locked Re: #JT4 and now #Q65 Decoding of very high S/N #JT4 #Q65

Andy Talbot

And now, having just modded the PIC code to generate Q65
Pure tones, no noise - NO DECODING whatsoever, ever, never,  to the extent I was pointlessly chasing up errors in the code and wasted over an hour of my life :-(

Add in some noise and straight decodes,
So the Q65 decoder, unlike the JT4 one,  needs background noise.

On Sat, 22 May 2021 at 17:09, Andy Talbot <andy.g4jnt@...> wrote:
If I add noise the frequency measurement works properly.
But the S/N doesn't

Is S/N reporting supposed to be implemented in JT4 wider modes?
I do very vaguely seem to recall a comment once that it was not done at all.

On Sat, 22 May 2021 at 16:41, Andy Talbot <andy.g4jnt@...> wrote:
I've been testing a new JT4G beacon source , a single PIC chip generating quadrature audio tones for directly feeding a quad-upconverter.  Testing by feeding the tone output directly into the soundcard, running Ver 2.4.0 RC4

In the past, with earlier versions of WSJT-X,  the JT4 decoder hasn't always been very happy with pure tones and very high S/N. and would often miss decodes.  The situation resolved itself when the S/N was degraded by deliberately adding in noise from an audio noise generator

Now, though, with the latest decoder behaviour with a pure tone input is different.    Decoding is always successful - there are no missed decodes - but the frequency estimation and  the S/N reported are way-out.   I use exactly 1000Hz for tone 0 , but the decoder reports a spread from 914 to 1050 at least, and S/N is reported as -15 or so, when it is really a very high positive value.

Getting the S/N wrong I can understand why it may be happening, with the S/N estimation going off scale with no noise to measure, but the random value of frequency reporting is a new one.   Why is this ?

See the  attached screen shots

Join main@WSJTX.groups.io to automatically receive all group messages.