HQPlayer Linux Desktop and HQplayer embedded

[matthias]

16 hours ago, Em2016 said:

The upcoming Anubis (with DXD/DSD DAC option) is rumoured to be just a little more than half the price of Hapi + DA8P... Rumours may be all over the place at this stage though.

 

Full info coming in a few weeks.

 

Thanks for the info.

I read the thread on Gearslutz, very interesting.

Do you know which DAC chip Merging use in the DA8P?

 

Matt

[Sigi]

Hi,

here is the log for 8 channels. There is a mistake in the assignment I think

In Aneman I am not able to connect signal 2 with channel 8

The other I can configure. And in the Matrix I think all is ok. What wonders me, is the fact that Filter 1 results in a wav. File described as 0

2019/03/04 19:01:27 Playlist clear
& 2019/03/04 19:01:27 Playlist add directory: /media/windows1/Ulf 22.1.2019/Diego Urcola/Mates
& 2019/03/04 19:01:27 Play
+ 2019/03/04 19:01:27 Playback engine running
 2019/03/04 19:01:27 Rate or blocksize change triggered
 2019/03/04 19:01:27 Rate: 44100, block size: 4704, frame size: 588
 2019/03/04 19:01:27 Block size: 4704 (sample: 4)
 2019/03/04 19:01:27 Playback engine ratio: 2.17687
 2019/03/04 19:01:27 Filter: poly sinc
 2019/03/04 19:01:27 Dither: triangular probability density function for 32-bit
 2019/03/04 19:01:27 Set volume: -3
 2019/03/04 19:01:27 Matrix pipeline 0: 0 -> 0 0/1
 2019/03/04 19:01:27 Matrix pipeline 1: 1 -> 1 0/1
 2019/03/04 19:01:27 Matrix pipeline 2: 0 -> 2 0/1
 2019/03/04 19:01:27 Matrix pipeline 3: 1 -> 3 0/1
 2019/03/04 19:01:27 Matrix pipeline 4: 0 -> 4 0/1
 2019/03/04 19:01:27 Matrix pipeline 5: 1 -> 5 0/1
 2019/03/04 19:01:27 Matrix pipeline 6: 0 -> 6 0/1
 2019/03/04 19:01:27 Matrix pipeline 7: 1 -> 7 0/1
 2019/03/04 19:01:27 Matrix pipeline 8: 8 -> 8 0/1
 2019/03/04 19:01:27 Matrix pipeline 9: 9 -> 9 0/1
 2019/03/04 19:01:27 Matrix pipeline 10: 10 -> 10 0/1
 2019/03/04 19:01:27 Matrix pipeline 11: 11 -> 11 0/1
 2019/03/04 19:01:27 Matrix pipeline 12: 12 -> 12 0/1
 2019/03/04 19:01:27 Matrix pipeline 13: 13 -> 13 0/1
 2019/03/04 19:01:27 Matrix pipeline 14: 14 -> 14 0/1
 2019/03/04 19:01:27 Matrix pipeline 15: 15 -> 15 0/1
 2019/03/04 19:01:27 Matrix pipeline 16: 16 -> 16 0/1
 2019/03/04 19:01:27 Matrix pipeline 17: 17 -> 17 0/1
 2019/03/04 19:01:27 Matrix pipeline 18: 18 -> 18 0/1
 2019/03/04 19:01:27 Matrix pipeline 19: 19 -> 19 0/1
 2019/03/04 19:01:27 Matrix pipeline 20: 20 -> 20 0/1
 2019/03/04 19:01:27 Matrix pipeline 21: 21 -> 21 0/1
 2019/03/04 19:01:27 Matrix pipeline 22: 22 -> 22 0/1
 2019/03/04 19:01:27 Matrix pipeline 23: 23 -> 23 0/1
 2019/03/04 19:01:27 Matrix pipeline 24: 24 -> 24 0/1
 2019/03/04 19:01:27 Matrix pipeline 25: 25 -> 25 0/1
 2019/03/04 19:01:27 Matrix pipeline 26: 26 -> 26 0/1
 2019/03/04 19:01:27 Matrix pipeline 27: 27 -> 27 0/1
 2019/03/04 19:01:27 Matrix pipeline 28: 28 -> 28 0/1
 2019/03/04 19:01:27 Matrix pipeline 29: 29 -> 29 0/1
 2019/03/04 19:01:27 Matrix pipeline 30: 30 -> 30 0/1
 2019/03/04 19:01:27 Matrix pipeline 31: 31 -> 31 0/1
 2019/03/04 19:01:27 Longest IR: 65536
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_0.wav for channel 0 = -311.293 dB
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_1.wav for channel 1 = -nan dB
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_3.wav for channel 3 = -nan dB
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_2.wav for channel 2 = -nan dB
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 0 = -0.926993 dB (-0.00132201 - 0.00167188)
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_4.wav for channel 4 = -317.263 dB
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 1 = -1.2488 dB (-0.00149285 - 0.00155767)
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_5.wav for channel 5 = -320.098 dB
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 3 = -2.4241 dB (-0.0097267 - 0.0123082)
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_6.wav for channel 6 = -nan dB
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 2 = -3.46833 dB (-0.0111449 - 0.0129099)
 2019/03/04 19:01:27 Impulse response read from file /home/audiolinux/.hqplayer/impulse_7.wav for channel 7 = -319.092 dB
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Impulse response resampled with ratio: 0.91875
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 5 = -6.6964 dB (-0.0292769 - 0.0294692)
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 4 = -6.80654 dB (-0.0291885 - 0.0276394)
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 6 = -0.270932 dB (-0.591976 - 0.465639)
 2019/03/04 19:01:27 Convolution filter of length 61234 initialized for channel 7 = -0.84319 dB (-0.621242 - 0.245483)
 2019/03/04 19:01:27 Matrix processing enabled
 2019/03/04 19:01:27 Offload: resampler=disabled convolution=disabled
 2019/03/04 19:01:27 Initialization complete, starting audio engine
+ 2019/03/04 19:01:27 ALSA output engine starting...
 2019/03/04 19:01:27 ALSA output set hardware parameters for requested rate: 96000
 2019/03/04 19:01:27 ALSA output channels: 2 - 128
 2019/03/04 19:01:27 ALSA output active channels: 8
 2019/03/04 19:01:27 ALSA output number of periods: 12
 2019/03/04 19:01:27 ALSA output period times: 5333 - 5334
 2019/03/04 19:01:27 ALSA output period sizes: 512 - 512
 2019/03/04 19:01:27 ALSA output period time: 5333
 2019/03/04 19:01:27 ALSA output period size: 512
+ 2019/03/04 19:01:27 ALSA output engine started at: 96000 (96000)
+ 2019/03/04 19:01:27 ALSA output engine running...
& 2019/03/04 19:01:27 Stop...
 2019/03/04 19:01:27 Parallel threads: 4
 2019/03/04 19:01:27 Nested parallelism: 0
 2019/03/04 19:01:27 Parallel pipelines: 8
- 2019/03/04 19:01:27 ALSA output engine stopping...
- 2019/03/04 19:01:28 ALSA output engine stop request...
# 2019/03/04 19:01:29 clALSAEngine::Stop(): snd_pcm_wait()
? 2019/03/04 19:01:39 snd_pcm_drain() failed: (-5) Unknown error -5
- 2019/03/04 19:01:39 ALSA output engine stopped
- 2019/03/04 19:01:39 Playback engine stopped
& 2019/03/04 19:01:39 ...stopped
 2019/03/04 19:01:39 Set volume: -3
& 2019/03/04 19:01:39 Playlist clear
& 2019/03/04 19:01:39 Playlist add directory: /media/windows1/Ulf 22.1.2019/Diego Urcola/Mates
& 2019/03/04 19:01:39 Play
+ 2019/03/04 19:01:39 Playback engine running
 2019/03/04 19:01:39 Rate or blocksize change triggered
 2019/03/04 19:01:39 Rate: 44100, block size: 4704, frame size: 588
 2019/03/04 19:01:39 Block size: 4704 (sample: 4)
 2019/03/04 19:01:39 Playback engine ratio: 2.17687
 2019/03/04 19:01:39 Filter: poly sinc
 

[Miska]

On Aneman, just map the channels 1:1, 1-1 2-2, 3-3, 4-4 and so on. So you get diagonal line on the routing matrix.

[Sigi]

I have done this yesterday, and the result was not ok. No influence of the filters in any of the 8 channels! 

I wonder why I can’t use easy connect anymore! It is just one command to connect the channels according to the matrix configured in HQPlayer. Why should this be different with Linux? 

Aneman is for professionals, who want to combine different equipments through Ravenna.

my and many applications of others I know, just use the 8 channels for active connection to amplifiers and/ use correction filters for room acoustic improvement.

best regards

sigi

 

[Sigi]

And, normally under windows, connection is automatic in 90 % of the cases. No need to start easy connect. Only in few cases when something has been manipulated in HQPlyer or Muso.

cheers

sigi

[Miska]

You pipeline setup looks correct, and I assume the matrix setup shows the filters correctly as impulse_0.wav, etc?

 

 

I've only used Aneman. I really have no idea of why for anything related to connections. That is well beyond my scope and internal to the Ravenna.

 

Matrix routing in  HQPlayer has no connection whatsoever with matrix routing in Ravenna. These two are totally independent and don't know anything about each other.

 

[Sigi]

My Matrix looks different:

Sorce channel: 12 12 12 12 for the input ( are the right and left stereo channels)

Mix channel is 1  2 3 4 5 6 7 8 . At least this was my original windows structure.

And it worked perfectly

In this way left convolved channels were 1 3 5 7 , right convolved channels were 2 4 6 8

[Miska]

12 minutes ago, Sigi said:

My Matrix looks different:

Sorce channel: 12 12 12 12 for the input ( are the right and left stereo channels)

Mix channel is 1  2 3 4 5 6 7 8 . At least this was my original windows structure.

And it worked perfectly

In this way left convolved channels were 1 3 5 7 , right convolved channels were 2 4 6 8

 

Yes, that's correct. Above is screenshot is my 3D headphone mix.

 

But I meant that you have the convolution filters correctly shown under "Process" as impulse_0.wav etc?

[Sigi]

Yes, they are impulse_0.wav , ......impulse_7.wav

cheers

Sigi

[Sigi]

I still wonder why easy connect doesn’t connect. Have no idea about Linux. But could it be that Ravenna is expecting from the matrix pipeline 1:1, 2:2 .....8:8 instead of 0:0 1:1 ......7:7 ? Perhaps stupid question but this was the way it configured in HQPlayer desktop , First Filter was 1, not 0

cheers

Sigi

[Miska]

59 minutes ago, Sigi said:

I still wonder why easy connect doesn’t connect. Have no idea about Linux. But could it be that Ravenna is expecting from the matrix pipeline 1:1, 2:2 .....8:8 instead of 0:0 1:1 ......7:7 ? Perhaps stupid question but this was the way it configured in HQPlayer desktop , First Filter was 1, not 0

 

That may have something to do with the driver, or actually Butler binary, but it is not related to the HQPlayer. There is normal Linux ALSA driver interface between the two.

 

Ravenna has no way of knowing anything about HQPlayer's matrix pipeline. It only knows HQPlayer is putting out 8 channels. You then need to connect those eight channels to corresponding DAC channels.

 

If you look at HQPlayer Desktop log file about the pipeline initialization you can see it looks the same with 0-based channels. If pipeline matrix is configured same way on both, it should work the same. Same filter files uploaded for the same pipelines.

 

But I cannot imagine anything automatic setting up the routing at Ravenna side any other way than 1:1 channel mapping... (you can anyway verify from Aneman)

 

[asdf1000]

21 hours ago, Em2016 said:

 

Me? No chance, I don't have a Mergic DAC. I've been wanting to get a Hapi + DA8P for a while, waiting for Linux support. This means I could use it at my desk for HQPe audio and MacOS audio and Windows PC audio.

 

All audio over ethernet cable - no USB devices at all.

 

The upcoming Anubis (with DXD/DSD DAC option) is rumoured to be just a little more than half the price of Hapi + DA8P... Rumours may be all over the place at this stage though.

 

Full info coming in a few weeks.

 

 

 

 

Btw: the source of the rumours (in addition to that official video teaser of course) is this Pro Audio thread:

 

https://www.gearslutz.com/board/remote-possibilities-in-acoustic-music-and-location-recording/1252514-merging-anubis-2.html#post13845805

 

For anyone interested.

 

Factual information to be revealed 21st March.

[bibo01]

9 hours ago, Miska said:

 

You get better sound quality by running individual modulator for each of those channels. It notably drops the noise floor. So instead output 8 channels from HQPlayer with matrix mapping from stereo to 8 channels.

 

But you get 4 times more processing and need a super computer for that.

[Miska]

50 minutes ago, bibo01 said:

But you get 4 times more processing and need a super computer for that.

 

No, not at all. It is just DSD256. You also need 4x processing power to do stereo DSD1024 compared to stereo DSD256...

Actually 8 channels of DSD256 is easier to achieve than stereo DSD1024.

 

[bibo01]

1 hour ago, Miska said:

No, not at all. It is just DSD256. You also need 4x processing power to do stereo DSD1024 compared to stereo DSD256...

Actually 8 channels of DSD256 is easier to achieve than stereo DSD1024.

Maybe I did not explain myself properly or I din't understand.

If I do upconversione 44.1->DSD256 on 2ch and you suggest to do it on 8ch, don't I need more processing power?

[Johnseye]

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

 

Thanks

[Miska]

5 hours ago, bibo01 said:

Maybe I did not explain myself properly or I din't understand.

If I do upconversione 44.1->DSD256 on 2ch and you suggest to do it on 8ch, don't I need more processing power?

 

Yes you do, but you don't need a supercomputer... DSD256 is not particularly heavy by today's standards.

 

[bibo01]

5 minutes ago, Miska said:

 

Yes you do, but you don't need a supercomputer... DSD256 is not particularly heavy by today's standards.

 

Have you actually tried it yourself on a Hapi or yours is deduction work?

[Sigi]

8 hours ago, Miska said:

 

That may have something to do with the driver, or actually Butler binary, but it is not related to the HQPlayer. There is normal Linux ALSA driver interface between the two.

 

Ravenna has no way of knowing anything about HQPlayer's matrix pipeline. It only knows HQPlayer is putting out 8 channels. You then need to connect those eight channels to corresponding DAC channels.

 

If you look at HQPlayer Desktop log file about the pipeline initialization you can see it looks the same with 0-based channels. If pipeline matrix is configured same way on both, it should work the same. Same filter files uploaded for the same pipelines.

 

But I cannot imagine anything automatic setting up the routing at Ravenna side any other way than 1:1 channel mapping... (you can anyway verify from Aneman)

 

 

Ok, and what can I do now? If I configure Aneman sometimes 8 channel starts, sometimes not. But what I get is not correct! Something is not working with the filters! Sound is asif the filters haven’t be used, every channel reproduces the whole frequency range from 20-20 KHz.  I checked them in Acourate and the filters are all OK!

 

[Miska]

16 minutes ago, bibo01 said:

Have you actually tried it yourself on a Hapi or yours is deduction work?

 

I've had exaSound e28 for a very long time...

 

[Miska]

Here's load when upsampling RedBook stereo to 8ch DSD256 with sinc-M filter.

 

[bibo01]

29 minutes ago, Miska said:

Here's load when upsampling RedBook stereo to 8ch DSD256 with sinc-M filter.

Thanks. This is with an i7-6950x.

[Miska]

8 minutes ago, bibo01 said:

Thanks. This is with an i7-6950x.

 

Yeah, it is a bit old already. Here running together with GTX1080. Also works with -2s filters without GPU. Probably newer 8-core (or more) CPU together with/without RTX2080 would do the same or better.

 

[michaelvv]

Hi

 

I tried the hqplayerd and it worked fine with the gentooplayer.

 

I would like to be able to play music directly from the hqplayerd command line. I found some info using curl

 

curl -v -H "text/xml" --data-binary "<?xml version="1.0" encoding="utf-8"?><GetInfo />" http://localhost:4321
* Rebuilt URL to: http://localhost:4321/
*   Trying ::1...
* TCP_NODELAY set
* Connected to localhost (::1) port 4321 (#0)
> POST / HTTP/1.1
> Host: localhost:4321
> User-Agent: curl/7.58.0
> Accept: */*
> Content-Length: 45
> Content-Type: application/x-www-form-urlencoded
> 
* upload completely sent off: 45 out of 45 bytes
<?xml version="1.0" encoding="utf-8"?><GetInfo platform="Linux" product="Signalyst HQPlayer 4 Embedded" version="4">GENERIC BUILD</GetInfo>

but It never returns to the command line after execution.

 

Any other solutions are welcome, saw the hqp-control source files, but don't know how to compile them.

 

Thanks for a great piece of software.

 

 

[bibo01]

On 3/4/2019 at 9:47 PM, Miska said:

On 3/4/2019 at 7:08 PM, bibo01 said:

By upmixing at drivers level and then downmixing 8ch->2ch at output with a specially made cable, you use all 8 D/As onboard ES9028 chip and achieve higher SQ.

 

You get better sound quality by running individual modulator for each of those channels. It notably drops the noise floor. So instead output 8 channels from HQPlayer with matrix mapping from stereo to 8 channels.

Sorry if I come back to this, but I don't understand the difference between my method and yours.

By "doubling up" the modulator that runs on ch1 onto ch3,5,7 through matrix mapping, aren't you effectively adding up four times the same result?! What's the difference with me copying the result of ch1 to ch3,5,7 - I still get a drop in noise floor?!

Afterwards, if you want stereo, you're still 8ch->2ch through wire at output, are you?!