I learned something interesting this evening: Turning on a CRT-based television raises the noise floor by about 5 dBm across the board.
As mentioned last post, I was monitoring a pretty interesting conversation between VE3SKY and CO2DC on Olivia in the 20 m band, getting about -13 dB SNR from CO2DC as reported by MultiPSK. That seems overly pessimistic. From the Excalibur spectrum view, though, at times the signal did appear to sink below the noise floor, and was only 2 or 3 dB above, best case.
Until the TV wiped out all but the strongest PSK31 signals, decode performance was solid, with only the odd word being completely unintelligible. VE3SKY mentioned how he rarely used Olivia, and CO2DC replied that he actually preferred it.
I've only witnessed two contacts using Olivia and I see the appeal. It's more robust compared to BPSK31, with lower latency than JT65A. From an audio point of view, it's also not as depressing as JT65A.
On the other hand, it seems as if you could go a lot faster given the bandwidths in question. While kicking around the idea of a faster mode with error correction, I stumbled across a comparison of QPSK31 vs BPSK31. Seems that even the creator is kind of disappointed over the popularity of BPSK31.
I think that whatever technical concerns about QPSK31 raised in 2002 are now moot in 2012. Even if your contemporary receiver or transmitter has a bit of drift, on the receiver end there's more than enough computing power to figure things out. If you can demod DRM with a laptop, then you can demod some low quality QPSK31 signal with the same.
But we should be able to do slightly better now. Use QPSK31 or QPSK63 with a constraint length 9 convolutional code. There's at least one open source library from KA9Q that supports it.
From there it's not far of a stretch to further fortify the transmission with a block code (Reed-Solomon, or BCH), or switch the convolutional code for a turbo code or LDPC code, computing power permitting.
No comments:
Post a Comment