Recording Impulse Responses for Speaker Virtualization

[Joe Bloggs]

This is an incorrect assumption of what ambiophonics is. By having the speakers closer together you are shortening the path of the contralateral and the ipsilateral. You then run the signal through a feedback delay network. In this manner, you are creating a cross-talk cancelation network.


Happy to see so much excitement about HRTF and spatial audio for headphones. Psychoacoustics, and HRTF's are my primary research responsibilities. Fun times indeed with all the new things happening here.

Great to see you here, but I don't see the contradiction here with "set up a speaker system that eliminates crosstalk". Unless you object to the "spending a LOT of time and money" part, if you consider that the crosstalk cancellation a done deal that can be realized on everything down to two speaker smartphones and stuff.

I also personally prefer upmixing stereo content to an actual conventional surround system to enhancing it using ambiophonics :]

 

[pfzar]

Great to see you here, but I don't see the contradiction here with "set up a speaker system that eliminates crosstalk". Unless you object to the "spending a LOT of time and money" part, if you consider that the crosstalk cancellation a done deal that can be realized on everything down to two speaker smartphones and stuff.

I also personally prefer upmixing stereo content to an actual conventional surround system to enhancing it using ambiophonics :]

The speaker system is not what eliminates the crosstalk. The understanding of the geometry applied to the algorithms is what provides the cues for the brain.

 

[jaakkopasanen]

I came across a very interesting presentation about binaural hearing and headphones. Strongly recommended read. It for example argues that around-ear headphones are not good for binaural rendering (speaker virtualization for example) because practically all around-ear headphones have 90 degree destructive interference which cannot be equalized away. The author also shows how blocked ear canal measurements can have correct localization but not timbre. The frequency response measured at blocked ear canal opening is not the same as at ear drum when wearing headphones. For best results the measurements should be done with probe microphones that measure at ear drum.

I picked this up from article A Deep Dive Into Harman Curves Part 1, which I also recommend for reading.

 

[johnn29]

Thanks for the links!

Half way through the presentation and my thoughts:

- He is correct on the mad man would put those mics so close to their ear drum!

- Is what we're interested in - speaker virtualisation - significantly different to dummy head recordings? We only have 7 point sources. We also hear all the room reflections too - so those 7 point sources become quite vast and we know from the trim tail script you've got in the research folder that the tail is very important for localisation. What the presentation is interested in is concert hall recordings that have a ton of sources that vary in distance, elevation, instrument type (string produces differently to a drum for example).


- Because of the above - the around ear headphone conclusion of the presentation doesn't seem valid? In my own experience I've used on ear (GW100, B&W p5) and a load of over ears (Bose 700, WH-XM3, DT990 and more). I agree that the ANC algorithm results in the most benchmarked result - especially with the Sony XM3's because they have a calibration button. So you can seat it on your head, calibrate it and you know that playback will be identical to when you took your recording, as long as you calibrated before recording. The Sony's and the Bose do mess with the bass/treble levels depending on volume though - but that's a simple boost/reduction. I do get the most realistic presentation with the GW100's (on ears) but I assumed that was because they were the most open per the rtings isolation measurements.

I've never had an issue with sound localisation with any of the recordings. Timbral differences yes - especially when trying to EQ in ear earphones via transform. With my Air Pods Pro, for example, I get the localisation but the sound signature sounds really off. The only way I've gotten them to sound right is by using Bose 700 and Air Pods Pro from rtings source, anything else sounds off. This is further complicated by firmware updates changing the sound signature - so Transform might not work as well.

While we're all in lock down I've not had to use Impulcifer as much and enjoying my actual loudspeakers. It's funny how now I appreciate head movement with a real set of speakers. But on weekends sometimes my wife can be reading a book next to me and I'm watching my OLED with a movie at reference volume with my headphones. Pretty sweet!

 

[jaakkopasanen]

I wouldn't say the claim about around-ears (vs on-ears) isn't valid in a room necessarily. Maybe more important is the question how much does the 90 intereference matter in the end. It's more or less a single notch in the frequency response and that doesn't mean the timbre wouldn't be good enough. It's interesting that your observation about GW100 seems to confirm the argument. Then of course there's the question of how important all this is vs comfort of around-ear headphones and all the technical abilities the higher end ones have.

 

[johnn29]

I'll do some more tests between the DT990 and the GW100.

The article also states

Regardless, due to the relatively unnatural “illumination” of the entire pinna by the radiated wavefront (and hence low fidelity to HRTFs that would generate a natural frontal soundscape), headphone playback (especially of stereo recordings but even for binaural) is intrinsically compromised.

I wonder if this has something to do with the fact that when you move rooms or locations - sometimes the HRIR I generate sounds "off". The visual cues are over-riding the auditory ones when you are in the same room - but when away from those visual cues your brain is only left with the sound.

 

[jaakkopasanen]

When you're listening one room (virtualized) in another room (physical) the visual and auditory cues contradict to some degree. How much depends on how similar acoustics the two rooms have. Visual cues are always dominant to auditory cues. This is actually how brains learn to localize sounds based on auditory cues in the first place.

 

[johnn29]

This will be of interest to people in the thread. JVC looks to be finally releasing it's Exofield tech: https://eu.jvc.com/audio/home-theater/XP-EXT1/

It was supposed to come out in April but I guess Covid set things back. It's a $1k price tag. But it uses mics inside the headphones for the customization and some sort of algorithm database. Includes DTS:X and Atmos. I'd be willing to give it a try purely for the convenience of an external box, rather than having to be tied into a PC.

Still not as flexible as Impulcifer due to being restricted to one set of headphones. I really like ANC for my use case.

 

[musicreo]

It was supposed to come out in April but I guess Covid set things back. It's a $1k price tag. But it uses mics inside the headphones for the customization and some sort of algorithm database.

I thought they want to use in-ear mics like shwon on their webpage EXOFIELD" Measurement. Using mics in a headphone and using a database is very dissapointing.

 

[johnn29]

The binural mics are a difficult issue for mass market products. It could be that they offer a proper binural measurement but for the sake of market adoption it's not the default. I think many audio companies are aware of the deconvolution process but in talking to DTS about their solution the biggest barrier is UX. That's why a lot of these products have remained very niche like the Smyth and the ones that are popularized rely on things like photos. I know Sony is looking at potentially playing a game to narrow your personal HRTF. We are just very lucky that Jaakko has given us such an elegant solution at 0 cost.

I just wish there was a box we could buy (like a Raspberri Pi) and it could use audio in over HDMI and output our HRIR. I don't mind using a PC in my office setup, but for my main theater I'd really like to just be non-PC. I still have a HTPC in there for MadVR's tone mapping but it's a bit of a pain to switch between watching on my Shield and resuming via headphone on my HTPC.

 

[castleofargh]

The AES people submit papers that are more often about all types of sound field and or HRTF/psychoacoustic related stuff. It's going to come, the only questions are when? And how far will they dare to go with the well needed customization?

On the other hand, I was already thinking "it's going to come soon" about 20 years ago. but the industry does seem motivated this time. I won't count on audiophiles, most start pulling out a cross and loading silver bullets when they read about equalizer. so DSPs... But I expect even them to slowly get into it after they get to try something that is properly customized for them, even if at first it goes against their deepest false beliefs.

 

[johnn29]

The gaming industry is where it's really going. What Sony is doing with the PS5 is a real big step forward. Once that's bedded in it'll filter down to other products I'm sure. Sony have a strong market in Japan where this sort of stuff is in demand due to the tiny apartments.

Agree on the audiophiles, it's a cult!

 

[mindbomb]

On the talk of headphone choice, rtings actually has been measuring pinna interactions as part of their passive soundstage measurements. (although they treat it as valuable rather than problematic).
https://www.rtings.com/headphones/tests/sound-quality/passive-soundstage

The sennheiser hd650 and grado sr125e seem like they do really well. Likely because the designs keeps the driver relatively close to the head compared to other headphones.

 

[jaakkopasanen]

I did a quick comparison of audio interfaces and binaural microphones is terms of signal to noise ratio. Behringer Uphoria UM202HD has clearly quieter mic pre-amps than Zoom H1n. 9 dB of difference by just using better audio interface. The Sound Professionals SP-TFB-2 mics beat Primo EM 258 capsules with a hefty margin of 11 dB when using Zoom H1n with both. This was unexpected because specs wise the EM 258 is a better capsule. Unfortunately I don't have adapter cables to connect The Sound Professionals mics to Behringer audio interface so can't test that at the moment.

Now it's easy for me to recommend The Sound Professional mics. They perform better and also are easier to work with because they have the silicone sleeve. These aren't even the master series mics! One user has also reported that his Primo capsules have developed significant high frequency roll off during in just a few months.

Behringer Uphoria UMC202HD with Primo EM 258 mics

Zoom H1n with Primo EM 258 mics

Zoom H1n with The Sound Professionals SP-TFB-2 mics

 

[musicreo]

How high was the voltage you used for the mics at the Behringer? There is probably a sweet spot for the primo capsules between 3-10V regarding signal to noise ratio. The other question is how constant is the build quality of the primo capsules or the Sound Professionals SP-TFB-2 ?