence (like a squelched radio or a weak
carrier) for a data carrier, thus holding off transmission
indefinitely. The new state machine design cures that problem, and
also gives better clock recovery.

A block diagram of the modulator portion of the modem was displayed.
The modem disconnect header (TAPR standard or PacComm extended) goes
into a data scrambler, to the transmit waveform ROM lookup table, to
a digital to analog converter, through a filter, and out to the
radio.  The clock can be obtained from the modem disconnect, or
generated on the 9600 baud modem board.  RTS from the TNC controls
PTT through a watchdog and an LED indicator.  Programmable logic
provides switching from the 9600 baud modem to the TNC's internal
modem.  It is still necessary to hook up the modem directly to the
discriminator and modulator inside the radio, not to the speaker and
microphone jacks.

A block diagram of the demodulator portion of the 9600 baud modem was
displayed.  The input buffer has a high input impedance, 100 kohm or
more.  A Butterworth 6kHz filter is tweakable to compensate for the
frequency response of the receiver, which is especially useful with
rigs with LC filters like Mitreks.  The data slicer feeds a
descrambler, state machine and DCD circuit, with an output for a bit
error rate test.  The DCD detects synchronous transitions, rather
than the lack of asynchronous transitions as before.

Block diagram of the bit regenerator.  It consists of a PAL and a
FIFO chip.  The FIFO inserts a nominal delay of 8 bits, and is
required to eliminate bit jitter and timing errors in the received
signal.  The result is that if the repeater can copy the input signal
at all, it will transmit a perfect signal.  A diagram of the
switching configuration shows how the bit regen is connected with the
TNC, so that the TNC can transmit on the channel instead of the bit
regenerator.

A sample of the 9600 baud modem was passed around the room.  It's a
four layer PC board, which helps cut the RF noise.  It's relatively
compact, but it has a lot of parts on it.  The documentation shipping
today is preliminary, and needs more information on hookup to various
TNCs and radios.  Updated documentation will be sent to early buyers.

Question: I'm interested in higher speeds than 9600.  Is this modem
planned to be scalable to higher baud rates?
Answer: It hasn't been tested, but it should work.  The op amps and the
ROM lookups are plenty fast.  The input analog filter would have to be
adjusted.  The transmit lookup table might help at higher speeds to
compensate for the nonlinearities of a wider filter.  We should try that
experiment.

Question: Has it been tested on the satellite?  How does its
performance compare to the G3RUH?
Answer: It hasn't been tested on the satellite yet.  Lab bench tests
aren't realistic, but they show that the new modem is no better than
the G3RUH, and 1 to 2 dB worse under some conditions.

Question: The PK232 limits the TBAUD (computer to TNC) rate to 9600.
Does this cause a problem when using the PK232 with a 9600 baud
modem?
Answer: We haven't noticed any problems in testing.  Probably the
worst thing is that you won't be able to keep the pipe completely
full on transmit, resulting in dead time.

On the PK232, the ALTMODEM 1 command permits the user to switch to
the 9600 baud modem from the keyboard.  This means you effectively
have a third radio port, because you can leave the PK232's two
existing ports hooked up to other radios.

TXD (delay between PTT and first data) is another issue.  If you have
the state machine DCD mod kit in your PK232, there's a small extra
delay that requires increased DCD at the other end.  With a TNC-2 or
a Kantronics DataEngine we could run TXD of 1 or 2.

Question: Has the modem been tried at 4800 baud on the 6m backbone?
Answer: No.

==================
Dewayne Hendricks, WA8DZP
Use of CDMA Spread Spectrum in the Amateur Service

Last year at the TAPR meeting, we talked about Part 15 spread spectrum
(SS) communications systems, and displayed a low-cost commercial
product capable of high data rates.  The year before that, N3FCT
presented a paper on license-free spread spectrum.  Folks in the San
Francisco Bay area were inspired to look at the Part 15 market.  The
results of a field test of units from Proxim was posted on Usenet.
One watt into a 6 dBi antenna gave 2 miles LOS tested and 8 miles LOS
predicted at 121 kbps.  There's been lots of activity in the wireless
LAN market and at the FCC since then.  We wanted to find out why the
amateur radio service isn't using SS techniques, and approach the FCC
for whatever rule changes are needed.  An STA (Special Temporary
Authority) for testing was sought and obtained.

A second generation Proxim unit was passed around.  It is a 121 kbps
data radio (data in, antenna out) in a very tiny box.  In the OEM
package for laptops, it costs $50.  The internal modules of the
Proxim radio were displayed - all very tiny.  Computer manufacturers
are starting to put these directly into laptops for wireless LAN
use.  Challenge to ham radio:  get coordinated with the computer BBS
folks and build a wireless Internet.

The ARRL Spread Spectrum Handbook is good for a basic tutorial and
for historical information on SS.  AMRAD did the experiments that led
to the current rules permitting SS.  The present rules are like
handcuffs.  In particular, they mean that amateurs won't be able to
use commercial SS products, because they don't happen to use one of
the few spreading sequences permitted by the rules.  We decided to
seek a rules change.  This turned into a long process.  It turns out
to be important to use connections in rulemaking matters.  After last
year's meeting, Paul Rinaldo, W4RI, at ARRL HQ was approached for,
and arranged, ARRL cooperation for an STA submission.  After several
months, and polishing by the League lawyers, the STA application was
submitted to the FCC, which promptly sat on it and did nothing.

About this time, the League managed to get FCC Chief Engineer Stanley
to speak at the Computer Networking Conference.  Stanley is a
proponent of spread spectrum, and he was interested in the amateur
proposal.  He assigned a staffer to the STA.  The STA requested a two
year authorization for any spreading code on any VHF or higher band.
Before the STA could be approved, the staffer had to get every agency
involved with the use of any of those bands to accept the proposal.
It was a lot of work, but with support for Stanley and the League is
was done.  The STA was granted.

There are plans for tests in at least the Northern and Southern
California areas.  People interested in serious experimentation with
SS can be added to the STA.  The intention is to eventually submit a
Petition for rulemaking to get a better set of SS rules.  The
restrictions on spreading codes and the requirement for narrowband
station ID are particularly onerous.

One test is starting in San Diego, under the California State Library
project for packet radio.  Using radios produced by SRI, the project
will interconnect libraries to the wide variety of online databases
available via the Internet, without the cost of a 56kbps landline
connection to the Internet.  The pilot project in San Diego is
sponsored by Apple Computer, and radios have been donated by
Tetherless Access (Hendricks's company) and by hams.  Parts of the
network are operated under Part 15 (license free), Part 97 (amateur),
and Part 5 (experimental).

Funding has been allocated in the Bay area to connect 100 libraries
from San Jose to Roseville, San Francisco to Sacramento, all in one
WAN.  Part 97 (amateur) radios are to be used for long haul links,
and Part 15 radios for intra-city links.  This experiment will last
through the end of the year. Phase II will involve redesigned 1.5
Mbps radios and associated networking software.  Currently the
project supports only Macintosh computers, partly because any Mac off
the shelf can handle up to 900 kbps data links.  Hams in northern and
southern California will be seeded with equipment to try out Phase
II.

There's a lot of work to do.  First, get the FCC rules changed.  That
will take 1 to 1.5 years, on the fast track.  It is hoped that the
rules will be changed before the STA ends.  Then hope that
manufacturers will go after the ham market.  We haven't done a good
job of keeping the FCC up to date.  We need to tell the FCC where the
public interest lies.  The FCC wants to help, but we have to play the
game: STAs, Part 15, waivers, and so forth.

Question: Is all the Part 15 activity at 900 MHz?
Answer: No, we're not using 900 MHz at all.  We are currently working at
2.4 GHz, and have plans for 5.7 GHz.

====================
Fried Heyn, WA6WZO, ARRL Southwestern Division Director

Read the Division newsletter for more about what's going on.  Some high
points:
There's a bill in Congress that can protect amateur frequencies from
further erosion.
A big effort funded by the ARRL membership was directed to preserving
amateur spectrum at WARC-92.  The results are not final yet, but it
appears that no amateur spectrum was lost at the conference!
The 1992 ARRL National Convention is at the Los Angeles Airport Marriott.


====================
Jon Bloom, KE3Z
TAPR DSP Project Report

KE3Z received one of the first set of beta test DSP boards from the
hardware designer, Lyle Johnson.  Procrastination set in: the board
has over 2000 holes, so assembling it is a bit of work.  Finally, got
it assembled.  The next step was to learn about how to do DSP
programming.  Highly recommended: the Computer Literacy Bookstores in
and around San Jose.  Good books for techno-weenies, including some
on DSP.

Some preliminary DSP applications are already written and working.  A
Bell 202 (1200 bps AFSK, like on 2m packet) modem is up and working,
with a driver for KA9Q NOS for packet use.  A RTTY modem (2125/2295
Hz AFSK) with a RTTY driver and receive-only AMTOR driver is working.
Dave Hershberger, W9GR, has written two audio-in/audio-out filter
programs.  One notches out tone interferers from the audio channel,
and works great.  The other tries to remove noise from an SSB signal,
and needs more work.  These filters were originally written for a
TMS32010 board of his own design, and have been mechanically ported
to the TMS320C20 on the TAPR DSP board (so they aren't optimized for
it).

The lesson is that DSP software isn't necessarily magic anymore.  It
is quite possible to write working modem software, for instance,
without getting heavily into sophisticated mathematics.  The basic
building blocks are simple, and design tools exist to handle filter
design.

Coming attractions:
    Bell 103 modem (HF packet)
    1200 baud PSK for Pacsats
    9600 baud FSK (K9NG/G3RUH/TAPR compatible)
    Spectrum display
    4800 baud PSK for Pacsat - the satellite has never been in this
mode, since no user modems have ever been built.
    2400 bps Kantronics-style
    improved W9GR "de-noiser" filter
    weather fax
    slow-scan television

Some of these applications exist for other boards, and just need to
be ported to the TAPR board.  Others need to be written from
scratch.  Many of these applications will be implemented over this
spring and summer.  PC software to support the applications is also
needed.  NOS is nice for some packet applications, but other
applications need other PC software.

A virtual-hardware block diagram of the RTTY demodulator is shown.
It's just two bandpass filters running into detectors, followed by a
comparator and lowpass filter.  Standard stuff.

DSP-oriented block diagram of the RTTY demodulator.  The bandpass
filters are straight out of a manufacturer's application note.  The
coefficients for the filters are computed by a computer program, so
no heavy math is needed.  The filter design is an 80-tap FIR (finite
impulse response) linear phase filter.  The detector is just absolute
value, then a peak detector followed by a decay.  The comparator is
just a subtraction.  The lowpass filter is another cookbook design.
That's it.  Notice the absence of any higher math in this
description.  This isn't the optimum demodulator, but it's as good as
most analog designs, and it works.

A similar approach was tried for the Bell 202 demodulator, but for
unknown reasons it didn't work very well.  A Bell 103 modem was taken
from another ap note, based on a discriminator design: the audio is
delayed by 90 degrees and mixed with itself, then lowpass filtered
and compared to zero.  The Bell 202 demodulator is just this design,
tweaked up for Bell 202 tones and bit rate.

The book _Digital Signal Processing Experiments_ by Alan Kamas and
Edward A. Lee contains a diskette with educational versions of
Burr-Brown DSP design tools, including the tool that generates
coefficients for FIR filters.  The book is quite inexpensive (about
$21) for a DSP tool set.  It's published by Prentice-Hall;
recommended.

Both DSP programmers and PC programmers are needed to work on
applications for the TAPR DSP project.  The PC level programmer sees
an environment much like a DSRI board plugged into a PC.  Contact Jon
Bloom if you're interested.

Tom Clark, W3IWI, spoke up:
The TAPR DSP project grew out of earlier development on the
Dalanco-Spry Model 10 DSP board.  About 30 of these boards were
purchased, and probably some could be made available to new
interested people.  Lots of applications were written for the
Dalanco-Spry board, and they're all available for the grabbing on
tomcat (Tom C's AT) by anonymous FTP via Internet, or by telephone,
or by floppy disk if necessary.  Another application that's needed is
a good adaptive HF modem (along the lines of Clover II), and HF
protocols that can use them.  AX.25 sucks on HF.  AMTOR has problems,
too.  Pacsat broadcast protocols are a bit like what's needed for an
HF protocol, but it needs changes for the HF environment.

Question: Who is doing satellite imaging?
Answer (W3IWI): The AEA box has it.  The Dalanco-Spry board had it,
and that version will be ported (or rewritten) for the TAPR DSP
board.  We may want to rewrite rather than porting Dalanco-Spry
applications, because the subset of instructions supported by that
processor was pretty brain-damaged compared to the instruction set of
the TMS320C20 used on the TAPR DSP.  The AEA box contains a Motorola
56001 DSP processor.  Its modems are superb compared to analog
modems.

Question: What sampling rate can DSPs handle?
Answer (W3IWI): The Dalanco-Spry board could do a spectral display at
50 KHz.
Answer (KE3Z): It's been suggested that we can just digitize the
antenna voltage.
Answer (W3IWI): Unintentional radiation can be a problem.  The DSP
processors can generate a lot of crud.
Answer (KE3Z): The TAPR DSP board is pretty well decoupled.

======================
Bob Nielsen presented an award to Chuck Green, N0ADI, for his
outstanding contributions to the development of packet radio through
ten years of TAPR.  Chuck accepted the award, but claimed that it
represents the teamwork that goes on behind the scenes everywhere.

Pete Eaton held the drawing for door prizes, then the meeting broke
for lunch.

======================
Mike Parker, KT7D
The Radio Workstation Concept

Block diagram:  Antenna connected to analog radio, connected to a
digital sampling and output box, connected to a general purpose
workstation.

The problem with DSP software is that it's so difficult to write for
special-purpose DSP processors.  By the time you've finished writing
the software, the special-purpose DSP processor you wrote it for is
obsolete.  Worse, by that time your general-purpose workstation has
improved in performance to the point where it outperforms the old
special-purpose DSP!  This whole problem can be bypassed by writing
the DSP application for the general purpose workstation in the first
place.  Portable languages can be used, so it should be easy to take
advantage of improved workstation technology as it comes along.

Block diagram of a sample application (not implemented): meteorology
satellite image reception and display.  The software can be written
in manageable small modules, like orbit prediction or overlay
generation, and the resulting modules can be patched together readily
to make a complex application.

The SPARCstation currently used as the workstation costs about
$10,000 after discounts.  Thus the TAPR DSP project and the Radio
Workstation approach address different (but overlapping) issues.  The
TAPR DSP is cost effective, IF you can get the software working
quickly.  Experimental applications written for the Radio Workstation
might serve as prototypes for DSP-board applications.

The current configuration uses a DAT (digital audio tape) machine
interfaced via SCSI to the SPARCstation or VAX, running Unix or VMS,
with signal processing software, an interactive display layer like X,
and FORTRAN with VMS extensions and C for widgets.  This isn't a
cheap configuration, but it's off-the-shelf.  The code developed for
this project is being made available free, on the condition that if
you add to it, you make your results available free also.

Why should TAPR get involved in this project?
  - to help promote research
  - to spend time doing research, not software development
  - to develop and debug applications for the TAPR DSP board
  - to help develop standards for file structures, datalink
structures, and so forth, before it's too late.

Question: How big is the publicly-available source code?
Answer: About 100,000 lines of code, comprising 300 processing primitives.

=====================
Tom Clark, W3IWI
Various Topics

Topic #1: 900 MHz
In 1985, Motorola and NEC were engaged in a battle to dominate the
market for cellular telephone base station equipment.  In 1990, NEC
gave up on the market, and the hardware they had managed to sell was
orphaned.  In 1991, the NEC hardware at cell sites in Richmond was
scrapped.  A total of over 100 45W radios already outfitted for 19.2
kbps data were made available surplus for $20 each.  The equipment
complement for a normal cell site was 16 transmitters and 16
receivers, all nicely racked up.  Six cell sites plus spares were
scrapped.  Each cluster weighs about 800 pounds and fills a pickup
truck.  The salvaged radios are all spoken for, but similar
opportunities may become available in other areas.

The radios are set up for 19.2 kbps data, with a digital interface,
used originally for signaling for billing purposes.  They are
designed for full duplex operation, and the receiver won't work
without the transmitter operating. An analysis of the filters
indicates that data scramblers will probably not be required.  The
transmitters are a very simple, conservative design.  They are
serviceable, understandable, and robust.

Block diagram of the receiver.  A buffer amp feeds a
synthesizer-driven mixer, followed by a standard IF and
discriminator.  A measurement of received signal strength goes to a
Z80 microprocessor.  The audio and demodulated data go out.  The 70
MHz IF filters are from the same line as those used in the Microsat
receiver. The RF filters will need to be replaced with ones that can
go up to the amateur 900 MHz frequencies.  High-side injection will
be needed, because low-side injection puts the IF image in a crowded
spot in the band.  The modification involves removing a chip
capacitor and trimming a microstrip.  It remains to be seen whether
the Z80 processor part of the board is useful.  Perhaps it could be
used for signal strength telemetry.

Block diagram of the transmitter.  A 15.36 MHz oscillator drives a
synthesizer.  A power controlled amplifier chain feeds the antenna.
The modulator frequency modulates the synthesizer.  One bandpass
filter in the RF path needs to be replaced to reach the amateur 900
MHz band.

All this cost us $20 per unit.  Be jealous.

Outstanding issues and problems:

 * The transmit frequency isn't easily moved.  Luckily, this is a
   relatively simple radio so modifications are easy.

 * Frequency stability.  In the cell site, the radios were driven by one
   common master oscillator at 15.36 MHz.  To use the radios individually
   and get 1 kHz error at 900 MHz, we need to provide an oscillator that's
   good to one part in 1e6.  Crystal manufacturers want $80 to $90 for
   such an oscillator, despite the quantity price of around $7.  If anybody
   knows a source for small quantities of 15.36 MHz oscillators, please let
   me know.

 * Antennas and preamps.  To use all 100 radios, we need to get 200 antennas
   and 100 preamps.  They have to be cheap and easy to replicate 100 times.

 * The vehicle locator service has priority over the amateur service in the
   900 MHz band.  The AVL (automatic vehicle locator) folks have been very
   aggressive about defending their allocation, even threatening civil suits
   against retail stores using theft alarms in the band.  The North Texas
   Microwave Group is also looking into this problem.

 * Network coordination and architecture.  This is mainly a political
   problem.  It has practical implications, like what kind of antennas are
   needed and where they have to be pointed.  The biggest problem is how
   get everyone to agree on something, anything.

 * Data pump.  The standard NET/ROM or TheNet stack of TNC's isn't suitable.
   Something like the Kantronics DataEngine, PacComm's V53 board, or the
   Gracilis PackeTen board is needed.  It has to be cheap, reliable, and
   robust enough to survive a mountaintop environment.


Topic #2: AO-13 orbit decay and Phase III-D

AMSAT OSCAR-13 is in a highly elliptical orbit, and its perigee height
has been decreasing steadily.  If this trend continues, the satellite
would be lost during 1992.  However, the perigee height is starting to
turn up, as predicted.

A graph shows the prediction generated using the Cray computer
running the NASA GEODYN theoretical model of deep space orbits, and
the NORAD tracking data obtained since the prediction was run.  The
real data tracks the prediction pretty well.  The prediction shows
that AO-13 will be lost in 1996.

The effect is NOT atmospheric drag.  The gravity field of the Sun and
the Moon are changing the shape of the orbit, making it more narrow
and moving it toward apogee.  When the eccentricity reaches 0.75, the
perigee will intersect the atmosphere.  The inclination is also
changing.

AO-13 is going to die.  There's no way to save it: there's no fuel on
board, and there's no way to dig a deep enough tunnel through the
Earth.  So, what we need is a replacement satellite.

A diagram of the Phase III-D mechanical design shows a *really* *big*
satellite.  It's roughly triangular, 8 feet on a side, with two solar
panel "wings" with a 17-foot span.  The configuration shown (one of
several proposed) has antennas for bands from 10 meters to 10 GHz.
The satellite will have receivers on 2 meters through 3 cm,
transmitters on 10 meters through 3 cm, and a programmable IF matrix
capable of selecting any desired combination of bands.

Users in urban areas are having more and more trouble installing
large antennas, so the satellite will have 10 dB to 20 dB more
performance on each link.  The 10m downlink will be capable of
several hundred watts, possibly using the long solar panel wings for
an antenna.  The gain antennas for 2m and 70cm consist of several
elements mounted on the sides of the spacecraft, each with its own
amplifier with controllable phase, giving many possible antenna
patterns.  The satellite will be 3-axis stabilized using momentum
wheels, so the antennas will always be pointing straight down at the
Earth.

This new satellite will also have an elliptical orbit.  We've learned
an important lesson with AO-13: elliptical orbits are chaotic.  It's
possible (though not easy) to predict what will happen with a
well-known set of initial conditions, but it's not possible to
compute a set of initial conditions that will result in a desired
orbit.  So, to ensure that Phase III-D will have a long life in the
desired orbit, it will be equipped with a motor that can be used many
times to made adjustments to the orbit.  The desired orbit makes
exactly 3 orbits in exactly 2 days, so the groundtrack repeats every
other day, with spectacular coverage.  The German AMSAT folks
negotiated a launch opportunity on the first experimental flight of
the Ariane 5 rocket (we can't afford a launch like this on a proven
vehicle).

Topic #3: TCP/IP and Internet

All sorts of packet-related goodies are available online on the
Internet on the computer ucsd.edu and tomcat.gsfc.nasa.gov.  Soon a
dedicated host for AMSAT mail, amsat.org, will be installed at UCSD
in San Diego under the supervision of network guru Brian Kantor,
WB6CYT.

Phil Karn, KA9Q, added a hack to his TCP/IP software package to
permit IP packets to be encapsulated inside IP packets.  This permits
packets from one amateur radio network to be sent to another amateur
network over the Internet, without the intervening hosts needing to
know about the amateur networks.  At least seven "encap" gateways
have been installed for this purpose: Honolulu, Sydney, Richmond,
Chicago, Las Vegas, Geneva, and Ottawa.  To install such a gateway
just takes someone with both radio smarts and a good Internet
connection.  Is this amateur radio?  Well, it sure is on the ends.

The internet has proven to be a valuable resource.  Encourage people
to find a way to get on the Internet.  The Internet powers that be
have blessed this kind of operation as a legitimate use of the
Internet.

Topic #4: World's Smallest NOS Box?

A HP-95LX palmtop computer was displayed running KA9Q NOS.  It has
both RS-232 and Infrared data interfaces.

Question: What funding is needed for Phase III-D?
Answer: The total commitment is about $3 million.  The ARRL and
AMSAT-NA have committed over $1 million, the rest comes from other
national groups around the world.

Question: Is that realistic?
Answer: I hope so.  If it isn't, and we can't get industrial
sponsorships to fill the difference, the AMSAT's ability to develop
bigger and better satellites is at an end.  The amateur satellite
program puts amateur radio in the limelight as a technical pioneer.

Question: Is there time to get the satellite built?
Answer: Yes.  But commitments have to be made soon.

Question: Is there a special fund for this?
Answer: Yes, the Phase III-D Spacecraft fund.
Followup: What do I write on the check?
Answer: "AMSAT" and lots of zeroes.  You can earmark any contribution
for particular projects if you want.

Question: Are there plans for a digital transponder?
Answer: Yes, in every path.  Imagine having T1 rates or better on some
of the microwave channels.

Question: How will the satellite know which way to point?
Answer: The idea of using GPS receivers to orient the satellite is
still being investigated.  The satellite will be outside the GPS
orbit much of the time, so we need to know the antenna patterns of
the GPS satellites (which aren't advertised).

Question: What data interface do the cellular radios have?
Answer: RS-422 differential.

Question: What about duplexers for all those radios?
Answer: We may just use separate antennas.  We did get some duplexers
and so forth with the cell hardware.

=====================
Lyle Johnson, WA7GXD, and Jack Davis, WA4EJR
Hardware Projects: Trakbox, Deviation Meter, Etc.

The Trakbox is a good example of a project that involved extensive
international cooperation.  Jack Davis, WA4EJR, was involved from the
beginning, and TAPR got involved recently.

The project began in Sweden with amateurs using building block circuit
boards from Micromint.  These boards were based on the 8051 family of
microcontrollers, and were originally published as projects in Byte
magazine.  Amateurs in Japan decided to make a special-purpose board
to cut costs.  Schematics and firmware were exchanged between amateurs
in various countries using UoSAT OSCAR-14.  The resulting board is now
available as a kit from TAPR.

The board is a standalone rotor controller.  This solves the problem of
pointing the antennas at satellites, especially fast-moving satellites
in low earth orbit, while also trying to do other tasks.  The Kansas City
Tracker has been available for a while now, but it requires an IBM PC,
and takes up a slot, and requires the PC to be on during the pass.  The
Trakbox eliminates these limitations.

The Trakbox is based on an 8051 microcontroller with RAM and program
memory, a realtime clock, and a LCD display.  The user provides
Keplerian elements through an RS-232 serial port, and then controls
the operation of the Trakbox using the LCD display and front panel
controls.  The box interfaces directly to Kenpro rotators, and can be
interfaced to other brands of rotator.  The Trakbox can also control
the receive frequency to compensate for Doppler shift, using either
the computer interface or the up/down step buttons on Icom, Kenwood,
or Yaesu radios.

A portion of the price of each kit sold is donated to the Phase III-D
project.  The documentation shipping now is preliminary.  The
assembly instructions are complete, but the operating manual is a bit
primitive.  Since the Trakbox is easy to use, this isn't too bad. The
manual will be updated, and software development continues to improve
speed and add features.  Firmware updates will be free, or TAPR can
reprogram the EPROM for the usual nominal fee.

A bare board identical to the one used in the RUDAK digital transponder
was passed around.

A piece of unusual looking hardware mounted on a big round aluminum
plate was displayed, and the audience was asked to guess what it
was.  It was a TAPR development project around 1985 that didn't get a
lot of publicity.  It was a data collection and experiment control
system for a payload built by a group of high school students in
Dallas.  It was flown in a GAS (Get-Away Special) Can on the space
shuttle Discovery in 1985.  Unfortunately, delays before and after
launch proved too much for the batteries, and all the data was lost
before it could be downloaded from the experiment.

The prototype of the TAPR Deviation Meter was passed around.  A block
diagram shows a 2 meter receiver with a 10.7 MHz IF followed by a
second IF, followed by peak detectors feeding an analog-to-digital
converter.  The A/D converter is read by a tiny microcontroller,
which also runs an LED display and an RS-232 port, and can output the
measured deviation to either.  The frequency synthesizer can tune any
frequency in the 2m band.  A calibration oscillator is on-board in
the 10.7 MHz IF to permit the board to calibrate itself.  Everything
on the prototype is working, except a single buffer amplifier in the
135 MHz local oscillator.  The prototype was used to test the 9600
baud modem kit.

Once the buffer amplifier problem is fixed, we can "turn the crank"
and make a batch of the deviation meters available.  TAPR has receive
mail claiming that it's impossible to make a product like this for
less than $100, but TAPR can!

Question: Does the TrakBox emit any RF at 2m?
Answer: Maybe a little, but your outside antennas probably won't hear
it.

Question: Does the LCD on the Trakbox show the time to the next pass?
Answer: No, it shows the present time.  It does have a (slow) future
prediction mode via the serial port.

Question: How was this nice circuit board created?
Answer: Chuck Green, N0ADI, did the layout using ProCAD.  TAPR uses
circuit board fab houses in Tucson and in Orange County, CA.

Question: How fast can a new board be manufactured?
Answer: That depends on the cost.  For a standard TAPR production run
of 20 to 100 multilayer boards, it costs about $800 to $1000 per lot.

=====================
Mark Oppenheim, KD6KQ
VITA - Volunteers in Technical Assistance

VITA was founded 30 years ago by a group of scientists in Arlington,
VA, to service as an information conduit to assist developing
countries with their technical infrastructure.  A staff of about 75
people presently answers requests free of charge.  Many VITA
volunteers are also registered with their areas of expertise,
including N6ARE, WA7GXD, NK6K, HB9AQZ.  In times of disaster, like
the earthquake in Armenia, VITA coordinates donors, using their large
communications setup at the office.  HF packet and Pacsat operation
are among the supported modes.

Since 1981, VITA has been working on using packet in developing
countries.  Local telephone systems are often like two tin cans with
a bad string between them, and are unusable for data.  Thanks to the
amateur radio work in reducing the cost of packet radio, packet (on
non-amateur frequencies) can be used to bypass the phone network. VHF
packet was used to coordinate aid during the Ethiopian famine.
Demonstration networks have been operated in Somalia, Lesuthu, and
Mozambique.  An HF network is operating in Sudan.

VITA first worked with packet satellites using UoSAT-2.  The software
was too manual, and the satellite was so deaf that a large water
buffalo would be needed to power the uplink transmitter.  UoSAT
OSCAR-14 was launched with partial VITA funding.  Software delays and
complexity were problems.  VITA hired a programmer to develop
simple-to-use groundstation software for their application, and they
use Quiktrak for satellite tracking.  The next step is to launch a
dedicated satellite, VITAsat.  There may be room for some amateur
radio payload as well.  A network with up to 500 groundstations in
planned.

UoSAT OSCAR-22 was originally to be mainly used by VITA and Satellife,
a group with similar purposes but oriented toward medical
technology.  The power amplifier on UO-22's non-amateur downlink
failed, so satellite allocations between UO-14 and UO-22 were
reshuffled.  UO-22 is now 100% amateur, and UO-14 is 100% VITA and
Satellife.

Numerous travelogue slides showed people, equipment, and camels.

Question: Are the locals generally able to operate the equipment?
Answer: Sometimes. Luckily, the equipment has been quite reliable so
far, so no maintenance has been required.

Question: How is all this funded?
Answer: The equipment is purchased by the "customer".  Some money is
also available through grants.  VITA does not charge for access to
the satellite.

Question: Do you have difficulties with importing technology?
Answer: Sometimes.

=====================
Masa Sawada, JF2GPF
TASCO

TASCO was founded by JA2AQO, the current president, in 1979.  TASCO
began by selling weather facsimile terminals and RTTY decoder.  In
the mid 1980's, TASCO began selling amateur packet radio into the
Japanese market under license from TAPR.  They are constantly working
on new technologies, and all technical staffers are amateur radio
operators.  Congratulations to TAPR on its 10th Anniversary.

=====================
Bill Henry
Clover II

Several articles about Clover II have appeared, and more are on the
way.  The last two Proceedings of the ARRL Computer Networking
Conference had papers by Ray Petit, W7GHM: 1990 about Clover I, and
1991 about Clover II.  A July 1990 QEX article featured Clover I.
Several articles have appeared in The RTTY Journal.  A series in
Communications Quarterly and another in QST will lead up to Clover II.

Clover is still experimental - it isn't a product yet.  It was
invented by Ray Petit a year and a half ago.  It was an outgrowth of
earlier coherent CW experiments, which used phase coherent detection
with very stable oscillators (a few parts in 1e8) and detection
bandwidths as narrow as 0.01 Hz.  Such equipment can demodulate
signals that can't be heard by ear.  The early Clover articles in QEX
caught Bill Henry's attention, and they teamed up to make a product.

The particulars of Clover II are derived to counteract what the
ionosphere does to corrupt a data signal.  One basic limitation of HF
propagation is signal to noise ratio.  Earlier RTTY designs have
concentrated mainly on this problem, by optimizing bandwidth and
using adaptive AGC or good wideband limiters.

Another difficulty with HF is the existence of multiple propagation
paths of different lengths.  The differential delays result in
selective fading.  On voice, this is merely annoying, but on data it
is a disaster.  Out of phase signals cancel.  On FSK, the mark and
space tones fade separately (a good RTTY demodulator handles this).
What's worse, the bits can get smeared out in time by up to several
milliseconds.  This is what's wrong with HF packet: 300 baud is too
fast.  Most experts agree that about 100 to 150 baud is the limit for
usual HF conditions.  Now and then, conditions are "like a wire" and
there's no problem with multipath.  Such conditions are quite rare -
and that's the only time HF packet works!  AMTOR users can tell you
that sometimes even 100 baud is too fast.

Clover copes with this problem by using a signaling rate of 31.25
baud.  That's pretty slow.  To get a reasonable data rate with such a
slow baud rate, Clover packs more than one bit in each pulse.  Clover
uses PSK with two, four, eight, or sixteen distinct phases to encode
1 to 4 bits of information on each pulse.  Clover then uses this
scheme in a time-staggered scheme with four different carrier tones,
resulting in a total bandwidth of 500 Hz, which is a good match for
available CW filters.  When conditions are good, Clover goes further
by adding two-level or four-level amplitude modulation, for even
higher maximum data rate, without changing the basic modulation rate
of 31.25 Hz.

PSK modulation can be a problem, because its bandwidth is usually
very wide.  Clover avoids this problem by pulsing the four tones on
and off with a very carefully-chosen pulse shape called a
Dolph-Chebychev function, and performing the phase changes while the
pulse is completely off.  The result is that the energy of a
four-tone Clover signal is very tightly contained within 500 Hz. With
a 60 dB limit imposed by the quantization in the digital-to-analog
converter, the Clover transmitter's sidebands are down 50 dB outside
the 500 Hz.  Two Clover signals can be spaced just 500 Hz apart (edge
to edge) with 55 dB rejection.

Graphs of the spectra of Clover, HF packet, and AMTOR show that Clover
is a *lot* tighter.  The standard rule of thumb says that two AMTOR
signals need to be 1 kHz apart, and HF packet signals need 2 kHz. The
spectra clearly show why: poor sideband suppression.  Clover is
effectively much narrower.  Not only that, but Clover is faster on
real channels.  Ignoring the question of 5-bit characters versus
8-bit characters for now, both HF packet and AMTOR have typical
real-world throughputs of 5 to 7 characters per second.  In tests on
the air, Clover typically achieves 50 to 80 characters per second.

Another weakness of HF packet is the error control scheme used.  With
packet, a long frame of up to 30 seconds is sent, and every single bit
in the frame must be received correctly, or it is discarded.  Because
of this limitation, HF packet operators must run small packet sizes
of 32 or 64 characters.  This makes the packets short enough to get
through (sometimes), but increases the cost of packet headers and
waiting between packets.  That's how a 300 baud mode gets down to 5
or 7 characters per second.  Clover uses a forward error correction
technique called Reed-Solomon coding.  This technique transmits a few
extra bits, and uses the carefully encoded redundancy in the data to
correct the received data without requiring a retransmission.  For
example, a R-S code that is 60% efficient can correct 25 errors in a
block of 255 bits.  Because the R-S code can correct some errors in
each frame, Clover is able to send longer frames without losing too
many to errors.  Of course, sometimes long frames just don't get
through due to fading conditions; in this case Clover can fall back
to shorter frame lengths.

As W3IWI has pointed out, HF calls for adaptive modems. Clover is
about as adaptive as you could want.  There are 8 basic modulation
modes to choose from (different numbers of phases and amplitudes for
each pulse), times 4 frame lengths, plus 4 different Reed-Solomon
codes of varying efficiency and error-correction capability, for a
total of 128 different modulations.  Every one of those 128
modulations has the same 500 Hz spectrum.  The Clover modem also
controls the transmitter output power.  Obviously, mode and power
selection has to be automatic!

The receiver measures the phase, time, and frequency dispersion of
the received signal and picks a mode.  It sends an order to the
transmitter specifying which mode it wants.  It can change modes
within a second if a short block length is in use.  The modes range
from 2.3 characters per second to 94 characters per second,
theoretical throughput.  The field tests have shown a typical range
of from 28 to 62 characters per second.  Note that the receiver
doesn't just move up to higher speeds when conditions are good and
down to lower speeds when conditions are bad.  It can figure out by
listening to the signal what exactly is wrong with the signal and
request the mode that best fits current conditions.  For instance, if
it notes that phase dispersion is bad, it can fall back to a mode
with fewer phases.  If it notes that it has excess signal to noise
ratio, it can command the other station to reduce power.  (This can
lead to the rather disconcerting situation where the transmitter's
meters are not moving, and the receiver's audio has no audible tones,
yet characters are still moving through the link!)

The implementation is DSP, DSP, DSP.  The input jack goes to a
transformer and a 16-bit A/D converter, and the rest is digital.  The
A/D converter is a 16-bit sigma-delta oversampling converter like the
ones used in digital audio applications.  It doesn't need any
anti-aliasing filter, and it has lots of dynamic range.  It currently
costs $20, but should get cheaper.  The transmit audio is also a 16
bit oversampling audio component, followed by a simple filter to get
rid of the residual 100th harmonic.  A Motorola 56001 DSP processor
supplies the signal processing horsepower.  The original design used
a 6809 microprocessor for general control functions, but it ran out
of gas.  The current prototype now uses a 68EC000 processor at about
30% utilization.  The board contains only bootstrap ROMs; the Clover
code is downloaded from the PC.

Vic Poor is writing a Clover driver for Aplink.  The Clover board has
FIFOs on the input and output to relax realtime requirements, which
is expected to be especially helpful for PCs running Windows.

Two working prototype Clover boards will be displayed at Dayton. On
initial release, the card will do only Clover.  If anybody wants to
write other modems for the board, the door is open.  A more pressing
need is a new protocol suitable for HF work.  Anybody who writes
network code and wants to write drivers for Clover, we can set you up
with hardware and provide assistance.  The command protocol will be
defined by next week.

Question: Isn't frequency accuracy and stability still a problem?
Answer: With the faster CPU, the DSP no longer has to do Reed-Solomon
decoding.  That means it has enough spare horsepower to do more
frequency acquisition and tracking.  Clover can now handle frequency
errors of up to 40 Hz, which is similar to the guidelines for HF
packet.

Question: What about intermodulation distortion in the transmitter?
Answer: Measurements of high-end rigs show very good IMD.  A worse
problem is broadband noise, which is about 40 dB down.  That doesn't
seem to hurt, either.

Question: What's the price?
Answer: The introductory price will be $995.  This is a lower price
than announced before, because it's now a PC plugin board rather than
a box.  The parts are expensive, even the socket for the DSP chip is
expensive.  The board is 4 layer.

Question: What is the peak to average power ratio?
Answer: 3 dB for all modes.

Question: Is it legal?
Answer: Yes.  It's not multiplex because the four tones are sent
serially rather than in parallel.  The emission designator is
500HJ2DEN.  The Chief Engineer and Chief of Enforcement of the FCC
have both agreed verbally that this modulation is legal.

=====================
Gwynn Ready, W1BEL
PacComm Topics

PacComm is currently working on a custom packet protocol for
commercial HF applications.  It runs long frames, but avoids the
problems of AX.25.  It uses a "superframe" containing multiple copies
of the address and multiple checksums.  It uses a selective nack
protocol so receiving stations need only nack missing pieces of each
superframe.

Topic #1: PACTOR

PacComm is exclusive licensee of PACTOR in the USA (for a few years),
and point of contact for PACTOR.  PACTOR is an ARQ protocol more like
AMTOR than like packet.  The PACTOR controller will also do RTTY and
AMTOR, and automatically falls back when talking to a non-PACTOR
station.  It stores partially-received frames in memory and tries to
combine them to get one good frame.  This technique is called "memory
ARQ".  The box is binary compatible with the original German
version.  Hardware is expensive to build in Germany, so PacComm is
building PACTOR controllers in the USA.  An Aplink driver is on the
way.

Topic #2: Baycom

Baycom (pronounced Bye-Comm) is a packet program for the IBM PC like
DIGICOM>64 for the Commodore 64, from the same German team.  They
have now formed a company, and have licensed PacComm
(non-exclusively) to distribute in the USA (and to work to enforce
their copyright in the USA).  Two modems are to be available: a
serial port modem based on the TCM3105, powered from the port and
physically inside the port connector, and a version based on the
AM7911 in a box with HF capability.  The PC-100 series will also be
upgraded for 4-port Baycom compatibility, 300/1200/9600 baud, with
modem disconnect and mounts for Tekk radios.  These products may be
available by Dayton.

Topic #3: NB-96 Product Line

The NB-96 line is a licensed version of the G3RUH 9600 baud modem.
The weaknesses of the original design for full duplex use have been
fixed.  The board has more groundplane and more bypass capacitors.
Receive and transmit circuits have been separated.  The components
that determine the modem speed are now on a header, so you can pick
whatever baud rate your radios can support.

The EM-NB96 external version of the NB-96 has been out of production,
but now it's back in production.  It is intended as a stopgap until
DSP modems become available and affordable, or for the operator who
already has most of the modems he expects to need and just needs to
add 9600 baud.  It is versatile: the user can switch to internal or
external clocks, route the push-to-talk, and choose the TNC's
built-in 1200 baud modem or either of two external modems.  An LED
indicates when you have the settings wrong -- this tends to cut down
on user support calls.

An integrated packet radio system containing the modem, the TNC, and
the RF circuitry all in a single box is in prototyping now.  The
original plan was to mount the radio on the circuit board, but
interference prevented that from working.  So the radio is now
mounted in the chassis, with the side benefit that there is now room
for two separate radios.

Topic #4: TNC Upgrades

When running 9600 baud continuously (like on a satellite downlink),
the TNC needs to support a terminal baud rate greater than 9600.  The
TINY-2 can run 19200 baud continuous, and it's the only TNC that can
do that out of the box.  TINY-2's are now shipping with 10 MHz Z80
microprocessors, which enables them to run at 38400 baud on the
terminal port.

A high speed data controller based on the V53 processor with an 85230
serial chip and a fancy power supply on-board is being marketed
mainly to commercial customers -- it's too expensive for the ham
market.  A PacComm HandiPacket is in use on the Russian space station
MIR, which has generated lots of publicity.  They are also used in
the ground terminals used with the DARPA Microsat project (not to be
confused with AMSAT's more sophisticated Microsats).

A 1200 baud modem about the size of your thumb is shipping 100 to 200
units a month to commercial customers, who are now asking for a tiny
9600 baud modem.

=====================

Pete Eaton made a few announcements, then the meeting broke for the day.
It reconvened at 9:00 AM on Sunday.

=====================
Fred Treasure, KE5CI
A Packet-Controlled Telescope

In 1987, Bill Neely, KC5ZG, purchased a large telescope mirror, with
plans to automate the pointing arrangements.  Since then, the system
has grown a little ...

Block diagram of the observatory system.  A control computer in town
communicates via a 9600 baud packet link, through a dedicated
repeater site, through another 9600 baud packet link to the
communication and storage computer in the "warm room" at the
observatory ranch.  That computer talks via a 2.5 Mbps ARCnet link to
the control computer, which in turn talks via 1200 baud asynchronous
RS-232 to the computer that controls the pointing.  Whew!

The NF/ Observatory tracks asteroids and other objects of interest,
capturing images using a CCD camera on the 17.5 inch telescope.
Congress has generated an initiative to shoot down an asteroid rather
than let it collide with Earth.  This observatory has volunteered to
gather data on near-earth orbit asteroids in support of this project.

The name "NF/" stands for "Neely Fraska Bar", the name of the ranch
on which the observatory is located.  The ranch is about 25 miles
from Silver City, NM, at an elevation of 5800 feet.  The packet
repeater is on Baldec Peak.

The 9600 baud links run on Mocom 70 radios with Texnet 9600 baud
modem.  The Texnet network control processor (NCP) has been modified
to run KISS with 2048-byte packets for TCP/IP.  The Silver City to
Baldec link is on 442.5 MHz, line of sight, with 4-element yagis on
each end of the link.  The link from Baldec to the ranch is on 447.5
MHz.  There's no line of sight path on this link, so there's a
16-element yagi on the tower on the mountain and an 11 foot dish with
21 or 22 dB gain at the ranch.

The system can transfer a 512x512 pixel image with 4 bits per pixel
(about a half megabyte of data) in about 10 minutes.  More typically,
it transfers a 320x200x8 VGA image in about 1 minute.  This kind of
image is suitable for previewing the results.  The full results are
stored at the site, and transported using magnetic tape.

The system includes three 80286-based computers and one XT-class
machine.  A request for a series of images is entered into the
computer in Silver City.  It goes through the links to the ranch
house, which turns on the telescope and the rest of the computers.
Lots of computers made more sense for this job than a larger
multitasking machine, because realtime control was needed.

The XT tracking computer can automatically calibrate the positioners,
starting from the weatherproof "parked" position.  Fixed sensors on
the telescope mount can get it pointed close enough to find a
calibration star.  Then the telescope automatically locks onto the
bright object near the center of the field.  Telemetry about the
weather and control voltages go back over the link to the in-town
computer.

The second computer provides the timing signals to the CCD sensor and
stores the resulting image on its hard disk.  Best performance from
the CCD is obtained by tuning the precise voltages applied to it.
This telescope achieves about 25 electrons of noise.  Compare this to
the best NASA sensors at 3 electrons of noise; pretty good for an
amateur effort.  About 60 images are stored on the hard disk drive,
then they are dumped in a batch to a magnetic tape drive attached to
the computer in the ranch house.  About once a week, the data is
fetched by car.  Once the tape is verified, the hard disk is erased.
The telescope dome automatically closes, the telescope parks, and
powers off.  The entire process is completely automated; no human
intervention is not required.  The telescope system has failsafe
interlocks: if moisture is detected, or if the link fails, the system
automatically parks and shuts down.

The system cost about $5000, plus about 5000 man-hours, mostly spent
on the programming.  Bill Neely, KC6ZG, and Lori Neely did most of
the programming; Fred Treasure and Barbara Treasure, N5HJN, provided
most of the hardware.  K2GNR and the Jet Propulsion Laboratory
provided the camera chips.  NASA provided the magnetic tape recorder.

Question: Are there other sites sharing the data format?
Answer: The data is recorded in a standard format.  However, there is
only one other telescope that's as completely radio-controlled as
this one: the Hubble Space Telescope!  The team gets invited to
professional astronomy events.  The professionals couldn't do a
project like this for $100,000.

Question: Is compression used on the links?
Answer: We tried LZW compression, and it gave only about 25%
compression on this type of data.  That's not worth the extra
trouble, since there is plenty of capacity on the links.

Question: What polarization are the links using?  My experience is
that vertical doesn't work in the mountains.
Answer: We chose horizontal, more or less by guess.  It works.

Question: What weather sensors are installed?
Answer: Anemometer, a cloud sensor that works by sensing the
temperature of the sky versus the ground, a moisture-detection grid,
and thermistors to measure temperature.

Question: What kind of cooling is used?
Answer: A three-state thermoelectric cooler and liquid cooling.

Question: Tracking rates?
Answer: Completely variable.

=====================
Lyle Johnson, WA7GXD
Bit Regenerative Full-Duplex Repeaters

Why full duplex?  It helps to cure the problems caused by hidden
terminals.  The coverage area can be controlled using antenna
patterns, or to a lesser extent by varying the repeater transmit
power.  The Tucson LAN has a large radius of 150 miles, and the full
duplex repeater helps throughput.  It helps stations at the edges of
the LAN to communicate.

With bit regeneration, the repeater demodulates incoming signals and
remodulates them before retransmission.  This allows the repeater to
control the deviation of the transmitted signal.  By using a bit
regenerator with some FIFO buffering (like the one on the new TAPR
9600 baud modem) to remove clock jitter and clock rate errors, the
transmitted signal can be perfect.  Most TNCs are crystal controlled,
so not much buffer is needed.

The bit regen repeater has some social impact as well: the repeater
operators get some control over the type of traffic that is
transmitted on the channel.  A station that habitually hogs the
channel can be throttled by the node TNC by simply generating a
glitch in the middle of each packet they transmit.  Technical
compliance can also be ensured.

Question: What are the pros and cons of checking the CRC before
regenerating the bits?
Answer: If the TNC waits for the entire frame to arrive so the CRC
can be checked before transmitting, it ends up acting just like a
digipeater.  That cuts the throughput by a factor of at least two.
The "lid filter" function mentioned above doesn't require checking
the CRC: if the node TNC glitches the wrong frame because of bit
errors in the received frame, the frame wouldn't have been accepted
at the receiving end anyway.  The other aspect is the need for the
output to come on the air quickly as soon as a signal is detected on
the input.  This is required so other users won't start transmitting
and colliding with the first input.  With the FIFO, the repeater only
sends a few bits of garbage data as a busy tone before the real bits
start to come out.

Question: Are the bit regen kits available now?
Answer: Yes.  It's just a couple of parts.  The bit regenerator could
be used on a 1200 baud regenerative repeater, too.

With a central repeater, the users can use directional antennas and
get better performance.  The single-point reliability is about the
same as a single central digipeater.  The digipeater at a good site
probably needs cavities anyway, so the additional cost of going full
duplex is minimal.

=====================
Mike Curtis, WD6EHR
Experiences with 9600 Baud

Working with many other contributors, Mike has created a beginner's
handbook for 9600 baud operation.  9600 isn't as hard as people think
it is.  The only big problem is getting hooked up to the radio.  On
transmit, the radio needs fairly linear response from 100 Hz to 5
kHz, low phase distortion, and a true FM modulator.  On receive, it
needs good linearity and [I missed something here -Ed.].

There are a few radios capable of doing 9600 baud out of the box.
The Tekk KS900L, the D4-10, and the Ramsey kits are all good.  The
Kantronics DVR2-2 works but isn't recommended where other strong
signals are a problem.  Multimode radios are generally good prospects
for conversion, because they use a separate circuit for FM.

Question: What about removing extra filtering from the discriminator?
Answer: Whatever works.

Question: What test bench setup do you use for bit error rate
measurements?
Answer: The G3RUH modem has a bit error detector, which generates a
click for each error.  Just hook up an audio amplifier and tune for
minimum clicking.

Question: What parameters do you recommend for 9600 baud links?
Answer: TXDELAY as low as 3 usually works, but 6 to 10 is better.
TXTAIL about 2.  FRACK (the T1 timer) at 4 or 5. These parameters
should be slightly more aggressive than at 1200 baud.

Question: Can 9600 be used co-channel with 1200?
Answer: Yes.  In fact, we have a repeater that is dual mode
9600/1200.  It seems to work, though there isn't much 9600 activity.

Question: Is 9600 used on 2m in Los Angeles?
Answer: Yes, on 2m and on 70cm.

Question: What about the Motorola Mostar 800 MHz radio?
Answer: Never heard of it.

Question: Thanks for writing the beginner's notes!
Answer: It was fun, and I had lots of help.  I hope 9600 baud will
take over as people try it and find it so much better that they want
to abandon 1200 baud.

=====================
[ sorry, I didn't catch this ad-hoc speaker's name or call.  -Ed.]
Experience with Mitrek Radios at 9600

We obtained some surplus Mitreks.  A technician from Motorola didn't
think they'd work on 9600.  They do, with both the G3RUH modem and
the K9NG modem.  We intended to support an application for appliance
operators, so we wrote a paper that describes the necessary
modifications in detail.  The Mitrek is fairly easy to modify.
Several people have successfully followed the paper.

It's important to use shielded wire like RG-174.  The original
instructions called for running the wires through a hole in the case,
but we found it possible to run the wires through the existing I/O
connector, which makes the modified radio much more serviceable.
About 10 are in use, with paths ranging from a few miles to 40 miles,
vertical and horizontal antennas.  When two of the radios are stacked
on top of each other, some IF crosstalk causes interference between
the two radios.

It's important to get 15 to 20 dB margin on the links.  We did bit
error rate tests, even though that isn't necessarily the best
indicator.  We use Comet and Diamond dualband antennas for
back-to-back VHF/UHF nodes.  Horizontal polarization with beam
antennas seems to work better.

The biggest problem with the Mitrek is bandwidth.  We replace the
4-pole filter in the front end with a wire, which is a little touchy
at 10.7 MHz.  We have avoided preamps because of problems with
turnaround time.  We run TXDELAY of 3 to 5.  We've tried TCP/IP, but
we usually run NET/ROM.  We have a Gracilis packet switch, but it's
not fully installed and working.


Question: Can you really run TXDELAY at 3 to 5 with K9NG modems?
Answer: Yes, even with the Mitrek's antenna relay.

Question: Did you modify the DCD circuit in the K9NG modem?
Answer: No.

Question: Did you make any modifications to the K9NG modem at all?
Answer: Some have done some work on the filters and DCD.

Question: Did you notice any performance loss when the front end filter
was removed from the Mitrek?
Answer: No.  We have three sites that are heavily infested with other
transmitters, and haven't had any real problems.

Question: How many hops work in the network?
Answer: 3, 4, 5.  The link to Chicago works, even though it's not 9600
all the way.

Question: How much difference did the custom EPROM in the G3RUH modem make?
Answer: We didn't make one.  The default modem is good enough.

Question: Do the Mitreks also have a low IF?
Answer: No.

=====================
Eric Gustafson, N7CL
More Experiences with 9600 Baud

The local oscillator the Mitrek generates about 1.5 watts of power,
which is applied to the mixer.  If you follow the Motorola
instructions for tuning the front end filters, you get a strong LO
signal everywhere, especially on the feedlines.  This signal will
leak into any other Mitrek, in both directions.  We had this problem
in a commercial installation in a mine in Nevada, resulting in
crosstalk between vehicles when they were close together.  We were
able to realign the front end filters, and the extra attenuation
helps substantially.


We use Tekk radios for 1200 bps telemetry.  Their front ends can't
take other strong signals.  We have found them reasonably useable;
others have had more luck.

Out of 10 radios, one was 15 kHz off frequency -- with 15 kHz
bandwidth filters!  Filters were misaligned, resulting in over 20 dB
of loss.  The Tekk radios perform very poorly under thermal cycling.
The transmitters perform pretty well, probably because they have to
pass FCC tests, but the receivers don't.  We went to crystal
manufacturer Hy-Q, sort of a Filters-R-Us house.  They provided
filters and crystals that cured the problems with the Tekk filters
and crystals, including frequency jumps and frequency errors.  So now
we order the Tekk radios without quartz, and get a cheaper price.

For many other radios, we found the filters weren't really quite wide
enough for 9600.  For a given radio, you can find an optimum FSK
shift but low signal to noise ratios.  Often, you have a choice
between high error rate with the best S/N performance or low error
rates but poor S/N performance.

Murata-Erie is now making 450 kHz ceramic filters specifically
designed for data transmission, in the standard package.  Two
bandwidths are available, about right for 9600 and 19200 baud
respectively.  So now the 1st IF in many radios can be opened up by
just replacing the filter with the right bandwidth unit.  The
resulting eye patterns look just like the transmitted signal.

Question: What's a Tekk radio?
Answer: A very small crystal-controlled transceiver that's both cheap
and inexpensive.  They are available for amateur frequencies as well
as the commercial units we use.  The transmitter tuning is very
broad.  Unfortunately, so is the receiver tuning.

Question: What application did Tekk originally intend for these radios?
Answer: Not sure.  They are advertised for telemetry applications.

Question: Do you know of a source for manuals for a 2m Mitrek radio?
Answer: If all Mitrek manuals are the same, I have one.

Question: There is a radio by Maxim (sp?) that's similar to the Tekk.
And the WA4ONG commercial product uses a Motorola Radius board set.
Do you have any experience with these?
Answer: We tried the Radius, and it worked well except for the
Motorola local oscillator problem.  It uses a MMIC mixer with no RF
stage, with the result that every harmonic of the LO out to at least
2 GHz is coming out strong.  They also have a spurious response 900
kHz away from the primary response (on 8 out of 8 units).

Question: Maxars?
Answer: Not tried.  We have used Maxtors successfully.

Question: Kantronics D4-10?
Answer: Not tried.  The Kantronics 2m version is just wretched.

Question: What is the part number of the filter you use in the Tekk
radio?
Answer: Custom part number Hy-Q QMF21MB30 roofing filter.  This mod
requires other component changes too.  The 450 filter is SFG or SFH
line.  The SFH450F works at 19200 or 9600 without compensation.

Question: What bandwidth?
Answer: +/- 12 kHz, which is a bit wide for 9600, but that allows for
some frequency error.

Question: So you would need a channel wider than 25 kHz?
Answer: It wouldn't interfere with, but might get some interference
from, adjacent channels at 25 kHz.  With the G filter, you get +/- 9
kHz, which is about right for 9600.  The skirts are a bit high, but
that's how they get flat group delay for data.

Question: What's a source for Murata filters is small quantities?
Answer: Maybe TAPR should provide this service.

Question: If we use a 455 kHz model instead of a 450 kHz model, it
would work in other radios than the Tekk.
Answer: So would the 450 kHz model, but it would be more painful.

Question: Was the 2nd LO frequency changed because of stability
problems?
Answer: The stock crystals were just on the wrong frequency.

Question: How much power does the Tekk radio generate?
Answer: Just under 2 watts.

Question: What quantity do you order to get prices of $40 to $50
each?
Answer: 8's.

Question: Any experience with Repco or Ramsey radios?
Answer: Not with Repco.  With Ramsey, my only experience is with
their test equipment, which was junk.
WA7GXD: Built a Ramsey kit.  The Ramsey people were cooperative about
filling shortages in the received kit.  Got it working for about an
hour, and then it died.  Haven't had a chance to work on it since.
KE3Z: Has a Ramsey radio working the ARRL Lab.  The coil in the
transmitter was way off, but could be tweaked.  The spectral purity
didn't meet the FCC requirements; it wasn't even close.  Ramsey has
said there is a new model coming out, and not to bother publishing a
review of the current model.  A user could provide postfiltering to
meet the rules.  The receive side wasn't great either.
N7CL: I prefer receivers that have a double-balanced mixer in the
front end.
Somebody: I just built a new 220 MHz Ramsey kit, and it's hot.  I have
it transmitting on 9600, still working on receive.
WA7GXD: Most radios work fine on frequency -- the question is how well
they work where they aren't supposed to work.

Question: What about the Alinco data radio?
W1BEL: It doesn't pass direct FM signal without modification.
W3IWI: I've heard that the only difference is the front panel and the
fact that it doesn't come with a microphone.
Somebody: The manual on my Alinco is the worst I've seen.
N7CL: The manual always comes last.

=====================

Pete Eaton said thanks to everybody who came.  If you're going to be
at Dayton, stop by the booth and say hi.

He then closed the TAPR annual meeting at 10:40 AM.

[Report Downloaded from UoSAT-OSCAR-22 by DB2OS]