HQ Player

[Jean Paul D]

I keep on exploring...

 

Seems that I could have no regret I can't feed HQP from my Audirvana library for Client seems fine BUT:

 

How do I select Beethoven by Richter ? I can list my recordings by Richer fine and all recordings of LvB works but the intersection?

How do I feed HQP with Qobuz?

How can I get France Musique web radio or podcasts processed by HQP ?

 

I'm also having a look at convolution. For the time being I'm using with great success the 3 first bands of a Meyer CP10 analog parametric equalizer to adapt bass response of my speakers in my room. I could duplicate these settings in digital domain and try them OK with HQP and I could probably obtain a (visually at least) better correction since I am not limited to 3 bands per channel in digital domain (but isn't multiplying eQpoints a liability?) while dedicating the Meyer to my vinyl rig. But, say I read Steve Hoffman recommends to notch or add a dB there with said BW or criticises this or that mastering and suggests shelving down the HF for it has been wrongly boosted IHO: can I keep the settings duplicated from the Meyer as a base and on case by case add a tweaking or should I create each time a new entry in Pipeline?

 

I have also explored MCH mixdown :

 

with success with  DVDAudio, ie America or Steely Dan, as it seems they offer remixes not mere remastering, much in the line of what Tony Visconti has done with remixes of Bowie : bass and drums sound more live (remember the days of freedom?) and less shied in the mixes. And the sound is perfectly consistent with enhanced but sound soundstage. But when I tried Bitches Brew 4.0 from SACD it was just n 'importe quoi : completely incoherent effects : the mixdown does not translate rightly.

Was it good luck/bad luck or some standards justify that MCH DVDA translate well as 2.0 and MCH SACD not ?

[Miska]

1 hour ago, Jean Paul D said:

How do I select Beethoven by Richter ? I can list my recordings by Richer fine and all recordings of LvB works but the intersection?

 

"+richter +beethoven" as search term.

 

1 hour ago, Jean Paul D said:

How do I feed HQP with Qobuz?

 

For example from Roon at the moment.

 

1 hour ago, Jean Paul D said:

How can I get France Musique web radio or podcasts processed by HQP ?

 

I don't know what these are, how they are accessed and what format are they in?

 

1 hour ago, Jean Paul D said:

while dedicating the Meyer to my vinyl rig

 

You can have HQPlayer process EQ also for vinyl...

 

1 hour ago, Jean Paul D said:

can I keep the settings duplicated from the Meyer as a base and on case by case add a tweaking or should I create each time a new entry in Pipeline?

 

You could use Pipeline profiles for this. Allows you to later easily pick one of those from the Clinet.

 

1 hour ago, Jean Paul D said:

Was it good luck/bad luck or some standards justify that MCH DVDA translate well as 2.0 and MCH SACD not ?

 

It should be similar for both, but the results vary depending on material. Since there are different kind of multichannel mixes.

 

[Jean Paul D]

10 minutes ago, Miska said:

 

"+richter +beethoven" as search term.

 

 

For example from Roon at the moment.

 

 

I don't know what these are, how they are accessed and what format are they in?

 

 

You can have HQPlayer process EQ also for vinyl...

 

 

You could use Pipeline profiles for this. Allows you to later easily pick one of those from the Clinet.

 

 

It should be similar for both, but the results vary depending on material. Since there are different kind of multichannel mixes.

 

adding pluses works, thanks

 

I don't want Roon, ie with Embedded Audirvana, much cheaper and less cumbersome, would work. I have Audirvana 

 

I access France Musique via Safari :https://www.francemusique.fr/concerts and don't know how to determine the format

 

processing vinyl via HQP would require the replacement of the Meyer by an ADC, wouldn't it? Might be objectively stupid or at least moot but I like the idea of keeping analog analog. Actually, I don't rule out euphony or better dynamic compression on LPs I unshelved recently, such as Kind of Blue as 45 rpm one sided discs by Classic Records, but comparisons with matched eQ were in favour of vinyl vs  my DSD 128 limited DAC that I find very very good HQP fed 7EC, when not comparing ; should I expect much from, say, a T+A DSD 8 fed 256 via a Mac M1 ? other suggestion? 

 

I was hoping it was possible to use the 

<plugin>:[arg1[=val]];[arg2[=val]];...;[argn[=val]] syntax on top of .wav or whatever format I would end with depending of the DRC solution I would opt for in order to create by base correction...

 

I had no miss when applying the mixdown formula you provide to DVDA (rather, I'd say it's gorgeous !) ; the process of ripping extracting tagging etc being pretty cumbersome, I stopped at the first SACD, that was a miss. Do you have other magic formulas I could try on SACD ? that would be mainly from Columbia Quadrophonic origin, maybe some 3 channels jazz too

[kelvinwsy]

Wish that is the case for my setup. When I was still running NAA the OS on the main server makes an audible difference. Win 10 Pro is just NOISY. The sound scape is Cluttered. Win Server is crystalline clear in comparsion. I was on Etheregen/SOTM Ultra Neo/ 2 BGTK?? Clocks with two more Netgear GS-105 switches and all LPSU's on everything with LPS1.2 and LPS 1 powering the important devices. 

Found that the sound was clean but soft transients and PRAT was just say 'a bit' slow. Spent a lot more effort and LT3045's psu's for everything and went back to USB direct with a split 2 headed USB cable - LH Labs aHve not gone back to the NAA mode for almost 6 months now. 

Dont have to keep the Etherregen/SOTM/Clocks/Switches all powered up which is very bad for Global Warming. room heating up. power utiltiy bill etc.  

BUt the effort to get USB mode to sound very good is very tedious. My New Aorus Ultra Z490.Intel 10700k HQPlayer PC is powered by3 ATX PSU's 3 LT3045 units. MB is Elfidelity filtered, SSD are externalised and shielded wtih Mumetal and case earth-linked to my Power Conditioner earth link. So are the verious parts of the PC case, PCIE slots etc. The AIO and cooling Fans are all flexible mounted and damped with vibration pads and powered by fan controller only taking the PWM signal from the MB and running at constant HALF Speed setting - minimize PWM fluctuations.  USB now thrumps NAA mode by far in PRAT, Dynamic Contrast, Drive, Bass Response, High Frequency Sound liveliness W/O hardness and mid range PRESENCE!!! ANd the Sonic background is ink black.

 

The above narrative explains that I have done extensive testing ( YMMV and IMHO on my system of course) NAA versus direct USB. But in my system, the OS sound difference is very very clear to hear. 5 -10 seconds and I can hear the comparative clutter of Win 10 PRO. So that is my finding is my system using SENN HD800/ Abyss AB1266 Phi CC

YMMV of course.

[Miska]

1 hour ago, Jean Paul D said:

processing vinyl via HQP would require the replacement of the Meyer by an ADC, wouldn't it? Might be objectively stupid or at least moot but I like the idea of keeping analog analog.

 

Yes it would, but allows more flexibility than analog EQ and likely at higher signal fidelity. But of course you can do it the way you prefer.

 

1 hour ago, Jean Paul D said:

I was hoping it was possible to use the 

<plugin>:[arg1[=val]];[arg2[=val]];...;[argn[=val]] syntax on top of .wav or whatever format I would end with depending of the DRC solution I would opt for in order to create by base correction...

 

Yes, you can combine the two that way.

 

[Luca72c]

12 minutes ago, kelvinwsy said:

Wish that is the case for my setup. When I was still running NAA the OS on the main server makes an audible difference. Win 10 Pro is just NOISY. The sound scape is Cluttered. Win Server is crystalline clear in comparsion.

 

Noisy? I think your PC or your setup (so complicated) surely have some problems...

I compared W10 HQP with Linux and Mac versions, on the same PC or in others (no Winserver as it is unsupported). Some friends were attending and sharing impressions.

We all heard differences, in terms of fullness/richness, soundstage width and depth, maybe some slight detail (but i cannot swear about this). But surely not in noise! How could noise be involved in such a thing?

[luisma]

21 hours ago, Quadman said:

With my issues with the Holo asio driver and DSD through HQP

Hi @Quadman

Just wanted to ask, so the May + HQP has interfacing issues due to drivers? is this something related to Windows, or Linux, or both? will the May accept 48k resampling from HQPe?

 

[Quadman]

5 minutes ago, luisma said:

so the May + HQP has interfacing issues due to drivers?

The issue is the May's Thesycon's ASIO driver does not like any other ASIO driver to be active.  If any of your audio outputs can see another ASIO driver then most likely HQP will not play DSD.  It will play PCM just no SDM.  Remove or un-enable any other ASIO driver and HQP will play DSD/SDM through it.  Yes 48K rates are supported.  I use desktop but enable 48K family and adaptive output so 44 and 48 up samples multiples of themselves.

[luisma]

40 minutes ago, Quadman said:

May's Thesycon's ASIO driver does not like any other ASIO driver to be active

Thank you, so under Linux the May will work with none of the issues of the Thesycon's, maybe others though

 

[Quadman]

29 minutes ago, luisma said:

Thank you, so under Linux the May will work with none of the issues of the Thesycon's, maybe others though

Probably.  You don't need a driver with Linux do you?

[luisma]

22 minutes ago, Quadman said:

You don't need a driver with Linux do you?

Yes you do, it is the ALSA driver which is integrated with the kernel API, but this driver it is "usually" pretty standard, I have seen it fail on such environments (original Allo USB bridge is one of them which stopped working with the Terminator DAC after Oct 2019 update)

[Luca72c]

@Miska, sorry if this has been asked already: if i have a stereo PCM file and i want to convert it to a crossovered 8ch DSD stream using matrix pipeline + convolutions, to be sent to a multichannel DSD dac, i obviously have to set both DSD conversion/upsampling work and matrix pipeline work. But which one is performed first?

To be clearer, i just want to know if being set such way, hqplayer is going to convert the PCM stereo stream to DSD and then apply convolution to that DSD stream (2 ch DSD conversion + channel copy in DSD + 8ch EQ on DSD), or is it going to apply convolution to the PCM stereo flow and then convert 8 channels of convolved PCM to DSD (channel copy in PCM + 8ch EQ on PCM + 8ch DSD conversion)?

I ask because the first case should be much less demanding for the PC Hqplayer runs onto than the second one, isn't it?

But anyway usually DSP are not able to work on DSD streams, were you able to overcome that limitation?

[Miska]

44 minutes ago, Luca72c said:

@Miska, sorry if this has been asked already: if i have a stereo PCM file and i want to convert it to a crossovered 8ch DSD stream using matrix pipeline + convolutions, to be sent to a multichannel DSD dac, i obviously have to set both DSD conversion/upsampling work and matrix pipeline work. But which one is performed first?

To be clearer, i just want to know if being set such way, hqplayer is going to convert the PCM stereo stream to DSD and then apply convolution to that DSD stream (2 ch DSD conversion + channel copy in DSD + 8ch EQ on DSD), or is it going to apply convolution to the PCM stereo flow and then convert 8 channels of convolved PCM to DSD (channel copy in PCM + 8ch EQ on PCM + 8ch DSD conversion)?

I ask because the first case should be much less demanding for the PC Hqplayer runs onto than the second one, isn't it?

 

For PCM -> PCM, PCM -> SDM, SDM -> SDM cases processing is performed at the source rate and then converted to the desired output resolution. And for SDM -> PCM case it is performed at 1/16th of the SDM rate in PCM domain. This is most of the time lightest way to do it and provides good results. The second case you describe is what happens and it is lighter than the first one.

 

44 minutes ago, Luca72c said:

But anyway usually DSP are not able to work on DSD streams, were you able to overcome that limitation?

 

HQPlayer can do all the same DSP for DSD streams as it can for PCM. Given that you have enough processing power to do it... So depending on case you may need GPU offloading. I consider having same functionality for DSD and PCM sources as very important feature of HQPlayer. Although it means HQPlayer has two DSP engines.

 

[Luca72c]

24 minutes ago, Miska said:

For PCM -> PCM, PCM -> SDM, SDM -> SDM cases processing is performed at the source rate and then converted to the desired output resolution. And for SDM -> PCM case it is performed at 1/16th of the SDM rate in PCM domain. This is most of the time lightest way to do it and provides good results. The second case you describe is what happens and it is lighter than the first one.

 

This is surely unexpected for me! So this means that performing EQ convolution on 8 DSD channels is heavier than performing DSD conversion (using EC modulators) on 8 PCM channels? Then performing convolutions on DSD is so heavier than performing convolutions on PCM?

 

28 minutes ago, Miska said:

HQPlayer can do all the same DSP for DSD streams as it can for PCM. Given that you have enough processing power to do it... So depending on case you may need GPU offloading. I consider having same functionality for DSD and PCM sources as very important feature of HQPlayer

 

Great, this is a unique feature in a DSP engine!

[Miska]

22 minutes ago, Luca72c said:

Then performing convolutions on DSD is so heavier than performing convolutions on PCM?

 

Yes, by roughly factor of sampling rate difference.

 

[jabbr]

32 minutes ago, Luca72c said:

1 hour ago, Miska said:

HQPlayer can do all the same DSP for DSD streams as it can for PCM. Given that you have enough processing power to do it... So depending on case you may need GPU offloading. I consider having same functionality for DSD and PCM sources as very important feature of HQPlayer

 

Great, this is a unique feature in a DSP engine!

Precisely. @Miska regarding conversion of PCM to SDM via EC vs SDM convolutions, the convolutions are CUDA accelerated, right?

[Miska]

4 minutes ago, jabbr said:

Precisely. @Miska regarding conversion of PCM to SDM via EC vs SDM convolutions, the convolutions are CUDA accelerated, right?

 

Yes, you can choose to offload either both filters and convolution. Or just convolution.

 

[Jean Paul D]

9 hours ago, Miska said:

 

Yes it would, but allows more flexibility than analog EQ and likely at higher signal fidelity. But of course you can do it the way you prefer.

 

 

Yes, you can combine the two that way.

 

any mixdown recipe for mch SACD originated from Columbia Quadrophonic ?

should I expect much from T+A or another DAC you would recommend @ 256 ?

I remember reading a post from a HQP user who was disappointed after he changed his BB1795 DAC for a RME 256 capable 

[Miska]

7 minutes ago, Jean Paul D said:

any mixdown recipe for mch SACD originated from Columbia Quadrophonic ?

 

The same mixdown matrix should work for all cases.

 

7 minutes ago, Jean Paul D said:

should I expect much from T+A or another DAC you would recommend @ 256 ?

I remember reading a post from a HQP user who was disappointed after he changed his BB1795 DAC for a RME 256 capable

 

You really need to give a listen and decide yourself. But note that T+A is not using the 1795 chip for DSD sources, only when you input PCM.

 

[Luca72c]

20 minutes ago, Miska said:

 

Yes, you can choose to offload either both filters and convolution. Or just convolution.

 

 

So isn't the ability to offload convolutions to CUDA able to move the balance in favour of the first case i guessed, performance wise?

Just because EC modulators calculations (needed to convert 8ch of PCM into 8ch of DSD) cannot be offloaded to CUDA, instead...

[Miska]

5 minutes ago, Luca72c said:

So isn't the ability to offload convolutions to CUDA able to move the balance in favour of the first case i guessed, performance wise?

Just because modulators calculations cannot be offloaded to CUDA, instead...

 

No, if you want 8 channels output in DSD, you need 8 modulators. Even if two channels have the same source data, they are independently modulated and the output data is different. (also two channels of PCM output with same source data won't have same output data)

 

Every time you want DSD output you need modulator. Same way as every time you want PCM output you need dither or noise shaper.

 

Also every time you run same source content through the processing, the output data is different. You practically never get the same data twice.

 

[Luca72c]

20 minutes ago, Miska said:

No, if you want 8 channels output in DSD, you need 8 modulators. Even if two channels have the same source data, they are independently modulated and the output data is different. (also two channels of PCM output with same source data won't have same output data)

 

Every time you want DSD output you need modulator. Same way as every time you want PCM output you need dither or noise shaper.

 

Also every time you run same source content through the processing, the output data is different. You practically never get the same data twice.

 

Interesting. This means there's some sort of random component in the modulation process? 

Are 2 modulated DSD versions from the same PCM source content substantially different or equivalent in practice? 

Isn't a modulated DSD stream (converted from a PCM source) equivalent to a normal DSD file stream? 

I ask because you say a normal DSD file can receive convolution as is, so why a modulated, converted-from-PCM DSD stream cannot? 

For example, a stereo DSD file obtained through conversion-from-PCM in Hqp4 pro cannot receive channel copy + convolutions in matrix pipeline to obtain a 8ch crossovered DSD file? 

[Miska]

47 minutes ago, Luca72c said:

Interesting. This means there's some sort of random component in the modulation process? 

 

Yes, very much on purpose.

 

47 minutes ago, Luca72c said:

Are 2 modulated DSD versions from the same PCM source content substantially different or equivalent in practice?

 

Equivalent in practice, but with different data.

 

47 minutes ago, Luca72c said:

Isn't a modulated DSD stream (converted from a PCM source) equivalent to a normal DSD file stream?

 

I'm not sure I understand the question. Can you elaborate a bit more?

 

47 minutes ago, Luca72c said:

I ask because you say a normal DSD file can receive convolution as is, so why a modulated, converted-from-PCM DSD stream cannot?

 

Of course both can have convolution. But if you start from 44.1k source, with DSD256 output. Why burn about 256x more CPU power doing convolution for DSD256 instead of doing it for 44.1k PCM?

 

47 minutes ago, Luca72c said:

For example, a stereo DSD file obtained through conversion-from-PCM in Hqp4 pro cannot receive channel copy + convolutions in matrix pipeline to obtain a 8ch crossovered DSD file?

 

Yes it can, it means you run 8 convolutions and 8 remodulators at SDM rates. Instead of 8 convolutions at PCM source rate and 8 modulators. Former is much much heavier because SDM convolution is so much heavier than PCM convolution. Rest of the process doesn't have significantly different load.

 

[Luca72c]

1 hour ago, Miska said:

I'm not sure I understand the question. Can you elaborate a bit more?

 

Yes it can, it means you run 8 convolutions and 8 remodulators at SDM rates. Instead of 8 convolutions at PCM source rate and 8 modulators. Former is much much heavier because SDM convolution is so much heavier than PCM convolution. Rest of the process doesn't have significantly different load.

 

Maybe this is what i didn't understand: is (re)modulation needed in convolution even if the stereo file already is in DSD format, no additional oversampling is requested and receiving only convolution? I was (probably wrongly) thinking that EC modulation was only needed in the process of converting PCM to DSD or DSD to different sampling rate DSD...

 

1 hour ago, Miska said:

Of course both can have convolution. But if you start from 44.1k source, with DSD256 output. Why burn about 256x more CPU power doing convolution for DSD256 instead of doing it for 44.1k PCM?

 

Because convolution can be offloaded to CUDA, modulation can't, and after convolving PCM you'd find yourself with 8 PCM channels to be modulated in the conversion to DSD process. That's what i was thinking. But if you say the same heavy modulation is needed anyway, each time a DSD stream is convolved (even if no conversion or resampling is requested), then my logic was totally inconsistent...

[Miska]

24 minutes ago, Luca72c said:

Maybe this is what i didn't understand: is (re)modulation needed in convolution even if the stereo file already is in DSD format, no additional oversampling is requested and receiving only convolution?

 

It is no different from requantization for PCM. By definition it is not a bit-perfect process.

 

26 minutes ago, Luca72c said:

But if you say the same heavy modulation is needed anyway, each time a DSD stream is convolved (even if no conversion or resampling is requested), then my logic was totally inconsistent...

 

You should preferably at least double the sampling rate when you process either PCM or DSD, to gain more headroom in both dimensions. Even if you do just volume control.

 

 

If I'd generalize, if you touch a single bit, like little bit of volume control, you are up to the same process. Be input or output be PCM, SDM, whatever. So you sort of have constant amount of effort, plus the extras you want. Depending on the case, the extras can be a lot less than the constant effort.