HQPlayer does not work (resolved, now some results)

[amirm]

You are asking this lazy person???

Seriously, I will give it a try if someone can post for me the typical options people use.

 

[Blumlein 88]

This one would work. Not sure what is typical. Depends upon whether you are doing everything DSD, or PCM or what your DAC will manage.

Here is one where I had the Audiophileo feeding a 96 khz capable DAC. I think Miska suggested his poly sinc filter as a good starting point and maybe his best filter for most purposes. This btw is an evaluation copy from maybe 18 months ago, but the settings I don't think are much different. I never purchased it. It will only work for 30 minutes in the evaluation version. I thought the interface could have been a little better, and while it works well, I wasn't convinced I was hearing a difference.

Usually you wish to use the highest rate available for your DAC. The manual gives a brief description of the various filter types for PCM. I wish I had kept the test files. Using the -4 db white noise, and an IMD twin tone, and a silent noise file plotting the way JA does at Stereophile (what most people call a Jurgen plot) resulted in lower noise floors, less aliasing even somewhat lower distortion than any other filtering I have seen. You are of course upsampling lower rate PCM and playing it back at a higher rate, but the analog output was cleaner than the native rates, and cleaner than other upsampling software I have used. So this is some good software and Miska really knows how to make some good filters.

It would be my opinion, if you want the best performing playback scientifically in a measurable way this is the software to use.

 

[amirm]

Thanks a lot. I will use that as starting point.

 

[mirekti]

Many claim DSD512 is where the magic happens. If I understood correctly, such high upsampling rate serves to help out the math behind the filters used.
There is also a Modulator specific for 512 AMSDM512+fs.
Everyone has his own preference, but from what I've read many like poly-sinc-short-mp filter which is supposedly good for jazz and acoustic music.

One thing to note is, when using DSD it was suggested to set the Vol Max to -3dB in order to prevent... ...well, to prevent something
You can also enable Log file in this window and see what HQPlayer negotiated with the DAC.

 

[Blumlein 88]

OOOPS!!!! Experimental mishap. Leaving this here for now, but may delete it.

I started out with a Dell laptop feeding the firewire lead to the TC Impact Twin. I was using a Lenovo with the Forte. After the first test I swapped places with the computers. I was checking other software for playback and upsampling. I swapped as that software was already on the Lenovo. There are some differences, but very, very minor. I ran the initial 48 khz natively in Foobar and it was also clean with low noise the second time on the Lenovo. Somehow the Dell firewire port is causing high jitter and a higher noise floor in the TC Impact Twin. With the Lenovo it is clean. Let people not complain about asynch USB.

--------------------------------------------------------------------------------------------------------


Okay had some time to kill so did a few measures with Foobar and HQplayer.

The ADC was the Focusrite Forte set to 192 khz/24bit. The DAC was the TC Impact Twin first playing a 48 khz test file with Foobar via ASIO. Next I played the same 48 khz test file via HQ Player using the TC set to 192/24 as HQ Player upsamples to that rate. I used the poly-sinc filter. In all of these FFTs the blue curve is the Foobar playback and green is HQ Player.

First up silence to see the noise floor. You'll see the HQ player is about 12 db lower in noise below 20 khz. The 31 khz spike is an idle tone in the Forte ADC, and the noise rise above 50 khz is the Forte.

Now a 1 khz tone. The noise floor is nastier around the Foobar tone while the HQ Player noise floor is not much effected. With HQP the harmonics rise cleanly out of an even noise floor. There also are some low level spikes at 43&45 khz and double that which are absent in HQP.

Now for some twin tone IMD signals at 18&19 khz. The Foobar playback has much more grass around the tones and more ultrasonic mirror tones as well.

Now a look at the Jtest signal at 12 khz. Would appear the HQ Player results in lower jitter around the Jtest signal. Presumably that is also why the general noise floor is lower.

Now some - 4db white noise from each device. The filter in the Foobar version rolls off starting at 20 khz and is pretty steeply filtered. The HQP version is much cleaner ultrasonically though not rolling off until near 22 khz.

So if one is comparing bit perfect playback at the native 48 khz rate and the upsampled playback via HQP it is clear HQP gives better results. Whether any of these artefacts are audible is another issue. HQP is measurably superior.

 

[Blumlein 88]

Sorry guys have to retract the above post.

I started out with a Dell laptop feeding the firewire lead to the TC Impact Twin. I was using a Lenovo with the Forte. After the first test I swapped places with the computers. I was checking other software for playback and upsampling. I swapped as that software was already on the Lenovo. There are some differences, but very, very minor. I ran the initial 48 khz natively in Foobar and it was also clean with low noise a second time on the Lenovo. Somehow the Dell firewire port is causing high jitter and a higher noise floor in the TC Impact Twin. With the Lenovo it is clean. Let people not complain about asynch USB.

So should I simply delete that post or leave it as a cautionary example?

 

[Liberal Democracy]

Sorry guys have to retract the above post.

I started out with a Dell laptop feeding the firewire lead to the TC Impact Twin. I was using a Lenovo with the Forte. After the first test I swapped places with the computers. I was checking other software for playback and upsampling. I swapped as that software was already on the Lenovo. There are some differences, but very, very minor. I ran the initial 48 khz natively in Foobar and it was also clean with low noise a second time on the Lenovo. Somehow the Dell firewire port is causing high jitter and a higher noise floor in the TC Impact Twin. With the Lenovo it is clean. Let people not complain about asynch USB.

So should I simply delete that post or leave it as a cautionary example?

Leave it, that's what we are about, honestly and integrity

Always questioning

 

[Blumlein 88]

Leave it, that's what we are about, honestly and integrity

Always questioning

Okay, I'll leave it. Unless the boss decides to expunge the thread with extreme prejudice.

 

[Blumlein 88]

As a follow up a few graphs once the correct and stable test setup was achieved.

First is a 12 khz tone. Green is HQ player which has resampled to 192 khz, blue is the native file at 48 khz played via Foobar, and red is Sox resampled to 192 khz played via Foobar. Nothing of note with essentially no difference.

Next is -4db of white noise. Again green is HQ player, blue is native 48 khz via Foobar, and red is Sox resampled to 192 khz via Foobar. Here you see some differences past 20 khz.

 

[Blumlein 88]

Now one final comparison. In this case we have HQ Player resampling to 192 khz vs HQ Player playing at the native 48 khz rate. Using poly sinc filtering as above.

First the 12 khz tone. Blue is native 48 khz and green is upsampled by HQ Player to 192 khz.

Next is the -4 db white noise where we see a little difference once again. Once again green is HQ Player at 192 khz upsampling and blue is the HQ Player at native 48 khz rates.

 

[amirm]

Thanks for running these tests. The net is that in audible band, i.e. < 20 Khz, there is no difference. Above that, HQPlayer has better response but is not something we are going to hear due to it being ultrasonic. Even there, the improvements are tiny.

 

[mirekti]

Thank you all for the results, I wish I was able to contribute!!!

One other thing comes to my mind, but I am not sure if this makes any sense from the measurement's perspective.
On CA forum many claim all the 'magic' occurs when the signal is upsampled to DSD512 not PCM, and then certain filters are applied to it. I believe I read that the math behind the filters works better with such higher (DSD512) upsampled file.
Is there a way to see that a file upsampled to DSD512 really is better than PCM 44.1 kHz through a "regular" DAC? I'd assume the best way to use HQPlayer is with a DAC that has NOS capabilities so only HQPlayer filter is applied.
I hope all of the above makes sense, if not, well, it don't so I am fine with that as well.

Thanks and cheers!!!

...maybe the topic's name should be updated as well.

 

[amirm]

I updated the title.

On using DSD 512, that is indeed what some people seem to do. Alas, that requires a DAC that accepts that rate. I don't think I have any.

 

[bibo01]

On DSD512, with the just released Amanero firmware update more people should be able to playback at that rate in Linux.

There is a marginal technical improvement between DSD256 and DSD512, however, going up in frequency rate it's not a guarantee of better output quality - it all comes down to your hardware implementation.

On the source side, in order to upsample to DSD512 with HQPlayer you need a pretty powerful PC, possibly through an NAA.
One thing to note (many people do not know about it), if you upsample DSD64 to DSDxxx, the conversion/noise filter combination that do apply are in DIFF/DSF setting menu. Desktop drop down menus do not apply.

On the DAC side, the performance between DSD4x and DSD8x is mostly influenced by the local oscillators' implementation and by analog filter output.

 

[Fitzcaraldo215]

I am really scratching my head about DSD512 or about ultra,ultra hi rez in general. I have never heard DSD beyond 256 or PCM beyond 384k, my DAC's limits. I tend to believe I am not missing anything, unless the laws of diminishing returns have been repealed. So, I do not exactly yearn for ever higher sampling rates or a DAC to handle them plus all the additional hard drives I would need on my NAS to store the files.

However, I do believe based on the informal comparisons I have done that hi rez can offer a small but noticeable and worthwhile sonic improvement over RBCD. That is if it is part of an end-to-end hi rez recording/playback chain using hi rez on the initial a-d. Downrezzing a recording to RBCD from native hi rez seems noticeably lower in sonic quality to me. I also think hi rez is much less convincing or worthwhile from analog masters or uprezzed from native RBCD recordings. Again, I have not tried HQPlayer and uprezzing plus conversion to DSD are probably HQPlayer's most widely used features.

I do have some excellent sounding files from a recording engineer friend natively recorded at DSD256, but their excellence could easily be the result of the special miking and engineering, and it likely is that rather than just the sampling rate.

I do not mean also to bring up the whole DSD vs. PCM thing, which has been overblown. I would rather not convert format, up or downsample anything, if possible. Some rave about HQPlayer converting RBCD to hiX DSD. I have not tried it since I am convinced that DSP Room EQ is much more sonically significant than DSD vs. PCM or extreme sampling rates. Also, other DSP features are crucial to me like bass management crossovers and Mch speaker distance correction.

So, my most usual listening mode is DSD64, 128 or 256 playback converted to PCM176 on-the-fly by my faithful JRiver library/player utilizing the Dirac Live plugin. HQPlayer claims to offer some DSP natively in DSD, but I am doubting it can support all my needs and the documentation seems elusive. I am also skeptical that, used just as a player and not uprezzer or DSD converter, it offers any sonic advantage, but it definitely has convenience disadvantages. I also do not think any bias controlled listening tests of HQP have ever been published.

But, any further insights would be welcome.

 

[amirm]

My DAC supports rates up to DSD 256. Is there a low cost DSD 512 that people advocate for use here?

 

[bibo01]

My DAC supports rates up to DSD 256. Is there a low cost DSD 512 that people advocate for use here?

People tend to use an iFi Dac for that. I don't remember which model does DSD512, but it's easy to look it up.

EDIT: I believe it's ifi Micro iDSD

 

[bibo01]

...
I do not mean also to bring up the whole DSD vs. PCM thing, which has been overblown. I would rather not convert format, up or downsample anything, if possible. Some rave about HQPlayer converting RBCD to hiX DSD. I have not tried it since I am convinced that DSP Room EQ is much more sonically significant than DSD vs. PCM or extreme sampling rates. Also, other DSP features are crucial to me like bass management crossovers and Mch speaker distance correction.

So, my most usual listening mode is DSD64, 128 or 256 playback converted to PCM176 on-the-fly by my faithful JRiver library/player utilizing the Dirac Live plugin. HQPlayer claims to offer some DSP natively in DSD, but I am doubting it can support all my needs and the documentation seems elusive. I am also skeptical that, used just as a player and not uprezzer or DSD converter, it offers any sonic advantage, but it definitely has convenience disadvantages. I also do not think any bias controlled listening tests of HQP have ever been published.

But, any further insights would be welcome.

In HQPlayer you can do convolution and Mch speaker management in both PCM and DSD natively. For ex. you do not need to down-convert from DSD to PCM to apply DSP.
In your case you would need to make new FIR filters (e.g. with free applications like DRC or REW) to convolve the signal because, unfortunately, Dirac filters are not exportable.

 

[amirm]

People tend to use an iFi Dac for that. I don't remember which model does DSD512, but it's easy to look it up.

EDIT: I believe it's ifi Micro iDSD

Thanks. I went ahead and ordered it.

 

[Fitzcaraldo215]

In HQPlayer you can do convolution and Mch speaker management in both PCM and DSD natively. For ex. you do not need to down-convert from DSD to PCM to apply DSP.
In your case you would need to make new FIR filters (e.g. with free applications like DRC or REW) to convolve the signal because, unfortunately, Dirac filters are not exportable.

Thanks. I was aware of that, but not the Mch support. But, I will stand pat with JRiver/Dirac Live for now.