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.
6 7 8 9 10 11 12 13 14 15
17 18
  1. johnn29
  2. lowdown
    Well this is a fine cup of tea. You've ruined my perfectly good, expensive audiophile speakers. Ok, not exactly, the speakers are just fine. But now I don't want to listen to them anymore because the sound I'm getting through my headphones is so much better!

    I got my Sound Professionals mics last week and the first attempt with Impulcifer even with less than ideal mic in ear positioning and no room correction was stunning. I've listened to binaural recordings for years and tried all the HRIR options in HeSuVi but have never gotten the full illusion of speakers 8 ft in front of me with the imaging and sound stage I heard on that initial measurement round. Was hard to wipe the smile off my face. Been doing measurement sessions daily since, trying to refine all the sonic aspects, with very mixed results. But last night I hit a great new plateau and it's almost unbelievable. The tonal balance, detail, imaging, intimacy to the sound as if the performers are in my living room. Now I've got to listen to everything I own again! I've spent hours sinking into the music, and I know this isn't even as good as it can get. There's no way I could thank you enough, and all those who've contributed to this virtual miracle.

    One tip, I had significant problems getting the Sound Pro mics to sit in my ears properly, and stay in position. As you documented I ended up gluing foam ear plugs onto the back and that made a huge difference in positioning, stabilizing, and getting much better recordings.

    I may be back with some novice level questions, and look forward to your future efforts, but wanted to offer an update and my extreme appreciation.
    johnn29 and jaakkopasanen like this.
  3. Joe Bloggs Contributor
    I downloaded and deconvolved FC from your samples

    Top is the raw results from it. It does seem that there is a very strong resonance at 65Hz, lasting for over half a second. The magnitude at that frequency is also high, but the main issue seems to be how long it drags on.

    Bottom is the impulse after manual editing (I split the impulse into three parts and EQed out the resonance tails in the latter parts with a parametric EQ. I'm actually surprised this worked :sweat_smile: )
  4. jaakkopasanen
    Awesome! I'm so glad to hear this. These kind of experiences people have is what keeps me working on this more than I should.

    Interesting. This should be visible in my plots but it's not. Probably my plots are not correct / perfect. I need to check those.

    I fed my room dimensions into a room mode calculator and indeed there is a lot going on just below 65 Hz: https://amcoustics.com/tools/amroc?l=830&w=360&h=268&r60=0.6

    Tracking filter and reverberation time management should then fix my issue.

    Ah, silly me. Thanks. I'll take a look (listen).
  5. jaakkopasanen
    @johnn29 I had a listen of your measurements. The problem sounds like the same I'm having. I noticed it best with side right and back right channels where the bass ringing is heard on the right headphone. I actually had the same ringing there even without room correction, although reduced in level.
    johnn29 likes this.
  6. johnn29
    Ok thanks for looking. I hope your tracking filter corrects it. I guess with the sub the room EQ I have on my AVR corrects for that room mode, so I don't get the issue with the sub
  7. lowdown
    As I've posted, the results I'm getting with Impulcifer are truly amazing. But I have encountered a couple of anomalies. These may be very obvious to you, but at the risk of embarrassing myself I'll offer them in case it could help someone else or provide some useful data.

    I have Sound Professionals mics, am using a Zoom H2N plugged into the PC, mics into the H2N, and Senn HD600 headphones into the PC headphone jack. The issue is I'm getting two very different headphone plots, neither of which looks good. The first seems sort of normal except for that severe trough:


    The second is virtually flat, except that it's a mess of tight peaks and troughs, like an extended earthquake plot:


    I assume I'm doing something different and wrong on both, but don't know what it is. But the key is when I use the 2nd flat but messy one to generate the hesuvi.wav file the results are spectacular. The first one yields really bad sounding results, which probably makes sense given how uneven it is. But whatever you're doing to smooth things out is working to perfection on the 2nd one. Again, I don't know if that's interesting to you, but there it is.

    Second, on several of my results I heard a slight but distinct low frequency thumping sound in my right ear on some music tracks. Not terrible, but noticeable and distracting. Everything else kept the illusion of coming from well in front of me, but those little thumps were right in my ear. Then on one run I included the room_target option using your Harmon file, and that thumping was completely gone. I didn't prefer the extra bass boost in the Harmon target so I edited the file and equalized everything from 20hz to 9khz flat, but left the Harmon settings below and above those two frequencies. The thumping was still gone, so I stumbled onto a solution.

    I'm so happy with the final result there's really not a problem to solve, I just wanted to thank you, and give some feedback that might be of interest.
    Last edited: Jan 14, 2020
  8. musicreo
    The headphone plots look very strange. How high was your headroom for the measurements? How was the Mic placement? The valley at 200 to 700 Hz is the reason why it sounds bad. Impulcifer will compensate it but this is not the real headphone response. The valley at 8 Khz is probably due to resonance between headphone and mic. This is how my headphone measurements look like. 1) AKG701 2) Sennheiser Hd555(with 595mod).
    AKG701_1.png hd555.png

    For me the best measurement so far was done with only one speaker in the center position. I guess that this a effect of the room accustics as I don't have a calibrated hearing position and for the center position the distance to reflecting walls is the highest of all.
  9. lowdown
    Yes, I agree mine look quite strange. On the 1st example the headroom would vary from -20 to -6 depending on the gain on the mic and output level on the PC volume control. The other odd thing I didn't mention is on the 2nd example the headroom was always -0 no matter what I did with the gain. I suspect there are some very basic things that I don't have right in both cases. As far as mic fit in my ears I had trouble getting a stable fit until I glued foam earplugs onto the back of the mics, but actually that had little to do with the measurements. I got nearly identical graphs with and without the earplug mod. What might be at play is a mismatch between the bit rate in Windows and on the H2N. I tried tweaking that on the last measurements and I'm pretty sure the graph switched from the 2nd version back to the first. But since I was getting such good final results with that 2nd graph I pretty much quit experimenting for the time being.

    As far as room measurements I have a calibrated UMIK-1 mic, and took a couple of different "specific" measurements, but again the results are good enough that I've paused making more. I'm only using stereo since my interest is in music vs movie surround. I may try using 1 speaker to see what effect that has but would kind of like to get that headphone measurement thing figured out first. Your graphs look great.
    Last edited: Jan 14, 2020
  10. lowdown
    Ok, I figured something out, and as I suspected it was very basic. Previously I had my headphones plugged in to the headphone jack on the laptop. I had originally tried using the headphone jack on the H2N but was getting feedback, and didn't realize "monitor" needed to be off, so I ended up using the laptop headphone connection. I believe that's what was causing most of the problems. I just now did another run with the headphones plugged into the H2N, with monitor turned off, and here's the graph:


    Much more normal looking. One other change is I cut the wings off of the Sound Professionals mics, so it's likely a better recording of my actual ears. No idea how much that was a factor, but it's certainly better now.

    Edit: So far the results from my original post, that flat/earthquake headphone graph, are better than this latest run. I'll keep playing and adjusting, but curious to me that it's so much better sounding. If I never get better results than that I'll still be ecstatic.
    Last edited: Jan 14, 2020
  11. musicreo
    That makes sense. If monitor was on, the sweep was mixed with the signal from the mics.
  12. lowdown
    Agree, novice mistake. But it also seems to have been a combination of having the headphones plugged in to the PC jack rather than to the H2N. Some of those runs were with monitor turned off as well. I'm also still puzzled why the sound using that flat messy scan is better than the newest ones. Sorting through various options to narrow it down.
  13. musicreo
    How does it sound without headphone equalisation?
  14. lowdown
    I haven't applied any specific headphone equalization. The flat fuzzy plot version sounds much better, much more tonally balanced bottom to top, and in HeSuVi it sounds best with no headphone equalization applied. The latest one, where the plot is more normal, is similar sounding in the low and mid frequencies but has a very bright treble section, which is reflected in the plot. The L/R balance isn't as good either. So far the best results I've had by far have come from what's obviously a mistake in the setup. It sounds so good I really don't know if it can be improved. The tonal balance, clarity, imaging and detail are stunningly good. A happy accident for me.
  15. sander99
    @lowdown: I have a feeling you don't understand what musicreo is saying. Normally impulcifer uses your headphone measurement to insert the headphone equalisation (or compensation) in the resulting hesuvi.wav file. So when using it with the HeSuVi software you don't use additional headphone compensation (because then you would compensate double).
    When you run python impulcifer.py you can use the option --no_headphone_compensation and then no headphone compensation is inserted in the resulting hesuvi.wav, then your headphone measurements are not used at all.
6 7 8 9 10 11 12 13 14 15
17 18

Share This Page