Jump to content
IGNORED

HQ Player


Recommended Posts

The response from Jussi was at least one core per channel (6 cores for 5.1) with 2 cores per channel preferred (12 cores for 5.1) and at least 3.2Ghz chip. Nothing mentioned about RAM.

 

With powerful GPU one can replace the second core per channel spec. Note that HyperThreading doesn't count, so the number of threads the CPU has doesn't really matter much for this use case - it only makes OS context switching faster. Clock speed spec also depends on the CPU generation, since newer CPUs with AVX2 and such are faster at equivalent clock speed.

 

Fast RAM is needed, but I don't have a specific definition for it, usually the faster multicore CPUs also have correspondingly faster and wider memory bus too. Same goes for cache size and speed. Amount of RAM depends on what one wants to do, if Roon is planned and such. But these days RAM is cheap, so I'd start with 16 GB. Just make sure all the CPU memory channels are populated.

 

My quiet Win10 machine with i5 6600T (Skylake quad-core 2.7 GHz base) and 32 GB of low latency DDR4-2133 can do DSD64-to-DSD256 upsampling for stereo, per-core load is slightly under 50%. It can also do DSD64-to-DSD512 with per-core CPU loads around 80%. But I know it cannot do DSD64-to-DSD256 for 5.1 channel, not even with GPU help. It can also do RedBook-to-DSD512 using poly-sinc-2s with similar 80% per-core load. So that's really close to it's limits and i7 6700T wouldn't make it much difference, since it's just 100 MHz faster and adds HyperThreading.

 

There's more about HQPlayer and hardware builds here:

http://www.computeraudiophile.com/f11-software/design-pc-server-roon-and-hq-player-25796/

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Well, that's my point though, given that the DAC takes over the oversampling (before generating the analog signal) from some frequency point, is there a benefit in having one type of upsampling (in software) to take it to that point and then allow hardware to take over from there in a different way?

My previous DAC, an entry level low performance Teac, upsampled everything with its ASRC to 176.4/192 and this sounded much worse than if I upsampled in HQP.

 

R

"Science draws the wave, poetry fills it with water" Teixeira de Pascoaes

 

HQPlayer Desktop / Mac mini → Intona 7054 → RME ADI-2 DAC FS (DSD256)

Link to comment
My previous DAC, an entry level low performance Teac, upsampled everything with its ASRC to 176.4/192 and this sounded much worse than if I upsampled in HQP.

 

R

Guys, everyone, I do hope I am not hurting anyone's feelings, but I'll say this very plainly: for my DAC, I am yet to hear any upsampling in software, be it in HQP or A+, make any positive impact. The best, the most natural sound I get is when I allow my DAC to do its job. This may be specific to my DAC and the system that follows but this is what I hear with my ears that are not a measurement system but a subjective instrument that I value as the best judge--again, for me, on my equipment. And the best sound I get is with A+, straight to DAC.

Ayre QX-5 Twenty + Oppo 205, Mac mini + Audirvana, Ayre KX-R Twenty pre, Ayre MX-R monos, Monitor Audio PL300 floorstanders, Shunyata cables

Link to comment
Guys, everyone, I do hope I am not hurting anyone's feelings, but I'll say this very plainly: for my DAC, I am yet to hear any upsampling in software, be it in HQP or A+, make any positive impact. The best, the most natural sound I get is when I allow my DAC to do its job. This may be specific to my DAC and the system that follows but this is what I hear with my ears that are not a measurement system but a subjective instrument that I value as the best judge--again, for me, on my equipment. And the best sound I get is with A+, straight to DAC.

 

I was merely responding to your question "is there a benefit in having one type of upsampling (in software) to take it to that point and then allow hardware to take over from there in a different way?"...

"Science draws the wave, poetry fills it with water" Teixeira de Pascoaes

 

HQPlayer Desktop / Mac mini → Intona 7054 → RME ADI-2 DAC FS (DSD256)

Link to comment
I was merely responding to your question "is there a benefit in having one type of upsampling (in software) to take it to that point and then allow hardware to take over from there in a different way?"...

Understood, but please note that I cannot disable the 16x oversampling that my DAC already does, regardless.

Ayre QX-5 Twenty + Oppo 205, Mac mini + Audirvana, Ayre KX-R Twenty pre, Ayre MX-R monos, Monitor Audio PL300 floorstanders, Shunyata cables

Link to comment
With powerful GPU one can replace the second core per channel spec. Note that HyperThreading doesn't count, so the number of threads the CPU has doesn't really matter much for this use case - it only makes OS context switching faster. Clock speed spec also depends on the CPU generation, since newer CPUs with AVX2 and such are faster at equivalent clock speed.

 

Fast RAM is needed, but I don't have a specific definition for it, usually the faster multicore CPUs also have correspondingly faster and wider memory bus too. Same goes for cache size and speed. Amount of RAM depends on what one wants to do, if Roon is planned and such. But these days RAM is cheap, so I'd start with 16 GB. Just make sure all the CPU memory channels are populated.

 

My quiet Win10 machine with i5 6600T (Skylake quad-core 2.7 GHz base) and 32 GB of low latency DDR4-2133 can do DSD64-to-DSD256 upsampling for stereo, per-core load is slightly under 50%. It can also do DSD64-to-DSD512 with per-core CPU loads around 80%. But I know it cannot do DSD64-to-DSD256 for 5.1 channel, not even with GPU help. It can also do RedBook-to-DSD512 using poly-sinc-2s with similar 80% per-core load. So that's really close to it's limits and i7 6700T wouldn't make it much difference, since it's just 100 MHz faster and adds HyperThreading.

 

There's more about HQPlayer and hardware builds here:

http://www.computeraudiophile.com/f11-software/design-pc-server-roon-and-hq-player-25796/

 

Thanks, Jussi. Much appreciated. Larry

Analog-VPIClas3,3DArm,LyraSkala+MiyajimaZeromono,Herron VTPH2APhono,2AmpexATR-102+MerrillTridentMaster TapePreamp

Dig Rip-Pyramix,IzotopeRX3Adv,MykerinosCard,PacificMicrosonicsModel2; Dig Play-Lampi Horizon, mch NADAC, Roon-HQPlayer,Oppo105

Electronics-DoshiPre,CJ MET1mchPre,Cary2A3monoamps; Speakers-AvantgardeDuosLR,3SolosC,LR,RR

Other-2x512EngineerMarutaniSymmetrical Power+Cables Music-1.8KR2Rtapes,1.5KCD's,500SACDs,50+TBripped files

Link to comment
My previous DAC, an entry level low performance Teac, upsampled everything with its ASRC to 176.4/192 and this sounded much worse than if I upsampled in HQP.

 

First step up from the source rate is always the most critical for the oversampling digital filter, and it becomes less important as the rate gets higher and the distance between original source rate Nyquist frequency and the new Nyquist increases.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Understood, but please note that I cannot disable the 16x oversampling that my DAC already does, regardless.

 

You don't have to, and you can reduce the amount of oversampling it does and increase the distance between it's filter corner frequency and the original Nyquist frequency.

 

Remember that 16x means "to 16x rate" which is 705.6/768 kHz. So you get 16x multiplier only for 44.1/48k input rates. If you input RedBook upsampled to 176.4k rate you get 4x oversampling by the DAC to 705.6k rate. And there's 66.15 kHz wide band between original Nyquist and the upsampled Nyquist. Thus the DAC's own filter has very minimal or no difference to the original pass-band.

 

And since the Ayre's "listen" filter is very leaky, instead of having strong ultrasonic images starting right above 22.05 kHz, you get the images starting much higher, only from 154.35 kHz onwards. These are also reduced to significantly lower level by the analog reconstruction filter, compared to the level of those right above 22.05 kHz.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Thanks Jussi. It is a standard established by cable companies (which I believe was set up) to let DVR and cable boxes communicate. It's pretty neat setup, using coaxial cable as ethernet cable. No worries here, Powerline modules / ethernet works great with Roon and HQPlayer.

 

http://www.mocalliance.org

My setup with moca adapters works great with HQplayer and an NAA.

 

Sent from my MI 2 using Computer Audiophile mobile app

Link to comment
You don't have to, and you can reduce the amount of oversampling it does and increase the distance between it's filter corner frequency and the original Nyquist frequency.

 

Remember that 16x means "to 16x rate" which is 705.6/768 kHz. So you get 16x multiplier only for 44.1/48k input rates. If you input RedBook upsampled to 176.4k rate you get 4x oversampling by the DAC to 705.6k rate. And there's 66.15 kHz wide band between original Nyquist and the upsampled Nyquist. Thus the DAC's own filter has very minimal or no difference to the original pass-band.

 

And since the Ayre's "listen" filter is very leaky, instead of having strong ultrasonic images starting right above 22.05 kHz, you get the images starting much higher, only from 154.35 kHz onwards. These are also reduced to significantly lower level by the analog reconstruction filter, compared to the level of those right above 22.05 kHz.

Thanks for this clarification. It is useful. I'm still not entirely clear why you are saying that the filter is leaky. I have not seen measurements for QB-9 DSD anywhere on the Web--only for the original QB-9, not even QB-9 192. Also, forgetting the science for the moment, why is it that it sounds to my ears much better ('natural', 'right') than any software upsampling I've tried so far?

Ayre QX-5 Twenty + Oppo 205, Mac mini + Audirvana, Ayre KX-R Twenty pre, Ayre MX-R monos, Monitor Audio PL300 floorstanders, Shunyata cables

Link to comment
Thanks for this clarification. It is useful. I'm still not entirely clear why you are saying that the filter is leaky. I have not seen measurements for QB-9 DSD anywhere on the Web--only for the original QB-9, not even QB-9 192.

 

AFAIK, the newer models use same or similar digital filters.

 

Also, forgetting the science for the moment, why is it that it sounds to my ears much better ('natural', 'right') than any software upsampling I've tried so far?

 

Cannot really comment about how you perceive the sound, if you like the sound of your Ayre as-is, then just stick to that.

 

But as I said before, I would recommend trying upsampling to DSD too. Although you are limited to DSD64. Then you get 64x oversampling in software and skip the oversampling step of your DAC.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
HQPlayer keeps ASIO driver loaded as long as it is running or until you go to the Settings dialog...

 

Jussi, thanks for the response. There is something different about how HQplayer handles Ravenna ASIO. I had some time and I tested Jriver 22 with Ravenna ASIO. Jriver does not drop the server connection when I skip to the next song or hit pause. OTOH, HQplayer does briefly drop the server connection when I skip to the next song or hit pause in Roon. I can see this happens because the server turns red on the NADAC display for a second like this.

image.jpg

 

When I tested with Jriver, I used standard gapped playback. I did not use any crossfade. This is the way I use Roon+HQplayer.

THINK OUTSIDE THE BOX

Link to comment
Jussi, thanks for the response. There is something different about how HQplayer handles Ravenna ASIO. I had some time and I tested Jriver 22 with Ravenna ASIO. Jriver does not drop the server connection when I skip to the next song or hit pause. OTOH, HQplayer does briefly drop the server connection when I skip to the next song or hit pause in Roon. I can see this happens because the server turns red on the NADAC display for a second like this.

 

What kind of settings are you using, does this happen without Roon? When you are upsampling to fixed output rate and skipping tracks or pausing playback, there are no actions done with the ASIO at all, it is playing all the time, non-stop. And does it work with Roon+JRiver?

 

OK, all the ASIO drivers are handled the same way in HQPlayer, it doesn't know anything about Ravenna or such. I have no idea what the Ravenna ASIO driver is doing and I don't have hardware to test with. Merging can take a look and they can contact me, and Dominique knows it... :)

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

I did send an email to Dominque and company last week so he's aware of it. But I haven't heard anything back in terms of a fix.

 

It only happens using Roon+HQplayer. If I just use HQplayer, it does not drop the server. If I just use Roon, it does not drop the server as described. I upsample to 384k. It also happens with no upsampling and no pipeline/convolution. So it seems to be something that happens only with Roon+HQplayer.

 

What kind of settings are you using, does this happen without Roon? When you are upsampling to fixed output rate and skipping tracks or pausing playback, there are no actions done with the ASIO at all, it is playing all the time, non-stop. And does it work with Roon+JRiver?

 

OK, all the ASIO drivers are handled the same way in HQPlayer, it doesn't know anything about Ravenna or such. I have no idea what the Ravenna ASIO driver is doing and I don't have hardware to test with. Merging can take a look and they can contact me, and Dominique knows it... :)

THINK OUTSIDE THE BOX

Link to comment
It only happens using Roon+HQplayer. If I just use HQplayer, it does not drop the server. If I just use Roon, it does not drop the server as described. I upsample to 384k. It also happens with no upsampling and no pipeline/convolution. So it seems to be something that happens only with Roon+HQplayer.

 

Could you email me a HQPlayer log file where this happens. It could indicate a where the problem is, at least give some direction...

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
Don't have Pipeline Matrix enabled.What is the purpose of this and what gain settings are people using ?

 

There are various different use cases it can be used for:

- Running digital cross-over filters, using for example 8-channel DAC allows 4-way stereo cross-over

- Performing stereo mix-down of multichannel material

- Performing cross-feeds or other headphone processing

 

So one can either copy one input channel to multiple output channels and perform processing for those. Or mix multiple input channels to fewer output channels. Or perform some mixture of the two, or something else...

 

If you don't need it or don't know how to use it, leave it disabled... :)

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

Could be there any issue running HQP on a 32bit win7 installation? I'm having trouble trying to convert pcm to dsd to a JLsounds card (+ their Asio). Under OSX 10.10 no problem.

thanks for any kind help.

g

MacMini 2012 i5 2.3| 4GB |Crucial M4 64 Gb SSD |10.9.1| NorthStar Driver | Optimization Script |HiFace 1| JRiver 19 Mac | Buffalo II DAC | Mastersound 220| Rega Naos

Link to comment

I have all my music ripped from CD's in FLAC bit perfect format. Which filter will be the best for up-scaling the tracks in 24/192? Poly-Sinc, FFT, FIR? CPU and memory are more than sufficient in the system.

 

Since HQ-Player is up-scaling the tracks, do I need to buy music from HD-Tracks in higher resolution? Will it make a difference?

Link to comment
Could be there any issue running HQP on a 32bit win7 installation? I'm having trouble trying to convert pcm to dsd to a JLsounds card (+ their Asio). Under OSX 10.10 no problem.

 

No, it should work. There are some limitations and some things are disabled on 32-bit builds, but most things should work. Since some optimizations are disabled on 32-bit buils, those are slower and thus need somewhat more CPU power for doing the same work. Is your Win7 installation on the same Mac, or is it some other computer?

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...