And this is why having the platform up an running, and infinitely programmable, gives us a chance to put the options to the test!
In multipoint use in Vancouver we didn't need FEC to get good throughput. We had a pure IP implementation, not AX.25 or similar. Fixed stations were 10-40 miles from the repeater, in my case even bouncing signals off a mountain (about 20 miles from the repeater) worked well - packet loss was little-to-none. These are real world tests.
I'm not saying the FEC isn't useful, but it's not a cure-all and incurs overhead. FEC is definitely useful in latency-sensitive applications, but overall throughput can be higher when latency is less important than throughput on a link with little loss. It can be a real throughput help on links with lots of loss, but then throughput suffers, so 56k can quickly turn into less than 9600. We talked about FEC as a possible need before building the 56 k network in Vancouver but it turned out to be moot in real world tests - FEC would have swallowed more overhead than it would have returned. One thought - almost all of the calculations that I've seen to predict error in high speed over-the-air networks have been far more pessimistic than in reality (meaning not all of the variable were taken into account). We were amazed that a low-level mobile station would perform so well at 56k. Commercial guys have the same issue, one (RadioLAN? can't remember exactly, 900 MHz ~56k network) said that their modems wouldn't work above about 8 MPH, but discovered that they actually worked up to almost full freeway speeds (I used mine up to about 55 MPH).
So - we'll see what happens - I'm willing to be proven wrong, if FEC improves throughput, it will win out, but having SDR means never having to say you're sorry (for long) - just reload and move on. Even if the first 56k multipoint implementation isn't perfect, it will certainly be much better than current 9600 technology (there's no reason to deploy it if it isn't). We might start with emulating the WA4DSY modem that we used in Vancouver (and other places), and innovate from there. I'd like to pursue an adaptive protocol, that maximises throughput by using FEC only when it improves throughput. We might also use a 9600 bps (or slower) multi-channel approach (since bitrates are regulated here, making sure that we stay within the regs is necessary) using various modulation techniques, using a variable number of channels, and bonding the channels together as needed. These are the kinds of innovations that ham radio can contribute to the world of communications, while still pushing data through our own stations.
I think I should start researching the WA4DSY modem further and see if I can come up with an SDR equivalent as a base to work from (or find another modulation technique). Another advantage of SDR is that we can emulate ideas in software even before the hardware arrives!
- Richard, VE7CVS