HQ Player

[Miska]

Is there any advantage to using this even-multiple of 48/96/192k PCM when using non-integer filters like poly-sinc-short?

 

A bit less CPU load...

[Miska]

All DSD DACs should play "non-standard" DSD frequencies, they are not common mentioned but they are the integer multiple for 48/96/192k PCM.

 

I love the integer and not the floating point conversion.

 

For example converting 44100 to 192000 is as much integer for poly converters as going from 44100 to 176400. Any conversion where the neither of the sampling rates is irrational number is "integer". Since almost all formats handle sampling rates as integers, the conversion is always also integer. What you describe as "integer" conversion is case where either numerator or denominator is equal to 1. Or in other words modulo (remainder) of the rates is equal to 0. Some algorithms/filters have this kind of restriction while others don't.

[Miska]

@ Miska question regarding poly sinc filter why does it tax the processor at 100% for about 8-10sec then drop down to 20% when it starts to play a track? Its the only filter that behaves this way to best of my knowledge?

 

When the playback is being started up, there's quite a bit of CPU time needed to initialize all the DSP algorithms. That's why there's a load peak at start. It also happens when source file sampling rate changes. Time needed for initialization phase depends on selected algorithm combination (including all the settings not in the main window).

 

Some of the initialization is cached, so stopping and starting another playback for the same source format doesn't have to re-do all the initialization.

[Miska]

I go back and forth between poly-sinc-short and closed-form-fast. Closed-form-fast is very smooth and relaxing but not boring. I'm still trying to decide whether the greater "bite" in poly-sinc-short is the "real" sound of the instruments coming through, or undesirable harshness.

 

QUESTION -- what is the difference between closed-form and closed-form-fast?

 

poly-sinc/poly-sinc-short (both linear and minimum -phase variants), and many of the others are apodizing filters. IOW, in most cases they replace "ringing" of the original decimation filter, either from ADC or from mastering (if the content was mixed at higher rate). So for example if you select poly-sinc-short-mp you get pre-ringing removed and replaced with short gentle post-ringing. Shape of the filter roll-off slope is also designed to avoid harsh sound while the short ringing and high attenuation keep transients snappy and clean.

 

For example poly-sinc-hb and closed-form are non-apodizing, so they leave the original ringing as-is and just compute more samples. Since they don't change the original data, the results are reversible back to the original data. Both could also be categorized "long" due to steep cut-off.

 

With apodizing filters you likely get more consistent sonic character regardless of source material, while the non-apodizing are much more sensitive to the source material properties and how it was produced.

 

I personally prefer apodizing ones, because lot of source material either has not so great sonic characteristics of low-resource ADC's digital decimation filter. Or characteristics of rate converter used at later production stage which are commonly not so great either, especially for non-audiophile RedBook productions.

[Le Concombre Masqué]

Guess that's why I associated Closed form with SACDs. How about poly sync ext ? On the samples I tested it was the willer for PCM...

poly-sinc/poly-sinc-short (both linear and minimum -phase variants), and many of the others are apodizing filters. IOW, in most cases they replace "ringing" of the original decimation filter, either from ADC or from mastering (if the content was mixed at higher rate). So for example if you select poly-sinc-short-mp you get pre-ringing removed and replaced with short gentle post-ringing. Shape of the filter roll-off slope is also designed to avoid harsh sound while the short ringing and high attenuation keep transients snappy and clean.

 

For example poly-sinc-hb and closed-form are non-apodizing, so they leave the original ringing as-is and just compute more samples. Since they don't change the original data, the results are reversible back to the original data. Both could also be categorized "long" due to steep cut-off.

 

With apodizing filters you likely get more consistent sonic character regardless of source material, while the non-apodizing are much more sensitive to the source material properties and how it was produced.

 

I personally prefer apodizing ones, because lot of source material either has not so great sonic characteristics of low-resource ADC's digital decimation filter. Or characteristics of rate converter used at later production stage which are commonly not so great either, especially for non-audiophile RedBook productions.

[Miska]

How about poly sync ext ? On the samples I tested it was the willer for PCM...

 

It is apodizing and otherwise similar to poly-sinc, but with different kind of roll-off curve.

[vinoboy]

Just downloaded 3.11.0. Cosed-Form filter sounds awesome. I do find one strange thing. My DAC (Geeek Pulse Infinity + LPS) is capable of playing DSD256 and I have set the SDM limit to 11289600. Before I play anything, the bit rate in HQP main window maxes out at 5644800. Not sure if anyone else has encountered this. Here are the screen shots:

 

 

Any idea as to what is going on?

 

TIA

[brightonjel]

I don't believe that LH has yet released a firmware update that supports DSD256, and therefore that 128 is as high as it goes. My Pulse supports 128DSD for 5.6 Mhz (integer scaled from 44.1) and 6.2 (scaled from 48).

[vinoboy]

I signed up for beta test...

[Miska]

Just downloaded 3.11.0. Cosed-Form filter sounds awesome. I do find one strange thing. My DAC (Geeek Pulse Infinity + LPS) is capable of playing DSD256 and I have set the SDM limit to 11289600. Before I play anything, the bit rate in HQP main window maxes out at 5644800.

 

HQPlayer can convert DSD rates up and down if you have DirectSDM disabled and/or matrix processing enabled.

 

If you set default/limit to higher than what hardware supports in Settings dialog it is anyway limited to the maximum supported by hardware.

[dean70]

One thing I have yet to try is if you have a mixed m3u playlist consisting of pcm 44100 & dsd 64 (from sacd) tracks, how is it handled with the filter selection?

 

I am reverting to poly-sinc if mixed tracks are detected to avoid up-sampling issues, but does HQPlayer handle mixed formats in the same playlist?

[Miska]

One thing I have yet to try is if you have a mixed m3u playlist consisting of pcm 44100 & dsd 64 (from sacd) tracks, how is it handled with the filter selection?

 

Depends on the case:

1) Output in PCM mode

1.1) Filter is set to "none" -> output follows the source

1.2) Input is PCM -> conversion using selected filter

1.3) Input is DSD -> conversion to 1/16th rate PCM using settings from DSDIFF/DSF Settings -> further conversion using selected filter

2) Output in SDM mode

2.1) Input is PCM -> conversion using selected filter

2.2.1) Input is DSD and Direct SDM is enabled -> output at source rate

2.2.2) Input is DSD and Direct SDM is disabled -> conversion to desired output rate using special SDM rate conversion method

 

I am reverting to poly-sinc if mixed tracks are detected to avoid up-sampling issues, but does HQPlayer handle mixed formats in the same playlist?

 

Yes, mixed playlists are OK, but may not be gapless when source format changes. If source, filter/modulator and requested output conflicts, then playback stops at that point.

[brightonjel]

Great! Means you get to debug it first and then I'll download it once it's working and a production release! Deal! :-)

I signed up for beta test...

[vinoboy]

HQPlayer can convert DSD rates up and down if you have DirectSDM disabled and/or matrix processing enabled.

 

If you set default/limit to higher than what hardware supports in Settings dialog it is anyway limited to the maximum supported by hardware.

 

Hi Miska... Thanks for the swift response ! I will give it a try.

 

Here is screen shot of the LH Control Panel. It indicates the sample rate is 11.2 mhz. which I take the DAC is capable of handling up to 11.2mhz? Now if you look at the main window of HQP, it still says 5.6mhz. Not sure if its a display issue or a true sample rate being played?

 

My understanding is that if you have 11.2 mhz file and if your DAC is capable of playing 11.2mhz sample rate, with the Direct SDM enable, it should play 11.2mhz as it is without any conversion, right?

 

Thanks!

[Miska]

My understanding is that if you have 11.2 mhz file and if your DAC is capable of playing 11.2mhz sample rate, with the Direct SDM enable, it should play 11.2mhz as it is without any conversion, right?

 

Yes, the main window settings are just ignored in that case... Those then apply when PCM content is encountered. Rate support is checked before changing, so in this case the 11.2 MHz rate should also appear as available in the sample rate drop-list in main window.

[Chodi]

I have a question for Miska. When I use closed form upsampling in pcm 352.8 DXD I am only using 6% cpu. When I do DSD256 upsampling using closed form I am using 60% cpu? DXD is basically 4 times DSD64 so I can't see why there would be such a resource difference? To me it is basically two sides of the same coin. Please explain?

 

(I7 processor on Windows 10 computer)

[kurb1980]

I have a question for Miska. When I use closed form upsampling in pcm 352.8 DXD I am only using 6% cpu. When I do DSD256 upsampling using closed form I am using 60% cpu? DXD is basically 4 times DSD64 so I can't see why there would be such a resource difference? To me it is basically two sides of the same coin. Please explain?

 

(I7 processor on Windows 10 computer)

 

When using DSD256 with closed form my CPU hovers at around 50% so I believe that's normal with DSD upsampling. I believe that the closed form filter is primarily used in the PCM realm I tried it with DSD256 not a big fan.

[Chodi]

I agree but I was more interested in the mechanics of why DXD and DSD256 with the same filter should have such a different use of resources.

[vinoboy]

Yes, the main window settings are just ignored in that case... Those then apply when PCM content is encountered. Rate support is checked before changing, so in this case the 11.2 MHz rate should also appear as available in the sample rate drop-list in main window.

 

Thanks Miska ! Makes sense now...

[Le Concombre Masqué]

the other way round here : pcm upsampled via poly sinc ext and DSD through Closed form

 

When using DSD256 with closed form my CPU hovers at around 50% so I believe that's normal with DSD upsampling. I believe that the closed form filter is primarily used in the PCM realm I tried it with DSD256 not a big fan.

[Miska]

I have a question for Miska. When I use closed form upsampling in pcm 352.8 DXD I am only using 6% cpu. When I do DSD256 upsampling using closed form I am using 60% cpu? DXD is basically 4 times DSD64 so I can't see why there would be such a resource difference? To me it is basically two sides of the same coin. Please explain?

 

Every time you double sampling rate you roughly double the load, just for the filters. So if you start from 44.1k to DXD you have "8x load" and to DSD256 you have "256x load" (or 11289600/352800=32x difference). In addition with DSD you have the load of the modulator which is about half of the work. With -2s and such the scaling is not so steep.

[Le Concombre Masqué]

Just listened to MFSL's remarkable SACD edition of Kind of Blue at really high level and it was really gorgeous setting rate at X2, filter as Closed Form, and DSD 7 as modulator ; an all time best

the other way round here : pcm upsampled via poly sinc ext and DSD through Closed form

[Serg.xa]

Hello! What settings to apply to files 44,1/16 - without transformations that on dac the signal in accuracy such how it was written down arrived? If perhaps send a screenshot. For what the closed form is necessary?

[Iain]

I have a question for Miska. When I use closed form upsampling in pcm 352.8 DXD I am only using 6% cpu. When I do DSD256 upsampling using closed form I am using 60% cpu? DXD is basically 4 times DSD64 so I can't see why there would be such a resource difference? To me it is basically two sides of the same coin. Please explain?

 

(I7 processor on Windows 10 computer)

 

Are you running 64 bit versions of Windows 10 and HQ Player?

 

If not you should be, as Intel Core I7 processor, 64 bit OS and applications will make a significant difference in performance over 32 bit OS and applications.

[Chodi]

Are you running 64 bit versions of Windows 10 and HQ Player?

 

If not you should be, as Intel Core I7 processor, 64 bit OS and applications will make a significant difference in performance over 32 bit OS and applications.

 

Yes, I am using 64 bit Windows 10. I don't really have any performance problem but closed form as DSD 256 is a stretch for my cpu. I find little difference in sound quality when using DSD 128 with closed form and the resource demand is much lower. Frankly I find that DXD has a bit of a digital edge to it with some recordings. DSD is more reliable in eliminating that edge. Some might complain the DSD is just smoothing things but I think it sound more natural with classical music. Both closed form and poly ext work well for me. I think the front end in use will find a sweet spot with one of these.