During the tests on 10GHz and 1.3GHz using JT4G mode last month, a few people reported difficulty with decoding the data in spite of a good S/N. In some cases the problem was solved by setting the clock properly, but not always... After modifying the GB3SCX 10GHz beacon yesterday to send JT4G, I tried to get local copy using a Dell Inspiron 6400 laptop fedd from a transverter / Rx while on site. Result absolutely zero, zilch, null, no local copy whatsoever. The Sync value was 0 in spite of the four frequencies being accurately tuned. I suspected it might be a timing problem, as laptops can have a bad reputation for sound card accuracy, and when another station some way away reported perfect copy of the new transmission, didn't try any more while 'up there on the hill top' Today, I fed an exact 1kHz signal into the laptop and measured the frequency - it came out at 1021Hz on all softare packages that had a spectrum function, suggesting the actual sample rate is 1000/1021*11025 = 10800Hz. Now, the resulting frequency shift due to this isn't enough to upset JT4G decoding by any means, but the sampling rate is also used for symbol timing. Over a 48 second transmission, the resulting error is just about one second - which covers more than four symbols at 4.375 Symbols/s - and would explain the complete inability to get any synchronisation. I then connected it back-to-back via audio patch lead with the desktop and tried to get the two computers to talk to each other in JT4G - again nothing. In the WSJT Options-Setup screeen is a couple of values for 'Rate In' and 'Rate Out' - which by default are each set to one. After guessing these may be for sample rate modification, searched all the documentation for reference to these terms, found one, and how to use them. When running, WSJT shows two numbers in the extreme bottom left of the main screen - both are close to unity. The instructions suggest copying these two values into the 'Rate In' and 'Rate Out' windows. On my (slow-running) laptop, the rate in was showing as 0.9797. Which just happens to be pretty damn close to 10800/11025. (The Rate Out was 1.006 ish). Entered these into setup, and the two computers could suddenly talk to each other, with mutual sync values in the very good 20 - 30 region. Interestingly, the modified rate values do not show up on the spectrogram display, this still showed 1kHz tone as being at 1021Hz, so I guess the correction is only applied to the symbol timing, or to the whole signal processing side and not JTSpec Anyway that's it solved -------------- QED ------------------ It would appear that WSJT actually appears to measure its own sample rate - presumably against the crystal controlled PC clock which it uses for schedule timing and relatively speaking, is 'accurate' The error is then reported as a ratio for input and output rates. I'd heard some laptops have notoriously bad soundcard sample rate errors, but would never have thought a respectable modern well known make would be that far out! I then injected a 1500Hz tone into the WSPR software, which unlike WSJT works at the more modern 12kHz sample rate. The tone was shown as being spot-on. Now, as 10800Hz has more factors in common with 12000Hz that does 11025 (*), its probably reasonable to surmise the soundcard is designed for 48/24/12kHz operation and is 'bodged' with a simple divider chain for an 'approximation' to get the old 11025 value.
(*) 11025 = 3 * 3 * 5 * 5 * 7 * 7 10800 = 2 * 2 * 2 * 2 * 3 * 3 * 3 * 5 * 5 12000 = 2 * 2 * 2 * 2 * 2 * 3 * 5 * 5 * 5 Andy www.g4jnt.com
