I thought I would send a note to the list about what I've built recently 
so that if anyone else is interested in building something similar we 
can trade notes.

I have a ham UHF repeater system in the San Francisco/Monterey Bay area 
( more info at www.wb6ece.org ) with 6 sites all sharing the same 
frequency. For years it has been CTCSS-select of which site you'll go 
through, but I've long wanted to link the sites together, use a voter to 
pick the best receive audio, and simulcast the results back out, but it 
took several clever (at least in my opinion :) ) ideas to get it to 
happen. (Not to mention a number of false starts with three different 
attempts to use 420 MHz analog linking, some self-designed and built PIC 
microcontroller-based audio delay and signalling boards, and other 
interesting and educational experiments that were ultimately dead ends)

At this point we have three of the sites online with the new linking 
system and it works even better than expected. Audio artifacts even in 
heavy-overlap areas are almost nonexistent. A "proper" simulcast system 
would use low-level transmit sites to reduce the overlap, but we've got 
mountaintop sites that have nearly complete overlap of, for instance, 
the San Jose area, and it really works well.

So what were the clever parts?

1. Transmit frequency control. We use Vertex VXR-5000 repeaters. The 
transmitter is modified by removing the internal 12.8 MHz TCXO and 
hooking the PLL chip reference up to a surplus HP Z3801A GPS-disciplined 
10 MHz reference oscillator. Then a PIC microcontroller is installed to 
jam the new divisor settings into the PLL chip over its control lines. 
No need to synthesize up a 12.8 MHz intermediate, as 441.300 MHz (our 
transmit frequency) is easily reached with a 10 MHz reference. Cost is 
the PIC chip, a capacitor, and some cable and connectors.

2. GPS-locked VoIP linking with software voting. The linking is all done 
over IP, by digitizing the audio at each site, computing the noise 
level, sending the audio and the noise level to a central site as VoIP 
packets, and sending the best choice (lowest noise level) packets back 
out to all the transmitters. But for that to work, the transmitter audio 
needs to be in phase, and for inaudible voter switching the receiver 
audio needs to be in phase as well.

So what we do is use a 4-channel 96 kHz sample rate audio card (M-Audio 
Delta 44) at each site, plugged into a Soekris net5501 (500 MHz Geode 
running Linux, 512 MB of RAM, 2 GB of Compact Flash instead of disk, no 
fans). One channel listens to the 1 PPS signal from the Z3801A GPS, one 
channel listens to a looped-back signal from the output of the card, and 
that leaves two full-duplex channels for linking. A software FLL/PLL 
algorithm to align the input audio sample time with the 1 PPS and the 
output audio sample time (as seen via the loopback connection) with the 
1 PPS and we get about 10 microsecond phase accuracy in both directions, 
which is adequate for simulcast. All in software, no specialized DSPs or 
hardware to clock the audio card or anything. And once the audio is 
digital, everything else (tone detection and generation, muting, 
filtering, input and output gain) can be done in software as well. So 
off-the-shelf hardware, plus a cable built to plug everything together, 
turns it into a pure software problem... and we've only just started on 
what's possible in software.

Feel free to comment or ask questions.

Matthew Kaufman, KA6SQG
[email protected]
http://www.wb6ece.org

ps. Once I catalog everything I have, I plan to sell off all the unused 
420 MHz radios and related gear I've accumulated in the learning process.

pps. We're also in the market for used VXR-5000 repeaters, and 
particularly, Pacific Research RI-210 controllers, for system expansion.





Reply via email to