DSDPlus Help - Cannot get clean decodes on DSDPlus from P25 with a strong signal

[rolypolyman]

This post is the culmination of two years of aggravation with P25 decodes on FMP24 (the paid version) and I am at the end of my rope. I am monitoring a municipal police repeater only 1/2 mile away. It is an extremely simple system that has all traffic one frequency, no control channel or anything, so reception is just purely decoding the P25 signal. As you can see from the screenshot I'm getting a good strong radio signal and the receiver is locked.

About 50% of the transmissions are good, but for the other 50% I get choppy and stuttering audio, from both dispatch and from their mobile units. I don't notice any particular pattern to it. The parties on the frequency all seem to hear each other fine. It sounds similar to what you might hear on an analog radio when the squelch is turned up a little too high. Squelch is not enabled, I checked by pressing Q, verifying it came on, then turning it off.

I am using a SDR Blog DSD receiver. Data is being fed from FMP24 to DSDPlus via "Direct FMPx linking" (port 20002). I have noise filter and shaping filter both off, and it has no effect either way. I have tried decoding with those filters on with similar results. In the menu tree I have all the decodes disabled except for P25. CPU use for DSDPlus.exe is 0.2% while idle and 1.1% while decoding, and for FMP24.exe it is consistently 1-2%.

The baseband raw file is here: fmp24_decode_error.zip (70.22MB) - SendSpace.com (75 MB ZIP) and it includes sample audio WAV files.

Can someone test the baseband file and let me know whether they're getting good decodes? I don't know what to do at this point or if there's a setting I'm missing.

 

[sonm10]

Unfortunately, DSD+ does not have a decent vocoder. What you're describing is pretty normal. I think most of us who use DSD use it for logging purposes, but does not make for decent monitoring. Might try SDRTrunk instead. I think it might have a superior vocoder.

One thought: the gain might be too high. Turn down the gain until signal is 3/4.

 

[hagensieker]

Do you have enough bandwidth there? Try 12.5 for bandwidth.

 

[slicerwizard]

CPU use for DSDPlus.exe is 0.2% while idle and 1.1% while decoding, and for FMP24.exe it is consistently 1-2%.

I'd say that's your problem. Your PC is snoozing and some USB I/Q data transfers are failing and they don't get resent. There are discontinuities in the FM demodulated raw audio, which indicates that the processed I/Q stream is missing samples. It's why the symbol phase graph in the DSD+ console window title bar keeps jumping around.

What versions of FMP24 and DSD+ are you running?

Have you run the lost I/Q data test in FMP24? Do you have extra CPU loading in FMP24 set to enabled, disabled or auto?

 

[rolypolyman]

Thanks, that gives me some leads. I'm not sure about PC snoozing as I'm using it and clicking things while collecting the baseband sample, and I hear the dropouts in realtime. I will pay more attention to the phase graph while this is happening.

Are you getting a choppy decode as well from the sample?

CPU loading is disabled. I did originally have CPU loading set to the default but I had to disable it because the waveform would go red after about a day of usage and the session would become corrupted. I posted about it fully in the thread below and it was recommended to disable CPU loading with -e0, which I did and the problem hasn't happened since then. BTW, it was speculated in that thread my computer goes to sleep, causing the corruption, but I actually keep it running 24/7 and have sleep disabled completely, the only thing that happens is I have the "enter password" window appear after an hour. The session becomes corrupted anyway.
DSDPlus - FMP24 keeps going into "evil mode" (high CPU load)

I haven't heard about the lost I/Q data test... where is it and what score is acceptable? What can I do about lost I/Q data?

I'll also put the bandwidth back on 12.5 as hagensieker suggested but I don't recall it making much of a difference as I had been using 12.5 until last month.

I am using FMP24 2.75 and DSD+ 2.347 and run Windows 10.

 

[rolypolyman]

Also just as a note if I press E in the spectrum graph and go to Extra CPU Loading Enabled, the white waveform turns red. I don't know if this is indicative of a problem. It goes away in Auto and Off mode. In the past, when I used Auto CPU and the session goes corrupt, this would always turn red.

 

[ChrisABQ]

I second SDRTrunk, it's superior audio for monitoring. It's an all-in-one easy install to get it running on Windows.

 

[rolypolyman]

I second SDRTrunk, it's superior audio for monitoring. It's an all-in-one easy install to get it running on Windows.

Thanks, unfortunately that's a no go unless it supports writing all transmissions to individual audio files coded with the timestamp and radio ID, like DSDPlus does. I frequently have to go back through archive files to hear traffic of interest, and that kind of functionality is essential.

 

[slicerwizard]

Thanks, that gives me some leads.

You already had leads. This was covered in your previous thread.

I'm not sure about PC snoozing as I'm using it and clicking things while collecting the baseband sample, and I hear the dropouts in realtime.

And? Your PC goes into low power / sleep states between the clicks.

I will pay more attention to the phase graph while this is happening.

Are you getting a choppy decode as well from the sample?

Yes. It's classic lost samples.

CPU loading is disabled. I did originally have CPU loading set to the default but I had to disable it because the waveform would go red after about a day of usage and the session would become corrupted.

What's wrong with red? Your previous posts on this mentioned nothing about corruption. What was corrupted?

I posted about it fully in the thread below and it was recommended to disable CPU loading with -e0, which I did and the problem hasn't happened since then. BTW, it was speculated in that thread my computer goes to sleep, causing the corruption, but I actually keep it running 24/7 and have sleep disabled completely, the only thing that happens is I have the "enter password" window appear after an hour.

Irrelevant. Your CPU still goes into low power states, especially when it's not being taxed.

The session becomes corrupted anyway.
DSDPlus - FMP24 keeps going into "evil mode" (high CPU load)

Again, there is nothing in that thread about corruption. FMP24 was dealing with the I/Q losses, but you disabled that functionality, so the I/Q dropping returned. Nothing mysterious going on.

I haven't heard about the lost I/Q data test... where is it and what score is acceptable?

Active keys:
...
L toggle lost I/Q data test mode

Ideally, there shouldn't be any losses detected, but a rate of one dropout per minute or less would have very little impact on voice monitoring/recording. You're getting them every few seconds.

What can I do about lost I/Q data?

You could let FMP24 deal with the problem instead of disabling that functionality. If that causes your machine to run too hard, slow down the clock. I doubt you need max performance all the time. My laptop can run at about 4 GHz, but I routinely keep it at about 0.8 GHz; FMP24 doing its job (aka "Evil Mode") isn't noticeable at 0.8 GHz - no extra fan action.

Or you could use some other SDR-based trunk tracking program; all of the others use far more CPU resources and your PC will definitely not be sleeping.

Or you could constantly run some useful background job, like SETI@home.

I'll also put the bandwidth back on 12.5 as hagensieker suggested but I don't recall it making much of a difference as I had been using 12.5 until last month.

It's obviously not your problem.

I am using FMP24 2.75 and DSD+ 2.347 and run Windows 10.

You haven't been keeping up with your updates, but that's also not the problem here. But you should get caught up.