HQPlayer HQPlayer Convolution Questions

[GMG]

Thanks Miska

indeed Dirac on PC goes up to 192. That’s what I used. 
 

[GMG]

On 7/25/2021 at 9:48 PM, Miska said:

owever, if you are using a tool like REW, it is better to use it's parametric EQ text file export instead of convolution filter. Because that way the whole process becomes much simpler and lighter (things like filter rate become totally irrelevant).

@Miska could you please advise how to perform this export from REW?

Is it done from by:

1. File-->Export-->Export Impulse Response as Text

Or,

2. Tools-->EQ-->Filter tasks-->Save filter coefficients to file

Or

3. Tools-->EQ-->Filter tasks-->Export filter settings as text

 

Also, any special care needs to be taken in save options? (format - MiniDSP/SigmaStudio, Normalise samples to peak, Apply IR window before export, export min phase version of IR)

 

MANY THANKS

 

[Miska]

On 8/25/2021 at 3:55 PM, GMG said:

3. Tools-->EQ-->Filter tasks-->Export filter settings as text

 

I don't have REW at hand right now. But IIRC this should do it and export compatible description file.

 

On 8/25/2021 at 3:55 PM, GMG said:

Also, any special care needs to be taken in save options? (format - MiniDSP/SigmaStudio, Normalise samples to peak, Apply IR window before export, export min phase version of IR)

 

Check what is your highest positive gain EQ and apply that amount of negative gain compensation.

 

 

[GMG]

Thanks @Miska

Just to verify if it will work in my case,

I created my impulse response by a loop with Dirac, so I didn't create any EQ filter in REW. I suspect that if I just load the recorded impulse response and export the EQ filter it will be empty. Am I wrong?

[Miska]

Just now, GMG said:

I created my impulse response by a loop with Dirac, so I didn't create any EQ filter in REW. I suspect that if I just load the recorded impulse response and export the EQ filter it will be empty. Am I wrong?

 

Ahh, OK, then you will want to export the impulse response itself as WAV (mono, preferably 64-bit floating point). Try with IR Window applied. But do not convert to minimum phase as it would spoil any phase correction done by Dirac.

 

[GMG]

thanks, I am using the WAV, but wanted to follow your recommendation to use the parametric version

Will have to check that I exported with IR window. 

Thanks!

[Miska]

2 minutes ago, GMG said:

thanks, I am using the WAV, but wanted to follow your recommendation to use the parametric version

Will have to check that I exported with IR window. 

 

If you don't have correction filter designed in REW, there's no parametric version to export. There's no way to export anything else than a regular convolution filter from a Dirac loop recording.

 

Just check for any applicable filter gain compensation needed. And check the result in HQPlayer's matrix dialog with the Plot-function to see if the result looks correct.

 

[GMG]

OK I will check

Here are the export options,

What do you recommend?

 

[Miska]

13 minutes ago, GMG said:

OK I will check

Here are the export options,

What do you recommend?

 

 

Either 32-bit PCM or 32-bit float are fine. Normalise and Apply IR WIndow.

 

[GMG]

Thanks.

 

Exporting with IR Window significantly changed the resulting IR gain. I used to have negative 2.8dB and now it is positive ~10dB. Does that make sense?

 

Where do I see the IR Gain in HQP Embedded? I had to check this in Windows version...

 

Now that you made me aware of Matrix, is it better to use it for the convolution if I have different gain compensation for the right and left channels? I assume it will benefit the balance of the stereo image. 

If it is preferred to do so in Matrix, How do channels 1&2 map to left and right?

 

And as long as we are on Matrix - will the loudness control work for Roon? or does it only work for HQPlayer direct and volume control by HQPlayer knob?

Will Adaptive Gain work with Roon?

 

(Sorry for getting a bit off topic...) 

[Miska]

3 hours ago, GMG said:

Exporting with IR Window significantly changed the resulting IR gain. I used to have negative 2.8dB and now it is positive ~10dB. Does that make sense?

 

Where do I see the IR Gain in HQP Embedded? I had to check this in Windows version...

 

Do not blindly trust that figure. Instead double-check with the plot function in matrix setup.

 

3 hours ago, GMG said:

Now that you made me aware of Matrix, is it better to use it for the convolution if I have different gain compensation for the right and left channels? I assume it will benefit the balance of the stereo image.

 

No, always use the same compensation for both channels, as the room correction impulses are already compensating any needed differences between the channels.

 

3 hours ago, GMG said:

If it is preferred to do so in Matrix, How do channels 1&2 map to left and right?

 

Yes, channel 1 is left and channel 2 is right. You don't need to change the mappings if you just want to use basic convolution stuff and such.

 

You don't need to run the processing under matrix necessarily, you can also leave it disabled and just use it for plotting the responses.

 

But make sure you don't have both convolution and matrix enabled at the same time. Because they are sort of mutually exclusive and mixing the two will result in confusing setup.

 

3 hours ago, GMG said:

And as long as we are on Matrix - will the loudness control work for Roon? or does it only work for HQPlayer direct and volume control by HQPlayer knob?

 

It works only with HQPlayer volume control, but Roon adjusts HQPlayer's volume when you output to HQPlayer.

 

3 hours ago, GMG said:

Will Adaptive Gain work with Roon?

 

Not at the moment, as there's no metadata coming from Roon. HQPlayer doesn't know where the track boundaries are or any metadata associated with the tracks. Roon would need to send relevant information to HQPlayer for it to work.

 

 

[GMG]

Thanks 

I guess I did something wrong but after a few attempts to plot the 2 different impulse responses (with and without IR window) the web interface is stuck on config tab and I can’t go back to matrix, convolution, main, etc. 

only log and restore

 

i checked the config page and I can’t spot any wrong setting. It’s same as I had it. 
 

?

[Miska]

4 hours ago, GMG said:

Thanks 

I guess I did something wrong but after a few attempts to plot the 2 different impulse responses (with and without IR window) the web interface is stuck on config tab and I can’t go back to matrix, convolution, main, etc. 

only log and restore

 

i checked the config page and I can’t spot any wrong setting. It’s same as I had it. 

 

When you get stuck to config page, HQPlayer usually cannot access your DAC. Possibly some other process has taken control of it (Roon?). Reboot your machine and maybe it will start working again.

 

[GMG]

here are the 2 plots, with (green) and without IR Window

Can you give brief explanation to what IR Window is doing and why it might be better to use with

 

[Miska]

4 hours ago, GMG said:

here are the 2 plots, with (green) and without IR Window

Can you give brief explanation to what IR Window is doing and why it might be better to use with

 

 

You can check out the REW manual. But it seems to remove the empty background noise stuff before and after the actual impulse response.

 

Green one seems to have over +10 dB gain while the red one is normalized to 0 dBFS.

 

[GMG]

On 8/26/2021 at 9:32 PM, Miska said:

 

If you don't have correction filter designed in REW, there's no parametric version to export. There's no way to export anything else than a regular convolution filter from a Dirac loop recording.

What if I use REW to design a filter on the recorded Dirac loop using a flat target curve on 0db line then calculate the eq filter and inverse all the calculated gain coefficients?

[Bob Stern]

On 8/26/2021 at 1:09 PM, GMG said:

Where do I see the IR Gain in HQP Embedded? I had to check this in Windows version...

On 8/26/2021 at 4:56 PM, Miska said:

Do not blindly trust that figure. Instead double-check with the plot function in matrix setup.

Is a bandpass filter applied before computing the IR gain?  I expect it would be more meaningful if bandpass limited to the frequency range on which we base our loudness perception, but broad enough to avoid being overly skewed by an intentional narrow peak or dip in the EQ.  For example, roughly 400 Hz to 4000 Hz.  (I just dreamed up those numbers without consulting any research.)

[Miska]

14 hours ago, GMG said:

What if I use REW to design a filter on the recorded Dirac loop using a flat target curve on 0db line then calculate the eq filter and inverse all the calculated gain coefficients?

 

No, don't do that, it doesn't make much sense. Then you could as well just design the filters straight with REW without Dirac and have a better result.

 

[Miska]

9 hours ago, Bob Stern said:

Is a bandpass filter applied before computing the IR gain?  I expect it would be more meaningful if bandpass limited to the frequency range on which we base our loudness perception, but broad enough to avoid being overly skewed by an intentional narrow peak or dip in the EQ.  For example, roughly 400 Hz to 4000 Hz.  (I just dreamed up those numbers without consulting any research.)

 

IR gain thing doesn't exist in Embedded. You are better off just using the plot function there.

 

Sorry, but I don't understand how rest of the thing is related to the discussion?

 

[GMG]

1 hour ago, Miska said:

Then you could as well just design the filters straight with REW without Dirac and have a better result.

Thanks Miska,

REW has many options and parameters and I don’t have the knowledge to properly design the filter (and don’t have the time to learn it right now).
When I tried a basic filter with REW in the past the result to my ears was not as good as Dirac. 
Dirac is very easy to use and now that I figured how to easily loop the filter back to REW and capture the impulse response I can use it also with network audio (Roon, NAA, etc.)

 

I assume that the benefit of using the parametric EQ from a not so well designed filter I will make with REW Vs. a WAV impulse response from a looped back well designed filter from Dirac will be negligible and maybe not worth the trouble. 
What do you think?

 

In any case, thanks you very much for your help, answers and shared knowledge. 

[jvvita]

Hello @Miska,

 

strange thing just happened today. Using Acourate generated filters, I loaded them into matrix pipeline (I wanted to use the profile feature to quickly switch and compare filters). But the pipeline introduce massive delay to my right channel (only) - about 1s-2s delay. The same filters work just fine using matrix - convolution option. Am I doing anything wrong?

 

always playing through Roon, with no dip in Roon. I’m HQPlayer, using poly ext2 - 7EC - dsd256.

[Miska]

6 hours ago, jvvita said:

strange thing just happened today. Using Acourate generated filters, I loaded them into matrix pipeline (I wanted to use the profile feature to quickly switch and compare filters). But the pipeline introduce massive delay to my right channel (only) - about 1s-2s delay. The same filters work just fine using matrix - convolution option. Am I doing anything wrong?

 

always playing through Roon, with no dip in Roon. I’m HQPlayer, using poly ext2 - 7EC - dsd256.

 

Check that you don't accidentally have more than one filter on one of the pipelines. And that you don't accidentally have both convolution and pipeline processing active simultaneously.

 

[jvvita]

I forgot to leave feedback. I just cleared all the filter and reinserted them from scratch. All is fine now. Thanks for the quick reply.

[jvvita]

hello @Miska,

 

getting back to the gain questions when using convolution. Acourate is saying that my filters result in aprox. -4db of gain. So, when I load then up in HQPlayer, I listen to music 4db lower in volume than without the filters. Can I compensate in the Matrix pipeline "adding gain" (lets say, about 3db)? the master volume control would be kept in -3db, as per your recommendation? Or, if I do not compensate, would it be possible to keep the volume control at 0db, given the filters 4db attenutaion?

[Miska]

2 hours ago, jvvita said:

hello @Miska,

 

getting back to the gain questions when using convolution. Acourate is saying that my filters result in aprox. -4db of gain. So, when I load then up in HQPlayer, I listen to music 4db lower in volume than without the filters. Can I compensate in the Matrix pipeline "adding gain" (lets say, about 3db)? the master volume control would be kept in -3db, as per your recommendation? Or, if I do not compensate, would it be possible to keep the volume control at 0db, given the filters 4db attenutaion?

 

Yes, assuming that there are no boosts at any frequency at all. The thing with many correction filters is that if there's any narrow frequency boost, you get perceived reduction in overall volume, but you still have same full band peak gain.

 

If you need to do any compensation, you should test it by playing full 0 dBFS frequency sweep and checking that Limited counter doesn't increase.

 

But I wouldn't bother over something as small as 4 dB.