HQ Player

[Chodi]

What is the sonic benefit of closed form, given the high tradeoffs that it brings cpus to theirs knees?

 

I will give you a reply in non-technical sense. I find that closed form presents a wider/deeper sound stage with greater separation of instruments. To be more specific, there is space between instruments that simply is not there with other filters. It also seems to have some effect on tonal accuracy although that is slight. I have experimented with upsampling 16/44 using closed form to DSD 128 and DSD 256. I honestly find little difference in the results although there is some. I cannot test DSD 512 as my dac supports DSD 512 but my computer is overwhelmed at that level of computation even with the CUDA offload. I have tested DSD 512 vs 256 with other filters and I find little difference.

 

Closed form fast brings nearly the same results but the difference is noticeable. I think that would depend on the resolution of your entire system as to if it would matter to you. Closed form fast requires far less computing power. Interestingly I do find that using closed form brings less difference between the choice of modulator provided you choose one that can handle high oversampling.

[ted_b]

Great feedback Chodi, thanks!

[Le Concombre Masqué]

I would list the tonal accuracy as first benefit and not by a slight margin. The sonic tradeoff being a leaner sound vs the poly sinc family. Closed form thus led me to consider that half an hour of continuous playing is required when I Corning/Green Regen connect my MBPr (everything else being 24/7 on and I had no noticeable warming period when using a regular hub instead of the Regen). If I don't jump on comparing stuff I get great listening sessions outside of classical and MFSL mastered recordings which first appeared as clear benefiters of closed form filter.

 

I also find myself using a single modulator , ASDM7, and not fiddling when using cff

I will give you a reply in non-technical sense. I find that closed form presents a wider/deeper sound stage with greater separation of instruments. To be more specific, there is space between instruments that simply is not there with other filters. It also seems to have some effect on tonal accuracy although that is slight. I have experimented with upsampling 16/44 using closed form to DSD 128 and DSD 256. I honestly find little difference in the results although there is some. I cannot test DSD 512 as my dac supports DSD 512 but my computer is overwhelmed at that level of computation even with the CUDA offload. I have tested DSD 512 vs 256 with other filters and I find little difference.

 

Closed form fast brings nearly the same results but the difference is noticeable. I think that would depend on the resolution of your entire system as to if it would matter to you. Closed form fast requires far less computing power. Interestingly I do find that using closed form brings less difference between the choice of modulator provided you choose one that can handle high oversampling.

[Chodi]

I did forget to mention the obvious in my post. You can run closed form in pcm/DXD mode (352K) without giving your computer a heart attack. That presents a very light load on the computer and still gives you the experience although not quite the same as the DSD version. You might even prefer it.

[Ipoci]

CUDA offload is also working here with my ASUS G501 jw.

 

I saw you own a MacBookPro 2.3 Quad -i7, could you please briefly describe the performance comparison of MBP vs. Asus G501 ?

 

I own a rMBP 11,4 2.2GHz quad -i7 and I'm quite sure I cannot upsample 16/44 to DSD256, so I'm trying to understand a possible alternative laptop platform that combines decent characteristics for business usage and enough power for music processing.

 

Thanks in advance, Massimiliano

[Miska]

Many listeners and designers prefer the sound of closed form filters. Miska is not one of them.

 

I can't tell you what the difference is, having listened only briefly, so I apologize for not answering your question.

 

I prefer apodizing filters for RedBook content because in many cases the decimation filters used to produce the content are sub-optimal.

 

Closed-form, by definition is not apodizing, so it may sound good with good source material where the digital filters used for recording and/or mastering stage are sonically good. But it cannot fix the defects of bad digital filters from the earlier stages unlike apodizing filters do.

 

So apodizing filters like poly-sinc (except poly-sinc-hb which is non-apodizing) usually give more consistent performance regardless of source material. Because you hear only differences in the actual source signal when material changes, but not the differences between digital filters used in the source material production phase (recording/mastering).

 

 

P.S. For some reason Metrum Musette (and maybe some other NOS R2R ladder DACs) seem to like closed-form filter. I first found out about it by listening and then also verified it with measurements.

[Jud]

Closed-form, by definition is not apodizing

 

That is one of those remarks where I say "oh, of course" after I hear it, but didn't think of it before. Thanks.

 

P.S. For some reason Metrum Musette (and maybe some other NOS R2R ladder DACs) seem to like closed-form filter. I first found out about it by listening and then also verified it with measurements.

 

"Some other...R2R ladder DACs" perhaps including Yggdrasil? (Obviously not NOS since it uses its own closed-form filter.)

 

Edit: Any thoughts as to why the combination of closed form filtering and no SDM seems to work?

[Mercman]

I saw you own a MacBookPro 2.3 Quad -i7, could you please briefly describe the performance comparison of MBP vs. Asus G501 ?

 

I own a rMBP 11,4 2.2GHz quad -i7 and I'm quite sure I cannot upsample 16/44 to DSD256, so I'm trying to understand a possible alternative laptop platform that combines decent characteristics for business usage and enough power for music processing.

 

Thanks in advance, Massimiliano

 

The Asus G501 jw is a much newer computer with a 4th Generation Haswell processor that was released in Jan 2015. Now there is Broadwell-the changes never end! For Business and music, the $1600 (price might be lower) Asus works very well. I needed Thuderbolt for my drives and the Asus has this feature.

 

The MacBook Pro I have is older and will have to work much harder converting PCM to DSD 256. You could purchase a new MacBook Pro that would be similar in performance to the Asus, but at a higher cost.

[Samuel T Cogley]

You do know that closed form is extremely processor intensive. You have a high end system but still most of us with good systems struggle to reach DSD 256 using closed form. In fact before the addition of CUDA in the beta it was impossible for me to get smooth results with DSD 256 using closed form. DSD 512 is massively demanding on computer resources and may be out of reach for all but the very best equipment available. Even then I'm not certain commercially available home systems can handle that. Closed form fast is a breeze to use and doesn't have the same demands.

 

Thank you for this feedback. It wasn't until I built that PC this year that I was able to hear DSD512 for the first time.

 

I'd be curious to hear what kind of system Miska thinks could handle closed-form/DSD512.

[57gold]

Read these last few posts on closed form filter...CD rip to DSD 128 sounded great on newly released, well engineered jazz...painful playing a 1960s analog to 1980s CD of the Byrds...tweeters sounded shrill. Switched back to poly-sinc-short-mp...much better!

[ted_b]

Read these last few posts on closed form filter...CD rip to DSD 128 sounded great on newly released, well engineered jazz...painful playing a 1960s analog to 1980s CD of the Byrds...tweeters sounded shrill. Switched back to poly-sinc-short-mp...much better!

 

Wouldn't it be great to have some cataloging system (maybe Roon could do this once it allows HQP setting changes) that pairs our favorite HQP filters/settings with actual tagged album names (or some unique identifier)? Cue up an album, HQP sets the best filters we choose last time we played the thing. Poly sinc for Byrds redbook, closed loop for Channel Classics Mahler 9. The possibilities are endless.

[Forehaven]

+1

[Miska]

Wouldn't it be great to have some cataloging system (maybe Roon could do this once it allows HQP setting changes) that pairs our favorite HQP filters/settings with actual tagged album names (or some unique identifier)? Cue up an album, HQP sets the best filters we choose last time we played the thing. Poly sinc for Byrds redbook, closed loop for Channel Classics Mahler 9. The possibilities are endless.

 

I already proposed it to them... And I've been planning to make HQPlayer remember the settings per album when playing back in album mode from the built-in library. Could be extended later to be more flexible.

[ted_b]

I already proposed it to them... And I've been planning to make HQPlayer remember the settings per album when playing back in album mode from the built-in library. Could be extended later to be more flexible.

 

Geniuses think alike (and me too).

[dean70]

Wouldn't it be great to have some cataloging system (maybe Roon could do this once it allows HQP setting changes) that pairs our favorite HQP filters/settings with actual tagged album names (or some unique identifier)? Cue up an album, HQP sets the best filters we choose last time we played the thing. Poly sinc for Byrds redbook, closed loop for Channel Classics Mahler 9. The possibilities are endless.

 

Have done this by dropping an XML file with the filter settings in the album folder & using hqp-control via a script to set hqplayer up before playback.

[ted_b]

Have done this by dropping an XML file with the filter settings in the album folder & using hqp-control via a script to set hqplayer up before playback.

 

How seamless is this? Meaning:

1) how easy to add or change XML file, or make one that can be added globally to X number of albums

2) does HQPlayer need to "restart" after each album?

Thx

Ted

[Miska]

I'd be curious to hear what kind of system Miska thinks could handle closed-form/DSD512.

 

I just managed to optimize the closed-form a bit further. Now, on Linux, it runs just fine to DSD512 on Xeon E5v3 + Quadro K620. GPU load peaks at 98% and I cannot touch web browser (or the desktop GUI in general) at the same time or it will have drop-outs (browsers use OpenGL for rendering pages these days).

 

Roon also uses OpenGL for rendering, so people who plan to use both together may need to take that into account when choosing graphics card solution...

 

Technically it is possible to also have multiple GPUs, but HQPlayer doesn't currently properly support that. It just picks up the first that has good match for the needed capabilities. Later on I can add support for balancing load on multi-GPU machines.

[Miska]

1) how easy to add or change XML file, or make one that can be added globally to X number of albums

2) does HQPlayer need to "restart" after each album?

 

Easiest is to make the setting changes through GUI and use the export/import functionality from File-menu to export and import different settings. Not very handy, but doable.

[Miska]

Have done this by dropping an XML file with the filter settings in the album folder & using hqp-control via a script to set hqplayer up before playback.

 

You can also control the active settings through hqp-control if you like. The hqp-control in 3.12 has some extensions on this front. You just need to make sure playback is in stopped state and then use following order for making settings:

1) Set mode (PCM/SDM)

2) Set filter

3) Set dither/modulator

4) Set output rate

This because previous selections change available selections for the next setting.

[Samuel T Cogley]

I just managed to optimize the closed-form a bit further. Now, on Linux, it runs just fine to DSD512 on Xeon E5v3 + Quadro K620. GPU load peaks at 98% and I cannot touch web browser (or the desktop GUI in general) at the same time or it will have drop-outs (browsers use OpenGL for rendering pages these days).

 

Roon also uses OpenGL for rendering, so people who plan to use both together may need to take that into account when choosing graphics card solution...

 

Technically it is possible to also have multiple GPUs, but HQPlayer doesn't currently properly support that. It just picks up the first that has good match for the needed capabilities. Later on I can add support for balancing load on multi-GPU machines.

 

Thank you!!!

 

I'm looking forward to trying the new beta!

[dean70]

How seamless is this? Meaning:

1) how easy to add or change XML file, or make one that can be added globally to X number of albums

2) does HQPlayer need to "restart" after each album?

Thx

Ted

 

Here is an example xml:

 

<?xml version="1.0"?>
<default>
   <mode name="pcm" />
   <filter name="poly-sinc-shrt" />
   <dither name="NS5" />
</default>

 

 

I have integrated this into my HTPC Shell application that controls all Audio Server / HTPC functions via Remote (which is still a work in progress).

 

Workflow:

1. Load playlist into HQPlayer via hqp-control.exe command

2. Read the xml file if it exists from the playlist folder

3. Set the filters in the order shown (left to right in the UI) via hqp-control.exe. I usually select the highest supported sample rate for the DAC

4. Send the -play option via hqp-control

 

The only thing is it will leave HQPlayer set to the last used filter settings, so need some sort of default to reset to.

[Ipoci]

I just managed to optimize the closed-form a bit further. Now, on Linux, it runs just fine to DSD512 on Xeon E5v3 + Quadro K620. GPU load peaks at 98% ...

 

What's the suggested rule for GPU selection on laptop ?

 

GeForce shows overall better computing performances but for CAD Quadro is suggested.

 

So, it should be better a Quadro K4000M (3.0) or a GTX965M (5.2) ?

 

Have a nice day, Massimiliano

[Mercman]

One interesting feature of the ASUS laptop is that you can select either the default Intel GPU or nVidia GPU to run an app.

[KMan]

Thanks to this forum, I tried HQPlayer and bought a license within few days through @bibo01 (which was a great deal). Thank you @Miska for creating HQPlayer. I have been using v 3.12.0 beta to checkout CUDA. Redbook to DSD256 works great with CUDA, Tried Redbook -> DSD512 for iDSD Micro, but my Macbook pro was overwhelmed.

 

Now for the real question, has anyone tried to upsample DSD64 -> DSD128 or DSD256 with CUDA? I have not had much luck, playback gets choppy and HQPlayer even becomes unstable. Playback will not start if I enable both CUDA and pipeline SDM. I have tried it both on Macbook Pro and a HP laptop both with Nvedia CUDA capable (with drivers installed).

 

MacBook Pro specs:2014, 2.5Ghz i7, 16GB RAM, GT 750M, 2048 MB Video RAM

HQPlayer DSF settings: Poly-short-lin/IIR/Standard

HQPlayer DSD playback: Ply-sinc-2s/ASDM7/(DSD128/DSD256/DoP 2xDSD)

Source Music file: DSD64 on a external HDD, USB3 as .dsf

 

I also use convolution filters (12K tap @96KHz) for L and R channels, I have tried to disable it does not seem to help much.

If I am doing something wrong here or you have a recommended setting please let me kno or it is possible that the Macbook does not have the chops to handle this.

 

Can't wait for Roon integration....

[KMan]

@Miska I want to report that DSD64 to DSD128/256 conversion may not be working on 3.12.0. I had the same issues on a windows desktop machine (Win 10, 3.5 GHz i7, no CUDA, 32 Gig RAM). I did not have the problem with the previous version on the Windows desktop. Then I reverted to old version (3.11.0), and I do not have the choppiness. I understand this is in beta, but I hope you can confirm if this is an issue.