HQ Player

[toddrhodes]

Hoping my Pi4 shows up today from Amazon. 

 

Beyond that, I built my new Ryzen 7 machine last night. 3700X CPU, 32G DDR4 PC3600 RAM. I thought it was fast *before* I discovered the RAM wasn't running at full speed and needed a BIOS update. And for whatever reason, "CUDA Offload" checked, not just grayed, slows it down now. If I just leave that grayed and let the CPU handle the work, I can do DSD256 ASDM7EC with about a 1-2 second lag before music starts playing. I don't recall the CPU load but it's very reasonable once playback starts, to my recollection. Now, it still had the dropouts but I blame that on the Pi3 that will hopefully be fixed today.

 

But my previous machine, an i7 6700k, wouldn't even start playing something on ASDM7EC though that may have been the same CUDA Offload thing (which I gather is likely user error - I'm new to this). Filter in use is poly-sinc-xtr2-2s.

[Miska]

16 hours ago, mirekti said:

I've finally got some spare time to try this out however, I am only able to find a quick guide on the web page.
Could you share a link to the manual where I could find an example how to do let's say 60Hz -6db Q-7, please?
Also, it would be great if the Pipeline Setup would offer a plot of what is actually being done with the signal. Any plans on adding it as a feature to release 4.xx?

 

Are you on Desktop or Embedded?

 

Desktop PDF manual is included in all copies of HQPlayer. On Windows and Linux you can find it under HQPlayer group in Start-menu. On macOS it is included in the DMG and you can place it anywhere you like.

 

Desktop includes also Plot-button to plot the frequency and phase response. This is not yet in Embedded, but is planned feature there as well.

 

[Miska]

12 minutes ago, toddrhodes said:

Hoping my Pi4 shows up today from Amazon. 

 

Beyond that, I built my new Ryzen 7 machine last night. 3700X CPU, 32G DDR4 PC3600 RAM. I thought it was fast *before* I discovered the RAM wasn't running at full speed and needed a BIOS update. And for whatever reason, "CUDA Offload" checked, not just grayed, slows it down now. If I just leave that grayed and let the CPU handle the work, I can do DSD256 ASDM7EC with about a 1-2 second lag before music starts playing. I don't recall the CPU load but it's very reasonable once playback starts, to my recollection. Now, it still had the dropouts but I blame that on the Pi3 that will hopefully be fixed today.

 

But my previous machine, an i7 6700k, wouldn't even start playing something on ASDM7EC though that may have been the same CUDA Offload thing (which I gather is likely user error - I'm new to this). Filter in use is poly-sinc-xtr2-2s.

 

What kind of Nvidia GPU do you have? If the GPU is slower than the CPU, then offload just makes things worse as CPU ends up waiting for GPU to complete it's tasks.

 

[toddrhodes]

1 minute ago, Miska said:

 

What kind of Nvidia GPU do you have? If the GPU is slower than the CPU, then offload just makes things worse as CPU ends up waiting for GPU to complete it's tasks.

 

 

It's a GTX2070 Super. Shouldn't be slow, but maybe compared to the Ryzen?

[Miska]

8 minutes ago, toddrhodes said:

It's a GTX2070 Super. Shouldn't be slow, but maybe compared to the Ryzen?

 

No, that should be pretty fine. What kind of filters are you running on it?

 

[toddrhodes]

12 minutes ago, Miska said:

 

No, that should be pretty fine. What kind of filters are you running on it?

 

Going from memory but I believe it's poly-sinc-xtr2-2s, ASDM7EC modulator and at the time, I was only trying DSD64 48k for the rate. I bumped it up after I determined it's faster to gray that out instead of check it, but that's my recollection. I can certainly do more testing on it, however I can help. And I'm not at all concerned about it, was just an observation.

[Miska]

6 minutes ago, toddrhodes said:

Going from memory but I believe it's poly-sinc-xtr2-2s, ASDM7EC modulator and at the time, I was only trying DSD64 48k for the rate. I bumped it up after I determined it's faster to gray that out instead of check it, but that's my recollection. I can certainly do more testing on it, however I can help. And I'm not at all concerned about it, was just an observation.

 

That should certainly work on that GPU without issues. Unless there is some other problem somewhere. Please check you have latest Nvidia drivers installed. If there is some other application loading the GPU a lot, then it could also affect HQPlayer operation. But this should be rare case (playing a heavy game while HQPlayer playing, etc)

 

[toddrhodes]

15 minutes ago, Miska said:

 

That should certainly work on that GPU without issues. Unless there is some other problem somewhere. Please check you have latest Nvidia drivers installed. If there is some other application loading the GPU a lot, then it could also affect HQPlayer operation. But this should be rare case (playing a heavy game while HQPlayer playing, etc)

 

 

I'll take a look. This was a totally fresh O/S install and I am almost positive I was doing this after the GPU driver install, but I don't mind playing :) Heck, I'm just impressed with how well Ryzen is doing on stuff that my old CPU was not handling well at all, and not just where HQP is concerned. It's just a much better experience now overall.

[mirekti]

1 hour ago, Miska said:

 

Are you on Desktop or Embedded?

 

Desktop PDF manual is included in all copies of HQPlayer. On Windows and Linux you can find it under HQPlayer group in Start-menu. On macOS it is included in the DMG and you can place it anywhere you like.

 

Desktop includes also Plot-button to plot the frequency and phase response. This is not yet in Embedded, but is planned feature there as well.

 

 

I am using Desktop. The filter itself works, but I get flat plot. Here are my test settings:

 

I guess I would need to see something similar like in Roon:

 

[mikel]

I was not successful by addition convolution files to the Pipeline.

What am I doing wrong?

Immediate crash when hitting play, and HQPlayer closes.

 

 

[bibo01]

On 12/17/2020 at 10:31 PM, stefano_mbp said:

The results of some tests:

- not using NAA (ie using ALSA Backend)

— flac 16/44.1 delivered as is by Minimserver (without transcoding to wave)

hqplayerd upsampling to

PCM 88.2 gapless ok

PCM 176.4 gapless fails 1 on 2 attempts

PCM 352.8 gapless fails

DSD x1 gapless ok

DSD x2 gapless ok

DSD x4 gapless ok

— alac 16/44.1 transcoded to wave 16/44.1 by Minimserver 

hqplayerd upsampling to

PCM up to 352.8 gapless ok

DSD up to x4 gapless ok

- using NAA (raspberry pi4 4GB/RoPieee XL)

— Alac/flac 16/44.1 transcoded to wave 16/44.1 by Minimserver 

hqplayerd upsampling to

PCM 176.4 and less gapless ok

PCM 352.8 gapless fails

DSD x1/x2/x4 gapless fails

- hqplayerd settings (with/without NAA)

PCM 1x/Nx poly-sinc-ext2

DSD 1x/Nx poly-sinc-ext2 Modulator ADSM7

 

 

@MiskaWhat is the final word on Stefano's report?

It seems to me that there are some odd results with gapless fails with/out NAA.

What's next step?

[toddrhodes]

 

 

So I made some progress today. It seems I had Multicore and CUDA offload set poorly? Since setting Multicore to gray and CUDA to checked, now I'm seeing good activity on both GPU and CPU and doing so, I can get ASDM5EC to run at DSD256. ASDM7EC at 128. Currently it seems to run fine at 256/48, but I've got it at 256/44 to give myself a little headroom, and for the fact that most of my collection is in 44/88, so it just seems to make the most sense for starting out.

 

I also learned a lot about exactly how much bandwidth gets taken up by streaming DSD256 to an NAA. I run gigabit in the house, and I have gigabit Fiber from AT&T. So normally, I have plenty of bandwidth. But unbeknownst to me, a cable I put in between my switch and the PC running Roon/HQP was throttling my autonegotiation down to 100/100. So in periods of high download (say, 95 Mb/s because I was grabbing something off my SFTP) or upload activity on the network, it was essentially saturated and that's where I was getting some of the really bad pausing and stuttering from. It wasn't due to HQ Player or settings or anything, it was just a fluke. Now that I have the cable fixed and there's plenty of bandwidth to go around from that PC, it's much better. 

 

And I haven't yet had a chance to sit down and really focus on it, but even playing around with settings at the PC outside my room, but hearing what was going on in the room - the first time I got DSD256 and ASDM5EC, I stopped what I was doing and just listened at the door to my room. It was pretty amazing, so I can't wait to actually focus on it here shortly.

[WilliamWykeham]

I’m using a trial version of Euphony and am unable to get HQP to see either of my NAA devices - a sms200 and an allo usbridge.  Any thoughts why that might be? Euphony finds the NAA name just fine, but not the NAA device.  I’ve checked that NAA is enabled on both.  

[Traktorist3d]

@Miska , please tell me, for DSC https://puredsd.ru/ in terms of pleasant sound, is the ADSM 7EC better or the ADSM 7EC 512FS better? Which is better to pick up a resampler? Poly sinc EXT 2?

[juanitox]

3 hours ago, Traktorist3d said:

@Miska , please tell me, for DSC https://puredsd.ru/ in terms of pleasant sound, is the ADSM 7EC better or the ADSM 7EC 512FS better? Which is better to pick up a resampler? Poly sinc EXT 2?

don't you have your own taste  in term of pleasant sound  ? 😀

[mikel]

Regarding my post above that the Pipeline Convolution was not working... found that there was a small mistake in the path.

Now that the path is correct, I can plot as well.

Now what happens is that when pipeline is enabled, nothing is playing. I hear a loud click initially, then silence.

Find a attached the filter plot.

[Traktorist3d]

59 minutes ago, juanitox said:

у вас нет собственного вкуса в плане приятного звука? 😀

No, I'm on the general average)

[bogi]

Miska, I am curious how you got this picture (with my IR files) containing both channels 0 and 1, because I am able to show plot for one channel only (otherwise my plot looks the same):

[AnotherSpin]

3 hours ago, juanitox said:

don't you have your own taste  in term of pleasant sound  ? 😀

 

People usually trust other people's opinions more than their own. And the opinions/beliefs of those whom they are willing to listen are borrowed or imposed the very same way. There is nothing "own" in minds. Illusions go in rounds.

[Traktorist3d]

34 minutes ago, AnotherSpin said:

People usually trust other people's opinions more than their own. And the opinions/beliefs of those whom they are willing to listen are borrowed or imposed the very same way. There is nothing "own" in minds. Illusions go in rounds.

As for me, there is simply no time to try everything. even more so convert to DSD 512 and listen

[Miska]

17 hours ago, bibo01 said:

@MiskaWhat is the final word on Stefano's report?

It seems to me that there are some odd results with gapless fails with/out NAA.

What's next step?

 

Remains a bit of mystery, as there are too many variables. But obviously using network makes difference here. Would be good to test on some Intel based NAA capable of full gigabit bandwidth.

 

[Miska]

5 hours ago, bogi said:

Miska, I am curious how you got this picture (with my IR files) containing both channels 0 and 1, because I am able to show plot for one channel only (otherwise my plot looks the same):

 

You can click Plot button for different pipelines without closing the plot window to plot multiple responses.

 

[Miska]

8 hours ago, mikel said:

Regarding my post above that the Pipeline Convolution was not working... found that there was a small mistake in the path.

Now that the path is correct, I can plot as well.

Now what happens is that when pipeline is enabled, nothing is playing. I hear a loud click initially, then silence.

Find a attached the filter plot.

 

Please check that you don't have convolution engine or something like that enabled that is not actually used. Also try with the default overlap-add and without Expand HF to see if this makes difference.

 

It should work if the plot is working. You could email me your log file from a playback attempt with this setup.

 

I wonder why there is 20 dB cut for 20 kHz here. Is there some strong metal dome resonance?

 

[il Carletto]

1 hour ago, Miska said:

 

Remains a bit of mystery, as there are too many variables. But obviously using network makes difference here. Would be good to test on some Intel based NAA capable of full gigabit bandwidth.

 

 

As I wrote in my post yesterday, I notice dropouts using my Mac Mini as Naa too.

 

[Miska]

11 minutes ago, il Carletto said:

As I wrote in my post yesterday, I notice dropouts using my Mac Mini as Naa too.

 

OK, I'm lost on the context who is talking about what issue. This is with HQPlayer Desktop and not about Embedded UPnP and no gapless. In your case it is likely something related to network. If it works without NAA with direct USB and doesn't with NAA, then issue is clearly network or the NAA.

 

Is your NAA running my NAA OS or HQPlayer OS image? If not, please as first thing try with those ones. RasPi prior to 4 is known not to work, so that can be ruled out.