DRC: comparing linear phase and minimum phase correction on Binaural Room Impulse Responses

[dasdoing]

I took BRIRs from here and corrected them: https://github.com/ShanonPearce/ASH-IR-Dataset/tree/master/BRIRs/R02
software used was: http://drc-fir.sourceforge.net/; the software give me an option to create linear phase and minimum phase correction filters with matchig frequency response

these are the virtual left speakers overlaid (the minimal diference seams to be because I can't align them 100% with REWs limitation to 0,1dB steps)

but there is a huge diference in the phase response:

better visibale in the step responses:

and the Group Delay:

the following folder contains various files processed with the corrected BRIRs.
the short ones are intended to compare the effect of the group delay diference.
the longer ones to compare overall effect of the phase correction

!!BRIRs are intended for headphones only!!!

https://drive.google.com/drive/folders/1-7aAGMWp9ahOFdImP2p5HExa-dlX9yaU

 

[Spyart]

Could you share please your edited Brir to test it with my own music or explain a bit how to edit available ones? That is very interesting topic for me (looking for nice mixing portable environment). Cheers

 

[dasdoing]

Could you share please your edited Brir to test it with my own music or explain a bit how to edit available ones? That is very interesting topic for me (looking for nice mixing portable environment). Cheers

actualy those BRIRs are part of a project of mine to create a "perfect mixing room" for headphones because all the comercial solutions I tried sound crap. I started editing BRIRs in 2019, and this is my 4th try.
it is not ready yet though since I have problems convoluting the originals with the corrections. these files were created with 2 convolvers in series for every of the 4 channels.
the convolution is problematic because of the time delay of the crossfeed channels.
I will post them once they are ready

 

[Spyart]

actualy those BRIRs are part of a project of mine to create a "perfect mixing room" for headphones because all the comercial solutions I tried sound crap. I started editing BRIRs in 2019, and this is my 4th try.
it is not ready yet though since I have problems convoluting the originals with the corrections. these files were created with 2 convolvers in series for every of the 4 channels.
the convolution is problematic because of the time delay of the crossfeed channels.
I will post them once they are ready

Same to me! My last hope was dearvr monitor but it completely crushed FR. Tried dsoniq realphones - crossfeed still narrowed stereo base (which is completely untrue for 0.8m triangle Dirac corrected near field in treated room). Still ambience there has dry wet amount but it's pretty static and doesn't help with locating sound (backward forward or up and down). An issue with transient translating in headphones still there. The only thing that helped a bit with setting up the compressors is expander in monitoring chain, otherwise drums are always behind sustained sounds and this is not true as well.
There is Slate Vsx on the market but charging 500 dollars for 50 bucks phones and impulses are unacceptable for me)
Shannon's BRIRs pretty much coloring a sound and take it to the reverb too much but I hope some day you would share your work with no coloration. Do for me the closest result to my near field in HD 650 (kiss moded, damped and no rear foam) is TB Morphit (slightly corrected studio target for HD 600!) and fab filter expander. Maybe TB Isone as well but still need some tweaking...

 

[Spyart]

Ok I tested your example on corrected headphones (eq only) and for me Pneuma was been a pretty much distant and with no lo end and missing of exact location of sounds. I found the first room from Shannon's collection is pretty much ok (my own room has around 0.26 RT from 100hz to 20khz) but still colorizing as hell. So I applied kind of inverse filter (like tilt to attenuate low end) and this thing was closest in terms of stereo location of sounds. Highly recommend such artist as Alva Noto (if you're into electronic music for sure) and his album 2018. Those short sounds are very demanding on excessive reverb and loss of location. Cheers

 

[dasdoing]

Ok I tested your example on corrected headphones (eq only) and for me Pneuma was been a pretty much distant and with no lo end and missing of exact location of sounds. I found the first room from Shannon's collection is pretty much ok (my own room has around 0.26 RT from 100hz to 20khz) but still colorizing as hell. So I applied kind of inverse filter (like tilt to attenuate low end) and this thing was closest in terms of stereo location of sounds. Highly recommend such artist as Alva Noto (if you're into electronic music for sure) and his album 2018. Those short sounds are very demanding on excessive reverb and loss of location. Cheers

what curve you corrected your phones to?

 

[Spyart]

what curve you corrected your phones to?

ok now it's something like a mix between diffuse field (because of Shannon project, this requires it) and bass half of harman then. Also my personal hrtf seems to need rejecting around 500hz but I believe it's personal thing.
So I'm relistening your files again and should said Pneuma has a beutiful sound with nice drums. Still distant and I can't identify each reverb on each instruments which is imo pretty important for mixing. Thus R02 and R04 are too reverberant for me. For some tracks (Dead Can Dance - Yulunga, where shakers nicely separated like L R in monitors) they are messing up the rhytms of any high frequency part. But overall dynamic is pretty fast and precise so I stick now with the R1 which imo need additionally pushing down around 170hz (or it's because of HD650, but hey, already corrected before).
Forgot to add - I tried 60 degrees files only because even 45 ones are too narrow for me. (maybe personal thing too but mixing defacto standard is 60)

 

[dasdoing]

ok now it's something like a mix between diffuse field (because of Shannon project, this requires it) and bass half of harman then. Also my personal hrtf seems to need rejecting around 500hz but I believe it's personal thing.
So I'm relistening your files again and should said Pneuma has a beutiful sound with nice drums. Still distant and I can't identify each reverb on each instruments which is imo pretty important for mixing. Thus R02 and R04 are too reverberant for me. For some tracks (Dead Can Dance - Yulunga, where shakers nicely separated like L R in monitors) they are messing up the rhytms of any high frequency part. But overall dynamic is pretty fast and precise so I stick now with the R1 which imo need additionally pushing down around 170hz (or it's because of HD650, but hey, already corrected before).
Forgot to add - I tried 60 degrees files only because even 45 ones are too narrow for me. (maybe personal thing too but mixing defacto standard is 60)

about the FR, it measures +- 1,5dB with var smoothing with all channels combined. a stereo pink noise is basicly perfect in comparison to no filter. a pink noise coming from one virtual speaker is not, though...can't have both with BRIRs. I still have to compare this behaviour to the real world.
about room 1 from the ash-ir database, it is realy the most dead room: https://github.com/ShanonPearce/ASH-IR-Dataset/wiki/Room-Information
but it is far from a real control room like R02 and R04 are. https://github.com/ShanonPearce/ASH-IR-Dataset/wiki/Room-Photos. but I might give it a try to hear it corrected

 

[dasdoing]

also I am studying a method to correct direct sound only over 1kHz on my real speakers. this might be intresting for the BRIRs, too...I still have so much to learn

 

[Spyart]

I think the key in too reverberant performance of r02 and r04 is because Hats was set 2 or 2.5 meters away. That's too much for near field, so that's way I can receive a way more reflections into my ears, even multiply by too. So the r01 is good compromise maybe. A bit deadly in sense of pushing air as on my monitors but FR is not damaged by excessive reflections. Today I mixed some rock and surprisingly I could hear 1db difference on eq, compression an volume veeery precisely than without brir. Amazing experience, I can build overall mix just for one hour!
Hope someone would steal IRs from slate vsx, yeah, it's could be illegal thing but I don't need closed back headphones... And they marketed as prepared for binaural. Oh man, what a nonsense. Just every hats or head should be calibrated by diffuse field response to achieve good result. But who knows, maybe closed back are good in LF while using BRIRs?

 

[Chromatischism]

I took BRIRs from here and corrected them: https://github.com/ShanonPearce/ASH-IR-Dataset/tree/master/BRIRs/R02
software used was: http://drc-fir.sourceforge.net/; the software give me an option to create linear phase and minimum phase correction filters with matchig frequency response

these are the virtual left speakers overlaid (the minimal diference seams to be because I can't align them 100% with REWs limitation to 0,1dB steps)

View attachment 101133View attachment 101134

but there is a huge diference in the phase response:

View attachment 101135

better visibale in the step responses:

View attachment 101136

and the Group Delay:

View attachment 101137

the following folder contains various files processed with the corrected BRIRs.
the short ones are intended to compare the effect of the group delay diference.
the longer ones to compare overall effect of the phase correction

!!BRIRs are intended for headphones only!!!

https://drive.google.com/drive/folders/1-7aAGMWp9ahOFdImP2p5HExa-dlX9yaU

Noob question. Is there a difference in distortion between them?

 

[MRC01]

... these are the virtual left speakers overlaid (the minimal diference seams to be because I can't align them 100% with REWs limitation to 0,1dB steps)
...

REW tip: The arrow buttons have 0.1 dB resolution. But you can type in any value, not limited to that resolution. Whatever value you type in, the text box will show it rounded to the nearest 0.1 dB. But the curve will actually move by the exact amount you type in.

 

[Propheticus]

Pardon my ignorance, but I'm not quite sure I get what you're trying to achieve here.
Isn't phase change one the effects that happen when sound reverberates in a typical room due to (out of phase) reflections from different walls and surfaces? Would you not kill/remove this spatial queues when you phase 'correct'?
Feels like you're taking an approximation of how a room behaves and then apply (partial) room EQ correction to again remove these characteristics.

Or are these phase issues you're correcting introduced by something else in the measurement path?

 

[dasdoing]

Pardon my ignorance, but I'm not quite sure I get what you're trying to achieve here.
Isn't phase change one the effects that happen when sound reverberates in a typical room due to (out of phase) reflections from different walls and surfaces? Would you not kill/remove this spatial queues when you phase 'correct'?
Feels like you're taking an approximation of how a room behaves and then apply (partial) room EQ correction to again remove these characteristics.

Or are these phase issues you're correcting introduced by something else in the measurement path?

it's not that FIR correction removes any reflections. it just aligns them where they are dominant; meaning in the region below 1000Hz-ish where the reflected sound dominates our perception. that is why the bass region is shifted, it has to travel further in realation to the direct sound we hear above 1000Hz. the phase correction "tells" the bass region to play before the rest, in order to compensate for that.
if you phase correct reflection in the region where direct sound dominates you would shift the direct sound forward, it would the be too early (pre-ringing)

 

[WeirdFM]

First time I heard about convolution was when I was building a linux box to use as digital (FIR) crossover nearly 20 years ago, and convolution kinda blew me away at that time, but I guess I just forgot about it after and didn't even think about the possibility of using it for simulating room response on headphones.

So after listening to some of the samples I downloaded equalizer APO and the wavs, and dang! This stuff works pretty well! One question though, after tinkering with APO myself and unable to make it sound very good - Either the effect was lacking in strength or I made everything too echoey - I found there were preset configuration files in the download and using those made the sound a lot better. Curious, I looked inside one:

BRIR convolution configuration for Equalizer APO
Room: Room Calypso, TU Berlin
Position: 1
Speaker setup: 2.0 Stereo
Dated: 11.07.2020

Copy: L_INPUT_L_EAR=L L_INPUT_R_EAR=L R_INPUT_L_EAR=R R_INPUT_R_EAR=R
Channel: L_INPUT_L_EAR L_INPUT_R_EAR
Convolution: ..\..\..\BRIRs\R01\BRIR_R01_P1_E0_A-30.wav
Channel: R_INPUT_L_EAR R_INPUT_R_EAR
Convolution: ..\..\..\BRIRs\R01\BRIR_R01_P1_E0_A30.wav
Copy: L=L_INPUT_L_EAR+R_INPUT_L_EAR R=L_INPUT_R_EAR+R_INPUT_R_EAR

I don't really understand why there are 4 channels used for stereo. It looks like 2 binaural HATS recordings were made and the full stereo filter from each of them is routed to one ear. And what does the A(Number) mean? Is it degrees, or maybe offset from the center listening position?

BTW, I quickly discarded the included correction filters for HD600's as well. They weren't doing a great job. Thankfully EQ:ing tonally neutral response was made very easy withing equalizer APO since you can see the effect on FR for the full stack of filters at the bottom.

@dasdoing I did take a listen to the samples you posted in your other thread about the subject and I could tell them apart easily enough I'd wager I'd get 100% right in ABX (I didn't actually bother so who really knows, right?). Samples in this one gave me a hard time though, I'm not confident at all. I mainly listened to the Tool one.