Recording Impulse Responses for Speaker Virtualization
May 14, 2022 at 9:29 AM Post #1,366 of 1,817
I’ve had problems in the past. Sometimes the windows update can mess with my hrir so I end up having to remove all sound drivers (except behringer), restart, delete the eqAPO folder with hesuvi inside, restart again for windows to automatically install sound drivers, then install vbcable, eqAPO and hesuvi then set everything back to how it was. It’s a pain.

If that doesn’t work try what I’ve posted earlier where I extract the headphone eq and used that in hesuvi. It gives more virtualisation for some reason.
Yes i should test that , but i'm more and more asking myself if the problem is ..me. I tried Out of the Head by fongaudio, and their demo, i only heard a dummy sound, with reverb, all IN my head, very very bad localisation... a lightyear from my 5.1 system.
No, in fact, whatever i tried i had very bad difficulties with front projection : at better i succeed to hear the sound coming just a little above my head, but never one or two meters from like my actual speakers are.
And i'm wondering if i would come to buy a smyth realiser (an used A8) if i could have the same problem
 
Last edited:
May 14, 2022 at 10:19 AM Post #1,367 of 1,817
Hi everyone,

Long-time lurker here. I'm also an owner of a Smyth Realiser A16 but have been using Impulcifer with a stock Sound Professionals MS-TFB-2 for quite a while. In case you ask, the reason for my ongoing interest in Impulcifer is its versatility for professional use (by having access to the raw IR, rather than re-convolute digitally from the Realiser). Also, the Realiser won't support CoreAudio (MacOS) due to restrictions of the driver structure of the built-in USB interface, so any application for mixing duties is out of question right now (well, if you want to stay in the digital domain, that is). This as well as teaching duties in sound/media engineering and psychoacoustics have pushed me back towards the Impulcifer route.

In my efforts to improve the capture process, I'm currently shopping for binaural microphones that are best suited for the task and ran across the MS-TFB-2-MKII (Link: https://soundprofessionals.com/product/MS-TFB-2-MKII/), apparently a new iteration of my current MS-TFB-2.

MS-TFB-2-MKII-3.jpeg


These look very interesting. Have any of you gathered experience with the MKII variant of these microphones? I assume the prospect to use them as part of an in-ear plug structure might improve captures (while avoiding a mess on the mic element as is to be expected in people with narrow ear channels = moi)? However, I'm not sure whether the mic elements will insert deep enough into the ear canal. Or is it more efficient to simply cut off the hooks of my old MS-TFB-2, put them in foam plugs and crush them into the ears in good old Realiser fashion?

Any ideas or recommendations for alternatives would be much appreciated.

Thanks!

EDIT: Another candidate might be this handsome fellow: https://soundprofessionals.com/product/SP-EAR-MIC/ or the https://soundprofessionals.com/product/SP-EAR-MIC-2/ (stereo set) as pictured below. Any experience with this one?

SP-EAR-MIC-2.jpeg
 
Last edited:
May 14, 2022 at 1:33 PM Post #1,369 of 1,817
Hi, i've finally received my umc 202 hd, no more "zigzag" in my plots.So i'm getting some results, but until know no good spatialisation. I'm beginning to think that my brain is not receptive to virtualisation. I've experimented a not so bad stereo virtualisation obtained at the end of the day (tired?) and next morning (awake!), with the exact same configuration no more spatialisation ! So i doubt... It brings me a question to the owners of smyth realiser : is is the same difficulties to obtain good virtualisation ?
There are so many variables I can only offer some ideas. If you are able to locate sounds in direction and depth normally without headphones then it doesn't seem likely that the problem is your hearing or brain. It's probably the localization cues in the sounds being delivered to your ears. Of course something in the measurement process, or Impulcifer command options could be the culprit, but also perhaps it's vision. That may sound odd, but what you're seeing or not when listening can have a big influence on sound localization. Impulcifer works extremely well for me when I'm sitting in front of my speakers in my normal stereo music spot. The illusion is exceedingly realistic, a soundstage 7 ft in front of me, pinpoint left to right imaging, realistic depth, everything as if I'm actually at a live performance. But I can use the same headphones and HRIR in my computer room in front of the monitor with a wall close behind it and the virtualization mostly collapses. If this isn't relevant to your situation then one thing you can check off the list, but it's worth being aware of.
 
May 14, 2022 at 2:37 PM Post #1,370 of 1,817
Yes i should test that , but i'm more and more asking myself if the problem is ..me. I tried Out of the Head by fongaudio, and their demo, i only heard a dummy sound, with reverb, all IN my head, very very bad localisation... a lightyear from my 5.1 system.
No, in fact, whatever i tried i had very bad difficulties with front projection : at better i succeed to hear the sound coming just a little above my head, but never one or two meters from like my actual speakers are.
And i'm wondering if i would come to buy a smyth realiser (an used A8) if i could have the same problem
Prof Choueiri mentioned that for some people, the center image simulation just doesn’t seem to work. Probably something to do with the visual cortex dominating their subjective experience even more than is usual for humans. So when your eyes and ears disagree because you're in a different room or you're looking at a wall instead of a band playing, your brain decides to trust what it sees but somehow rejects the audio cues even more than is "normal". Things get better on the sides because you're eyes are in front, and also maybe because unlike the center image, side stuff also have interaural cues(variations in timing and frequency response between the 2 ears that define at least the direction in a much clearer way than mono stuff where only the general FR has a little impact on vertical elevation and that's about it for localization cues.
At least that's my hypothetical explanation. I have no evidence of anything ^_^.

A few options to try:
Listen for a while in the dark, nothing visual must disturb you. Most people end up with an impressive space from audio after a few minutes. While doing this, try to not move your head at all(if you don’t have a chair high enough to maintain your head, maybe lie down on a bed). After a while, can you get frontal audio at some distance? If not, I'm not optimistic that there is a virtual solution for you.
If things go well, find out if moving your head kills the distance while in the dark. You need to be in real darkness so you eyes can't lock onto something to get spatial cues.
Last thing to test is to listen to the same convoluted audio while sitting where you’d usually sit to use your speakers. And then do the same somewhere else. That is to check how important the room and seeing the speakers will affect you. there's a paper I posted somewhere suggesting that at least at a statistical level it does make a difference to be in the original conditions, but again, how much it does for you is what counts. If the difference is marginal, you don't have to bother with that. If it's not, then you'd better record from where you'll be sitting(I imagine in front of your computer at a desk).

To me, head movements ruin everything and I really need head tracking. I also find a great benefit to listening to PRIRs I made with the monitors on my desk(at about 1m from me). Some will be amazed to have the recordings of a big room with fancy speakers, but I tried and ended up going back to my actual room(even thought it's an acoustic nightmare!!!!!). Somehow I get the same reverb, I can anchor the sound source to the speaker monitors I see in front of me(that also makes a huge difference for me). It just feels reals and I tend to keep the center image between those 2 speakers, maybe because I hear it there? Maybe because I'm expecting the center image to be between speakers so my brains wants it there? I know how I feel and what works for me, but I don't have anything beyond guesses when it comes to why.
On the A16's thread, at least 3 people have said that they didn't use the head tracking(didn't like it or didn't see the benefit). I cannot understand, but I also can't ignore that I exist and they exist. So which one are you? Who knows? hopefully my weird experimental suggestions can let you get some ideas even without actually testing head tracking+PRIRs(headtracking modeled on some dummy head were a disappointment for me, and then again, some other people loved them... subjectivity is such a bother :wink: ).


About difficulties in making measurements, I would argue that it’s easier to fail with the A16 because you have to record all speakers at several look angles for the head tracking. I can't count how many times I ended up having the mic move in one ear while I turned to a different position, or while I tried to read something on the A16. Having a super clever part time slave to do the all thing for you while you focus only on being there and turning when asked to, that makes all the difference IMO.
About the final frequency response, the uncertainties are the same as with impulcifer . You probably could benefit from some EQ(on top of the measurements) with one of the methods discussed in this thread somewhere, that are mainly based on equal loudness contour and some luck. For your center image problem, the FR as I said can impact elevation. That in turn may or may not help trick your brain about setting a distance for the virtual center. That's really a case by case thing and it's near impossible to know when you get it right beside feeling for yourself that the center is at eye level(or whatever level it's supposed to be based on where the speakers were).

As for the A8, it seems to require solid knowledge of the manual. Most people seem to give up on making prirs on the A16 because they find it overwhelming(the Smyth guys said so). I sure was like that for the first 2 weeks, but then I finally pulled my fingers out of my butt and started struggling toward a result(the first step is the hardest, like always). Anyway, people who owned both, say that the A16 is much easier:sob:. Then again, you're here. Meaning that along all the 7 other people on this thread, you have the willpower to put in the work and RTFM if it helps get away from the failed stereo that is normal headphone use. You being part of that elite group of binaural mic warriors(too much?^_^), makes me hopeful that you'd deal with whatever Realiser complexity.


Where are you in France? You should definitely try to find a Realiser owner near you who would agree to demo it. There is really no substitute to trying it yourself. Il y a toujours Paris et le revendeur Francais(Gilles Gerin) qui propose une demo avec calibration, si jamais. C'est sur RDV, il faut lui demander sur av-in.com .


edit: OMG I wrote a book! Sorry about that.
 
May 14, 2022 at 3:00 PM Post #1,371 of 1,817
@ castleofargh
Thanks! Don't be sorry, "au contraire", your answer is fascinating, i will have to read it several times to try everything you've said. I've got several hours of trials that are coming :slight_smile:
Thanks again :clap:

@ lowdown : and when you made the measurements, were yours speakers at 7 feet ?
 
Last edited:
May 15, 2022 at 5:41 AM Post #1,372 of 1,817
@jaakkopasanen a thought came to me. Hearing aids are a huge market and the ones my mum has are around the back of the ear and don’t pick up sound too well or give positional ques that well. She need to wear two and I guess it would only work for people who wear two hearing aids. Could impulcifer be useful? Have the mic (instead of a mic it’s a probe tube) be in ear unlike the current hearing aids and have that give information to amplify any sounds that need amplifying for the patient.

So the sound information it’s collecting just like impulcifer is tailored to exactly what is being received inside the ears.

The hearing aids are expensive around £2000 I’m sure we could make a lot of money with specific hrir hearing aids.
 
May 15, 2022 at 5:45 PM Post #1,373 of 1,817
There are so many variables I can only offer some ideas. If you are able to locate sounds in direction and depth normally without headphones then it doesn't seem likely that the problem is your hearing or brain. It's probably the localization cues in the sounds being delivered to your ears. Of course something in the measurement process, or Impulcifer command options could be the culprit, but also perhaps it's vision. That may sound odd, but what you're seeing or not when listening can have a big influence on sound localization. Impulcifer works extremely well for me when I'm sitting in front of my speakers in my normal stereo music spot. The illusion is exceedingly realistic, a soundstage 7 ft in front of me, pinpoint left to right imaging, realistic depth, everything as if I'm actually at a live performance. But I can use the same headphones and HRIR in my computer room in front of the monitor with a wall close behind it and the virtualization mostly collapses. If this isn't relevant to your situation then one thing you can check off the list, but it's worth being aware of.


maaaan 7ft distance, i don't have a room that big, i can go around 4ft max the results are already impressive, i can't imagine in a room that big
 
May 15, 2022 at 6:00 PM Post #1,374 of 1,817
@jaakkopasanen a thought came to me. Hearing aids are a huge market and the ones my mum has are around the back of the ear and don’t pick up sound too well or give positional ques that well. She need to wear two and I guess it would only work for people who wear two hearing aids. Could impulcifer be useful? Have the mic (instead of a mic it’s a probe tube) be in ear unlike the current hearing aids and have that give information to amplify any sounds that need amplifying for the patient.

So the sound information it’s collecting just like impulcifer is tailored to exactly what is being received inside the ears.

The hearing aids are expensive around £2000 I’m sure we could make a lot of money with specific hrir hearing aids.
I think there are also hearing aids that are very small and placed completely with the mic inside the ear canal. That way the localisation cues are largely preserved.

Impulcifer is of no use here. What should it do? Impulcifer hrirs can be used to give localization cues to sound corresponding to one of a limited number of known locations (where the loudspeakers are for example) and that doesn't have any localisation cues yet (just an audio channel). The sound picked up by mics in the ears already have all the cues so nothing needs to be changed, just amplified. And that sound can be in fact a summation of different sounds from many different directions, and those directions can be any of all possible directions - infinitely many!
 
May 16, 2022 at 1:34 PM Post #1,376 of 1,817
:L3000:It's sort of like sitting in the 1st row at the concert. But 4 ft could be like resting your chin on the front of the stage. Not a bad spot.
ahahahah if wasn't for the fact that i get too much high frequencies that's annoying sometimes, i should try recording the brir with the speaker in the corner to get more distance while still being in the center of the room
 
May 16, 2022 at 1:51 PM Post #1,377 of 1,817
ahahahah if wasn't for the fact that i get too much high frequencies that's annoying sometimes, i should try recording the brir with the speaker in the corner to get more distance while still being in the center of the room
Experimenting with different speaker locations and distances is a good idea. A corner placement is likely to boost the bass frequencies, but if it's too much it can be fixed with EQ, and there may be other advantages with your speaker and room. As others have suggested using EQ to adjust those bothersome high frequencies can be very helpful. I've made some EQ tweaks to my best BRIR with very good results.
 
May 16, 2022 at 3:13 PM Post #1,378 of 1,817
@ castleofargh
ok i've tried with this video (and others of this channel) :

Right/left separate: perfect virtualisation : amazing !
When it comes to center stereo image with both left and right playing, at first a big part of the sound come back in my head, but another part stay on the virtual speakers - strange. But i've got the fealing that more i listen these video, the more the center sound seems to get out of my head. Do you think that a sort of training of my ears/brain could make the virtualisation work for me ?
 
Last edited:
May 16, 2022 at 9:00 PM Post #1,379 of 1,817
Experimenting with different speaker locations and distances is a good idea. A corner placement is likely to boost the bass frequencies, but if it's too much it can be fixed with EQ, and there may be other advantages with your speaker and room. As others have suggested using EQ to adjust those bothersome high frequencies can be very helpful. I've made some EQ tweaks to my best BRIR with very good results.
actually i'd like some more bass, i think the brir i'm making have too much clarity, very much compared to the vanilla headphones (HD660S), i listen to some audio tracks and is like i'm missing some details under 1000hz or so, i can hear very clear voices and stuff happening but when it comes to some deep stuff it's literally almost inaudible if compared from before... i don't know if this is because i was using dolby hrir since a lot of years so now i'm really used to them at the point i'm getting the feeling that the real fidelity is what i was getting before; maybe it's because i was used to hrirs with large spaciality and now that i'm using impulcifer recordings at 4ft distance i feel this weird satisfaction mixed to unsatisfaction; i applied a peak filter to lower the peak i was having at 4000hz frequency, it's better than before but still there's no reason to keep recording at 4ft
 
May 17, 2022 at 3:43 AM Post #1,380 of 1,817
I recorded at 3.3 ft, 4 ft 5 ft, 6 ft and 7 ft. My best measurements where obtained at 5ft. (recording room size 6.3m*4.1m). I think the optimal distance is strongly depending on the room and the used speaker.
 

Users who are viewing this thread

Back
Top