We'd like to tell you about some inexpensive antenna,
radio, and computer interface hardware which allows communication of digital
data at rates up to 2 megabaud (1 megabaud = 1 million bits per second)
on an Amateur Radio band. In addition to the data link, an analog voice
channel is provided. It requires only an external microphone and speaker
for simultaneous full duplex audio communication. The link operates in
the 10-GHz Amateur band and uses an inexpensive commercial parabolic antenna
along with a Doppler radar transceiver module to provide medium range communications
at low cost. We'll discuss modifications to surplus networking interface
cards that let you use this high speed data in Amateur Radio service with
IBM-style personal computers.

The Amateur accustomed to conventional AX.25 packet
operation might wonder why anyone would want to go to the trouble of building
a digital radio approximately 1000 times as fast as the 1200-baud systems
prevalent on the VHF bands. Although many metropolitan areas are experiencing
severe congestion on some of the packet channels, it's also true that many
keyboard-to-keyboard QSOs are taking place. A great deal of traffic is
also being handled on the worldwide bulletin board systems using today's
equipment. The success of AX.25 packet radio has suggested the need for
faster systems to improve current performance and has spawned some fundamentally
new ideas for Amateur Radio.

A whole spectrum of new user applications and
the possibility of a nationwide or even worldwide digital Amateur network
are two major areas made possible by faster hardware.

New applications

Amateurs will also be able to share resources.
A station with an interesting database will be able to make it available
to others. On-line call directories, QSL information, and technical data
-- not to mention computer programs and even the computers themselves --
can be shared. It's possible for one Amateur to actually run programs and
applications on someone else's computer as though it were located in his
own shack. Remote control of equipment and remote sensing are other possibilities.
Remote digital control of repeaters or even complete stations, including
audio or video uplinks and downlinks, can be supported. Conventional voice
repeaters (analog) may be replaced by digital hardware for completely digital
round tables. Since this data can be transmitted anywhere the network permits,
there can be multistate, national, or even worldwide voice nets. If the
data rate permits, all of these different applications could conceivably
be going on at the same time.

An Amateur Radio network

Goals

This was necessary to support some of the applications
previously described. It was also built to help advance Amateur use of
the microwave spectrum.

Fortunately, microwaves and high speed communication
fit together very well. In fact, if the data rate is increased significantly
it is absolutely necessary that wider and wider bands be used. As frequency
is increased, antennas of reasonable physical size are better able to focus
the transmitted beam without wasting signal in different directions. The
Amateur microwave bands, through 24 GHz, offer the best available performance
and cost for such communication. In order to be widely useful our link
needed several attributes:

• 
  To be inexpensive -- competitive with present TNC/radio
  combinations

Problems

At these speeds the data is too fast for normal
serial ports on most computers, for the internal bus operations of many
computers, and for TNCs. Similarly, the software to process data at these
speeds can no longer operate on a character by-character basis. Any solution
we developed for these problems also needed to work with commonly available
hardware, most notably the IBM PC and its clones.

Microwave hardware, propagation, and high speed
data are new ground for many Amateurs. This means that any high speed link
hardware needs to be relatively easy to work with.

What we built

The two ends of a link operate "split." One transceiver
oscillator typically operates on 10,450 MHz while the other end is 105
MHz lower, on 10,345 MHz. The difference between the two transmitter frequencies
corresponds to the receiver first IF frequency. The receiver first IF on
each end is generated when the remotely transmitted signal (frequency modulated
by the data to be transmitted) is mixed with the local transmitter. Each
end uses its own transmitter as a receiver local oscillator, and each unit
transmits continuously. Therefore, each receiver sees the same IF. This
is the same full duplex arrangement used for many years by Amateur microwave
enthusiasts. The transmitters run 5 to 10 mW of output power. The transmitter
is frequency modulated as its bias supply is varied and the frequency voltage
dependency of the transceivers is used for tuning. This same technique
was used previously to phase lock such oscillators.1 The 2-foot dish shown
here has a gain of about 33 d8, or 2000 times at 10.5 GHz. When driven
by the microwave transceiver, the effective radiated power (ERP) is about
the same as that of a 10-watt 2-meter radio driving a quarter-wave whip.
We selected 105 MHz for the receiver first IF, with provision for tuning
+/- 10 MHz to accommodate differential frequency drift with time or temperature
of the free-running microwave transceivers. Using an IF in this range also
lets you do some simple troubleshooting and testing with commonly available
commercial FM broadcast receivers. No correction is necessary if both ends
drift in the same direction because the IF doesn't change. Automatic Frequency
Control (AFC), implemented by tuning the second LO nominally at 150 MHz,
is provided to keep the receiver tuned correctly. This conversion produces
the second IF at the point where detection takes place at 45 MHz in a Motorola
MC13055 FSK receiver chip. This chip is specified to operate at data rates
up to 2 megabytes per second (Mbps), but has actually been used as high
as 10 Mbps.

Automatic frequency control circuits keep each
receiver correctly tuned, even when the first IF deviates from 105 MHz.
A search oscillator is also provided to allow the receiver to "find" the
incoming signal when the link is first powered up, or if you lose signals
temporarily. The searching is controlled by the Data Carrier Detect (DCD)
circuitry. Once the signal is found, the oscillator is shut off and the
AFC tunes the receiver correctly. Because the data is digital, an appropriate
offset is introduced to the tuning depending on whether the data is a "0"
or a "1" We added the audio channel as an afterthought. It provides for
human communication, particularly while debugging the link and operating
it with digital data the first time. An electret microphone produces the
transmit audio signal. This is amplified and limited by high and low pass
filters before modulating the transmitter. Levels were selected to provide
only small deviation compared with that of the digital channel. This allows
the audio channel to operate without significantly interfering or degrading
the digital data. A volume control and speaker amplifier sufficient for
driving headphones or a small speaker are provided on receive.

An analog meter displays strength as carrier-to-noise
(C/N) ratio or discriminator output to aid when manual frequency control
(MFC) is used. The MC13055 FSK receiver hip has a built in logarithmic
amplifier which can give a pretty accurate measure of C/N. A switch lets
you select manual or automatic frequency control.

The microwave transceiver, modulator, and receive
Preamplifier are mounted together in a box located at the Prime focus of
the parabolic antenna. A horn antenna was designed to illuminate the dish
antenna efficiently so that ear maximum gain could be obtained. The rest
of the receiver, as well as circuits for the audio channel, are located
,a separate enclosure. This lets you place the antenna and microwave hardware
a considerable distance from the receiver for tower or mast mounting.

We used Emitter Coupled Logic (ECL) for incoming
and outgoing data. These are differential lines and can be used even when
there is considerable line length -- for example, when the microwave portion
is high on a tower or the receiver is located some distance from the host
computer. A standard 15-pin connector is used as the interface to the radio
hardware. Those familiar with Local Area Networks (LANs) may recognize
this connector and pinout as identical to that of a Media Access Unit (MAU),
the device used to link a computer to a coaxial cable connecting a building
or area wide network. You need just 12 volts DC at approximately 350- mA
data input and data output to operate all the radio link hardware.

Antenna

The horn feed and microwave assembly attach to
a plate and are held at the dish focal point by four 1/4-inch diameter
struts made from soft aluminum rod. We found this material at a local home
supply store. First cut the aluminum rod to length; then drill and tap
it at each end. Use a tubing bender to shape the rod properly. Figures
2A, 2B and Figure
3 show the mounting bracket and feed support struts, respectively.

Feedhorn

The feed/microwave assembly is shown in Photo
A. The mounting plate has a short section of waveguide at its center.
The feedhorn and transceiver mount on opposite sides and a small Bud box
encloses the electronics. Make the waveguide section by milling (or drilling)
and fitting a 0.40 by 0.90-inch rectangle in the plate. Great precision
of construction for either the feed or the mounting plate isn't necessary
to obtain good performance. Figure 5 shows
the feed mounting plate.

Microwave assembly

Receiver assembly

Computer interface

The original system design called for standard
off-the shelf EthernetTM adapter cards. Normally Ethernet lets a number
of computers intercommunicate within a local area at a data rate of 10
Mbps. N3EUA suggested we use the same widely available cards and the associated
IEEE 802.3 protocol with the adapter card clocks slowed to 1 Mbps. Using
a standard Ethernet adapter gives you access to an existing range of networking
software, including NetBios/PC Network and TCP/IP implementations. It's
an interface familiar to the general ham community, one that the advanced
packeteer probably already knows and works with.

Stock Ethernet cards can't be slowed from 10 to
1 Mbps without extensive reworking. The serial data interface portion of
the cards must produce a clock rate within 0.001 percent of 10 Mbps to
conform to the IEEE 802.3 specification. By the time we discovered this
we had working "RF bit pumps:' but found ourselves without a suitable digital
interface.

Several years ago IBM produced a digital communications
adapter known as the PCLANA or SYTEK 6120. This adapter was designed to
communicate using FSK signals transmitted on a coaxial cable at a 1-Mbps
rate. The card implemented Ethernet framing with a NetBios interface directly
on the card. The card consists of a local ľP (80186), an Ethernet chip
(82586), custom 802.3 serial data interface, RAM, ROM, PC bus interface,
gnd an RF modem. While the RF modem is interesting, it was unnecessary
for this project and was disconnected.

We discovered that this card had all the right
pieces: 1-Mbps Ethernet frames, defined PC interface, and because 10-Mbps
Ethernet was displacing these cards, good availability on the surplus market
-- sometimes just for the asking.

To use the PCLANA card, we needed to gain access
to the TTL signals directly (before modulating the RF modem), build an
adapter card onto the PCLANA for converting TTL to differential ECL, generate
DCD, and route Transmit Data (Txd) and Receive Data (RxD). You'll need
a software driver if you want something other than a NetBios interface.
A schematic of the card is shown in the IBM PC Technical Reference Manual,
"Options and Adapters" section.

How the adapter card works

The microwave modem interface closely matches
the IEEE 802.3 MAU which normally connects the digital interface to a coaxial
cable. A MAU is also known as an Ethernet transceiver (XCVR). We chose
this interface because it allows long (120 foot) cable runs, good common
mode noise immunity (0.6 volts), and a standard interface.

TxD is converted to differential ECL levels with
an open collector (OC) NAND gate and a resistor totem pole. This produces
the right voltage level for one end of an MC10116 differential line driver
input. The other differential input is tied to the MC10116 Vbb. The output
of the MC10116 is differential ECL. Because the ECL drivers are open ended,
each line is pulled to ground with a 470-ohm resistor. The polarity of
ECL lines to the pins of the DB15 connector is important. Switching these
lines will result in a bit sense inversion.

RxD is converted from differential ECL to TTL
using the differential drive from an MC10116 biasing the base emitter junction
of a 2N3906 PNP transistor. This drives a standard TTL NAND input.

TxD is qualified by Ready to Send (RTS) from the
PCLANA (U1b). This is necessary because the SYTEK SIC doesn't clamp the
TxD line to logical zero when RTS goes false.

The SYTEK SIC interface chip is designed to work
on a cable with multiple stations. It monitors the DCD line, and if it
finds that it has been true too long, the SIC declares the cable jammed.
Because the microwave RF modem provides continuous DCD with or without
data, we used a retriggerable one shot (U2a) connected to the incoming
RxD signal to generate an appropriate non continuous DCD. On the first
low-to-high transition, DCD will be asserted and the one shot will start
timing. Each low-to-high transition in the incoming data will retrigger
the one shot and keep it from expiring. When data stops and the RxD line
clamps to a low, the DCD will fall to a zero when the one shot expires.
Because each end of the link expects to hear itself, the DCD is ORed with
the local RTS (U1a).

Building the interface adapter and 

PCLAN adapter modification

The PCLANA modification can be simple or time
consuming, depending upon how elegant you want the finished project to
be. The adapter daughter board must be mounted on the PCLANA. The best
place to do this is over the long metal cover housing the actual RF modem.
Decide on a place and prepare your mount.

You might consider removing enough of the RF modem
components to make room for the daughter adapter board. If you do, be aware
that the RF modem reports to the onboard ľP (80188) as to the validity
of the -12 volt DC power line (for historical reasons). If you remove the
RF modem without connecting this signal, the ľP will report an error whenever
you attempt any operation. Locate Q20 and R121 on the left side of the
pc board next to the lower left corner of the RF modem cover (if it's still
installed). The left end of R121 is tied to the base of Q20. Cut the trace
to the right end (or unsolder R121) and reconnect it directly to -12 volts
DC on the board (bus pin B7).

You need four signal lines to connect the daughter
card to the PCLANA. They are: TxD, RxD, DCD, and RTS. You also need three
power lines: +5 volts, +12 volts, and GND.

You'll find the connection to ground underneath
the screw holding the mounting bracket to the rear of the card. Positive
5 volts DC is taken from a trace leading to edge connector Al (component
side of the card, on the right). Positive 12 volts DC is also taken from
a trace leading from B9 (solder side of the card, on the left). This trace
also "comes through" to the component side of the card and is easier to
solder to. The current demand is low, so you can make contact by cleaning
a portion of a wide trace, tinning it, and soldering to it.

Make signal connections to the SYTEK SIC chip
by bending pins up out of the socket and soldering to them. Find IC U16
in the upper left of the board. It's designated SIC and is a wide 28-pin
device. Pin 1 should have a red dot and be on the lower left. Remove the
device carefully; you probably won't be able to find a replacement easily.
Bend pins 12, 13, 17, and 18 up so they won't make contact with the socket
when the IC is reinserted, and so you can solder to them. Reinsert the
IC. Connect the daughter board signal leads directly to the exposed leads.
Pin 12 is RxD, 13 is DCD, 17 is RTS, and 18 is TxD. It's a good idea to
use shrink tubing on each lead. Be careful not to heat the device unnecessarily
while soldering.

With the daughter card mounted on the PCLANA and
all seven interface lines hooked up, measure the resistance from the +5
and +12 lines to GND. Find the cause of any reading less than 700 ohms
before installing the card in the PC.

If everything checks out, install the board in
a PC bus slot and power up the computer. If the microwave hardware isn't
attached to the 15-pin daughter board connector, the PC will delay for
about 15 to 45 seconds during the boot cycle. It may display an error 3015.

To complete testing, install a similarly modified
card in another PC, connect the microwave hardware, power up both computers,
and check to see that both FSK receivers show DCD. It may be necessary
to reboot the PC by selecting ctl-alt-del after DCD has been established.
When the PC is booted, the PCLANA runs through a series of diagnostics
which include frame loopback. The loopback will fail if the microwave hardware
doesn't have DCD. Depending on the PC, the PCLANA card will declare itself
inoperative until you run diagnostics again.

If you have a copy of the IBM PC network program,
you can now start it on both machines and share disks. Playing with the
network can provide lots of creative fun. With the link running to a fellow
ham, you can access each other's hard disks. Suppose you just finished
some nice graphics and you want to show them off. Instead of reaching for
your An/ camera, bit dump the screen image to disk and do a DOS copy file
from your disk to your friend's. Your friend will be able to see your work
in seconds. A driver is also available to provide packet interface support
for Phil Karn's (KA9Q) TCP/IP package for the PC. (Contact the authors
for further information.)

Tune-up and testing

Once the two ends are operating 105 MHz apart,
you can align the receivers. Select MFC and midrange control setting, and
use a counter or 2-meter receiver to monitor the VCO frequency. Set it
to 148 to 150 MHz at midrange. You should be able to tune several MHz on
either side of this center with the manual tuning control. You can use
AFC to tune it even further once the other circuits are operating. Tune
the VCO to 45 MHz above the previously measured frequency of the microwave
IF and adjust the 45-MHz bandpass coil for maximum C/N reading. It may
be necessary to separate the units or use conductive foam material to keep
the signal strength reading on scale. Once the receiver is peaked on a
45-MHz IF, tune the discriminator coil to center the detector output voltage
on pin 10 or 11 of the MC13055. With the squelch control set to maximum
resistance of 5 k, a 0.35-volt change on pin 12 corresponds to 10-dB change
in signal strength. Keeping the receiver tuned to center with MFC, adjust
the position and absorbers to produce about 10 dB of C/N. Then adjust the
squelch control so the DCD light just extinguishes. Measure the squelch
control resistance again and calibrate your C/N reading by calculating
sensitivity:

V(p12) = 0.070 * Rsquelch (1)
Rsquelch = squelch resistance in kilohms. 
V(p12) = voltage change at MC13055 pin 12 in volts for 
        10-dB change in signal.

Adjust the discriminator output sensitivity to give
full scale on your meter as you use the MFC to tune across the incoming
signal. Finally, verify that the search oscillator runs when you select
AFC and that there's no incoming 105-MHz IF signal. This should appear
as a sawtooth oscillation on the VCO tuning line.

The audio channel should work without further
adjustment. Some background noise may be audible even when signal strength
is high due to the phase noise of the microwave oscillators, but the level
shouldn't be objectionable.

Use an oscilloscope to verify data throughput
and correct transmitter deviation setting. Monitor the discriminator output
with the scope and set the scope sensitivity to give full screen display
as you use the MFC to tune through a signal. When you reach this stage,
the transmitter may be modulated with data and the deviation adjustment
may be used to set the discriminator output to slightly less than full
screen. It may be necessary to iterate with the bias setting to keep the
transceiver bias centered on 6.25 volts.

At this point, both the transmitter and receiver
should be functioning properly and may be used for audio or digital communications.
When units are separated by a great distance, or signals are otherwise
weak, it may be beneficial to monitor the audio channel as an aid to link
adjustment. Audio communication will be possible even when noise is causing
excessive errors on a data channel.

Performance, results, and 

remaining problems

Aligning the microwave dishes requires some skill.
If you haven't done this before, allow plenty of time for alignment, have
solid mounts, and don't expect it to be like 2-meter work. Use the audio
channel to listen for receiver quieting and get a feel for the narrow beamwidth.
Don't try to hand hold the antennas at both ends; both must be pointed
correctly before either end hears anything. It may be useful to use manual
frequency control at first. If you can put one end of the link at a high
elevation temporarily, you can power the other end from 12 volts DC in
your car and drive around to see what microwave communication feels like.
The experience you gain doing this will help to make you a good judge of
final locations for the digital link. Because these are low budget systems,
line-of-sight transmission is probably necessary for anything other than
fairly short links. An exception would be if a good planar reflector were
located close to one end and used as an efficient mirror. You can try this
technique to keep the hardware at ground level using a mast or tower-mounted
"billboard" reflector. This has environmental advantages for the hardware,
too.

Measurements indicate that the unit passes data
with a low bit error rate (BER) down to signal strengths below 15-dB C/N.
It's important to use direct paths; severe distortion can occur when multiple
paths exist between the ends of the link. Such multipath conditions can
cause link failure, even with very large C/N. This sensitivity should provide
low error data transmission on a line-of-sight path of more than 40 miles
with well-stirred air. In many locales, marine air layers and other causes
of fading and ducting may require shortening the path to guarantee high
linkup time.

With the link installed at two locations 13.5
miles apart in northern California, C/N measurements show that there is
at least 10 to 15 dB more signal strength than the minimum necessary. This
indicates that more than 40 miles should be possible with this hardware
as shown. However, because longer paths are more likely to experience propagation
anomalies and heavy rain could decrease signal strength temporarily, it's
desirable to use slightly larger dishes for longer paths. The audio link
has proved useful in system troubleshooting too.

As with any ham project, there's certainly room
for improvement. Because the original RF design was for a 2-Mbaud link,
you should be able to improve DX by optimizing the receiver detector bandwidth.
If you operate with a mast, the equipment needs to be waterproofed for
all- weather use. As an alternative, you could mount a coax waveguide adapter
to the feedhorn and locate the microwave circuitry remotely in a more protected
environment. Use low loss semi- rigid coax cable for connecting to the
antenna if you do so. A suitable coax waveguide adapter was described in
an earlier article.2

If you want to use them, almost any of the surplus
radar detector, motion detector, or burglar alarm Gunn transceivers should
work well. M/A-COM GunnplexersTM, although somewhat more expensive, will
work too. They have built-in electronic tuning that permits modulating
at higher rates for full 10-Mbaud data links or An/ uses. It should be
possible to frequency modulate them by driving the electronic tuning input
from the ECL output, properly scaled and offset with a resistor network.

The first prototype of this link was built using
24-GHz radar transceivers and metal lamp reflectors for antennas. This
arrangement works very well. Because of the modular system design, you
may substitute these microwave assemblies for the 10-GHz ones described
here without making any other adjustments. The higher gain available for
a given dish antenna size at 24 GHz can actually provide better performance
over some paths.

You can use a larger antenna for greater DX or
more difficult paths. If you use something other than a 0.5 F/D reflector,
you'll need to design a new feedhorn to produce maximum performance. Other
diameters are available from the indicated supplier.

Where to go from here

We are also working to build inexpensive 250 to
500-Kbaud 900 and 1200-MHz radios to give the individual user access to
other hams and to a "backbone:' using this higher speed microwave hardware.
We hope to have a fledgling moderate speed network in place in northern
California and Colorado by the time this article goes to press. As the
hardware is put into place, the platform for some really exciting applications
and a whole new era of Amateur Radio becomes a reality.

The authors would like to thank WN6I, N3EUA, K3MC,
and N6TTO for their encouragement and perseverance during testing. We'd
also like to thank our XYLs, Sharon and Lynn.


 


 

REFERENCES


1. Glenn Elmore, N6GN, "Designing A Station For
The Microwave Bands. Parts 1-3," Ham Radio Magazine, February, June, and
October 1988.

2. Glenn Elmore, N6GN, "Designing A Station For
The Microwave Bands. Part 2," Ham Radio Magazine, June 1988, page 35-37.