1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Recording Impulse Responses for Speaker Virtualization

Discussion in 'Sound Science' started by jaakkopasanen, Oct 9, 2018.
1 2 3 4 5 6
8 9 10 11 12 13
  1. jaakkopasanen
    I actually got the surround setup room correction working. I just had renamed two room measurement files wrong. Now the demo contains recordings for 7.1 room response corrected HRIR.

    I'll need to write the measurement guide still but for now the most eager ones can read through the updated processing documentation and play around with webcam mic placement helper. That thing leaves a ghost behind when taking a picture which can be used to place the room measurement mic in the same location where the binaural mic was. The seven rectangles in right side are slots for different measurements. Select a slot by clicking it and the picture will be saved to that slot. Clicking another slot will take it's ghost to the big picture. Like so: https://i.imgur.com/vKwrEGI.png. Here the microphone stand is in front of me and you can see both my face and the measurement microphone. It's not the best of all solutions for the problem but serves for now. It would be better to have the webcam above the listening position looking down.
  2. johnn29
    Excellent! I read the documentation yesterday - I must confess I'm a little confused with how it'd work so I'll wait for the full documentation.

    Initially I just expected room correction to flatten the frequency response. In the bass range I thought it'd just be a simple flat line until around 80-100hz and after that it'd be some sort of average response for each speaker/ear so that the peaks and nulls are removed. That way you'd maintain the HRTF but obviously it's way more complicated than that.

    I have a UMIK-1 so I'll try it out when the documentation is live.
  3. jaakkopasanen
    I'm afraid it's not possible to know how the HRTF affects the measured frequency response and how much is from room acoustics. This requires a reference measurement with calibrated microphone in the precisely the same location as the in-ear microphone. You place the measurement microphone in the same spot where left ear in-ear mic was and run record sweeps. This produces room-FL,FC,FR,SR,BR,BL,SL-left.wav on 7.1 speaker system. Same thing has to be repeated for right ear and obviously the measurement microphone has to be moved to the position of right ear.
  4. johnn29
    Oh ok that doesn't seem complicated as I thought. I'll give it a go soon, very excited about this and it's a real way your solution is better than the realiser
  5. arksergo
    jgazal and sander99 like this.
  6. jaakkopasanen
    I've been using the HRIR with room correction for several hours now and I have to say I'm very pleased. So much in fact that I've started to do critical listening again. I was of the opinion that I might even prefer my Custom Art FIBAE 3 CIEMs for music listening to HD 800 or my speaker setup because they provide very laid back and pleasant listening experience when EQd neutral. Now however the FIBAE 3s start to sound boring compared to what HD 800 can do with room corrected HRIR. Harman target equalized headphones are very close to my neutral but it's lacking a bit of the excitement the room corrected HRIR offers.

    Compared to my speakers, with which the HRIR was recorded, virtualization offers a lot better detail retrieval. It's on par with the HD 800 without virtualization and in some ways better. Imaging is also significantly better than with my speakers. This was a complete surprise for me because I didn't expect virtualized speakers to be able to provide better imaging than the physical speakers and room they try to simulate. Soundstage is similar and maybe even a bit wider. Vertical placement of the instruments is definitely more accurate. On the physical speakers instruments and vocals seem to shift upwards but with virtualized speakers they are centered around the vertical level of the physical speakers with clear variation. I'm not sure if the vertical places of the sounds are correct per se but it's nice to have more two dimensional sound image.

    Good imaging with an actual speaker soundstage combined with the natural detail level of the headphones make it even easier to detect all the detail in the recording because now that the sound "blobs" are further apart it's easier to focus on individual instruments. I think it would be exceedingly difficult to achieve this level of detail retrieval with speakers alone. Same seems to apply to dynamics. HD 800 are very dynamic headphones and the speaker virtualization doesn't compress the dynamics too much. Maybe there is a slight reduction but I actually prefer it this way because on some songs the HD 800 can be a bit too much. All in all the technical ability of the headphones seems to translate very well to speaker virtualization.

    To me the speaker virtualization with headphones is the way to go for ultimate hifi. It combines the best of both worlds retaining the technicalities of headphones while providing the sound stage of speakers and even exceeding them in imaging. Tonality with room correction is the best I've heard in headphones or speakers. Having the frequency response tailored to my ears makes for extremely natural and effortless listening experience.

    I think I should be able to improve on the bass speed and punch with decay management. Currently the room corrected HRIR has a bit slow bass because the bass decay is so long. I'm also hoping that the improved signal-to-noise ratio from tracking filter would improve imaging even further. At least the early reflection management should do this if I manage to pull it off. I got an idea of detecting the reflected sound direction the same way as human brains localize sounds and then cancelling the reflected sounds that are unbalanced or come from undesired directions. We'll see how this works out.
    johnn29 and jgazal like this.
  7. johnn29
    Sounds exciting - I'm still waiting to try out the room correction when I get some time.

    I did notice in some of the OOYH presets that work particularly well for me the stereo center image sounds like a real center. I assume because the room that was measured has bang on frequency response for the left and right channels so I'm very much looking forward to trying that out on my own.

    Had the perfect use case for Impulcifer just now - there's construction work going on in the cellar beneath my room. I wanted to watch a show in peace in my office so don the Bose 700's, ANC to max and my 7.1 channel measurement from my theater in HeSuVi. Can't get over how you can just do that now days. After not using it in sometime I always have to check the speakers aren't on too - testimony to the re-creation of a loud speaker.
  8. Joe Bloggs Contributor
    What does your room correction consist of?
  9. jaakkopasanen
    Currently it's just a minimum phase EQ to Harman room target but I intend to implement mixed phase filter. Tips for that are welcome if anyone here has any insights.
  10. jaakkopasanen
    I did some quick testing with different mic insertion depths and heapdhone placements. The results are wild. As in there is a huge variation in frequency response depending on both of these factors and because of this I'm not getting a correct channel balance consistently. Inserting the mics deeper into the ear canal will boost treble by some 10 to 20 dB and this is most likely due to ear canal resonance. I'm suspecting that the ear canal resonance plays a role even when the mics are at the ear canal opening and the ear canal hasn't been blocked as is the case with The Soundprofessionals' mics. Basically all literature says that the best way is to use ear canal blocking mics but I kind of assumed that this only meant that the mic needs to be at ear canal opening and the actual blockage isn't that important. Perhaps I was wrong to assume. I'm going to test this hypothesis by blocking the ear canal with an regular foam earplug.
    jgazal likes this.
  11. sander99
    (I decided to post this reply here as I think that is more appropiate than in the "To crossfeed or not to crossfeed?" thread.)
    (@ironmine: I assume you meant where you can buy these, not when?)
    I am trying to decide which mics I may buy. I actually think the ones in ironmine's picture look the most suitable I have seen - shape and size like I mean - because there is minimal material in the earflaps and the mic is really in the opening of the ear canal, and not a little bit further to the outside like many others. Also these resemble the Smyth Research mics more than any of the others and Smyth sure know something about HRIR/PRIR measurements. The plastic clips - even more with the cable running around - inside the earflaps of the soundprofessional worry me because how could they not change the reflections/influence of the earflaps? I would sooner suspect those plastic clips cause problems. But it is just an intuitive feeling.
    Last edited: Nov 4, 2019
    jgazal likes this.
  12. jaakkopasanen
    My next mics will definitely be Primo EM258 capsules in either IEM shell or simply glued to an ear plug. This will allow me to make them so that the mic sits exactly at the ear canal opening and the ear canal is blocked. Plenty of cheap IEM options in AliExpress for less than $2. And as a matter of a fact I just ordered 3 pairs of IEMs for $3.46 :D
    Last edited: Nov 4, 2019
  13. jaakkopasanen
    Hmm. Now I also found these:

    Add IEM housings or ear plugs and that's a complete binaural mics for around $100 with very little work. Those are Primo EM172 which are 10mm capsules so I don't know how well they would fit at the ear canal opening. Need to test with some dummy capsules first.
  14. johnn29
    I had that channel balance issue with many of my recordings - I figured I messed up something on my DAW. I got lucky and got a bang on recording off one of my attempts that I actually A vs B'd the real speakers and it sounded very similar, so I'm happy with those. But I need to do more recordings for the room correction and other rooms I have.

    If blocking the ear with a regular foam plug works I can continue to use the sound professional mics I have - otherwise I'll need a new solution.
  15. johnn29
    I've run 4 recordings today and each of them had channel balance issues with the sound professional mics. It seems I've only got one recording that's bang on for channel balance from another day. It's very hard to get right - not sure if it's the headphone compensation or the speaker measurement. My setup makes it harder because I'm trying to use the Bose 700s which have a high clamp force.

    On the positive you can fix the issue with HeSuVi's attenuation. If you play the white noise Atmos test tones for the center channel - you can adjust attenuation until both L and R are equal in voicemeter. But obviously this is having an impact on the fidelity of the HRIR.
    Last edited: Nov 5, 2019
1 2 3 4 5 6
8 9 10 11 12 13

Share This Page