HQPlayer4 EC modulator tips and techniques

[Solstice380]

22 hours ago, bodiebill said:

 

Perhaps, but I am somewhat puzzled as the dropouts are tiny and only occur after 40 seconds or so. And they also occur with dsd128 and temperatures remain moderate (<60C). So I suspect it is something else.

 

Have you tried adjusting the Buffer?  

 

My liquid cooled i7-6700K OC’d to 4.2GHz with 2400 G. Skill Ripjaw RAM and RTX2080 OC GPU can smoothly play polysincshrt-mp, xtr and ext2 filters with EC7 from up to 96/24 to DSD256.  Stuttering starts with 192/24 I tried.  The CPU cores get up to the 70s so I don’t think I want to play it all night but not an issue for a whole album.  Time to build a new machine, this one is 5 years old but the GPU is new and can go right in. 😎

[bodiebill]

13 hours ago, Miska said:

What kind of CAS latency does this memory have compared to earlier one? Both configurations with 4 memory modules?

 

This CPU requires that it is running at high enough boost clocks for the relevant cores in order to perform ASDM7EC successfully.

 

What kind of filter are you using?

 

One module of Apacer DDR4 RAM 2666MHz 8GB. Also tried two modules but that did not remove the dropouts. I believe CAS latency is 19 and the Ballistix I used earlier has lower CL. Could that be it?

Filter used is poly-sinc-ext2.

 

I did get rid of the dropouts by going from Load-Line Calibration level 4 to level 3 in the BIOS (ASRock Z390M-ITX/AC). This applies more voltage. I did not increase the CPU frequency, so I am not sure whether one would consider my tweak as overclocking.

 

[bodiebill]

12 hours ago, Solstice380 said:

Have you tried adjusting the Buffer?

 

HQPlayer buffer is set to maximum: 250.

[Solstice380]

31 minutes ago, bodiebill said:

 

HQPlayer buffer is set to maximum: 250.

 

Oh OK.   So increasing the CPU voltage stopped the dropouts?  That’s interesting.  

[bodiebill]

8 hours ago, Solstice380 said:

Oh OK.   So increasing the CPU voltage stopped the dropouts?  That’s interesting.  

 

Well, I thought so, but then I only checked with redbook files. When I play a 96/24 file I again have dropouts.

@Miska Is it advisable to go back to my previous RAM which has lower CAS latency?

[Miska]

14 hours ago, bodiebill said:

Well, I thought so, but then I only checked with redbook files. When I play a 96/24 file I again have dropouts.

@Miska Is it advisable to go back to my previous RAM which has lower CAS latency?

 

For best performance, both memory channels should be populated with low CAS latency memory. Doesn't necessarily solve the problem, but at least helps.

 

[bodiebill]

22 minutes ago, Miska said:

For best performance, both memory channels should be populated with low CAS latency memory. Doesn't necessarily solve the problem, but at least helps.

 

Thanks Miska. Playing PCM 96/24 material, I tried many changes to the following basic setup:

 

Windows PC with control point (UPPlay or JRiver) =>

Audiolinux Server with HQPlayer embedded upsampling to DSD256 with EC7 =>

Audiolinux endpoint with NAA

 

This I tried:

• DSD128 instead of DSD256
• non-EC7
• no upsampling at all (PCM)
• playing from different source (NAS instead of local HD)
• removing ISO Regen
• changing DAC
• trying different NAA endpoint
• substituting HQPlayer OS for AL on server
• attaching DAC directly to server (ALSA setting)
• trying normal (non-fiber) ethernet connection
• limting ethernet speed to 100Mb on both server and endpoint
• changing kernel in AL
• going back to 2 modules of lower latency Crucial Ballistix RAM
• several overclocking options (hardly relevant as the problem even occurs without upsampling)

but in all cases I have tiny dropouts (1st one after 14 seconds, then some 30 seconds later etc., irregularly). I checked the original file: playing on a mac with VLC there are no dropouts.

 

I hope it can be solved somehow, but I do not know what else to think of...

Any advice will be greatly appreciated!

 

 

[Miska]

30 minutes ago, bodiebill said:

 

Thanks Miska. Playing PCM 96/24 material, I tried many changes to the following basic setup:

 

Windows PC with control point (UPPlay or JRiver) =>

Audiolinux Server with HQPlayer embedded upsampling to DSD256 with EC7 =>

Audiolinux endpoint with NAA

 

This I tried:

• DSD128 instead of DSD256
• non-EC7
• no upsampling at all (PCM)
• playing from different source (NAS instead of local HD)
• removing ISO Regen
• changing DAC
• trying different NAA endpoint
• substituting HQPlayer OS for AL on server
• attaching DAC directly to server (ALSA setting)
• trying normal (non-fiber) ethernet connection
• limting ethernet speed to 100Mb on both server and endpoint
• changing kernel in AL
• going back to 2 modules of lower latency Crucial Ballistix RAM
• several overclocking options (hardly relevant as the problem even occurs without upsampling)

but in all cases I have tiny dropouts (1st one after 14 seconds, then some 30 seconds later etc., irregularly). I checked the original file: playing on a mac with VLC there are no dropouts.

 

I hope it can be solved somehow, but I do not know what else to think of...

Any advice will be greatly appreciated!

 

Best way to figure it out is to simplify everything as much as possible, so local content played by controlling HQPlayer with HQPlayer Client, upsampling turned off to a locally connected DAC (without Regen or other USB gadgets, with a cheap standard USB cable). And get it first working with such setup before adding anything more. Because otherwise there are way too many variables in the picture to figure out where the problem is.

 

But sounds like some I/O issue. Did you try the same DAC with the same cable from a Mac? Have you tried changing the USB cable? You could try booting up HQPlayer OS from a memory stick on another computer with DAC connected to that one, upsampling turned off again.

 

If the dropouts are small, then it may be a lost or broken USB packet (which are 125 µs long).

 

[bodiebill]

11 hours ago, Miska said:

Best way to figure it out is to simplify everything as much as possible, so local content played by controlling HQPlayer with HQPlayer Client, upsampling turned off to a locally connected DAC (without Regen or other USB gadgets, with a cheap standard USB cable). And get it first working with such setup before adding anything more. Because otherwise there are way too many variables in the picture to figure out where the problem is.

 

But sounds like some I/O issue. Did you try the same DAC with the same cable from a Mac? Have you tried changing the USB cable? You could try booting up HQPlayer OS from a memory stick on another computer with DAC connected to that one, upsampling turned off again.

 

If the dropouts are small, then it may be a lost or broken USB packet (which are 125 µs long).

 

I just drastically simplified things by starting another PC (a spare laptop) with HQPlayer embedded. I played the 96/24 PCM file locally without any upsampling or conversion, trying two different DAC's with two different cables (without USB gadgets). In both cases I get the dreaded small dropouts (13s, 55s, etc.).

This happens with HQPe version 4.12.1 as well as 4.13.

 

The same file plays without problems directly on a mac, Windows PC or Chromebook with an applicable software player.

 

Here is a link to the mono 96/24 file (Hungarian String Quartet, Beethoven 1st movement of 1st string quartet):
https://drive.google.com/file/d/15GiDDTk-6bs5Q8K0bSLNLJUeCS9vpYOB/view?usp=sharing 

[Miska]

20 hours ago, bodiebill said:

I just drastically simplified things by starting another PC (a spare laptop) with HQPlayer embedded. I played the 96/24 PCM file locally without any upsampling or conversion, trying two different DAC's with two different cables (without USB gadgets). In both cases I get the dreaded small dropouts (13s, 55s, etc.).

This happens with HQPe version 4.12.1 as well as 4.13.

 

The same file plays without problems directly on a mac, Windows PC or Chromebook with an applicable software player.

 

Here is a link to the mono 96/24 file (Hungarian String Quartet, Beethoven 1st movement of 1st string quartet):
https://drive.google.com/file/d/15GiDDTk-6bs5Q8K0bSLNLJUeCS9vpYOB/view?usp=sharing 

 

I can reproduce the drop-out at 13s position with this file, seeking around it always produces it at the same point, and reason can be found from the log file:

# 2019/12/26 21:35:46 ReadFLACErrorCB(): CRC error

IOW, the file is broken at that position. With VLC on Ubuntu 18.04 I get even worse sounding drop-out at that point.

 

If it doesn't happen on all systems with the same file, I'm not sure what is the reason. Maybe different version of FLAC decoder. Ubuntu has FLAC version 1.3.2. HQPlayer OS has FLAC version 1.3.3.

 

[bodiebill]

43 minutes ago, Miska said:

44 minutes ago, Miska said:

I can reproduce the drop-out at 13s position with this file, seeking around it always produces it at the same point, and reason can be found from the log file:

# 2019/12/26 21:35:46 ReadFLACErrorCB(): CRC error

IOW, the file is broken at that position. With VLC on Ubuntu 18.04 I get even worse sounding drop-out at that point.

 

 

Thanks Jussi for checking, appreciated!

 

So now I am a bit worried as I had dropouts with quite a few files lately. Is it possible that my media disk got corrupted, and hence some files got broken? Is there a program to check integrity of PCM files?

 

I will see if I can locate some of the source files to check their integrity.

 

The good news is that the problem does not seem to arise from lack off juice of the server, I guess.

[Miska]

7 minutes ago, bodiebill said:

So now I am a bit worried as I had dropouts with quite a few files lately. Is it possible that my media disk got corrupted, and hence some files got broken? Is there a program to check integrity of PCM files?

 

I will see if I can locate some of the source files to check their integrity.

 

It is possible, but files rarely get corrupted on disk unless the disk is old. And in such cases it is usually reported by the disk, you could check the disk's SMART data and logs if there have been errors.

 

It is more likely that the corruption has happened when files are being written to the disk, due to RAM problems or other hardware problems when the data is on the way to the disk.

 

FLAC files have some amount of CRC checking for correctness. WAV/AIFF and DSD files don't, so those are more tricky to check. Best way would be to process checksums (MD5, SHA1, SHA256 or similar) against the original ones to see if the checksums match. But that may be hard to achieve if there are no multiple copies of the files.

 

Do you have copies of the files elsewhere? "rsync" command line utility can be used to do a "dry run" 

 

17 minutes ago, bodiebill said:

The good news is that the problem does not seem to arise from lack off juice of the server, I guess.

 

Yes, very unfortunate that it took so much effort to root cause at least this one...

 

[bodiebill]

1 hour ago, Miska said:

Best way would be to process checksums (MD5, SHA1, SHA256 or similar) against the original ones to see if the checksums match. But that may be hard to achieve if there are no multiple copies of the files.

 

I managed to locate the source file and indeed: it had a different md5 checksum and it has no dropouts! It also sounds slightly different, so I am not sure what happened. Maybe it got corrupted, while downloading or later in my system. I hope to find out before long whether this applies to other files also.

 

A relieve though that we were able to rule out most other causes that I suspected earlier. Thanks for helping out!

[jamesg11]

Anyone achieved dsd256 via EC modulators with the sonicTransporter i9?

[seatrope]

@shahed99 any updates in how your 3950X build worked with HQP? 
 

I see @teera_ash had some issues with pauses that weren’t there with 3700X...

[scintilla]

What about the 3.2 Xeon W 8-core that ships with an iMac Pro? It looks like no CUDU offload with the AMD video boards. i'm looking at the improved cooling with the ability to dissipate 500W and comparative cost vs. a fully loaded iMac and if that W processor could be overclocked at all might be able to handle the ECs unles it truly is just a matter of clock cycles/speed.  How has the i9 iMac been working for you Jussi? I need a new box to replace my late-2012 i7. 

[Miska]

44 minutes ago, scintilla said:

How has the i9 iMac been working for you Jussi? I need a new box to replace my late-2012 i7. 

 

My older 27" quad-core i5 iMac was quiet under full load. This new 27" iMac with i9-9900K works fine for running ASDM7EC to DSD256. But down side is that it sounds like hair dryer when doing so. On idle it is quiet, but under CPU load the fans speed up and it becomes noisy. When doing ASDM7EC to DSD256 it is sounding "only" like a hair dryer, not like a vacuum cleaner which it does when building HQPlayer software...

 

[scintilla]

Interesting.  So, I guess the i9 is an option, but my machine is 5 feet from my listening chair, so that could be a problem with fan noise.  I'm inclined to think the Pro is going to be available at discounts soon as Apple prepares for a new product cycle too.  A refurb pro isn't much more than a loaded i9 right now and if it also has enough clock speed to handle ASDM7EC at DSD256, that's enough for me. And it would be silent. Big if tho until someone tries it. The i9 clearly bests the Xeon 8 core on single-core benchmarks. How that plays out in real life remains to be seen. 

[StreamFidelity]

9 hours ago, Miska said:

iMac with i9-9900K works fine for running ASDM7EC to DSD256. But down side is that it sounds like hair dryer when doing so.

 

That's why a fanless Audio PC is a good thing. This can also be done well with an Intel i9-9900K. 😉

[Miska]

7 minutes ago, StreamFidelity said:

That's why a fanless Audio PC is a good thing. This can also be done well with an Intel i9-9900K. 😉

 

It is more about the kind of fans used. For example my big audio PC is silent although it has CPU fan, four case fans, PSU fan and three GPU fans. It would be interesting to see such passive cooled, with 1000W PSU, overclocked GeForce RTX2080Ti and CPU with >120W TDP.

 

Physical design of iMac makes silent cooling difficult. It cannot house large heatsinks and several slowly rotating 15 cm fans. If you have a small heatsink and small high speed fan, you are bound to have lot of noise.

 

But OTOH, one can house noisy HQPlayer server in an air conditioned machine room and just have a fanless NAA in the listening space.

 

[StreamFidelity]

1 hour ago, Miska said:

It would be interesting to see such passive cooled, with 1000W PSU, overclocked GeForce RTX2080Ti and CPU with >120W TDP.

 

It's true. I don't see passive cooling in this configuration either. In my fanless Audio PC DSD 512 is maximum possible. But I can easily run ASDM7EC with DSD 256x48 and convolution filter.

 

1 hour ago, Miska said:

But OTOH, one can house noisy HQPlayer server in an air conditioned machine room and just have a fanless NAA in the listening space.

 

The problem of moving parts of a fan with electromagnetic fields remains. 😉

[bodiebill]

12 minutes ago, StreamFidelity said:

The problem of moving parts of a fan with electromagnetic fields remains. 

 

But is, in my experience, less of a problem, if the server is well (optically) isolated.

[4est]

3 hours ago, StreamFidelity said:

 

That's why a fanless Audio PC is a good thing. This can also be done well with an Intel i9-9900K. 😉

My i9 9900ks is nearly silent using a Noctura cooler, but as Jussi stated, an NAA with the server in another room is the easiest way to accomplish this.

[craighartley]

27 minutes ago, 4est said:

My i9 9900ks is nearly silent using a Noctura cooler, but as Jussi stated, an NAA with the server in another room is the easiest way to accomplish this.

Which Noctua cooler do you use? I’m looking for a quieter option for my 9900ks. 

[Miska]

4 hours ago, StreamFidelity said:

The problem of moving parts of a fan with electromagnetic fields remains. 😉

 

I have never seen that as an issue. Fans run from independent PWMs and there are much bigger fields from other parts. If you worry about such, NAA is there to fix that as well as acoustic noise.