HQ Player

[ted_b]

17 minutes ago, asdf1000 said:

 

You've heard the latest release , 4.2.1 ?

 

I posted about 4.21.1

[austinpop]

54 minutes ago, Miska said:

While the code between the versions in that case is absolutely the same without any changes. Since there are no changes involved it is some black magic outside of computer science and I cannot do anything about it.

 

Respectfully asking: even if the code is the same, are there any deployment or runtime differences, or parameter differences, between the 4.90, 4.91, and 4.92 versions? For example, number of threads, thread priorities, pool sizes, thread order?

 

None of us believe in black magic, but our ears aren't lying either. We do know that things like processor affinity, threading parameters etc can make profound sonic differences, without changing a line of code.

 

We are offering you our ears to get to the bottom of this mystery. Without revealing any proprietary information, could you look at possible root causes, and build some test versions - say 4.9a, b, c etc. Then let us try them out and give you feedback on SQ. Perhaps this will isolate the cause.

 

It's your call, of course.

[guiltyboxswapper]

13 minutes ago, austinpop said:

We are offering you our ears to get to the bottom of this mystery. Without revealing any proprietary information, could you look at possible root causes, and build some test versions - say 4.9a, b, c etc. Then let us try them out and give you feedback on SQ. Perhaps this will isolate the cause.

 

 

To be fair to Miska, he did clearly state he only changed a single line of code.  Now I know from past experience that compiler systems aren't entirely deterministic, but that is a whole can-o-worms (read: massively time consuming) over a single line of code.

[Miska]

17 minutes ago, austinpop said:

Respectfully asking: even if the code is the same, are there any deployment or runtime differences, or parameter differences, between the 4.90, 4.91, and 4.92 versions? For example, number of threads, thread priorities, pool sizes, thread order?

 

No... Only bug fixes in the NAA code and one fix in the manual PDF.

 

17 minutes ago, austinpop said:

Without revealing any proprietary information, could you look at possible root causes, and build some test versions - say 4.9a, b, c etc.

 

And how would these be different from each other?

 

[jimdukey]

"None of us believe in black magic, but our ears aren't lying either."

Or are they?

You "hear" it, Miska doesn't.

I think ( believe ) that Belief plays a role,

You believe there is a difference, he doesn't.

I remember when the EC Modulators were introduced.

People said the DSD 128 with EC was better than DSD 512 without.

I don't think That lasted very long...

Making Audio Comparisons is very difficult,

stopping and starting, you lose the memory of the first test before the second one starts.

I've even changed my mind about what I preferred, only to change back later.

It's VERY subjective, and maybe even System Dependent.

I wonder what people will be thinking about HQ in 2 weeks, 2 months?

I'm stuck at 4.8.1 until I get a new Computer, which won't be soon, knock on wood.

I ilsten at DSD 128, and am very happy with HQ, as i have always been.

It's fun to sit back and read the 48 Audiophile Style emails in my box each morning recently!

I'm sure this will all shake out before I need to upgrade!

Maybe there will be Secret Sauce Packets distributed soon...

 

[Account Closed]

8 minutes ago, austinpop said:

 

Respectfully asking: even if the code is the same, are there any deployment or runtime differences, or parameter differences, between the 4.90, 4.91, and 4.92 versions? For example, number of threads, thread priorities, pool sizes, thread order?

 

None of us believe in black magic, but our ears aren't lying either. We do know that things like processor affinity, threading parameters etc can make profound sonic differences, without changing a line of code.

 

We are offering you our ears to get to the bottom of this mystery. Without revealing any proprietary information, could you look at possible root causes, and build some test versions - say 4.9a, b, c etc. Then let us try them out and give you feedback on SQ. Perhaps this will isolate the cause.

 

It's your call, of course.

I of course would also like to get to the bottom of this as I am hearing the same thing as you and others posting here. But, I don't think he can or will be able to go much deeper into this for a lot of reasons, most of which I am sure you are well aware of given your background. There are too many variables to even begin to contemplate and testing them would get into the realm of ears vs scopes which is not something I would want to take on if I was in his position. My guess is that this is the end of his involvement in this issue and that he will move forward with technical code based development.

 

As you said in a prior post, if like 4.90 then hang on to it so you can always go back if future releases are not to your liking. 

 

[Miska]

4 minutes ago, guiltyboxswapper said:

To be fair to Miska, he did clearly state he only changed a single line of code.  Now I know from past experience that compiler systems aren't entirely deterministic, but that is a whole can-o-worms (read: massively time consuming) over a single line of code.

 

There are total two bug fixes between 4.9.0 and 4.9.2. 4.9.1 fixed one and 4.9.2 fixed second. The 4.9.1 -> 4.9.2 change is not even near the code path that gets run during playback. And none of the changes are near code paths that get run when not using NAA.

 

[drjimwillie]

Earlier, some users with low amounts of RAM said that the 4.9 was increasing their ram use to where they had an issue.

possibly, 4.9 is doing something different with RAM?

[Miska]

3 minutes ago, jimdukey said:

I remember when the EC Modulators were introduced.

People said the DSD 128 with EC was better than DSD 512 without.

 

EC modulators are objectively clearly better, with the cost of 3x heavier work. Something I can clearly analyze from the data coming out of HQPlayer.

 

OTOH, good thing is that DSD256 is sweet spot for many DACs (including the newest AKM's 4499 flagship) and that happens to be possible with EC modulators as well.

 

8 minutes ago, jimdukey said:

I'm stuck at 4.8.1 until I get a new Computer, which won't be soon, knock on wood.

 

Hmm, why is that?

 

[austinpop]

Yes, the message is loud and clear.

[Miska]

4 minutes ago, drjimwillie said:

Earlier, some users with low amounts of RAM said that the 4.9 was increasing their ram use to where they had an issue.

possibly, 4.9 is doing something different with RAM?

 

Couple of megabytes more consumption when using NAA output. But no big differences. There is quite a bit of difference between earlier versions and 4.9 in terms of NAA output. But between 4.9 versions the differences are just bug fixes.

 

[seeteeyou]

Instead of connecting the USB output to USB DAC, I guess that maybe we could connect that to the OTG port of Raspberry Pi 4 instead

 

https://www.diyaudio.com/forums/pc-based/342070-linux-usb-audio-gadget-rpi4-otg.html

 

Then run stuff like arecord that's included in alsa-utils, repeat the same steps for 4.9.0 / 4.9.1 / 4.9.2 etc.

 

Finally compare 3 files with DeltaWave

 

https://deltaw.org

https://deltaw.org/DeltaWaveSetup.zip

 

Here's are some examples

 

https://www.symphonic-mpd.com/forum/topic/126/deltawave

https://www.symphonic-mpd.com/release/files/deltawave_comparison_20200918.zip

https://www.symphonic-mpd.com/release/files/deltawave_comparison_20201018.zip

http://mimizukobo.sakura.ne.jp/articles/articles029.html#003

 

 

 

[drjimwillie]

Hi Jussi,
 

Will you please help me understand how Roon communicates with HQPlayer?

 

I assume Roon, with HQplayer as it’s engine sees HQplayer as an internal device?

 

There are two instances when this happens:

1. Roon and HQplayer on the same device.

2. Roon core on one computer and HQ player on your server.

 

also, there are two different file/location instances.

1.  Local files stored on your server OR stored on the local network on your NAS.

2.  Streamed data from i.e. Qobuz. 
 

Are these presented to HQplayer as a file or a stream or a URL?

 

Is it the same if Roon core and HQ player are on the same or different computers?

Is it the same if it is a local file or if it is streamed?

 

Does Roon do anything to the file or the stream before it hands it to HQplayer?

 

Does Roon do anything to the file or the stream or the data after HQplayer is done with it?

Thank you,

Will

[Miska]

18 minutes ago, drjimwillie said:

I assume Roon, with HQplayer as it’s engine sees HQplayer as an internal device?

 

There are two instances when this happens:

1. Roon and HQplayer on the same device.

2. Roon core on one computer and HQ player on your server.

 

Yes, both cases work the same, but in case (1) you can use virtual "localhost" IP that avoids bunch of protocol stack and actually sending something like ethernet packets anywhere. The difference is at OS level, for the applications it doesn't make a difference.

 

20 minutes ago, drjimwillie said:

also, there are two different file/location instances.

1.  Local files stored on your server OR stored on the local network on your NAS.

2.  Streamed data from i.e. Qobuz. 
 

Are these presented to HQplayer as a file or a stream or a URL?

 

For HQPlayer it always looks the same, a stream from Roon. (by the way, URL is just something that turns into stream)

 

22 minutes ago, drjimwillie said:

Is it the same if Roon core and HQ player are on the same or different computers?

 

From point of both applications yes. From actual network traffic point of view there is of course difference. Since when both are on the same computer there's nothing being sent out of the computer.

 

23 minutes ago, drjimwillie said:

Is it the same if it is a local file or if it is streamed?

 

Yes, from HQPlayer point of view it looks the same.

 

23 minutes ago, drjimwillie said:

Does Roon do anything to the file or the stream before it hands it to HQplayer?

 

Depends on your settings. But at least it is decoded. So you can play MP3, AAC or what ever Roon supports, but HQPlayer doesn't.

 

24 minutes ago, drjimwillie said:

Does Roon do anything to the file or the stream or the data after HQplayer is done with it?

 

No, for Roon, HQPlayer is like a black hole for data. Once the data comes into HQPlayer it is out of Roon's scope.

 

[Miska]

40 minutes ago, seeteeyou said:

Then run stuff like arecord that's included in alsa-utils, repeat the same steps for 4.9.0 / 4.9.1 / 4.9.2 etc.

 

Data, apart from dither noise is the same in all these cases, unless 4.9.0 happened to skip a block...

 

If you get some notable differences, then there is something wrong in your test setup.

 

P.S. Note that on every playback on the same version you get different dither noise too.

[drjimwillie]

46 minutes ago, Miska said:

For HQPlayer it always looks the same, a stream from Roon. (by the way, URL is just something that turns into stream).

 

Thank you for the answer.

if I have Roon core on a  separate computer and a file/song on my NAS and HQplayer on my server, then If I am understanding you correctly Roon open that file on the NAS and turns it into a stream, then streams is it over ethernet to my server for HQplayer to process. It does not tell HQ player where to find that file and then HQplayer opens the file?

 

and, it is the same if the file is streamed from Qobuz?

[Miska]

13 minutes ago, drjimwillie said:

if I have Roon core on a  separate computer and a file/song on my NAS and HQplayer on my server, then If I am understanding you correctly Roon open that file on the NAS and turns it into a stream, then streams is it over ethernet to my server for HQplayer to process. It does not tell HQ player where to find that file and then HQplayer opens the file?

 

Yes, exactly. So you see the difference between playing files with HQPlayer natively.

 

14 minutes ago, drjimwillie said:

and, it is the same if the file is streamed from Qobuz?

 

Yes...

 

[glitchesbrew]

7 hours ago, Miska said:

How long did you wait? It may take a while to initialize, especially if you are asking it to do conversion between rate families.

 

You could try poly-sinc-xtr-mp-2s instead, it likely starts much faster.

 

A couple of minutes. Again I was just surprised because I thought that filter was lighter thatn let's say Closed Form 16M.

The 2s version works fine. 

[Miska]

5 minutes ago, glitchesbrew said:

A couple of minutes. Again I was just surprised because I thought that filter was lighter thatn let's say Closed Form 16M.

The 2s version works fine. 

 

May be lighter to process, depending on case. But initialization may be heavier. closed-form is very quick to initialize.

 

[tonym]

Miska , as the Devialet Dac upsamples to 24/192 , but as you observed there is an issue with the Devialet Air Ethernet driver , should I set HQPlayer to 22bit. Or is there a better solution ? Its odd because I've never seen anyone raise the issue before

[Lbameule]

4 minutes ago, tonym said:

Miska , as the Devialet Dac upsamples to 24/192 , but as you observed there is an issue with the Devialet Air Ethernet driver , should I set HQPlayer to 22bit. Or is there a better solution ? Its odd because I've never seen anyone raise the issue before

Tony, you shouldn't need to use the Air driver. Are you sure that if you connect directly it doesn't work? I have a Dev too, and I don't use it. I never had an issue...

[tonym]

My server is in a different room to the Devialet , the server is connected to a switch.

The Devialet is connected to the ethernet wall socket in the other room.

I have HQplayer enabled in Roon and the Devialet Air driver selected in HQPlayer

[AudioDoctor]

@Miska Would your HQPlayer OS image run and work on a 2011 Mac Mini?

[Lbameule]

Please, do try turning off the Air driver, and seeing what appears in your HQP settings. My setup is different, I use a MicroRendu between my server and the Dev, so what is seen is the mR, but please do try and see if the Dev is seen directly, without Air working.

[Miska]

13 minutes ago, tonym said:

Miska , as the Devialet Dac upsamples to 24/192 , but as you observed there is an issue with the Devialet Air Ethernet driver , should I set HQPlayer to 22bit. Or is there a better solution ? Its odd because I've never seen anyone raise the issue before

 

I have not heard about it before either. You don't need to touch DAC Bits, you can leave it at "Default" or set it to 24 for future. HQPlayer doesn't allow setting DAC Bits to more than it has detected, only less. So the result is now 22.

 

But it would be good to notify Devialet about this if they'd possibly fix it in future release.