DSD+ voice sounds horrible, lots of errors

[thedrizzle]

(((EditToAdd: Newbie replies are in purgatory, but upped a new pic of CPU running pretty low. Also, I wasn't clear, but I am using UniTrunker to feed DSD+.
Also: With great thanks to mtindor I think I am wrestling with this: https://wiki.radioreference.com/index.php/Simulcast_digital_distortion
Also,also: Becoming more accepting that mtindor is right and I'll have to dig in to FMP docs and figure it out. But if I lose TRecorder function I'm toast. That is the software that made this fun again lol. Maybe I can run UniT next to FMP to just feed TRec while FMP controlls. Ok back to reading.)))

Banging my head against this wall trying to figure out why DSD+ Fastlane gives tons of errors and voice sounds horrible. I am very RF amateur, but pretty good with computers. I think (maybe?) it is in my VAC setup or Windows recording levels, but can't nail it down and as last resort I hope maybe someone can point me in a good direction.

Basics: 4 RTL dongles. A single no-name which is the signal stick, and 100% health and I think out of scope.
Then I have 3 x rtl-sdr.com sticks, each doing voice duty. Each voice stick feeds into an even numbered VAC and DSD+ outputs to the odd numbered VAC above it. VAC #1 locked up a few weeks ago, and instead of fight it, I just made a new one and VAC1 is dead to me. Assume I have the correct VAC# feeding into DSD+, its confusing maping VAC# to whatever arbitrary device table DSD+ uses which changes each boot-up

Voicemodel ---> VAC2 ---> DSD+---->VAC3
2Voicemodel ---> VAC4 ---> DSD+ --->VAC5
3Voicemodel ---> VAC6 ---> DSD+ --->VAC7

I am using the command line arguments from dsdtune ran on the same input file I have shared.

Screenshots and input.wav from hitting R in dsd+: https://drive.google.com/drive/folders/1jz4rF-zke6K-BRPzHDrZpKzQoN-JfQQS?usp=sharing

After spending many hours getting the signal stronger with some better antennas, I'm realizing something is off and I don't think it is signal. I get a 40ish SNR when testing in SDR# via mouse hover while tuned to control channel on all Voice dongles. So assuming that I have spent this evening trying to figure out why so, so, so many errors pouring out of DSD+ and why the voices sound horrible.

I think I have narrowed down to these symptoms:
- When I put UniScope on an even # VAC coming out of Unitrunker, and switch it to CQPSK, the 3-eyes are constantly scrolling right to left. I can kind of control the speed of the scroll and slow it down by adjusting the recording level of that VAC upward, but it never gets to a static point.
- Just the normal scope view in DSD seems to have the waveforms moving horizontally, instead of other videos I have seen of functioning systems, where it seems static waveforms pop up and down in the same spot.

Any thoughts??? Please??? I have it so close to working awesome, and the voice quality is really just horrible. I can take video if needed but hoping to avoid.

Also,Trunking Recorder screenshot is just to show how much I love that program and it's simple web server, etc.

EditToAdd: hell here's a link to the TR if you want to hear a crappy sound version of Baltimore PD lol, (I'll probably remove this later, don't have bandwidth to serve the public, change to http): hteeteap://weakerthan.dyndns.org:888

Thanks,
Dan

 

[dlwtrunked]

Banging my head against this wall trying to figure out why DSD+ Fastlane gives tons of errors and voice sounds horrible. I am very RF amateur, but pretty good with computers. I think (maybe?) it is in my VAC setup or Windows recording levels, but can't nail it down and as last resort I hope maybe someone can point me in a good direction.

Basics: 4 RTL dongles. A single no-name which is the signal stick, and 100% health and I think out of scope.
Then I have 3 x rtl-sdr.com sticks, each doing voice duty. Each voice stick feeds into an even numbered VAC and DSD+ outputs to the odd numbered VAC above it. VAC #1 locked up a few weeks ago, and instead of fight it, I just made a new one and VAC1 is dead to me. Assume I have the correct VAC# feeding into DSD+, its confusing maping VAC# to whatever arbitrary device table DSD+ uses which changes each boot-up

Voicemodel ---> VAC2 ---> DSD+---->VAC3
2Voicemodel ---> VAC4 ---> DSD+ --->VAC5
3Voicemodel ---> VAC6 ---> DSD+ --->VAC7

I am using the command line arguments from dsdtune ran on the same input file I have shared.

Screenshots and input.wav from hitting R in dsd+: https://drive.google.com/drive/folders/1jz4rF-zke6K-BRPzHDrZpKzQoN-JfQQS?usp=sharing

After spending many hours getting the signal stronger with some better antennas, I'm realizing something is off and I don't think it is signal. I get a 40ish SNR when testing in SDR# via mouse hover while tuned to control channel on all Voice dongles. So assuming that I have spent this evening trying to figure out why so, so, so many errors pouring out of DSD+ and why the voices sound horrible.

I think I have narrowed down to these symptoms:
- When I put UniScope on an even # VAC coming out of Unitrunker, and switch it to CQPSK, the 3-eyes are constantly scrolling right to left. I can kind of control the speed of the scroll and slow it down by adjusting the recording level of that VAC upward, but it never gets to a static point.
- Just the normal scope view in DSD seems to have the waveforms moving horizontally, instead of other videos I have seen of functioning systems, where it seems static waveforms pop up and down in the same spot.

Any thoughts??? Please??? I have it so close to working awesome, and the voice quality is really just horrible. I can take video if needed but hoping to avoid.

Also,Trunking Recorder screenshot is just to show how much I love that program and it's simple web server, etc.

Thanks,
Dan

I ditched VAC for free VBCable which seems to work better but your need for so many virtual audio cables makes that not a choice. What happens if you feed a single frequency from say HDSDR to DSD+? Do you still have a problem? What is the CPU load on the PC like when you are doing what you are doing? I would check these things to try to pin-point the issue.

 

[thedrizzle]

CPU Screenshot attached. I thought about that too, I do have some low level background processes running that are optional, but unless it is super sensitive I think I'm ok to keep em going. Good point, I should have mentioned I checked that and seems ok. I think I will try your suggestion to pipe the SDR# on a busy voice channel into a VAC and bypass Unitrunker to isolate. Maybe if that is still bad, replace the VAC cable with the VBCable just to test/isolate that single connection. Good idea.

 

[mtindor]

Banging my head against this wall trying to figure out why DSD+ Fastlane gives tons of errors and voice sounds horrible. I am very RF amateur, but pretty good with computers. I think (maybe?) it is in my VAC setup or Windows recording levels, but can't nail it down and as last resort I hope maybe someone can point me in a good direction.

Basics: 4 RTL dongles. A single no-name which is the signal stick, and 100% health and I think out of scope.
Then I have 3 x rtl-sdr.com sticks, each doing voice duty. Each voice stick feeds into an even numbered VAC and DSD+ outputs to the odd numbered VAC above it. VAC #1 locked up a few weeks ago, and instead of fight it, I just made a new one and VAC1 is dead to me. Assume I have the correct VAC# feeding into DSD+, its confusing maping VAC# to whatever arbitrary device table DSD+ uses which changes each boot-up

Voicemodel ---> VAC2 ---> DSD+---->VAC3
2Voicemodel ---> VAC4 ---> DSD+ --->VAC5
3Voicemodel ---> VAC6 ---> DSD+ --->VAC7

I am using the command line arguments from dsdtune ran on the same input file I have shared.

Screenshots and input.wav from hitting R in dsd+: https://drive.google.com/drive/folders/1jz4rF-zke6K-BRPzHDrZpKzQoN-JfQQS?usp=sharing

After spending many hours getting the signal stronger with some better antennas, I'm realizing something is off and I don't think it is signal. I get a 40ish SNR when testing in SDR# via mouse hover while tuned to control channel on all Voice dongles. So assuming that I have spent this evening trying to figure out why so, so, so many errors pouring out of DSD+ and why the voices sound horrible.

I think I have narrowed down to these symptoms:
- When I put UniScope on an even # VAC coming out of Unitrunker, and switch it to CQPSK, the 3-eyes are constantly scrolling right to left. I can kind of control the speed of the scroll and slow it down by adjusting the recording level of that VAC upward, but it never gets to a static point.
- Just the normal scope view in DSD seems to have the waveforms moving horizontally, instead of other videos I have seen of functioning systems, where it seems static waveforms pop up and down in the same spot.

Any thoughts??? Please??? I have it so close to working awesome, and the voice quality is really just horrible. I can take video if needed but hoping to avoid.

Also,Trunking Recorder screenshot is just to show how much I love that program and it's simple web server, etc.

EditToAdd: hell here's a link to the TR if you want to hear a crappy sound version of Baltimore PD lol, (I'll probably remove this later, don't have bandwidth to serve the public, change to http): hteeteap://weakerthan.dyndns.org:888

Thanks,
Dan

I always throw this out first - If using SDR# for decoding digital, be basolutely sure that you have "Filter Audio" unchecked.

Now, onward to CQPSK. If you're monitoring CQPSK/LSM sites, there is nothing in your setup that is meant to decode that properly. SDR# as audio input will not allow DSDPlus to decode LSM/CQPSK. Unitrunker doesn't do it. FMP(A) + DSDPlus will decode simulcast.

Any SDR software + DSDPlus = no CQPSK support
Any input + Unitrunker = no CQPSK support (Rick can correct me if I'm wrong)
FMP(A) + DSDPlus = CQPSK support

So if you are trying to decode simulcast, you need to use FMP(A). In your case, with dongles, FMP.

Mike

 

[thedrizzle]

Hmmmm interesting. I didn't think that this might be a job that UniTrunker + DSD+ couldn't handle. I will read up on FMP, although the bit I had seen seems scary heh. Actually Trunking Recorder pretty much requires UniTrunker best I can tell But, can I point to the fact that it is working, all 3 voice dongles are tracking and switching, and it is decoding. Just with a lot of errors and with a waveform that looks kind of "off" to an untrained eye fwiw.

I was doing a little more browsing on VAC alternatives, and stumbled into some screenshots of others SDR setups, and I think locking my SR to 48,000 might have been a bad move. I'm going to experiment with that a bit. Also going to continue to look into VAC alternative, and reading on up CQPSK and how relates to FMP, et al. TYVM

 

[thedrizzle]

Also just found your posts on this thread regarding the same system I am trying to monitor.... appreciate your helpfulness sir, then and now heh. https://forums.radioreference.com/v...-software/354092-sdr-unitrunker-dsd-help.html

So due to the CQPSK this system is using, it is difficult or not possible to get clean audio? I'm still reading up, but had to post after chuckling at finding your old posts, on same system.

 

[thedrizzle]

So I've spent many hours this morning studying and learning about LSM, CQPSK, Simulcast, etc. My head is swimming in these new terms/technologies, but I want to post a summary of my thoughts and inquire on what my best path might be. I also want to thank mtindor because every new thread I find that is relevant to my struggle, has a very informative post by him, among others.

Now that I have the right terms/nomenclature I am finding a lot of people in my same position, some relating to the exact same system I am trying to monitor. Thank you again for getting me on a good path (although a path that seems to frustrate a lot of people in same situation).

Current thoughts:
1) Did I shoot myself in the foot by increasing my signal quality? It does seem it became worse after upgrading antennas. Along same thinking, should I look in to reflector/directional to try to ignore other towers? I kind of think all of the towers are in one direction (all towers are downtown, and downtown is SE of me).

2) A lot of people say that switching to OP25 and Linux works a lot better for LSM/Simulcast. Should I go that direction? (however see next thought)

3) The Trunking Recorder software has a dependency on Unitrunker, and TR is the frontend that really makes this "look right" to me and my other end users.

At the end of the day I think I'll settle on "this is a crappy situation but at least other people think it's crappy too", misery loves company heh.

If you were in my situation and didn't want to drop multiple hundreds of $$$ to buy a real scanner (or those G4/G5 "pagers"), which way would you go?

Thanks for your time,
Dan

 

[thedrizzle]

Eureka! I knew something I had done lately made it worse! Seeing the signal just dancing around in the scope when I suspected the ideal was a static waveform sitting there ready to carry data, I was so frustrated! I had the scope up, switching antennas, killing gain to near zero, and it still was all over the place, even while fuzzy as hell.

While I was in Unitrunker seeing how settings affected the scope, I changed my sample rate back down to 2.048 and the scope just froze dead center showing a beautiful pattern just as I imagined it lol. I'm posting a video to recreate showing the drastic change. I foolishly thought more sample rate = better, but whoa was I wrong. I can hear ppl again. Give me about 10 mins to record this and then look in the link in my OP.

Update: Added Video SAMPLERATEOMG.mp4 to the Google drive in original post. Its aVERY drastic difference IMHO. Thanks a million times for the thoughts & ideas.

 

[mtindor]

Eureka! I knew something I had done lately made it worse! Seeing the signal just dancing around in the scope when I suspected the ideal was a static waveform sitting there ready to carry data, I was so frustrated! I had the scope up, switching antennas, killing gain to near zero, and it still was all over the place, even while fuzzy as hell.

While I was in Unitrunker seeing how settings affected the scope, I changed my sample rate back down to 2.048 and the scope just froze dead center showing a beautiful pattern just as I imagined it lol. I'm posting a video to recreate showing the drastic change. I foolishly thought more sample rate = better, but whoa was I wrong. I can hear ppl again. Give me about 10 mins to record this and then look in the link in my OP.

Update: Added Video SAMPLERATEOMG.mp4 to the Google drive in original post. Its aVERY drastic difference IMHO. Thanks a million times for the thoughts & ideas.

I would not have expected the change in sample rate to make that much of a difference. But I guess there are multiple things that could have been at work there.

First off, I mix up the proper nomenclature all the time regarding simulcast systems. One minute I say CQPSK, the next LSM. The diehards would/should correct me when I don't use the right terminology, but I suspect they have just given up on me. In the end I'm always referring to a "simulcast" site. So don't feel bad if all of the verbage is confusing -- it still is for me after many years.

And I keep forgetting that just because the site you are listening to might be a simulcast, that doesn't mean you will be _so_ affected by distortion that it would ruin your day. Depending upon where you are (versus all of the towers in the simulcast), it's entirely possible that you will have very minimal distortion and thus can copy nearly 100%.

For many people, they might pick up multiple sites in the simulcast with sufficient strength that the signals are mixing (but not in time) at their antenna. Those people are the ones usually plagued with the distortion issue. Then there are others who might be in a location where they pick up only one of the towers in the simulcast, or they pick up one with an SNR so much higher than the other towers that the brute force signal strength of that one tower allows them to copy with minimal distortion.

Based upon what you're saying, it sounds like in your particular location you are lucky and may not have much distortion -- and that your ultimate problem was a sample rate issue.

And, I must admit, i would have _never_ thought of checking / trying different sample rates when troubleshooting an issue with simulcast systems. I always make the assumption that the problem is distortion, always. Guess i learned something new

I'm glad your probably was one that was easily resolved. With the particular applications you want to use, you wouldn't have any alternative options if the sample rate change didn't fix things for you.

Mike

 

[thedrizzle]

Just an update:

DSDPlus Fastlane working great, new version slipped right in and seems it might be better sounding. Added 2 more dongles, so now 1 signal, and 5 voice. Somebody stop me lol Anywho, here's some pictures showing the CPU load with UniT + VACx10 + DSDPlusx5 + TrunkingRecorder. I'm at the top of my comfort level for that CPU%, so I think this box is done with expansion.

Thanks for the previous help, and sorry ahead of time for the hacks I had to do with cabling because of mistake in ordering adapters hah. Maybe this helps someone else.