Surround Virtualization for Headphones

[zeerround]

What about 1st order ambisoncs (4 ch.) so you can get immersive sound?

Do you have a way to use the Waves NX Quad Stereo plugin in Audacity or Reaper or Plogue Bidule or something?

 

[zeerround]

I never listened to Quad headphones during the Quad era although I listened quite often to the stereo folddown of my Quad albums (matrix and CD-4) with stereo headphones (tried the SQ "decoding" w/stereo headphones only, didn't notice much of a surround effect).

Skimming this thread, does the preprocessing to create a surround sound effect for stereo headphones modify the frequency response of any of the input channels (to simulate what the shape of the human ear does to sounds from various directions)?


Kirk Bayne

Yes, as well as the delay.

 

[Dale H.]

Here is version 2.0 of drag and drop scripts to convert multichannel surround to virtual surround for headphones:

https://drive.google.com/file/d/1uVFbdcCHb0T6kfzq9CiXPdG9WKXeusYc/view?usp=sharing
Edit: 03-14-2021

There is a small bug in the script for converting to Binaural, with Headphone equalization, that only affects 7.1 input files.​

​

There is an erroneous file delete command at line 48 in the .bat file.​

​

You can just delete it, or use the attached corrected version​

​

From the included ReadMe.txt file:

This is version 2.0 of virtual surround for headphones. Now supporting 3d / Immersive
Surround Formats.

This converts stereo, 4.0, 5.0, 7.1, 7.1.4, or 7.1.4.4 multichannel surround
files to binaural audio for headphones, with or without headphone equalization.

The resulting binaural flac files (2 channels 48Khz 24bits) can be played back using any stereo
player with your headphones.

Drag and drop scripts are used with ffmpeg and sox (included) doing the hard work for you.

The ffmpeg and sox included here are self contained, and won't change any existing ffmpeg or sox
install you have.

Drag and drop one or more stereo, 4.0, 5.0, 7.1, 7.1.4, or 7.1.4.4 multichannel surround
files onto:

Convert_to_binaural.bat

to convert the input files to virtual surround (binaural), using the My-IR.wav room+ears
impulse response file

Convert_to_binaural headphones.bat

to convert the input files to virtual surround (binaural), using the My-IR.wav room+ears
impulse response file AND equalize for your specific headphones model using My-Headphones.wav

The output files will be in the same directory as your input files. "_binaural" will be added to the input
file name. For the script inducing headphone response, "_binaural_headphone" will be added to the input
file name.


In the Headphones-Files folder are more than 1200 headphone impulse response files. Copy the one
for your headphones model to this folder and rename it My-Headphones.wav (you can delete the existing
one, it's for Sony MDR 7506 Headphones).

If there is more than one Impulse Response for your Headphone Model, you can try each and pick the one
that sounds best.

The source for the headphone response files is:

ShanonPearce/ASH-IR-Dataset

In the IR-Files folder are 22 different ear+room impulse responses to try.

BBC_IR.wav is from here:

bbc/bbcrd-brirs

Recorded in a BBC room with a Neumann KU100 dummy head microphone.

A1_IR.wav is that same room response, but adjusted for my ears and my HD 800S headphones by
my Smyth Realiser A16

The remaining 20 Impulse Response ear + room files are from here:

SADIE | Spatial Audio For Domestic Interactive Entertainment

The first two being recorded with two different dummy head mics, and the rest with actual
humans

To use the above scripts you'll need to delete the My-IR.wav file in this directory (It is a copy
of BBC_IR.wav) and copy your favorite IR file to this directory and rename it My-IR.wav

In order to help you pick the best room+ears IR file for you, drag and drop on or more
stereo, 4.0, 5.0, 7.1, 7.1.4, or 7.1.4.4 multichannel surround
files onto:

Test_ALL_IRs.bat

For each input file, and output file will be created for each of the 22 IRs in the IR-Files
folder. The name of the IR file will added to the input file name. e.g. for an input file
called "Love Alive Stereo.flac":

"Love Alive Stereo_A16_IR_binaural.flac"
"Love Alive Stereo_BBC_IR_binaural.flac"
"Love Alive Stereo_D1_IR_binaural.flac"
etc.

Test_ALL_IRs headphones.bat

Same as above but adding the headphone response file My-Headphones.wav. Output filenames will
have the name of the IR file and "_binaural_headphone" added.

The folder "Test and Example Source Tracks" has Channel ID tracks in each of the supported surround
formats, 2 through 16 channels, as well as noise test tone track, for 16 channels.

These are all intended to help you pick the best IR file for your ears. Drag and drop one or more of them onto
on of the Test_ALL_IRs, as described above, to check it out.

You should also try some music tracks.

Lastly, the script "Add Headphone Response to Stereo (or Binaural).bat will add the My-Headphones.wav response
to any stereo (or already converted to binaural, withOUT headphone response) file.

Cheers,
Glenn

Thank you so much for this great program! I've been doing a lot of research for the past while trying to find a usable solution for listening to my surround music on headphones. I don't have a multichannel speaker set up anymore because of a recent move. Trying all of the plugin solutions with Reaper, or even VST Plugin hosts requires a steep learning curve (for me anyways), and has been very frustrating. The last few days I was looking for a real-time/live means of listening with headphones and stumbled upon this thread. I'm so glad i found it.

I do have a question about what i just discovered about binaural audio, the crossfeed aspect. I believed that in order to have very discreet sound, that I shouldn't be hearing the right channel (or even the rear channels) bleed into the left channel. After reading up on it I now understand that there has to be a certain amount of this "bleeding" to achieve a natural spatial sound with headphones. It looks like some sophisticated software does allow one to control how much crossfeed is in the final mix. I'm curious to experiment with this issue, to see what it would sound like to have little or no crossfeed, and yet have the rear channels sound like they are behind me in the mix. Is there any way that this could be done with your solution? I do see that when I look at the wave files for the different Impulse Responses that some seem to have a lower volume in some crossfed (?) channels. This is more evident when using the ChannelIDS 7.1 test rather than a music track.
I hope you can enlighten me. Thanks.

 

[zeerround]

I would think that crossfeed is already baked in. The virtualization works by convolving the hearing via binaural mics (in human ears or a dummy head) with the source channel, so you always get 2 channel out.

However if you want to experiment something Like Plogue Bidule would be perfect for that, or insert your favorite audio software here, that could mix a little L intro R and vise versa. I guess sox or ffmpeg could probably be made to do that too, if you prefer the command line.

 

[Dale H.]

I would think that crossfeed is already baked in. The virtualization works by convolving the hearing via binaural mics (in human ears or a dummy head) with the source channel, so you always get 2 channel out.

However if you want to experiment something Like Plogue Bidule would be perfect for that, or insert your favorite audio software here, that could mix a little L intro R and vise versa. I guess sox or ffmpeg could probably be made to do that too, if you prefer the command line.

Thanks for your reply. Actually I would want to do the opposite of what you say, I would like to be able to mix less of L into R and vice versa. But if as you say, the crossfield is "baked in" to binaural to begin with, that may be difficult to do. I did read something though about Logic Pro mixing in Binaural that may allow one to control the crossfield in some way. I guess I'll have to look more closely at the different IR examples.

 

[zeerround]

Well in that case I guess you would want to tweak the IR samples, before convolution. Maybe this would effectively be making the IR head bigger? More shadowing on the further ear (but not more time delay, unless you tweak that as well).

You might get a more informed response here: Recording Impulse Responses for Speaker Virtualization

 

[MidiMagic]

There are several factors in binaural sound:

1. Level differences between the two channels (ears)
2. Delay between the near ear to the far ear (depends on angle of incidence of the sound)
3. Filtering of the signal to the far ear due to head shadowing removing higher frequencies
4. Phase change to sounds from the back half of the room caused by the pinnas (earflaps)

 

[jagxjr15]

Auro-3D for headphones is available on Storm Audio ISP pre/pro electronics (and the Bryston knock off), which are advertised to upmix mono/stereo/5.1/7.1 to Auro-3D over headphones. The Auro-matic up-mixer for Auro-3D over speakers is arguably a state-of-the-art up-mix solution, and only Storm Audio and Bryston pre/pro electronics have it for use over headphones, as far as I know. The positive reactions to Auro-matic over speakers bodes well for the same over headphones if cost is no object.

The Smyth Research A8 and A16 realisers can work sometimes, but from my experience with the A8, it can be a frustrating exercise -- the multichannel image is sometimes seemingly randomly unstable, possibly because the head tracking is done by IR and does not work in the vertical plane, meaning if your head is tilted slightly up or down the image can collapse. When it works it can be great, but it's not a reliable solution to the problem of multichannel playback over headphones, in my experience.

 

[zeerround]

Well maybe we are going a little of topic but I have to disagree with @jagxjr15's comments on the Smyth Realiser (I have an A16 and use it every day). I don't know if the A8 has this setting but a key setting for A16 is:

Menu --> Settings --> System --> HT Settings --> Stabilization = Optical

That should solve "image is sometimes seemingly randomly unstable"

Also I don't find that the (lack of, at this time) vertical tracking is a problem.

Now I almost never use my 13 Genelec speakers, instead using the A16, so as not to disturb others in the house with my surround music. Also, with the A16, I can experiment with 16 (virtual) channels, so 9.1.6 or setups with speakers on the ground (vs. ear level or ceiling) etc.

 

[jagxjr15]

Well maybe we are going a little of topic but I have to disagree with @jagxjr15's comments on the Smyth Realiser (I have an A16 and use it every day). I don't know if the A8 has this setting but a key setting for A16 is:

Menu --> Settings --> System --> HT Settings --> Stabilization = Optical

That should solve "image is sometimes seemingly randomly unstable"

Also I don't find that the (lack of, at this time) vertical tracking is a problem.

Now I almost never use my 13 Genelec speakers, instead using the A16, so as not to disturb others in the house with my surround music. Also, with the A16, I can experiment with 16 (virtual) channels, so 9.1.6 or setups with speakers on the ground (vs. ear level or ceiling) etc.

 

[jagxjr15]

The Smyth A8 has no such Stabilization setting. The fact that Smyth have had to add such a setting to the A16 means they acknowledge having had problems with the A8. The A16 has other issues such as locking up and front/surround imaging swapping, etc. There are serious A16 problems reported on the head-fi.org forum. Smyth have addressed the A8 head tracking by changing to a wired system on the A16 rather than the flakier A8 optical IR head tracking. It's such a small company (Smyth used to brag about having 500 users!) that supporting users with problems can be difficult for them (turnaround can be slow, meaning a serious problem can take months to resolve).

Also, the "experimenting" you mentioned is not something one would need to do if the Realiser worked properly out of the box. And configuring "virtual" speakers is always problematic; the better solution would be to use an up-mixer to generate the image without any in-room speaker-placement "configuration" required by the end-user. And all of the fuss about HRTF makes no sense to me because binaural recording works with a one-size-fits-all dummy head without a personalized HRTF measurement.

 

[zeerround]

Can't speak to the A8. Yes it's a small company. Never the less their product is the best in class. Waves NX, etc. doesn't come close.

In my experience, the A16 surround image is 100% convincing. Just like in the videos, other than the physical sensation of the headphones on your head, you can't tell the your speakers aren't on (from the listening position anyway. They don't get louder as you get closer - not yet 6DOF like VR). Other users here have said their results are not as good, pending professional personal impulse response recording, but perhaps I've been lucky.

Based on its price and limited availability (and the price of best in class headphones on top) it is a pro tool for pro users (rack mount, balanced outputs, etc.).

What I've offered in this thread is an intro to what can be done without spending the money, and without headtracking, but yeah it all depends on your head size, ear shape, etc. matching one of the IR files.

If you're happy with binaural recordings, and they exist for the music you want to listen to, great!

My goal is help more people enjoy immersive surround, so headphone surround virtualization is one part of that. Immersive up-remix is another, and > 7 channel encoding and playback is a third. All of which I am posting about in this forum.

 

[jagxjr15]

Can't speak to the A8. Yes it's a small company. Never the less their product is the best in class. Waves NX, etc. doesn't come close.

In my experience, the A16 surround image is 100% convincing. Just like in the videos, other than the physical sensation of the headphones on your head, you can't tell the your speakers aren't on (from the listening position anyway. They don't get louder as you get closer - not yet 6DOF like VR). Other users here have said their results are not as good, pending professional personal impulse response recording, but perhaps I've been lucky.

Based on its price and limited availability (and the price of best in class headphones on top) it is a pro tool for pro users (rack mount, balanced outputs, etc.).

What I've offered in this thread is an intro to what can be done without spending the money, and without headtracking, but yeah it all depends on your head size, ear shape, etc. matching one of the IR files.

If you're happy with binaural recordings, and they exist for the music you want to listen to, great!

My goal is help more people enjoy immersive surround, so headphone surround virtualization is one part of that. Immersive up-remix is another, and > 7 channel encoding and playback is a third. All of which I am posting about in this forum.

Like you, I am dedicated to surround virtualization over 2-channel headphones. I want to eliminate the costliest, most error and distortion-prone part of the sound reproduction-chain, the speakers and the room. Let's instead concentrate on electronic construction of an ideal room, which is possible only with headphones. Trinnov have shown that it is possible to electronically reposition speakers in a room, and Smyth, with their direct-bass option, have shown that room modes and bass distortion can be eliminated by using headphone surround virtualization.

Right now I have bought a store demo Storm Audio ISP at a big discount and I am hoping the Auro-3D over headphones lives up to the hype. Widescreen Review swears by the Auro-Matic up-mixer as the best available. They need an up-mixer because most UHD blu-ray discs make poor use of Atmos, so they put the native 7.1 Dolby or DTS through the Auro-Matic up-mixer when reviewing movie discs. And they love it. Without the up-mixer all of their expensive speakers, assiduously tweaked for correct placement in their listening rooms, would be wasted. I am betting that the Auro-3D headphone up-mix will work just as well. And (look, Mom!), no HRTF personalization!

My comment about binaural goes deeper than was clear from my post. There are almost no binaural music recordings, so I was not praising binaural for music recording. My comment was to point out the fact that none of the people selling surround virtualization (with an HRTF personalization component) ever mention that binaural recording works PERFECTLY without any such HRTF personalization. This simple fact, that you can effectively capture the whole 360-degree sound-field without any HRTF personalization, calls into question the belief that HRTF personalization is necessary for surround virtualization over headphones. The fact that binaural recordings in general (not binaural recordings of classical music, which does not effectively show what is possible with binaural -- binaural recordings where sound sources are moving and are deployed all around the dummy head are a much better reference) give a near perfect 360-degree sound-field replication without any recourse to HRTF personalization, puts the lie to the belief that HRTF personalization is necessary.

I am coming to believe that all the attention paid to HRTF by Smyth, JVC, et al, is a red herring. It's a marketing ploy to convince high-end audio consumers that they won't miss their big-bucks speakers when listening over (much less expensive) headphones. So you go through this little procedure, and that's why it's just as good as your $100K speakers: it's personalized to your the shape of your head/pinna. But I don't think it matters, in fact. My point was that binaural recordings virtualize a 360-degree sound environment perfectly without any special attention to anybody's personal HRTF idiosyncrasies.

 

[zeerround]

"My point was that binaural recordings virtualize a 360-degree sound environment perfectly without any special attention to anybody's personal HRTF idiosyncrasies. "

My guess is that works for you because your head & ears are a good match for whatever recording head was used for the binaural recording.

Decades of research into sound localization and HRTFs would not have happened if what you posit was true for all, although I will say that a lot of virtualization techniques hope for OK results with some kind of average HRTF, and/or by emphasizing "room" acoustics vs. head an ear effects. Dolby headphones, etc...

Then there's ambisonics... Ambisonics is currently enjoying it's widest usage ever, as the main surround virtualization for VR headsets. This is mainly because it's computationally easy to rotate etc. as your head moves.

 

[jagxjr15]

"My point was that binaural recordings virtualize a 360-degree sound environment perfectly without any special attention to anybody's personal HRTF idiosyncrasies. "

My guess is that works for you because your head & ears are a good match for whatever recording head was used for the binaural recording.

Decades of research into sound localization and HRTFs would not have happened if what you posit was true for all, although I will say that a lot of virtualization techniques hope for OK results with some kind of average HRTF, and/or by emphasizing "room" acoustics vs. head an ear effects. Dolby headphones, etc...

Then there's ambisonics... Ambisonics is currently enjoying it's widest usage ever, as the main surround virtualization for VR headsets. This is mainly because it's computationally easy to rotate etc. as your head moves.

No, binaural works for almost everybody, not only me, and engineers can produce all of the HRTF data they want, but the proof is in the listening. Even in the Smyth A8, the supplied test signal gives you (right out of the box) great in-room virtualization without any HRTF measurement/correction: it nearly perfectly places the signal at each speaker position, one at a time, with no measurement of your room, your speakers or of your HRTF.

There has been lots of work done, even at accredited universities, on so-called 'paranormal' phenomena, but the fact that lots of people, who should know better, are working on it doesn't mean it's not a waste of time, money, and effort. It makes sense that the HRTF would need to be taken into account for surround virtualization, and I believed it did, but then I started thinking about my experiences with binaural audio (and related experiences) and my experience was not consistent with the Smyth party line. How could generic binaural produce such detailed, extremely specific 360-degree sound imaging with NO personalization, if the personalization was critical to the experience

I have been an obsessive audiophile since around 1979, and I've been to lots of high-end dealers and attended high-end hi-fi shows in the USA and Europe. I remember sitting next to Martin Colloms at the Heathrow Show as he tried to explain to a skeptical listener how one power amp was better than another because of its "rhythm and pace" and that in time the truth would be revealed to him if continued to believe, without evidence, that power amps had a "sound."

The Smyth Brothers advised against using the default A8 config because, they said, it would sound bad compared with your own personalized setup. I actually tried using the default config and I thought it sounded pretty good, and I've heard plenty of systems that I could not afford to buy! I didn't think my personalized config sounded any better than the default. And it's kind of an open secret that many A16 users are using the default config and liking it; they thereby get their immersive headphone audio without going through the prescribed elaborate religious ritual of the cult of Smyth personalization.

BTW, there is a low cost way I happened upon to get decent surround virtualization on CATV, using only the RCA outputs on the CATV set-top box. The Sennheiser RS175-U does remarkably well at getting the sound out of your head and into the room. It won't give you a full 360 sound field (seldom are sounds projected behind you) but it makes TV watching (with not exactly pristine stereo audio) much, much better than it should at such a modest cost ($200+s/h on Sennheiser.com). The center channel moves out of your head into the TV and the L and R have detail, width and depth and image integrity. I cannot recommend the RS175 (without the -U suffix) because my sample had very weak bass, even with the bass boost engaged, but the -U version corrected that. If you want to try surround over headphones, but are reluctant to spend time or money experimenting, you might want to try the RS175-U. It uses a wireless RF transmitter, but to get the 'surround' to gel properly, you need to be seated in front of the TV, as you would expect. The RCA input worked much better for me than the TOSLink. And there is no setup/calibration. My old wireless TV headphones (which lacked any 'surround' effect) died a few months ago and I didn't expect much from the new RS175-U, but I was very pleasantly surprised by this very inexpensive system.

 

[stevendive]

No, binaural works for everybody, not only me, and engineers can produce all of the HRTF data they want, but the proof is in the listening. Even in the Smyth A8, the supplied test signal gives you (right out of the box) great in-room virtualization without any HRTF measurement/correction.

Just wondering, do you always have visual stimulie with the binaural?

For me so far, any binaural has been disappointing and especially poor at the front with anything I've tried with excellent 'phones or earbuds.

 

[Bob Smith]

I've found a gizmo from EPOS | Sennheiser has made me very happy with rendering 7.1 through stereo headphones: their GSX 1000 Audio Amplifier. This thing's intended for gamers, so there's an intercom channel that's easy to ignore. On the Mac you do need to delve into Audio/MIDI Setup to get the Mac to understand it takes multichannel on first use; after that, it needs no special prep.

No doubt the Smyth A16 Realizer is far superior, but I get full surround (including rear channels) that's entirely satisfactory for casual listening or movie viewing from the GSX 1000. Evidently it's got a Cyber Monday deal going for it today... $189.00

https://www.amazon.com/dp/B01LDTP484/ref=cm_sw_em_r_mt_dp_YVZGJRR0VZW59PATJ7E5

Sure, it's running on some population-averaged HRTF. That having been said, it works well for my friends and I. My expectation is that there will be a lot of more sophisticated devices at lower price points joining the A16 in the next year.

Bob

 

[Soundfield]

No, binaural works for everybody...

Oh no it doesn't! It doesn't work for quite a lot of people. It certainly It doesn't work for me. Never has. I've tried lots of incarnations of it over the years. I can discern no great improvement over stereo headphone listening (which I also find unpleasant). All I get is 'in the head' sound placement with absolutely no front to back separation whatsoever.

 

[jagxjr15]

Just wondering, do you always have visual stimulie with the binaural?

For me so far, any binaural has been disappointing and especially poor at the front with anything I've tried with excellent 'phones or earbuds.

Try closing your eyes when listening. Binaural only works for me if I close my eyes while listening -- and it's night and day -- eyes open, stereo only -- eyes closed, the full 360 immersive binaural effect! Also, it may depend on the choice of headphones: the Stereophile review of the AKG 1000 headphones said that they produced the binaural effect for people who didn't get it from any other headphones.

 

[jagxjr15]

Oh no it doesn't! It doesn't work for quite a lot of people. It certainly It doesn't work for me. Never has. I've tried lots of incarnations of it over the years. I can discern no great improvement over stereo headphone listening (which I also find unpleasant). All I get is 'in the head' sound placement with absolutely no front to back separation whatsoever.

Well, it works for the vast majority of people. It didn't work for me either, until I closed my eyes while listening. Also, it may depend on the choice of headphones: the Stereophile review of the AKG 1000 headphones said that they produced the binaural effect for people who didn't get it from any other headphones.