HQ Player

[r_g]

Here is the Logfile.

I called hqp-control.exe exactly at 00:03:00

HQPlayer.log

[kelvinwsy]

On 3/2/2019 at 6:05 PM, Miska said:

 

So LKS is Amanero?

 

Because then I'd suspect the reason for ASIO driver not switching to DSD mode could be because the Amanero may have been flashed with one of the Linux-compatible firmware versions. Meaning that it won't work correctly with the Windows driver...

 

But overall, I don't think logs tell much more that what you said. If the driver doesn't switch to DSD mode it is sort of black box why it happens.

Hi - I am having a similar issue with my new SMSL VM D1 Dac - Connected to my HQPlayer Server USB direct to the VM D1 Dac, DSD 512 appears No Problem

1. When I use the SOTM SM200 Ultra Neo NAA Endpoint, NO DSD appears at all in Native Mode. Only when I ticked DOP, then DSD appears but ONLY up to DSD256.

2 . Using my other Win10 Pro NAA minipc, connected to the VM D1 DAC, the DSD option appears on the Setttings menu BUT again ONLY UP TO DSD256.

 

So I have contacted SMSL Support on the inability of the D1 DAC to happen native DSD on Linux and  on Windows DSD 512/ Waiting for an answer. 

 

So what does it take an End-user to get Amanero to get full Linux compatibility with DSD 512?

Kelvin

 

[Miska]

1 hour ago, kelvinwsy said:

Hi - I am having a similar issue with my new SMSL VM D1 Dac - Connected to my HQPlayer Server USB direct to the VM D1 Dac, DSD 512 appears No Problem

1. When I use the SOTM SM200 Ultra Neo NAA Endpoint, NO DSD appears at all in Native Mode. Only when I ticked DOP, then DSD appears but ONLY up to DSD256.

 

This is not similar issue though...

 

This tells that the Linux kernel in sMS-200 is missing native DSD support that is compatible with your DAC.

 

1 hour ago, kelvinwsy said:

2 . Using my other Win10 Pro NAA minipc, connected to the VM D1 DAC, the DSD option appears on the Setttings menu BUT again ONLY UP TO DSD256.

 

Did you remember to turn off DoP? Is your HQPlayer Server also running Win10 Pro with the same driver version?

 

Please try booting my NAA image instead on this computer and see if you get DSD512 without DoP that way?

 

[Miska]

9 hours ago, r_g said:

Here is the Logfile.

I called hqp-control.exe exactly at 00:03:00

HQPlayer.log 9.04 kB · 1 download

 

Seems like it crashes somewhere on the way from control interface to loading the configuration file, because there's no report about loading of the file... However there's one earlier successful configuration load before...

 

I'll look into this again.

 

[kelvinwsy]

HI Jussi

No need to do the last suggestion. This Minix Neo Minipc NAA endpoint has been my workhorse for at least 2 years. No problem getting native DSD 512 hooked up to the iFi iDSD Micro BL. It has to be the linux Eunhasu OS for the SOTM SM200 Ultra. There is a list of DSD512 capable DAC's listed on the Eunhasu manual webpage. Obviously the SMSL VM D1 is not there But i will check again. I have asked the local dealer whom I bought the SM200 through. Let him ask SOTM Audio for the answer. Maybe I need to flash a new version of the Eunhasu OS which is up to 4.56 from m 4.32 but there were so many bugs that  tried up to 4.32 and gave up.

 

Yes I will try your bootable image on the MINIX Minipc this evening. 

[bibo01]

@Miska Just a minor problem...

I have a folder.jpg in each of all my album folders.

Somehow, one cover does not display when I load the album. This folder.jpg is only 400x400 and it's not corrupted (it opens easily in other applications).

I opened HQPlayer.xml to check and the cover path is correct, which obviously is the same of the tracks. The tracks playback fine, but the image does not display (even in Touch mode). Moreover, if I open the tracks with MP3Tag, I noticed that they already have the cover inside, but still it does not display in HQPlayer.

Any suggestion?   

[Miska]

5 minutes ago, bibo01 said:

@Miska Just a minor problem...

I have a folder.jpg in each of all my album folders.

Somehow, one cover does not display when I load the album. This folder.jpg is only 400x400 and it's not corrupted (it opens easily in other applications).

I opened HQPlayer.xml to check and the cover path is correct, which obviously is the same of the tracks. The tracks playback fine, but the image does not display (even in Touch mode). Moreover, if I open the tracks with MP3Tag, I noticed that they already have the cover inside, but still it does not display in HQPlayer.

Any suggestion?   

 

Does the directory name(s) have anything special? The embedded cover should be displayed when playing regardless, if the playback succeeds then the metadata has also been read. Cover flow in touch mode uses detached covers for performance reasons. While the play view displays embedded cover when such is available and if not, falls back to detached.

 

[bibo01]

9 minutes ago, Miska said:

 

Does the directory name(s) have anything special? The embedded cover should be displayed when playing regardless, if the playback succeeds then the metadata has also been read. Cover flow in touch mode uses detached covers for performance reasons. While the play view displays embedded cover when such is available and if not, falls back to detached.

 

Yes, that's what I thought too.

Anyhow, I just found out the culprit. Inside the album there was also front_cover.png file which probably was disturbing HQPlayer because it was over 10MB! I simply renamed it and all it's back to normal.

[Hensema]

On 2/27/2019 at 6:45 PM, Miska said:

You could also compare against poly-sinc-mqa which begins to roll off earlier, but not as steep and doesn't have as much attenuation. But gives relatively very short impulse response in time domain.

 

I never tried poly-sinc-mqa-mp because I don’t have mqa files. I was interested in how a less steep filter sounds. Very natural and easy listening to me!  It is my new favorite! PCM 44.1 to 352khz on a R2R dac.

[r_g]

23 hours ago, Miska said:

 

Seems like it crashes somewhere on the way from control interface to loading the configuration file, because there's no report about loading of the file... However there's one earlier successful configuration load before...

 

I'll look into this again.

 

 

Thank you so far.

 

As I said, I can reproduce the crash on my Computer because it's happening very regularly..

So if you have a HQP-version with an 'extended logfile' I could run it and send you back the extended log.

It's just a thought.

 

 

 

 

 

[Miska]

11 minutes ago, r_g said:

As I said, I can reproduce the crash on my Computer because it's happening very regularly..

So if you have a HQP-version with an 'extended logfile' I could run it and send you back the extended log.

It's just a thought.

 

I don't have anything more than what is already in there, rest would need to be run under debugger.

 

I tried yesterday to reproduce it on Windows by loading different configurations with hqp-control but I didn't manage to reproduce the crash. The code has not been changed since it was introduced either... But anyway I made some changes to the code in question and that will be included in next release. Just have to hope it helps...

 

[521ztz]

Hi Miska,

 

I was using i9-9900k @5Ghz to run poly-sinc-xtr, ASDM7,  44.1k x 512, the sound quality was terrific and my PC could play music smoothly without stuttering, but the CPU usage was around 50%. I tried to turn on cuda offload in Hqplayer but my GTX 1080 could not handle it (with 100% GPU usage), the music stuttered.

 

So I decided to bought a Quadro K6000 for CUDA offload.

 

Now I have both GTX1080 and K6000 running in my PC. But when I turn on CUDA offload in HQplayer, the GPU usage of K6000 is 100% but the sound still stuttered (same as cuda offload to GTX 1080 before).

 

It's weired since the K6000's double-precision should be much more powerful  than the 9900k and GTX 1080.

 

I've downloaded and installed Cuda Toolkit from Nvidia.

 

Did I miss something? I tried to turn on K6000's double-precision in Nvidia control panel but I could not found this option. Does it matter?

 

Appreciate if I can have you input.

 

Many Thanks.

ztz

 

[521ztz]

It's strange that HQplayer often ceases responding on my PC. 

Miska could you please have a look at my log file?HQPlayer.log

[r_g]

5 hours ago, Miska said:

 

I don't have anything more than what is already in there, rest would need to be run under debugger.

 

I tried yesterday to reproduce it on Windows by loading different configurations with hqp-control but I didn't manage to reproduce the crash. The code has not been changed since it was introduced either... But anyway I made some changes to the code in question and that will be included in next release. Just have to hope it helps...

 

 

OK I wait till the next release and will report if it helps.

 

btw.. when calling hqp-control it's not important to really change any setting. Even if the settings in the new xml-file are the same as already loaded, HQP is crashing.

[Yviena]

@Miska Regarding dac bits that you recommended i set to 17 bit, wouldn't sign magnitude architecture  affect linearity differently, soekris mentioned that the dac1541 is linear down to -140db, and setting it to 18 bits sounds bettter.

[Miska]

8 minutes ago, Yviena said:

@Miska regarding dac bits that you recommended i set to 17 bit, wouldn't sign magnitude architecture  affect linearity differently, soekris mentioned that the dac1541 is linear down to -140db.

 

Those measurement plots on the site you referred was showing linearity to about -100 dB. I don't myself have Soekris DAC to test with.

 

But I doubt it would be linear to -140 dB, so far I have not seen any R2R doing that. Holo Spring is linear to -120 dB (20-bit) and it has additional R2R linearity compensation network. -140 dB would require resistors that are more accurate than 0.00001% (combined errors with any input code combination may not exceed this).

 

[Miska]

23 minutes ago, r_g said:

btw.. when calling hqp-control it's not important to really change any setting. Even if the settings in the new xml-file are the same as already loaded, HQP is crashing.

 

I didn't manage to crash it. But it could be that it is somehow related to some particular setting.

 

[Miska]

3 hours ago, 521ztz said:

Now I have both GTX1080 and K6000 running in my PC. But when I turn on CUDA offload in HQplayer, the GPU usage of K6000 is 100% but the sound still stuttered (same as cuda offload to GTX 1080 before).

 

What settings are you trying to run there?

 

First GPU I have that can do poly-sinc-xtr from 44.1k and 48k to 44.1x512 is RTX2080. What particular improvement in the GPU architecture made that happen, I don't know. (Turing is sort of cost-optimized version of Volta) Maybe it is related to this new feature although I don't do so much integer, but memory address calculations are part of integer though:

• concurrent execution of integer and floating point operations

Not sure how much latency they had earlier for switching between the two. Also memory access speeds have been improving all the time.

 

3 hours ago, 521ztz said:

It's weired since the K6000's double-precision should be much more powerful  than the 9900k and GTX 1080.

 

Problem is just that synthetic benchmarks don't correlate very well with real world applications. Kepler is older architecture, so it doesn't have all the features of for example Volta. In benchmarks for example there's usually single task running on GPU with fairly small data set and no communication with CPU. While HQPlayer has somewhat larger data set and runs multiple GPU tasks in parallel and actively communicates with the CPU. Multitasking is one area where Nvidia has been improving in recent generations.

 

3 hours ago, 521ztz said:

I've downloaded and installed Cuda Toolkit from Nvidia.

 

This is not necessary... You only need latest drivers.

 

3 hours ago, 521ztz said:

Did I miss something? I tried to turn on K6000's double-precision in Nvidia control panel but I could not found this option. Does it matter?

 

I don't think so, if HQPlayer runs on the GPU then it is all fine from that perspective. There's no such option, all the features are always enabled.

 

[Yviena]

24 minutes ago, Miska said:

 

Those measurement plots on the site you referred was showing linearity to about -100 dB. I don't myself have Soekris DAC to test with.

 

But I doubt it would be linear to -140 dB, so far I have not seen any R2R doing that. Holo Spring is linear to -120 dB (20-bit) and it has additional R2R linearity compensation network. -140 dB would require resistors that are more accurate than 0.00001% (combined errors with any input code combination may not exceed this).

 

Hmm i found another another measurement of the dac1421, it's the unbalanced version so it should be applicable for the dac1541 too:

In this one the soekris looks linear to 18 bits correct me if i'm wrong.

[Miska]

6 minutes ago, Yviena said:

Hmm i found another another measurement of the dac1421, it's the unbalanced version so it should be applicable for the dac1541 too:

In this one the soekris looks linear to 18 bits correct me if i'm wrong.

 

Yes, looks like linear to -100 or -110 dB, so 17 or 18 bits. Either way, not a performance limiting factor when combined with noise shaping at high enough rate. Yggdrasil seems to just reach RedBook's -96 dB so that would work best with 16-bit and noise shaping.

 

[Johnseye]

On 8/16/2018 at 2:34 PM, Miska said:

 

If you want something really long that is like M-Scaler, then closed-form or poly-sinc-xtr would be closest.

 

 

Hey there Jussi, can you please explain the differences between poly-sinc-xtr, sinc-M and closed-form-M filters?

and by really long, do you mean taps?

[Miska]

47 minutes ago, Johnseye said:

Hey there Jussi, can you please explain the differences between poly-sinc-xtr, sinc-M and closed-form-M filters?

and by really long, do you mean taps?

 

Yes, number of taps = length.

 

ext2 and sinc-M are similar, but with different length. xtr cuts a bit higher up (see plots earlier in the chain). All these go into apodizing category. While closed-form is strictly non-apodizing.

 

[Johnseye]

1 hour ago, Miska said:

 

Yes, number of taps = length.

 

ext2 and sinc-M are similar, but with different length. xtr cuts a bit higher up (see plots earlier in the chain). All these go into apodizing category. While closed-form is strictly non-apodizing.

 

 

xtr has 1M taps in PCM but ext2 doesn't have that many correct?  Does sinc-M have 1M taps?  I've been assuming the M stands for million.

[freesteve]

Miska,  Is there an audible difference with HQ Player & HQPlayer Pro ?

[Yviena]

3 hours ago, Miska said:

 

Those measurement plots on the site you referred was showing linearity to about -100 dB. I don't myself have Soekris DAC to test with.

 

But I doubt it would be linear to -140 dB, so far I have not seen any R2R doing that. Holo Spring is linear to -120 dB (20-bit) and it has additional R2R linearity compensation network. -140 dB would require resistors that are more accurate than 0.00001% (combined errors with any input code combination may not exceed this).

 

Soekris mentions here https://www.audiosciencereview.com/forum/index.php?threads/soekris-dam1021-r2r-dac-measurements.2324/post-64278 that

 

Quote

You need to distinguish between absolute linearity (which create the distortion) and level linearity, basically in a Sign Magnitude DAC the distortion is relative to the actual signal level, while in a non Sign Magnitude DAC the distortion is relative to the full signal level, increasing as the signal level goes down.

Does this change things or is linearity still 17-18 bits ?