HQ Player

[stefano_mbp]

On 2/25/2021 at 10:38 AM, Miska said:

For me, this is certainly gapless. (Audirvana running on macOS, content from local Thunderbolt HDD)

today I tried again with Audirvana + HQPlayer Embedded and this time I must admit that the gapless works .... only difference is that last night I installed version 4.22.2 .... the other tests were with version 4.22.1 and earlier .... can there be a correlation?

[asdf1000]

21 hours ago, Miska said:

Not sure what you are asking...

 

Ringing is part of the impulse response.

 

Thanks Jussi. I try to ask again with pictures, see attached. But I think I get what you mean from you reply.

 

I know ringing is part of real world IR but ideal transient response should have zero ringing? Analogue signal?

 

So the ringing is "noise" is a sense? And the sample is the "signal" ?

 

So in the attached, the fact that "x" is larger than "y", does that have any relevance or significance to anything?

 

Is the IR with "x" closer to the analogue waveform because it is the highest "signal to noise" ratio?

 

I don't think SNR is a term you DSP gurus apply to impulse responses but hopefully you get what i mean - the ratio between sample peak and the highest ringing peak.

 

 

 

 

 

 

[Miska]

8 minutes ago, asdf1000 said:

I know ringing is part of real world IR but ideal transient response should have zero ringing? Analogue signal?

 

That would be analogue signal with unlimited bandwidth. With digital, you are always limited to bandwidth of fs / 2. Which is the reason for all the filter stuff in first place. If you want to get closer to ideal transient response, you need to stick to DSD domain.

 

8 minutes ago, asdf1000 said:

So the ringing is "noise" is a sense? And the sample is the "signal" ?

 

No... :D

 

8 minutes ago, asdf1000 said:

So in the attached, the fact that "x" is larger than "y", does that have any relevance or significance to anything?

 

Not really. The sum of the impulse response is filter gain, which is most cases is 1.

 

8 minutes ago, asdf1000 said:

I don't think SNR is a term you DSP gurus apply to impulse responses but hopefully you get what i mean - the ratio between sample peak and the highest ringing peak.

 

This is kind of dangerous waters... Because you should really also look into frequency domain version of the impulse response in parallel. Because it is same thing seen from a different angle. Closest to that you can get is https://src.infinitewave.ca (sorry, it is horribly outdated in regards of HQPlayer, latest responses are from 2.9.1 version).

 

[asdf1000]

2 minutes ago, Miska said:

 

That would be analogue signal with unlimited bandwidth. With digital, you are always limited to bandwidth of fs / 2. Which is the reason for all the filter stuff in first place. If you want to get closer to ideal transient response, you need to stick to DSD domain.

 

 

No... :D

 

 

Not really. The sum of the impulse response is filter gain, which is most cases is 1.

 

 

This is kind of dangerous waters... Because you should really also look into frequency domain version of the impulse response in parallel.

 

 

LOL thanks.

 

So in summary - the fact that x is larger than y in the images I showed doesn't actually mean anything, when talking about transient response ?

 

[Pogballistics]

Sinc S doesn't play with Hires 88.2, 96 and so on?

In this case what should I do? I can't enable 48k cos it gives some popping.

 

I really like this filter tbh.

Other preferences are poly-sinc-short-mp-2s and poly-sinc-ext2.

[Solstice380]

My failing memory needs a little help, if anyone would be so kind.  Where is the X64 HQPlayer OS image for NAA located on the Signalyst site?  I just looked and could only find the ones labeled “...embedded...”. Thanks in advance.  

[Holzohr]

3 minutes ago, Solstice380 said:

My failing memory needs a little help, if anyone would be so kind.  Where is the X64 HQPlayer OS image for NAA located on the Signalyst site?  I just looked and could only find the ones labeled “...embedded...”. Thanks in advance.  

 

https://www.signalyst.eu/bins/hqplayerd/images/

[ericuco]

8 minutes ago, Solstice380 said:

My failing memory needs a little help, if anyone would be so kind.  Where is the X64 HQPlayer OS image for NAA located on the Signalyst site?  I just looked and could only find the ones labeled “...embedded...”. Thanks in advance.  

 

https://www.signalyst.eu/bins/naa/? is this what you are looking for? This link is located at the very bottom of the “consumer” page.

[Miska]

4 hours ago, asdf1000 said:

So in summary - the fact that x is larger than y in the images I showed doesn't actually mean anything, when talking about transient response ?

 

It actually depends a lot how you look at it.

 

Step response is ultimately what we are talking about.

 

If you look at it in linear scale like in pictures you quoted:

 

Minimum phase filter (at the bottom), doesn't look so great.

 

However, if you look at it in logarithmic scale, like hearing works, it looks quite different:

 

You can see in the last one (minimum phase), that the non-existence of pre-ringing becomes major compared to amplitude of the post-ringing. Compared to linear phase pre-ringing.

 

Then overall phase response is another factor.

 

This is why the filter choice also so much depends on the source content. Are you listening to progressive rock recorded in small separate studio cabinets and mixed into one piece on a mixing desk. Or are you listening to classical music recorded with just two microphones in a real acoustic space.

 

Former doesn't doesn't have any real space, but has very tight transient response requirements. While latter one is all about the real space and doesn't really contain any sharp transients (or just one or two, compared to hundreds or thousands in the former case).

 

[OzarkMtn]

I am using the Oppo Sonica DAC that employs the ESS chips so I understand that I should employ DirectSDM in Mac version of HQP4D. After changing any settings I get this enclosed message. What do I need to modify to eliminate the message?

My volume settings are:

Vol Min                   Vol Max

-60.0db        ---    -4.0db

 

In DSD source settings 6db gain box is checked.

Thanks,

Sorry about the gargantuan image size.

[Miska]

43 minutes ago, OzarkMtn said:

I am using the Oppo Sonica DAC that employs the ESS chips so I understand that I should employ DirectSDM in Mac version of HQP4D. After changing any settings I get this enclosed message. What do I need to modify to eliminate the message?

My volume settings are:

Vol Min                   Vol Max

-60.0db        ---    -4.0db

 

In DSD source settings 6db gain box is checked.

Thanks,

Sorry about the gargantuan image size.

 

Do you have a specific reasoning for Direct SDM vs non-direct (through DSP pipeline)?

 

When you enable Direct SDM you get fixed -3 dBFS volume for PCM. And if you have set volume maximum to -4 dB, it means that your latest volume setting must be lower than what you are asking now (-3 dBFS), so you get warning to avoid nasty surprises.

 

[jamesg11]

On 2/24/2021 at 8:35 AM, jamesg11 said:

Very pleased with the newly arrived mac mini M1.

 

Does ASDM7EC/48 to 256 using polysinc ext2 with everything so far, except 24-192 which needed either ASDM5EC or easier filter.

Remains to try dsf128 & dsf256 files.

 

Then 2 channel convolution ...

Update with new macmini M1 -

 

now able to do all pcm to 256 with ASDM7EC/48, polysinc ext2, including 24-192  ... care of simply changing buffer from default to 100.

Go figure ...

 

Also dsf128 to 256 fine, leaving only the puzzle of dsf256 files which don’t work (tried 5EC etc). Odd, this final hurdle.

[OzarkMtn]

7 minutes ago, Miska said:

 

Do you have a specific reasoning for Direct SDM vs non-direct (through DSP pipeline)?

 

When you enable Direct SDM you get fixed -3 dBFS volume for PCM. And if you have set volume maximum to -4 dB, it means that your latest volume setting must be lower than what you are asking now (-3 dBFS), so you get warning to avoid nasty surprises.

 

I thought that I read when using the ESS chip DACS to use Direct SDM???

I am upsampling everything (Tidal-Roon-HQP) 5EC to DSD256. So I should not use Direct SDM at all for my current purposes?

[ted_b]

9 minutes ago, OzarkMtn said:

I thought that I read when using the ESS chip DACS to use Direct SDM???

I am upsampling everything (Tidal-Roon-HQP) 5EC to DSD256. So I should not use Direct SDM at all for my current purposes?

If you are using DirectSDM then all DSD stuff is NOT upsampled, just PCM.

[Miska]

31 minutes ago, jamesg11 said:

Also dsf128 to 256 fine, leaving only the puzzle of dsf256 files which don’t work (tried 5EC etc). Odd, this final hurdle.

 

Re-process/re-modulation of DSD256 and higher is more demanding. Sometimes even if you go just DSD256 -> DSD256. Load-wise not so much different from DSD128 -> DSD256. But the amount of data on DSD256 sources is double. So it likely winds down more to the amount of sheer data (memory bandwidth) rather than just the processing itself.

[Miska]

30 minutes ago, OzarkMtn said:

I thought that I read when using the ESS chip DACS to use Direct SDM???

I am upsampling everything (Tidal-Roon-HQP) 5EC to DSD256. So I should not use Direct SDM at all for my current purposes?

 

If you can process everything with ASDM5EC, including DSD sources, it is better. So without Direct SDM.

 

What in most cases you'd want to avoid is sending DSD64 to any DAC. If you can get DSD256 with EC modulators - perfect.

[OzarkMtn]

25 minutes ago, Miska said:

 

If you can process everything with ASDM5EC, including DSD sources, it is better. So without Direct SDM.

 

What in most cases you'd want to avoid is sending DSD64 to any DAC. If you can get DSD256 with EC modulators - perfect.

Thanks, I now understand. Had things interpreted half-a** backwards. Every filter that I have tried has no problem @ DSD256 with the recommended ASDM5EC.

 

Jussi,

What are you doing up so late/early on a Saturday/Sunday?

[jamesg11]

1 hour ago, Miska said:

 

Re-process/re-modulation of DSD256 and higher is more demanding. Sometimes even if you go just DSD256 -> DSD256. Load-wise not so much different from DSD128 -> DSD256. But the amount of data on DSD256 sources is double. So it likely winds down more to the amount of sheer data (memory bandwidth) rather than just the processing itself.

So then, with the macmini M1, I assume down-sampling to 128 from 256 is similarly demanding?

Otherwise, a 2-stage down- then up- could work ... inelegant though.

[Uni]

@Miska

 

I am using loopback on Mac to upsampling spotify.

 

I need to scroll the textbox up and down and choose "audio:default/192000/2" and click run to make it happen.   Is there any simpler method to do it?  Is there any command or shortcut to run "audio:default/192000/2" ?

[Miska]

8 hours ago, jamesg11 said:

So then, with the macmini M1, I assume down-sampling to 128 from 256 is similarly demanding?

Otherwise, a 2-stage down- then up- could work ... inelegant though.

 

Doing DSD128 from DSD256 is somewhat demanding task, but modulator at DSD128 output rate is lighter than at DSD256.

 

[Miska]

6 hours ago, Uni said:

@Miska

 

I am using loopback on Mac to upsampling spotify.

 

I need to scroll the textbox up and down and choose "audio:default/192000/2" and click run to make it happen.   Is there any simpler method to do it?  Is there any command or shortcut to run "audio:default/192000/2" ?

 

Spotify is 44.1k, so you shouldn't be doing 192k input. Because then CoreAudio is doing rate conversion from 44.1k to 192k and that is unnecessary.

 

No, not much simpler, either doing that from Client or from the server main window.

 

[asdf1000]

14 hours ago, Miska said:

Are you listening to progressive rock recorded in small separate studio cabinets and mixed into one piece on a mixing desk.

 

Modern electronic, pop, urban music etc would all fit in this same category right?

 

poly-sinc-short-mp would be ideal for these styles?

 

These genres are mostly mixed entirely on computer these days (they call it "in the box")

 

[Solstice380]

1 hour ago, asdf1000 said:

poly-sinc-short-mp would be ideal for these styles?

I find that one very good for classic rock, pop, prog, etc.  But, I get caught up a little with checking ext2 and Sinc-S and M and they are just different.  Not necessarily better or worse, just different!  

[Pogballistics]

1 minute ago, Solstice380 said:

I find that one very good for classic rock, pop, prog, etc.  But, I get caught up a little with checking ext2 and Sinc-S and M and they are just different.  Not necessarily better or worse, just different!  

What difference did you observe in Sinc S and Sinc M?

 

I haven't tried Sinc M but Sinc S is my favourite filter currently and goes beautifully with my chain.

 

I use Sinc S with ASDM7EC at DSD64 and 48k enabled.

[skipspence]

3 hours ago, Miska said:

 

No, not much simpler, either doing that from Client or from the server main window.

 

@Miska Is there any possible to install HQP Client on a x86 OS? Guess not but asking here just for sure? Got spare Windows x86 tablet but using Win RDP to select source URI just unbearable☹️... I asked @Ales Prochazka if he could add the feature like source URI in HQPControl app but not sure if it's possible for him either?