[bogi]

The pre-ringing is not masked by hearing, while hearing has a masking period shortly after a transient.

I didn't do this kind of experiments before and I am surprised how that psychoacoustic masking of post-ringing works! In fact I hear only the pre-ringing with long.flac, including clear detection of 3kHz tone (I checked that against 3kHz played through this online tool). I hear no such thing after the pop. Therefore I opened long.flac in Audacity to look how pre- and post-ringing looks. Both files look similar, just the ringing in long.flac is longer. How much is the filter used in long.flac longer? From the picture it looks that maybe only about 3x or 4x, but the picture resolution may be too low.

The surprise for me is how diferently these files sound to me. Only from signal level / time graph I would not expect such a difference. How short should be pre-ringing in order not to be directly detectable, like in short.flac?

 

[bogi]

really good song to test for preringing if you play around with fir vs iir... it became a mess with a highpassfilter at bass frequencys and FIR

Even without 'highpassfilter at bass frequencys and FIR', when I compared long filters with few short ones, I preffered much cleaner sound of short filters with that recording (thanks for reminder, I like R.E.M. much). With my D300 DAC in direct DSD mode (and the rest of my chain) I liked most poly-sinc-xtr-short-mp with that track.

I am playing youtube videos through PotPlayer WASAPI out, through VAC to HQPlayer Desktop and this way even youtube content sounds surprisingly well.

 

[jlaako]

How much is the filter used in long.flac longer? From the picture it looks that maybe only about 3x or 4x, but the picture resolution may be too low.

Quite a bit, I don't remember the exact figures right now. This linear amplitude scale is not really representative, but in Audacity you switch the vertical scale to logarithmic (dB) and that gives you better visual idea of what you hear.

How short should be pre-ringing in order not to be directly detectable, like in short.flac?

It needs to be reasonably short not to have any "flavor", like medium length. Certainly nothing NOS-like. We are talking about few milliseconds length, few hundreds of taps at RedBook.

When filter becomes longer, it begins to gain more and more recognizable "zzap" tone for linear phase, and more kind of "blip" tone for minimum phase. This because minimum phase has all the ringing moved to the tail side. While the transient itself should be "toneless". Minimum phase better preserves the sharp attack nature of the transient.

I made couple of samples with medium length that nicely demonstrate this flavor:

Linear phase:
https://www.sonarnerd.net/tmp/medium-lp.flac

Minimum phase:
https://www.sonarnerd.net/tmp/medium-mp.flac

These would be 1/8th in length in time at RedBook rates, so this is like 8x accentuated. With short filters even this 8x emphasized case sounds clean.

So one really needs to put quite a bit effort and focus into making a filter that is short enough to keep time domain sharp, not add flavor, and still have excellent performance as a filter. Since when increasing length, it quite quickly begins to introduce such effects.

 

[Ghoostknight]

Even without 'highpassfilter at bass frequencys and FIR', when I compared long filters with few short ones, I preffered much cleaner sound of short filters with that recording (thanks for reminder, I like R.E.M. much). With my D300 DAC in direct DSD mode (and the rest of my chain) I liked most poly-sinc-xtr-short-mp with that track.

hmmm i have to test this, i think sincM is one of the longer filters right?

IIR (minimum phase) definitely sounds smoother, the thing is just that FIR with somewhat minimal ("non-audible-ish") preringing sounds more correct, if you hear a difference between absolute phase inverted or not you also hear the linear phase property of FIR (linear phase)

the REM song is not "your usual audiophile song", i just "found" it while testing around with some kind of subsonic filter, i just wish there would be a way to implement a subsonic filter without getting huge amounts of preringing.... the REM song was unlistenable ... i would say its a somewhat transient rich song with higher dynamic range, which makes the preringing more obvious... specially with these fast following transients after each other... it becomes a smeared mess instead of clean transients...

the only thing that was acceptable was a -3db high shelf at 8hz... its still worse than non if i focus on transients but the effect is slight i would say... the highpass was like 10 times more audible

hmm i just got an idea.. i could actually test this with EasyEffects, but not with hqplayer i think... doing my EQ as FIR as usual... beside the subsonic filter which runs in a seperate eq in IIR mode... this could be actually the compromise im looking for...

 

[Ghoostknight]

hmm i just got an idea.. i could actually test this with EasyEffects, but not with hqplayer i think... doing my EQ as FIR as usual... beside the subsonic filter which runs in a seperate eq in IIR mode... this could be actually the compromise im looking for...

hmm i just tested it, this works surprisingly well...well, it seems to be the only option if you wanna filter subsonics and use FIR
is there any way todo this in hqplayer? probably not or?

btw if someone is curious... actually my studio monitors already have a subsonic... not sure at what frequency but i would guess sub 20 ... i can still clearly hear a subsonic at 14hz ... the speaker sound "thinner" and there is less "heft" to bass notes, not sure how much is distortion and how much subsonics i perceive tho.. these monitors just go down to 40hz... a subsonic at 14hz still clearly has influence on the sounds...
my guess is its mostly distortion, the sound also clears a bit up with subsonic.. i guess it just takes more of the intensive "load" from the speaker away...

 

[bogi]

hmmm i have to test this, i think sincM is one of the longer filters right?

Yes ...

It is very easy to switch filter to poly-sinc-xtr-short-mp and to compare the two ...

I wrote in the previous post "this way even youtube content sounds surprisingly well". That was with poly-sinc-xtr-short-mp. No subsonic FIR needed. Of course, the result depends on the whole chain. With direct DSD path no more interpolation filter appears in my signal path.

it seems to be the only option if you wanna filter subsonics and use FIR
is there any way todo this in hqplayer? probably not or?

You can convert any IIR EQ filter to FIR in HQPlayer with possibility to choose it to be minumum or linear phase.

 

[jlaako]

the only thing that was acceptable was a -3db high shelf at 8hz... its still worse than non if i focus on transients but the effect is slight i would say... the highpass was like 10 times more audible

hmm i just got an idea.. i could actually test this with EasyEffects, but not with hqplayer i think... doing my EQ as FIR as usual... beside the subsonic filter which runs in a seperate eq in IIR mode... this could be actually the compromise im looking for...

How about just a normal 2nd order parametric high-pass in HQPlayer matrix? With corner somewhere between 1 and 5 Hz? Or what kind of thing do you have in mind? Not a DC-block but something else?

My Accuphase RIAA for example has second order filter with 25 Hz corner as subsonic (switchable).

 

[jlaako]

You can convert any IIR EQ filter to FIR in HQPlayer with possibility to choose it to be minumum or linear phase.

Yeah, one can run any combination of multiple convolution filters and IIR PEQ's and choose of those IIR's are converted into FIR. So for example combination of convolution FIR plus IIR EQ is also possible.

 

[Ghoostknight]

How about just a normal 2nd order parametric high-pass in HQPlayer matrix? With corner somewhere between 1 and 5 Hz? Or what kind of thing do you have in mind? Not a DC-block but something else?

i use a file like this:

Equaliser: Generic
CLP R Speaker+Sub 70hz 40% Jan 14
Preamp: -3.0 dB
Filter 1: OFF HP Fc 14 Hz
Filter 2: ON LS Fc 10 Hz Gain -3.00 dB Q 0.800

im actually unsure what order the "HP" (highpass) is in this eq file and if there are other options but with easyeffect even with a 6db order highpass i hear the pringing with FIR

would reducing the corner freq under 5hz help reduce preringing? the lowest i went was 8hz and yes basicly a dc block filter (unfortunaly easyeffects is locked to minimal 10hz...)

Yeah, one can run any combination of multiple convolution filters and IIR PEQ's and choose of those IIR's are converted into FIR. So for example combination of convolution FIR plus IIR EQ is also possible.

but i can just choose one eq per matrix right? so either the whole matrix runs in FIR or all in IIR

i guess i could make an convolution file with the subsonic filter with REW and load it in aditionally to my matrix setup, i will test this thanks! i might share the convolution file here then, so other can test it if they want

You can convert any IIR EQ filter to FIR in HQPlayer with possibility to choose it to be minumum or linear phase.

this i know but the question was how to run part of EQ in FIR while other part (the subsonic) in IIR

 

[Ghoostknight]

File is attached, unfortunaly REW also just allows minimal 10Hz High pass.... it will filter quite a bit of low end, dont be surprised

Do i set the preamp on the matrix or convolution pipeline?

 

[jlaako]

im actually unsure what order the "HP" (highpass) is in this eq file and if there are other options

It is second order. There's also first order, but for this purpose I would suggest second order.

but i can just choose one eq per matrix right? so either the whole matrix runs in FIR or all in IIR

No, you can have any number of IIR EQ's, either inline, or text files. Any number of convolution filters. Or any combination of the two. You can have both convolution FIR and parametric IIR at the same time.

Every time you click Browse and select a file, one more is appended to the Process entry. All processing items are comma separated on the line.

I also recommend to use the Plot-button to check the results if they match what you are expecting.

i guess i could make an convolution file with the subsonic filter with REW and load it in aditionally to my matrix setup, i will test this thanks! i might share the convolution file here then, so other can test it if they want

If you use REW, it is better to export the .txt file which is completely sampling rate agnostic, and you can choose in HQPlayer if you want to convert it to FIR (convolution filter), with original minimum phase or convert to linear phase, or keep it as IIR. REW's convolution filter conversion just generates minimum-phase conversion of the IIR.

Only restriction is that if you ask HQPlayer to convert IIR to FIR, all PEQ's are converted same way.

this i know but the question was how to run part of EQ in FIR while other part (the subsonic) in IIR

You can have something like "iir:type=hp;f=5,MyFancyDRC.wav". This will run the IIR as IIR (unless you ask for something else) and the FIR WAV as such.

But yes, if you have all PEQ defined for HQPlayer, then it is either IIR or FIR conversion (minimum or linear phase).

 

[jlaako]

File is attached, unfortunaly REW also just allows minimal 10Hz High pass.... it will filter quite a bit of low end, dont be surprised

HQPlayer doesn't have such restriction.

If you use matrix, put everything there and disable the simple convolution engine from the convolution dialog. Otherwise it becomes confusing and error prone.

 

[Ghoostknight]

It is second order. There's also first order, but for this purpose I would suggest second order.



No, you can have any number of IIR EQ's, either inline, or text files. Any number of convolution filters. Or any combination of the two. You can have both convolution FIR and parametric IIR at the same time.

Every time you click Browse and select a file, one more is appended to the Process entry. All processing items are comma separated on the line.

I also recommend to use the Plot-button to check the results if they match what you are expecting.



If you use REW, it is better to export the .txt file which is completely sampling rate agnostic, and you can choose in HQPlayer if you want to convert it to FIR (convolution filter), with original minimum phase or convert to linear phase, or keep it as IIR. REW's convolution filter conversion just generates minimum-phase conversion of the IIR.

Only restriction is that if you ask HQPlayer to convert IIR to FIR, all PEQ's are converted same way.



You can have something like "iir:type=hp;f=5,MyFancyDRC.wav". This will run the IIR as IIR (unless you ask for something else) and the FIR WAV as such.

But yes, if you have all PEQ defined for HQPlayer, then it is either IIR or FIR conversion (minimum or linear phase).

Thanks! so if i understand correct im doing it the right way for my "setup"
all PEQ in hqplayer matrix converting to FIR (so i can change the majority of eq's still easily and they run in FIR, REW would be much more complicated here with rePhase...)
and one IIR convolution file with the subsonic as IIR


---

this setup is much better than the highpass in FIR... "i think" i can somewhat hear the phaseshift but it honestly doesnt matter much because other sideeffects of FIR are gone that made it far worse imo

this is also probably "the" test if you wanna expierence preringing... setting up a subsonic at 10Hz and changing between FIR and IIR processing, tho dont get confused because of the lost lowend... listen for transients

i also have this filter:

Filter 10: ON PK Fc 130 Hz Gain -12.00 dB Q 10.000

i never noticed this one in terms of preringing... the subsonic highpass as FIR is far worse

 

[Ghoostknight]

BTW @jlaako thanks for the great software it seemed kinda restricted (well every player kinda does for me, pipewire and windows just are more convenient, but also not bitperfect soo...) at first but it kinda offers more the deeper i dive into it ... and its pretty much monopoly software if we go into the higher end filters etc (i knew there had to be some audiophile devs out there )

also thanks for providing such a restriction free trial version... other companys would have give us max 10 minutes with restricted features... im gonna buy it if finances allow it

i also have 2 Feature requests

1. Did you already thought about VST support? (or atleast some linux supported format?) that could open up some interesting features ... like different crossfeed plugins, oldschool equalizer (if someone wishes so), compressors (which i sometimes like mostly for movies...) or SonarWorks perhaps? tho i have to admit most of the plugins would be kinda gimmicky and you probably wanna avoid running audio trough a couple of badly made plugins with all kind of resampling etc going on

2. Please make it a proper "player" .. like with album view, library managment, playlist managment that is less cumbersome and stuff like this, it feels kinda stupid to just search the files and drop them into hqplayer each time i wanna play them, it might be me, im kinda lazy

now i just need to figure out how to route my pipewire audio into hqplayer but i cant get this damn alsa loopback device working.... i already tried but it just wont transmit audio, it even shows up under devices
i guess this would also make my feature request 2 obsolete since i could use each player i want and let the audio route trough hqplayer...

 

[Ghoostknight]

If you use matrix, put everything there and disable the simple convolution engine from the convolution dialog. Otherwise it becomes confusing and error prone.

does that mean better not use convolution and matrix at the same time?