Recording Impulse Responses for Speaker Virtualization
Nov 20, 2020 at 9:42 AM Post #451 of 1,816
Nov 21, 2020 at 7:40 AM Post #452 of 1,816
I finally, finally found the missing link from the measurement chain! The 3.5mm female stereo to 2x male mono Y-splitter cable: https://www.amazon.com/gp/product/B0785VKZW4. It's required to connect binaural mics with a 3.5 mm stereo male connector to the RODE VXLR+ adapters but for some reason this type of cable is extremely rare. I bought it from Amazon just to test it out for you and can confirm that it works as intended. This particular cable even has shielding so there won't be any concerns about the low mic signal picking up electrical interference. It's also available in other Amazon stores, not just the north american .com one.
 
Nov 21, 2020 at 11:09 AM Post #454 of 1,816
How is your algorithm calculate perceived frequency response? Is it just a sum of direct sound and all reflections? Researches shows that calculation of perceived FR can be tricky.
There is no perceived frequency response in Impulcifer, only what the binaural microphones capture. That is of course affected by the listener's own ears so the end resulryis very neutral. Research shows people prefer neutral.

Room correction could be benefit from perceived frequency response since that might make it possible to fix speakers and whatnot but at the moment I have no tools for estimating such a thing.
 
Nov 30, 2020 at 12:28 PM Post #455 of 1,816
Hello guys.

Pretty interesting project. I just tried to play with it.
I have new Corsair HS70 Pro Wireless headphones and I build my own In-ear microphones with WM-61A.
Can I use the resulted hrir.wav simply in EQ APO as Convolution with impulse response? I tried it and I always got result which is very left sided (and even each side have different sound). I tried --no_room_correction --no_equalization --no_headphone_compensation but the result is almost the same.

Thanks!
 
Nov 30, 2020 at 12:47 PM Post #456 of 1,816
Hello guys.

Pretty interesting project. I just tried to play with it.
I have new Corsair HS70 Pro Wireless headphones and I build my own In-ear microphones with WM-61A.
Can I use the resulted hrir.wav simply in EQ APO as Convolution with impulse response? I tried it and I always got result which is very left sided (and even each side have different sound). I tried --no_room_correction --no_equalization --no_headphone_compensation but the result is almost the same.

Thanks!
Welcome aboard!

You better install HeSuVi and use hesuvi.wav. There is a way to use the hrir.wav with EqualizerApo without HeSuVi but I haven never tried that so cannot tell you which additional steps need to be taken.
 
Nov 30, 2020 at 2:30 PM Post #457 of 1,816
For EqualizerApo I use following code for Brirs that use the common channel configuration. For the HESUVI files you have to change it as it uses: L-l L-r LS-l LS-r LB-l Lb-r C-l R-r R-l RS-r RS-l RB-r RB-l C-r .

#Common preamp
Preamp: -16 dB
L=Left ; R=Right C=Center; SUB=LFE; SL=Left Surround; RL=Rigth Surround, RL=Rear Left; RR=Rear Right
#Create virtual speaker channels
Copy: L0=L R1=L L1=R R0=R C1=C C2=C SUB0=SUB SUB1=SUB SL0=SL SR1=SL SL1=SR SR0=SR RL0=RL RR1=RL RL1=RR RR0=RR
#Mute Input Channels
Copy: L=0 R=0 C=0 SUB=0 RL=0 RR=0 SL=0 SR=0
#virtual channels that are filtered
Channel: L0 R1 L1 R0 SL0 SR1 RL0 RR1 C1 C2 SUB0 SUB1 SL0 SR1 SL1 SR0 RL0 RR1 RL1 RR0
#Folder of Convolution files complete folder name or if files in config "\...."
Convolution: convolution\BRIRs\CR7_KU_ROTM_(L_R_C_LFE_LS_RS_LB_RB).wav
# Copy the virtual speaker channels to Left & Right Headphone Channel
Copy: L=L0+L1+C1+SL0+SL1+RL0+RL1+SUB0 R=R1+R0+C2+SR1+SR0+RR1+RR0+SUB1
Channel: L R
Convolution: convolution\BRIRs\KH_Filter\KU100_CDFC_minphase.wav



@Benik 3 If you build your own mics I would suggest to use the Primo em258 which is the better capsule.
 
Last edited:
Nov 30, 2020 at 3:19 PM Post #458 of 1,816
Thanks for reply guys!
I will test it :)
I compared the DIY microphones with Omnitronic MM-2USB and it looks pretty OK, but thanks for the tip on the EM258!
1606767534053.png


EDIT: Both options works, but the sound in hesuvi is different from just APO (suggested by musicreo). In Hesuvi it's not so harsh.
Anyway I found that I have swaped back speakers :D
 
Last edited:
Dec 2, 2020 at 10:14 PM Post #459 of 1,816
Room measurement guide is here! It was a long time in the making but I guess better later than never. Please do give feedback if you have any ideas about improvement or something is confusing.

https://github.com/jaakkopasanen/Impulcifer/wiki/Measurements#room-correction

Is impulcifer good to remove the reverb? I tried many hrir and overall they are with much reverb

I'm trying to achieve good spacial hrir like the dolby home teather dh2 in hesuvi but with almost no reverb, but i'm new to these kind of stuff so i need a hand to understand
 
Dec 3, 2020 at 4:32 AM Post #460 of 1,816
Is impulcifer good to remove the reverb? I tried many hrir and overall they are with much reverb

I'm trying to achieve good spacial hrir like the dolby home teather dh2 in hesuvi but with almost no reverb, but i'm new to these kind of stuff so i need a hand to understand
Impulcifer has tools to shorten the reverbant tail but not from existing BRIRs. You should do your own measurements with binaural mics. The results will be so much better.
 
Dec 3, 2020 at 4:53 AM Post #461 of 1,816
I tried to play with it. The sound is really so wide, but I have problem, that it's too wide :D
When I tried to play e.g. Mass Effect Andromeda with enabled hesuvi, I almost wasn't able to identify, from where the sound come. It was like when I'm on some big place and sound came from long distance...

Does the room measurement is there to remove effect of the room where you measure or on the other hand to make the sound similar to room where you are?
Also what does the Equalization do? (the switch --no-equalization).

Thanks!
 
Dec 3, 2020 at 5:04 AM Post #462 of 1,816
I tried to play with it. The sound is really so wide, but I have problem, that it's too wide :D
When I tried to play e.g. Mass Effect Andromeda with enabled hesuvi, I almost wasn't able to identify, from where the sound come. It was like when I'm on some big place and sound came from long distance...

Does the room measurement is there to remove effect of the room where you measure or on the other hand to make the sound similar to room where you are?
Thanks!

This is exactly my issue with 99% of algos like this. I don't need an artificial recreation of a closed space (of any size), because most of my source material is NOT supposed to sound like I'm in an enclosed space. For open-space spatial-direction accuracy most algos fail (esp. without head-tracking, which enables you to move your head and pinpoint the direction more accurately).

If somebody solves this - and no, so far it's not Dolby (Atmos) or Creative (Super X-fi) or DTS or any of the other players - they will be successful and hopefully rich :-D
 
Dec 3, 2020 at 5:17 AM Post #463 of 1,816
I have now new Corsair HS70 Pro and I still don't know if I should keep them because of this (without effect the stereo is too much "separated"). Otherwise the sound is pretty good (especially for gaming headphones).
Before I had SoundBlaster tactis 3D Sigma where is simply a slider in their settings, for setting the "3D sound effect". It was perfect, I have it on 15% so it's not clean stereo but It's not also like this.

I also tried of curse Dolby Atmos and DTS:Headphone X and it also sound weird. Mainly the DTS was really bad.
 
Dec 3, 2020 at 11:00 AM Post #464 of 1,816
This is exactly my issue with 99% of algos like this.

Impulcifer is not an algorithm that simulates speakers. Impulcifer is a tool to measure and deconvolute your own "BRIR" (smyth calls this personalized impulse response PRIR ).

I tried to play with it. The sound is really so wide, but I have problem, that it's too wide :D

From my experience with Impulcifer I can tell that for me the two most critical parts are the mic placement and the SNR of the measurement. The mics should be placed as close as possible to the ear canal. This improves for me the speaker localization and the headphone compensation. Measurements with a low volume and a lot of noise sound more
distant to me and have more reverb. After trying different mountings I now use the construction like shown in the image below for 3 of my Primo pairs. It is working very well and I can even place the speakers without taking out the mics in the room.

p1050558-procmics_1052101.jpg
p1050549-proc-mic1-r11_1052100.jpg


I tried --no_room_correction --no_equalization --no_headphone_compensation

Headphone compensation is very important for biaunral audio.

Strange I do not see "--no_equalization" in the readme?
 
Dec 3, 2020 at 11:27 AM Post #465 of 1,816
I tried to play with it. The sound is really so wide, but I have problem, that it's too wide :D
When I tried to play e.g. Mass Effect Andromeda with enabled hesuvi, I almost wasn't able to identify, from where the sound come. It was like when I'm on some big place and sound came from long distance...

Does the room measurement is there to remove effect of the room where you measure or on the other hand to make the sound similar to room where you are?
Also what does the Equalization do? (the switch --no-equalization).

Thanks!
What kind of speaker setup did you use for the measurements? A regular stereo speakers perhaps? If so, do the speakers sound too wide to you? You could try narrowing the stereo speaker triangle if you're not happy with the current results. When you run the speaker test in HeSuVi, can you accurately locate the virtual speakers and are they in their expected locations?

The sense of distance is affected by a couple of things. You could take another measurement while sitting closer to the speakers. You could also try playing with the decay time parameter. Shorter decay times will usually bring the sound closer to you.

Room correction is to fix problems in room acoustics. No room is perfect and the sound of the room is captured by the measurement. You can measure room response with a calibrated room measurement microphone and then Impulcifer will create a correction filter which it will bake into the BRIR output.

No equalization disables equalization. The equalization option is not the headphone compensation but the additional eq curve which you can bake into the BRIR. You need to have eq.csv in the folder to use this. --no_equalization is just a shortcut for disabling this in case you want to try compare with and without. The main use case for additional equalization curve is to transfer the results to another headphones than what was used to do the measurements, for example IEM.
 

Users who are viewing this thread

Back
Top