Threads of Binaural virtualization users.

[NotInYourHead]

+++
Also, writing this has made me curious about your situation—seeing that you’re calibrating your headphone response to a DF target.
1. Are you using the binaural room response you recorded at your ears together with your headphone response to perform the virtualization?
2. Or Are you trying to take a response that was made for another user or a dummy head (even though the headphones are the same model) and adjust it to better match what you hear?
3. Or Are you simply listening to a standard binaural recording or video through your headphones, rather than using a personalized virtualization captured from your own ears?

I believe it's a bit of both option 2 and 3.
My first foray into speaker virtualization began with ASH Listening Set, a collection of pre-recorded BRIRs using a dummy head in different environments. The creator explicitly said headphones should be equalized to the DF target curve for best results, which I agree with. The set also comes with pre-made HP correction filters (they don't sound right to me personally).
Then he released ASH Toolset, which in a nutshell is a tool that allows you to customize your own BRIR, from "acoustic space", levels of direct sound, room target curves, and most importantly, "Listener type" which can be a dummy head or public databases of SOFA files. Since then I've wondered if strictly following DF curve equalization is necessary or "right" anymore because responses can vary wildly depending on the setup. So far I still think DF sounds great, but I tweaked my existing curve, adding a +3db bass shelf filter for 200Hz and below to make the sound fuller. As mentioned before, I also think DF is way too aggressive on treble reduction and now there's a shelf filter @6Khz, providing a gentle cut. In a way, I think I ended up with tone controls lol
Hence why I believe in-ear mics can solve the problem: I can tailor the FR to my ears and get it right from the get-go instead of fiddling with theoretical headphone targets.

 

[Lion♡]

I believe it's a bit of both option 2 and 3.
My first foray into speaker virtualization began with ASH Listening Set, a collection of pre-recorded BRIRs using a dummy head in different environments. The creator explicitly said headphones should be equalized to the DF target curve for best results, which I agree with. The set also comes with pre-made HP correction filters (they don't sound right to me personally).
Then he released ASH Toolset, which in a nutshell is a tool that allows you to customize your own BRIR, from "acoustic space", levels of direct sound, room target curves, and most importantly, "Listener type" which can be a dummy head or public databases of SOFA files. Since then I've wondered if strictly following DF curve equalization is necessary or "right" anymore because responses can vary wildly depending on the setup. So far I still think DF sounds great, but I tweaked my existing curve, adding a +3db bass shelf filter for 200Hz and below to make the sound fuller. As mentioned before, I also think DF is way too aggressive on treble reduction and now there's a shelf filter @6Khz, providing a gentle cut. In a way, I think I ended up with tone controls lol
Hence why I believe in-ear mics can solve the problem: I can tailor the FR to my ears and get it right from the get-go instead of fiddling with theoretical headphone targets.

Yes, I think most of your views are correct.
Here's a simple and concise statement of my position on your points:

When applying EQ for the purpose of HPCF, the goal is for the headphones to disappear. If the room response you're hearing belongs to someone else (or a dummy head), then another EQ will be applied to compensate for the inconsistent response. This can make it sound somewhat plausible, but fundamentally, it's unavoidable because your physical cues aren't the same. You are not another person (or a dummy head).
So, I'll refrain from commenting on the process and results of EQing to match another person (or a dummy head).
There's a difference between trying to make something that doesn't fit, fit, and making something that already fits, fit even better....
But does "not fitting" mean it's wrong? No, it doesn't. It just means it doesn't fit.
If your definition of "the right sound" is how it compares to the way you actually hear, then you should be hearing the right sound without EQ.
Conversely, if you need to add EQ or seek out and add something else, it means the sound doesn't match how you actually hear at all.

 

[Lion♡]

I understand that everyone participating in this thread uses their preferred DSP individually.
If anyone happens to be using impulcifer, I’ve personally applied a few modifications.
Most of these are tasks that can be done directly in REW and are fairly simple, but I’ve made it so that with just a simple command word, they can be processed instantly within impulcifer whenever needed.

There are a few changes, but two of them are as follows.
The first is virtual bass synthesis.

Of course, this applies to all channels, not just the front ones. Each opposite-ear channel is synthesized according to its own ITD, which means the ITD varies depending on the angle of each speaker.
The concept of mono bass has existed for a long time, but the best scenario is when all speakers at different angles have their own ITD and reproduce the full range.
The virtual bass (using a 15 Hz 4th-order Butterworth filter) also includes ILD.
The crossover uses a Linkwitz-Riley 8th-order filter up to about 200–250 Hz; going beyond that is not recommended.
If you just want to replace the subwoofer region, you can set -vbass=100

And the second is automatic ITD matching.
There are options for e (early), l (late), and a (average), with the default being disabled.
Of course, these small corrections were also possible in REW or Audacity, and surprisingly, there are some differences in how they’re handled. But it’s important to consider that if the common delay is off and the sound shifts to one side, it naturally sounds shifted, which can be annoying (though in reality, real-time brain correction usually covers this).
Whether or not it’s not correct, the response is accurate based on the orientation of the face and body. response is not wrong.
From the moment you adjust it, the perceived spectrum may improve or be restored, but it can also cause a canceled or muffled sound.
High frequencies rely less on ITD compared to low frequencies, but depending on the test tone, the difference can still be significant.
In severe cases, spatial perception such as distance and depth can feel compressed.
Adjusting by 1–2 samples (about 20–40 microseconds) is generally fine.
If the difference is larger (for example, one ear is 250 microseconds and the other channel is 333 microseconds), it’s better to re-record.
The best approach is to record well from the start rather than relying on processing later.

Of course, advanced synthesis is actually easier and more accurate to do in REW, since you can control everything exactly as intended.
However, I included these basic features in impulcifer to make it easier for beginners starting with BRIR to get started more smoothly.

Also,
An outstanding user from Korea has created a True HD Atmos sweep, and I’d like to thank them for it. (Also, thanks to @OCA for providing advice on creating the Atmos sweep in advance.)
It has been confirmed to work well up to versions 7.0.4 or 7.0.6, so AVR Atmos users can easily record using this sweep when using impulcifer.
(The LFE channel is excluded because it’s more convenient to generate it virtually, and it’s unclear how impulcifer handles the low-passed LFE impulse. Also, not many users employ the 9.1.6 layout, and the layouts clearly verified by Korean users go up to 7.0.6, so the sweep was created accordingly. However, if 9.1.6 is needed, we can ask that user for assistance.)
https://drive.google.com/drive/folders/1UoZWQVNONW1GSoxIb7BUwvGqxR82zGv6

 

[NotInYourHead]

Yes, I think most of your views are correct.
Here's a simple and concise statement of my position on your points:

When applying EQ for the purpose of HPCF, the goal is for the headphones to disappear. If the room response you're hearing belongs to someone else (or a dummy head), then another EQ will be applied to compensate for the inconsistent response. This can make it sound somewhat plausible, but fundamentally, it's unavoidable because your physical cues aren't the same. You are not another person (or a dummy head).
So, I'll refrain from commenting on the process and results of EQing to match another person (or a dummy head).
There's a difference between trying to make something that doesn't fit, fit, and making something that already fits, fit even better....
But does "not fitting" mean it's wrong? No, it doesn't. It just means it doesn't fit.
If your definition of "the right sound" is how it compares to the way you actually hear, then you should be hearing the right sound without EQ.
Conversely, if you need to add EQ or seek out and add something else, it means the sound doesn't match how you actually hear at all.

Update: After some A/B tests (not blindfolded), I ditched the DF EQ filters entirely, and it sounds much better. Realism is enhanced, and everything sounds less distant and tinny. I believe it's due to DF's tampering with the 4Khz region (there should be a hump in 3-4Khz region due to ear gain) but ASH Toolset has likely already included ear gain into its BRIR output.
All I have to do now is tune the bass and treble levels

 

[Lion♡]

Update: After some A/B tests (not blindfolded), I ditched the DF EQ filters entirely, and it sounds much better. Realism is enhanced, and everything sounds less distant and tinny. I believe it's due to DF's tampering with the 4Khz region (there should be a hump in 3-4Khz region due to ear gain) but ASH Toolset has likely already included ear gain into its BRIR output.
All I have to do now is tune the bass and treble levels

It’s normal to have ear gain. More precisely, it’s the typical binaural room response—including ear gain—multiplied by the response that flattens the IEMs or headphones you’re currently wearing.
But,

So, I'll refrain from commenting on the process and results of EQing to match another person (or a dummy head).

As I said, I have nothing to say about that situation...

If you’ve tweaked it in various ways and you like it, then that’s good for you.

 

[fcserei]

Finally I could convince my wife to scan my head and I've created a medium resolution (~10k triangles) model and simulated my personal HRTF with mesh2hrtf (9h computation time on a M1 Mac)

I have to admit, personalization adds a lot to the binaural experience.

Previously experimented with lots of sofa files from different databases, and what was apparent, that while all more or less can externalize the source, they have wildly different perceived tonalities.

Below the default generic head HRTF simulation and my personal results from the simulation.

Default:

Personal:

Default:

Personal:

The HRTF is used in the SPARTA binauraliser in 5.0 mode with the added ambisonic room simulation. I don't want to simulate speakers in a room like Impulcifier or the Realizer, I want to put the performance in a real space.
The rest is Supperware head tracker, and Yamaha HP1or Sennheiser HD559 headphones, both EQ-d flat at my blocked ear canal. The HRTF is generated flat at the entrance of the ear canal. There is no need for diffuse field EQ, Harman curve or other sorcery, because my personal diffuse field curve is basically the weighted average of my full sphere HRTF. The frontal stereo channel gets the proper HRTF treatment, and the front and rear reflections are created from the ambisonics room simulation, so they are arriving with the correct weight and HRTF through the binauralizer.

So far I have only a limited time to listen, but the results are encouraging. The personal HRTF solidifies the out of head experience compared tho the general HRTFs, but what's more important, it makes the music somehow tonally correct. I don't feel the need fiddling with the EQ so far.
I feel headtracking is still very important to maintain the illusion, without it the experience collapses if you move, but the big change is that the personal HRTF can maintain not only the position of the stage but also no shifts in perceived tonality in case of big head movements. The whole sound is easy and natural being there experience with regular stereo recordings.

Next step: I will try a 40k triangle full resolution simulation...

 

[katamana]

Just curious, to what extent can you simulate speakers using Impulcifer (with Hesuvi)?

I've tried dozens of measurements with Impulcifer, using a standard stereo speaker set (equilateral triangle placement, 1 meter from ear to speaker). The result is amazing, however, I feel the simulation is kind of wider and closer than real speakers., lacking solid image like real speakers but more like the feeling of headphones (I'm using HD800s btw) without simulation (i.e., left signal only exists on left side, making soundstage wider but to some extent, 'fake'). My ears and brain will definitely tell me it's just a simulation and differs from my speakers (and I feel more reverb than real speakers, or the reason may be a bad capture of reflections...). I tried to reduce the early reflections, but doing this will make the sound drier (of course...) and even more like headphones without simulation. Will it perform better if I set the speakers to, for example, 2 meters, or just narrow the distance of the speakers while keeping the listening point the same? My room is a little bit messy and I think I have to move some stuff out to test...

I've also tried other binaural virtualization solutions like APL Virtuoso (with my customized sofa generated by mesh2hrtf), dearVR, Waves NX, etc. Impulcifer performs the best among all methods, as others feel like dragging the center of sound just out of my head and placing it near my nose. I guess the performance of virtualization really depends on early reflections but I wonder if Impulcifer will do much better as I feel mismatch of soundstage/image between the simulation and real speakers currently.

Btw, thank you for adding the virtual bass function, my speakers will roll off from 40hz and I do need more bass after processing by the original procedure.

 

[myHRTF]

The result is amazing, however, I feel the simulation is kind of wider and closer than real speakers., lacking solid image like real speakers but more like the feeling of headphones

There are several possible reasons for that:

• The microphone capsule needs to be inserted deeper inside the ear canal.
• Issues with measurement repeatability, for example caused by the fixture used to keep the capsule inside the ear canal.
• Limitations of the blocked ear HRTF measurement method as pointed out by Griesinger ("convenient untruth").

Will it perform better if I set the speakers to, for example, 2 meters, or just narrow the distance of the speakers while keeping the listening point the same?

I guess it's a more fundamental problem as pointed out above.

 

[katamana]

There are several possible reasons for that:

• The microphone capsule needs to be inserted deeper inside the ear canal.
• Issues with measurement repeatability, for example caused by the fixture used to keep the capsule inside the ear canal.
• Limitations of the blocked ear HRTF measurement method as pointed out by Griesinger ("convenient untruth").

I guess it's a more fundamental problem as pointed out above.

Thanks! I have three binaural mics, sound professional, microphone madness and samson lm10x. sound professional can be placed deeper than microphone madness but performs worse, still wondering if it's because the angle (I attached a small piece of foam earplug but when inserting the capsule it will not be strictly tangent to ear canal). The lm10x doesn't work very well in my ears as I find it hard to place (and fix) it in ear canal. I'll try more tests and hope i can find some improvements.

 

[myHRTF]

I took a look at the microphones that you have listed: lm10x is far too big to put inside the ear canal. The sound professionals mics also look too big for the job but I might be wrong on this. The microphone madness mic is suitable in terms of size.

Philipp

 

[Curvature]

The most complex part of this is capturing a personal HRTF. I am looking into trying @fcserei's method.

I took a look at the microphones that you have listed: lm10x is far too big to put inside the ear canal. The sound professionals mics also look too big for the job but I might be wrong on this. The microphone madness mic is suitable in terms of size.

Philipp

I don't know of any accessible ear canal microphones. Do you? I have the SP mics as well. They sit inside the concha, just behind the tragus.

 

[katamana]

I took a look at the microphones that you have listed: lm10x is far too big to put inside the ear canal. The sound professionals mics also look too big for the job but I might be wrong on this. The microphone madness mic is suitable in terms of size.

Philipp

Hi! Sorry for not specifying clearly. Actually the lm10x may be the smallest among what I listed. It's only 3mm size and other two mics are both 6mm. The problem of lm10x is the cable is too hard and I find it difficult to place in my ear canal (since it's not designed as a binaural mic but lavalier microphone). For the sound professionals I use ms-tfb-2, and cut the entire silicone hooks/shell so it can be placed within ear canal rather than just outside of opening. Microphone madness ms-bs-8 works well in stability as it comes with a nice foam earplug, however this also stops it being placed at a deeper position in ear canal (should be right at the ear canal opening)

 

[fcserei]

For ear canal mic i use a modified Dayton iMM6. Removed the capsule from the endpiece, extended the wire and put the capsule in a fitting IEM earplug.
It comes with calibration and i made a couple of measurement, and it seems it retains the FR quite well in free field compared to the original setup.
Tried to use it for the HRTF measurements but the room is too noisy, too much room interaction even though I measure < .3 s RT60, and to make a full sphere measurement I need a serious geometry setup.

I will try to use it for validate the simulated hrtf with measured ones in some specific directions.

What is the mic perfect for is headphone correction measurements. It helped me to understand why I like the Yamaha HP1 for binaural. It takes eq very well, distortion is low and very low sensitivity for repositioning both the mic and the headphone.

 

[GregHouse]

Hi.
Is there an alternative to Impulcifer?
Another program for creating binaural room impulse responses (BRIR)?

 

[myHRTF]

Hi.
Is there an alternative to Impulcifer?
Another program for creating binaural room impulse responses (BRIR)?

We happen to do exactly this but without acoustic measurements being required. Instead, we adapt the sound of specific headphone models to your individual HRTF online. Currently, we are in beta release stage and plan to launch stereo within the next two months. Surround is expected to follow by end of the year, too.
You may try it online for free, you just need your headphones and a ruler or similar:

myHRTF Personalized Filter Designer

If required, PM me to get your favorite songs filtered offline for testing with known music.
Looking forward to your feedback.

Philipp

 

[myHRTF]

I don't know of any accessible ear canal microphones. Do you?

Sorry for the delay. The SP mic you have shown is not suitable for HRTF/BRIR measurements since it is too far away from the entrance of the ear canal. In the past, company Knowles provided for suitable mics but I'm not sure if they are still available.

 

[GregHouse]

I found SOFAs included in the Realiser as they are public and "recreated" a "true stereo" crossfeed (without head tracking).

Is it possible to download the SOFAs used in Realiser?

 

[GregHouse]

We happen to do exactly this but without acoustic measurements being required. Instead, we adapt the sound of specific headphone models to your individual HRTF online. Currently, we are in beta release stage and plan to launch stereo within the next two months. Surround is expected to follow by end of the year, too.
You may try it online for free, you just need your headphones and a ruler or similar:

myHRTF Personalized Filter Designer

If required, PM me to get your favorite songs filtered offline for testing with known music.
Looking forward to your feedback.

Philipp

I own a headphone that is in the database.
I tested it and found it very interesting.
Much easier to use than the Impulcifer.

 

[Curvature]

I found this microphone using a Knowles capsule: http://feichter-audio.com/produits/captations/m1/

Very small. Flattish frequency response, but I can calibrate that against my Earthworks M30.

I have contacted the company to request the price.

Edit: Typo.

 

[Zekbone]

I'm not currently a binaural user, but I want to be.

I've been researching this issue for the past week when I decided to finally learn what "Windows Sonic for Headphones" meant on the taskbar. I use an LG OLED TV as my PC monitor and use the streaming apps and my 4k bluray player to watch movies and shows, and it made me question how exactly Dolby Atmos on the native apps or bluray player were actually being downmixed, if it was doing the same atmos binaural stuff as on music streaming sites, or if it was just downmixing to stereo. Given that my TV was not flashing the "Dolby Atmos" notification in the top right, I'm almost certain its the latter.

For most gamers, it probably makes the most sense to just rely on the game's engine based postitional audio and included HRTF mode than to force output in Atmos and then downmixing to binaural with Windows or HeSuVi, but with movies that doesn't work because streaming apps on PC are garbage (Disney+ is 720p and Stereo on the officiall MS store app) and piracy is basically required to view 4k bluray discs on PC.

I've been on a quest to find some kind of gadget that can do in "hardware" what Windows and the Xbox Series X can do in software for movies that doesn't cost 4 thousand dollars like the Smyth Realiser A16. It needs to be able to convince the TV to pass on the full atmos signal, then do some kind of binaural conversion. This video is a good demonstration of the plug and play generic HRTF solutions on Windows. I've found all 3 (Sonic Atmos and DTSX) more or less convincing, so I'm not concerned about which of those 3 standards such a gizmo would do, or measuring my own personal HRTF or whatever for a custom one.

Some Denon and Marantz AVRs describe a virtualization mode for specifically headphones in their manuals, but it would be tedious and expensive to actually buy one then verify its not just doing DSP to an already stereo downmix or if its a "true" binaural downmix. Also, these virtualization modes don't work with bluetooth, and I'd like it to work with my Qudelix 5k which acts as hardware agnostic EQ for easy switching.

Incidentally, Qudelix sells a (channel based) 7.1 surround sound processor in the T71, but I'm not aware of any TV or AVR that actually supports USB DACs. Most devices that can do this are designed for PCs.

The Smyth seems like the only truly bulletproof solution, but 4k dollars for an effect I might not necessarily even like in every movie is obviously out of the question.