HQ Player

[semente]

Hi @Miska, would it be possible to have a remote 'Drop Folder' in a Mac or Windows machine to remotely load HQPlayer Embedded?

 

It would be great if I were able to drag-drop files or albums from iTunes/Music into such a folder in a remote laptop instead of having to use Screen Sharing to control my headless Mac mini and HQPlayer Desktop.

I don't know how many people would use such a feature though, but could run a poll if that is helpful.

[stefano_mbp]

4 minutes ago, semente said:

Hi @Miska, would it be possible to have a remote 'Drop Folder' in a Mac or Windows machine to remotely load HQPlayer Embedded?

 

It would be great if I were able to drag-drop files or albums from iTunes/Music into such a folder in a remote laptop instead of having to use Screen Sharing to control my headless Mac mini and HQPlayer Desktop.

I don't know how many people would use such a feature though, but could run a poll if that is helpful.

If you share a folder on your network and define it as the HQPe library you can drop into that folder what you like, scan the HQPe library and you’re done 

[Argon]

Thanks all for the help with the FIFO drops.  It has been happening randomly for months somewhere between 90 min and 12 hours after starting the stream; never able to stream for more than 12 hours w/o a Push to FIFO error..

 

+ 2023/03/29 01:11:33 NAA output network engine started at: 96000

! 2023/03/29 01:14:14 NAA output clNetEngine::PushPCM(): clNetEngine::SendStreamPCM(): clSemaphore::Wait() (1438)

! 2023/03/29 01:14:14 clHQPlayerEngine::Execute(): push to FIFO failed

  2023/03/29 01:14:14 Stop request (reset)

& 2023/03/29 01:14:14 Stop...

- 2023/03/29 01:14:14 Playback engine stopped

 

My system was

 

SonicTransport i9 (Roon) -> HQPlayer -> Cat6 -> SOTM Switch -> Fiber / Finisar SFP -> Sig Rendu -> DAC

 

In the end I tried:

 

2 different dacs (PS Audio DS & AQ Dragonfly) - both drop

2 different USB cables ($3 Amazon basic & the amazing aerogel clad solid silver from Happa) - both drop

3 different SFP modules (TPLink, Finisar SFP, Finisar SFP+) - all drop

3 different fiber cables (single mode and multimode depending on optical module)- all drop

2 SFP equipped switches (SOTM and TrendNet) - both drop

Switch -> Cat6 -> Sonore Optical Module FMC -> Sig Rendu (using all above optical combos)  - no change in drops

Changed the micro-SD in the Sig Rendu to a new one ordered from Sonore.

 

(for the scientists on here, the above matrix is an ideal factorial DOE, 40 runs + 10 replicates)

 

BTW: All devices on the network are 1000 MB/s negotiated rate.

 

There is no issue with my SonicTransport i9 Roon source, I am streaming to a B&O device in my office constantly, weeks on end w/o interruption and Roon ARC doesn't drop anytime I'm in the car or traveling.

 

Yesterday I converted an old Mac to act as an NAA endpoint and used the a multitude of the SFP or SFP+ networking systems described above with the AQ Dragon DAC into one of Mac's USB ports.; I have been streaming for 2 days now w/o issue.  I wont declare successes for a few more days but it is pointing to my Sig Rendu as the root cause.  Hard to say the Rendu Sig is "broken" it works fine for xx hours before it craps out.  It's been my experience that  electrical systems either work or the don't.  It's almost as if a buffer fills or a component over-temps (IR pyrometers shows no sign of heat other than the SFP transceiver that runs 35-39C)

 

Is there anything I am missing, the Sig Rendu has such a good reputation???

 

 

 

[Miska]

 

21 minutes ago, Argon said:

BTW: All devices on the network are 1000 MB/s negotiated rate.

 

Remember that the Rendu can do max 400 Mbps in ideal conditions (flow control functional). This is bandwidth limitation of the internal bus between the ethernet controller and CPU. If flow control is not functional, it will have occasional stalls where it drops to practically 0 for some duration. This is because the ethernet controller's packet buffer overflows without flow control. Which in turn causes packet loss, which in turn causes re-sends, which in turn causes more overflow/packet loss and so on. This causes complete traffic jam.

 

Flow control is handled by the ethernet controller hardware, based on it's internal packet buffer status. So it is practically 100% effective.

 

21 minutes ago, Argon said:

Yesterday I converted an old Mac to act as an NAA endpoint and used the a multitude of the SFP or SFP+ networking systems described above with the AQ Dragon DAC into one of Mac's USB ports.; I have been streaming for 2 days now w/o issue.  I wont declare successes for a few more days but it is pointing to my Sig Rendu as the root cause.  Hard to say the Rendu Sig is "broken" it works fine for xx hours before it craps out.  It's been my experience that  electrical systems either work or the don't.  It's almost as if a buffer fills or a component over-temps (IR pyrometers shows no sign of heat other than the SFP transceiver that runs 35-39C)

 

Is there anything I am missing, the Sig Rendu has such a good reputation???

 

It is possible the issue is that flow control doesn't get negotiated on one of the hops which causes this. Or it misbehaves one way or the other. Usually though lacking the flow control causes only frequent drop-outs in audio, but no connection losses or sporadic long term random issues. It would need some low level network analysis of possible causes to figure out the root cause.

 

Of course it could be some malfunction in the Rendu too.

 

Flow control issues have been most typical reason for playback problems with Rendu though.

 

[semente]

3 hours ago, stefano_mbp said:

If you share a folder on your network and define it as the HQPe library you can drop into that folder what you like, scan the HQPe library and you’re done 

 

Not ideal but it's a step in the right direction.

 

What I would like is a folder that will work as a virtual HQPlayer window, allowing me to drag-drop files but these would not be transfered to the (end) computer hosting HQPlayer Embedded, just like I do now when I drag tracks from iTunes/Music into HQPlayer desktop. The dragged file would direct HQPlayer to the original file location (I don't know the correct terminology for that - pointer file?).

[Miska]

6 minutes ago, semente said:

What I would like is a folder that will work as a virtual HQPlayer window, allowing me to drag-drop files but these would not be transfered to the (end) computer hosting HQPlayer Embedded, just like I do now when I drag tracks from iTunes/Music into HQPlayer desktop. The dragged file would direct HQPlayer to the original file location (I don't know the correct terminology for that - pointer file?).

 

I have plan on my TODO-list to support drag-and-drop with HQPlayer Client as well. So you would drop content on HQPlayer Client window to queue it for playback.

 

[vortecjr]

1 hour ago, Argon said:

Thanks all for the help with the FIFO drops.  It has been happening randomly for months somewhere between 90 min and 12 hours after starting the stream; never able to stream for more than 12 hours w/o a Push to FIFO error..

 

+ 2023/03/29 01:11:33 NAA output network engine started at: 96000

! 2023/03/29 01:14:14 NAA output clNetEngine::PushPCM(): clNetEngine::SendStreamPCM(): clSemaphore::Wait() (1438)

! 2023/03/29 01:14:14 clHQPlayerEngine::Execute(): push to FIFO failed

  2023/03/29 01:14:14 Stop request (reset)

& 2023/03/29 01:14:14 Stop...

- 2023/03/29 01:14:14 Playback engine stopped

 

My system was

 

SonicTransport i9 (Roon) -> HQPlayer -> Cat6 -> SOTM Switch -> Fiber / Finisar SFP -> Sig Rendu -> DAC

 

 

In the end I tried:

 

2 different dacs (PS Audio DS & AQ Dragonfly) - both drop

2 different USB cables ($3 Amazon basic & the amazing aerogel clad solid silver from Happa) - both drop

3 different SFP modules (TPLink, Finisar SFP, Finisar SFP+) - all drop

3 different fiber cables (single mode and multimode depending on optical module)- all drop

2 SFP equipped switches (SOTM and TrendNet) - both drop

Switch -> Cat6 -> Sonore Optical Module FMC -> Sig Rendu (using all above optical combos)  - no change in drops

Changed the micro-SD in the Sig Rendu to a new one ordered from Sonore.

 

(for the scientists on here, the above matrix is an ideal factorial DOE, 40 runs + 10 replicates)

 

BTW: All devices on the network are 1000 MB/s negotiated rate.

 

There is no issue with my SonicTransport i9 Roon source, I am streaming to a B&O device in my office constantly, weeks on end w/o interruption and Roon ARC doesn't drop anytime I'm in the car or traveling.

 

Yesterday I converted an old Mac to act as an NAA endpoint and used the a multitude of the SFP or SFP+ networking systems described above with the AQ Dragon DAC into one of Mac's USB ports.; I have been streaming for 2 days now w/o issue.  I wont declare successes for a few more days but it is pointing to my Sig Rendu as the root cause.  Hard to say the Rendu Sig is "broken" it works fine for xx hours before it craps out.  It's been my experience that  electrical systems either work or the don't.  It's almost as if a buffer fills or a component over-temps (IR pyrometers shows no sign of heat other than the SFP transceiver that runs 35-39C)

 

Is there anything I am missing, the Sig Rendu has such a good reputation???

 

 

 

Go to the Rendu homepage and add /diag to the units IP address and post that complete output here so we can see what is going on. 

[tcode]

@Miskain settings, Buffer time changes to "Default" after selecting buffer time. It won't remember the choice.  4.21.1 Desktop version.

[semente]

2 hours ago, Miska said:

 

I have plan on my TODO-list to support drag-and-drop with HQPlayer Client as well. So you would drop content on HQPlayer Client window to queue it for playback.

 

 

Thanks, that would be amazing.

[Miska]

2 hours ago, tcode said:

@Miskain settings, Buffer time changes to "Default" after selecting buffer time. It won't remember the choice.  4.21.1 Desktop version.

 

Which OS/build and backend is this with?

 

[pavi]

On 3/29/2023 at 9:03 PM, leManu said:

Also, what do I need to upsample to DSD 1024?  I'll buy a new computer for this, so I need the required specs.

 

the ryzen 7950 / 64g with the 4090 does dsd1024 quite comfortably with the amsdm7ec 512+fs modulator. the 4090 is overkill, likely. that’s what i use. ubuntu 22.10

[AKira]

59 minutes ago, pavi said:

 

the ryzen 7950 / 64g with the 4090 does dsd1024 quite comfortably with the amsdm7ec 512+fs modulator. the 4090 is overkill, likely. that’s what i use. ubuntu 22.10

What is oversampling?
 

[Chodi]

Time for an off the wall question. Does Nvidia Resizing Bar video setting have any effect on HQP performance. This is only available in 30 and 40 series cards.  I am aware that it can have a positive effect, gain, in some current games. I am not aware if this technology has any effect on performance as it relates to HQP? Perhaps Jussi could comment.

[tcode]

9 hours ago, Miska said:

 

Which OS/build and backend is this with?

 

Oh, it was very late yesterday for me, I was adjusting buffer time in Input Device Settings, where I don't use any backend. Please ignore and sorry for bothering you!

[Miska]

5 hours ago, Chodi said:

Time for an off the wall question. Does Nvidia Resizing Bar video setting have any effect on HQP performance. This is only available in 30 and 40 series cards.  I am aware that it can have a positive effect, gain, in some current games. I am not aware if this technology has any effect on performance as it relates to HQP? Perhaps Jussi could comment.

 

Possibly it can help, although HQPlayer transfers between CPU and GPU are not that huge. Biggest impact is likely when you perform for example convolution together with upsampling for DSD sources.

 

[manishex]

So I was playing around with r26 and enabled DSD mode, played some songs (windows 2016) and now whenever I go into matrix pipeline or settings and save it now takes 10-15 seconds to complete whereas before it took 2-3 seconds even with DSD mode disabled again. I've tried opening up the same hqplayer on windows 10 with the same settings more or less and it takes 3 seconds. I have 50kb log files for both. attached. I reinstalled hqplayer to no avail.

windows 10 - HQPlayer4Desktop.log Windows 2016 - HQPlayer4Desktop.log

[Hensema]

Once I had stuttering while I oversampled 512fs on the GPU, but not on 256fs. The cpu and gpu load where nice and low. Turned out that I had my GPU in the wrong slot (always use the nearest to the cpu) and was operating at pci-e 2.0 x4. That was not enough bandwidth.

Since pci-e 3.0 x16 has far more bandwidth, it would never be an issue for for only oversampling.

Something else…

When using 32fs pcm  oversampling on the gpu it always sounds better to me than on the cpu. Has anyone else noticed this? 

 

 

[Kalpesh]

@Miska I used to eQ in pipeline with .txt files. Then I added, coma, .wav crossover add ons limited to 65 K and it works like a charm

 

I just went through the process of issuing the corrections as a single  .wav per channel, limited to 65 K too. What is best for HQP ? The new .wav include the crossovers so it should be the same I guess but last time I compared, I preferred to have the Pk in textfiles rather than issuing .wav : right or wrong I was under the impression it was lighter to handle for HQP and more precise too.

[Miska]

5 minutes ago, Kalpesh said:

@Miska I used to eQ in pipeline with .txt files. Then I added, coma, .wav crossover add ons limited to 65 K and it works like a charm

 

I just went through the process of issuing the corrections as a single  .wav per channel, limited to 65 K too. What is best for HQP ? The new .wav include the crossovers so it should be the same I guess but last time I compared, I preferred to have the Pk in textfiles rather than issuing .wav : right or wrong I was under the impression it was lighter to handle for HQP and more precise too.

 

Parametric EQ .txt is processed separately as parametric EQ. And convolution is then separate item.

 

Parametric EQ is lighter to process yes.

 

For both types, there are some optimizations.

[Kalpesh]

29 minutes ago, Miska said:

 

Parametric EQ .txt is processed separately as parametric EQ. And convolution is then separate item.

 

Parametric EQ is lighter to process yes.

 

For both types, there are some optimizations.

Thank you, so, bottom line, I leave things as are : .txt with Parametric EQ, coma, .wav with crossover ?

[Miska]

6 minutes ago, Kalpesh said:

Thank you, so, bottom line, I leave things as are : .txt with Parametric EQ, coma, .wav with crossover ?

 

Yes, that is perfectly fine.

[GMG]

Is this measurement of the Gustard r26 a concern for direct DSD with HQPlayer?

is it a shortcoming of Gustard or does Holo or other R2R implementation have similar measurements?

 

https://www.l7audiolab.com/f/gustard-r26/

 

[Hensema]

From the measurements I’ve seen of the Holo Spring 3 the DSD section is at least as good as PCM. DSD might have a little bit more noise but is otherwise cleaner. 
 

Always trust your ears on what sounds best. But from the measurements of the Gustard R26 I wouldn’t use DSD. Try PCM 768 at 16 bit with noise shaper instead.

[kelvinwsy]

On 3/14/2023 at 9:17 PM, kelvinwsy said:

Yes Sir @Miska it has arrived and opening it for photos!

And the good news , the Gustard A26 with factory firmware plays Hqplayer NAA DSD512 , EC7v2 smooth as butter!

SQ hard to say ad I hv not hooked ip my fibre optic converters and the isolation netgear switch(😱actually the 12v unit burnt, and I seem to be unable to locate the other 5v unit!

So both Hqplayer server and A26 straight into the modem that is already downstresm of the IP provider's modem

A quick sound check.. fr memory .. it is comparable to direct USB

BUT against DDC..Master clock .. usb/i2s I think HMMMH🤔🤔🤔

 

l have gone full fiber connection of my DAC and Hqplayer PC Server to my modem..

At DSD512EC7v2 there are noise like static in the bkgrd! Music stutters just a jot every 15 secs or so

W/o further investment in a clocked switch , I can v quickly say that SQ is a distinct 2nd to my clocked USB signal chain

 

 

 

[Argon]

On 4/1/2023 at 5:14 PM, vortecjr said:

Go to the Rendu homepage and add /diag to the units IP address and post that complete output here so we can see what is going on. 

Transitioning from the NAA-endpoint (running on a Mac) to my Rendu SE (changing nothing else upstream) and starting the music stream results in the attached diag report from the Rendu.  I was able to stream for ~6 hours before replicating the push FIFO error as before.  Checking in over those hours I wasn't able to see any significant changes, just the quantity of data Rx and Tx increased as expected.  I did not see an accumulation of errors or drops.  @vortecjrHopefully this is insightful for you, TIA.

sig-rendu-start.pdf