Recording Impulse Responses for Speaker Virtualization

[conql]

Have you guys tried comparing headphones and loudspeakers by quick switching? I was amazed by my measurements before but now I feel quite the opposite. The simulation sounds too bright even though the localization is quite accurate. I haven't done careful tweaking now but I think it's about 5db more with a q value of 0.5 in the highs. I use @musicreo 's setup where ear canals are partially blocked, and I tried to insert the mics as deep as I can so it's less than 1cm from ear drum. This problem doesn't come from impulcifer, because when I play binaural recordings of speakers, it still sounds the same bright. I tried the manloud method proposed by David Griesinger, but I cannot get consistent results as he does and it doesn't sound good.

One big problem of the measurement is to get the correct timbre. This talk from David Griesinger does explain the problem. But I have some measurements were the timbre was ok although the measurement was not taken at the eardrum. Still I believe that deeper mic insertions improve the final result.

I measured the hptf after equalizing headphones, which is very flat during multiple wearing except for some small dips at high frequencies. So it can only be the problem of measuring. But if my understanding is correct, measuring from open ear canal should not cause large systematic error like this. So which step did I possibly go wrong?

 

[jaakkopasanen]

Have you guys tried comparing headphones and loudspeakers by quick switching? I was amazed by my measurements before but now I feel quite the opposite. The simulation sounds too bright even though the localization is quite accurate. I haven't done careful tweaking now but I think it's about 5db more with a q value of 0.5 in the highs. I use @musicreo 's setup where ear canals are partially blocked, and I tried to insert the mics as deep as I can so it's less than 1cm from ear drum. This problem doesn't come from impulcifer, because when I play binaural recordings of speakers, it still sounds the same bright. I tried the manloud method proposed by David Griesinger, but I cannot get consistent results as he does and it doesn't sound good.


I measured the hptf after equalizing headphones, which is very flat during multiple wearing except for some small dips at high frequencies. So it can only be the problem of measuring. But if my understanding is correct, measuring from open ear canal should not cause large systematic error like this. So which step did I possibly go wrong?

Did you do headphone compensation? Could you share your plots with us?

 

[conql]

Did you do headphone compensation? Could you share your plots with us?

The next two plots are what I get after applying headphone equalization and put on everything again.

The shp 9500 sounds significantly brighter than speakers, while the mdrv6 sounds only a bit brighter, probably just around 1 dB more high. But 9500 are open back headphones, why do they sound worse?

Apart from timbre, localization is good for both headphones.

 

[jaakkopasanen]

The next two plots are what I get after applying headphone equalization and put on everything again.



The shp 9500 sounds significantly brighter than speakers, while the mdrv6 sounds only a bit brighter, probably just around 1 dB more high. But 9500 are open back headphones, why do they sound worse?

Apart from timbre, localization is good for both headphones.

What are these plots? They don't quite look like what Impulcifer produces?

 

[conql]

What are these plots? They don't quite look like what Impulcifer produces?

No they aren't. I have many headphones to measure so I don't want impulcifer to embed the headphone compensation. So I write some code to call autoeq to do it. It's not smoothed because I find it makes no difference.

 

[castleofargh]

No they aren't. I have many headphones to measure so I don't want impulcifer to embed the headphone compensation. So I write some code to call autoeq to do it. It's not smoothed because I find it makes no difference.

Are you measuring the FR of the headphones at your ears, or are you just relying on the database?

 

[conql]

Are you measuring the FR of the headphones at your ears, or are you just relying on the database?

I am measuring FR of headphones by mics that are inserted really deep into ear canal. And I also measured the FR after headphone equalization which is flat enough as expected.

 

[jaakkopasanen]

I am measuring FR of headphones by mics that are inserted really deep into ear canal. And I also measured the FR after headphone equalization which is flat enough as expected.

And you measured with microphones in the exact same position as you did the speaker measurements ie. not taking microphones off in the meanwhile?

 

[Joe Bloggs]

I mean... I tried this in the past,
1. measure BRIR of loudspeakers using in ear mics
2. measure BRIR of over-ear headphones using same in ear mics
3. take final BRIR to be 1-2

and yes it sounded nowhere near what would be expected in terms of tonality. Much brighter and sharper on the headphones after the steps than the loudspeakers

I could only put it down to the mics not being my eardrums and the *FR difference between the speakers and the headphones* being much different on the mics than on my eardrums

 

[conql]

And you measured with microphones in the exact same position as you did the speaker measurements ie. not taking microphones off in the meanwhile?

I did take off microphones in previous measurements, but I think my mic setup is quite consistent and I do get steady measurements similar to musicreo. However, to verify this I measure loudspeakers and headphones again without taking the mics off and try to keep them in the same place. The results are sadly the same, some headphones are just much brighter than the others. And every headphone sounds less smooth than speakers. I give up now and decide to tweak the highs manually. Apart from tonality, I'm very satisfied with my current best measurement. I get pretty good imaging and localization and the sound is better than any headphones I've heard. Thank you for your great work! @jaakkopasanen

Recently, I write a small program as a graphical interface for impulcifer to help speed up measurement and make evaluating different profiles much easier. I really want to make it public but it's very buggy now because I don't have much programming experience and was aimed at fast development. If someone finds it helpful, I would love to keep developing it and upload it to GitHub when I have time.

 

[reter]

I mean... I tried this in the past,
1. measure BRIR of loudspeakers using in ear mics
2. measure BRIR of over-ear headphones using same in ear mics
3. take final BRIR to be 1-2

and yes it sounded nowhere near what would be expected in terms of tonality. Much brighter and sharper on the headphones after the steps than the loudspeakers

I could only put it down to the mics not being my eardrums and the *FR difference between the speakers and the headphones* being much different on the mics than on my eardrums

can i ask what headphone your tried? open back?

 

[jaakkopasanen]

I did take off microphones in previous measurements, but I think my mic setup is quite consistent and I do get steady measurements similar to musicreo. However, to verify this I measure loudspeakers and headphones again without taking the mics off and try to keep them in the same place. The results are sadly the same, some headphones are just much brighter than the others. And every headphone sounds less smooth than speakers. I give up now and decide to tweak the highs manually. Apart from tonality, I'm very satisfied with my current best measurement. I get pretty good imaging and localization and the sound is better than any headphones I've heard. Thank you for your great work! @jaakkopasanen

Recently, I write a small program as a graphical interface for impulcifer to help speed up measurement and make evaluating different profiles much easier. I really want to make it public but it's very buggy now because I don't have much programming experience and was aimed at fast development. If someone finds it helpful, I would love to keep developing it and upload it to GitHub when I have time.

Have you tried Impulcifer's headphone compensation? I think it's doing the same but just for eliminating potential error sources.

That being said, I do tweak the treble myself. I measure with blocked ear canal mics though.

 

[conql]

Have you tried Impulcifer's headphone compensation? I think it's doing the same but just for eliminating potential error sources.

I use the same parameters as impulcifer except smoothing. But none of those parameters affects tonality from my experiment. Anyway, I will still give it a try when I have some time.

 

[Joe Bloggs]

can i ask what headphone your tried? open back?

Closed back ones

 

[morgin]

Ignore if this a dumb idea but would there be a way to use ray traced audio to simulate a room with speakers and have each channel play in those virtual speakers? I’m thinking something like the vr cinema apps you can get to watch movies.. have a full screen video playing in a virtual environment but also in that have virtual speakers with ray traced audio for each channel.

Another idea would be instead of physical measurements with in ear mics. Use ray tracing audio in a virtual environment with a virtual head (with our personal head/ear distance measurements) We could play a sound from virtual 7.1 speaker and listen to audio ques and adjust the virtual in ear mics in real time for best and accurate results.

Idea 3 use lidar same as what you get on new iPhones to scan your ears place the scan on a virtual head same size as your own. Then use the virtual speaker and raytraced audio to take measurements

With a program like this




With those we might be able to get height channels if I’m not pulling ideas from my backside