Stereo Crosstalk Cancellation (XTC): A review by STC. And trying out uBACCH, ambeo.one, and AmbiophonicDSP VST plus Foobar foo_dsp_ambio plugins.

Hey everyone, it has been very busy as expected through September here at the Archimago household. As a result, I haven't been able to spend as much time on the audio hardware hobby (still very much enjoying the music of course!).

Over the years, you might have come across the term "crosstalk cancellation" (XTC) for speaker playback, words like "ambiophonics" or acronyms like "BACCH" may have passed by in your readings (for example great discussion on Reddit). For this blog post, let me ask an outspoken advocate of XTC to join for some guest content to discuss the reasons and techniques, and later on, we'll talk about a few DSPs one could try at home to experiment...

Take it away STC for Part I, I'll be back with STC for Part II, then I'll talk about trying some of the DSP options in Part III:

Part I: An introduction to XTC (by STC)

Crosstalk cancellation is an attempt to deliver stereo sound as it was meant to be played without the interaural crosstalk errors with loudspeakers (see Keele, 1986 for a technical exposition). Many listeners may not realize how much realism their high-end stereo hi-fi system is lacking until they hear a system with crosstalk-cancelled (or XTC for short) playback. It is like comparing a photo of a window with a view to an actual window looking out at the scenery. Or like watching Avatar in 2D versus 3D.

Is XTC the same as immersive or 3D audio? No, it is not, but the distinction is now slowly blurring with spatial audio incorporated with immersive audio. Immersive or 3D audio produces sounds from different directions providing envelopment with the aid of additional channels and loudspeakers while XTC produces the sound as if there are tens of speakers creating the frontal stereo stage extended across 180° if it's there in the recording. Most newcomers expect a bigger stage with XTC and often wrongly adjust the parameter to force one. XTC only applies to the interaural crosstalk errors in stereo reproduced via loudspeakers. It is a correction system and can enhance the sound unlike MQA or Dolby Pro Logic. It can be used with immersive/multichannel content as well.

In my opinion, immersive audio is good when you have visual cues to help your brain create the sonic scene. Formats like quadraphonic can create a very natural 3D sound with XTC, that would be great for home theater setups. It can produce sounds from anywhere in the 360° circle around you, as if the sound sources are in the room with you. Sony's 360 Spatial Sound Mapping with their 4 speakers system is heading towards that direction.

Crosstalk cancellation is not a new concept. The most basic way to do crosstalk cancellation is just to put a physical partition between the two speakers, but that is not practical. Atal & Schroeder's "Apparent sound source translator" patent filed in 1962 suggested a technique with signal processing. Despite being around for a long time, audiophiles have ignored it presumably because some think that any signal processing is "bad" for high fidelity sound.

[Ed: of course, rational audiophiles here at the Musings don't hold on to such silly phobic myths.😉]

One example of an early consumer XTC device is Carver’s Holographic line of processors, preamps, and amplifiers which came out in the early 1980's based on his 1979 patent (see Carver's Q&A ad on Sonic Holography on page 69). 

XTC's intent is not to change the original sound. The goal is to conserve the sound waves that are intended to be presented to the ear on the same side, but reduce the crosstalk that reaches the opposite ear. You can see the illustration here.

The impractical physical mattress-between-the-ears approach. ;-) On the right we see some old photos from back in the day when they were experimenting with this (Keele, 1986 paper).

In some ways this is like listening with headphones since each ear only hears the corresponding speaker. You may think then that the sound of speakers with XTC is like the experience of listening with headphones, but that would be wrong. Listening to stereo sound with headphones is actually unnatural because it bypasses the role of the pinna, thus creating a soundstage inside the head.

To understand why, we need to know how our ears and brain work and how stereo creates an illusion of sound images.

The purpose of stereo and multichannel formats is to re-create the sound scene of an event as if you were there in person. The most accurate way to do this is to record the sound with microphones inside your ears and listen to the recordings with headphones. These recordings are called binaural. Assuming the recordings are done in high fidelity, they will sound exactly like the original event to you and only you. Some users may perceive a closer realism if their Head-Related Transfer Function (HRTF) is similar to the person's (or dummy head's) HRTF that made the binaural recording. HRTF is a result of how one's head and ears (pinna, ear canal) interact with the sound that arrives towards us, therefore for each of us, the effect will be somewhat different. Individual parameters include Interaural Time Difference (ITD), Interaural Level Difference (ILD) and pinna frequency shaping among perhaps other complexities. Readers are advised to look up the HRTF topic elsewhere (such as this paper) for a better understanding.

An example of the variations in frequency response among 8 individuals due to HRTF. See here for source. Notice most of the variation happens in the higher frequencies.

For any one of us, when sounds are presented around us horizontally, the head/pinna changes frequency response to help with localization starting from ~1.5kHz.

Stereo attempts to re-create the scene by capturing the ILD (and sometimes ITD). Since the sound waves are coming from loudspeakers, they will be modified by the pinnae of the listener which give additional cues to the brain to localize the direction of the sound. This sound is more natural compared to stereo listening with headphones since three separate cues can be delivered to the brain - ILD, ITD and frequency shaping by the pinnae. Headphones do not replicate the pinna's role as the sound is delivered directly to the ear canal without natural modifications by the pinna. Unlike headphones though, the crosstalk errors limit the realism due to a constant secondary sound heard by the opposite ear, again, this is the interaural crosstalk error which will always present as a delayed sound of ~220μs (with maybe -3dB attenuation), damaging the localization and blurring the sonic image.

There are stereo width enhancement techniques that aim to create a more spacious and realistic sound stage from two loudspeakers. However, most of the existing methods, such as SRS, compromise the fidelity of the sound and introduce artifacts or distortions. Rather, crosstalk cancellation/XTC has emerged as a high-fidelity solution that leverages digital audio processing. Crosstalk cancellation eliminates the unwanted interference between the left and right speakers, allowing each ear to hear only the intended sounds. This creates a binaural effect that simulates the natural hearing of sound sources from different directions as captured in the stereo recording.

There are various products that implement crosstalk cancellation using different algorithms and names. Many products use the Recursive Ambiophonic Crosstalk Elimination (RACE) algorithm, which was developed and made public for free by Ralph Glasgal and Robin Miller in their 2006 paper. The RACE algorithm is based on the principle of "ambiophonics", which is a term coined by Glasgal at least since the late 1980's to describe the binaural reproduction of two-channel sound using loudspeakers. The term ambiophonic therefore is meant to distinguish this crosstalk cancellation approach from the conventional stereo ("stereophonic") playback, with its inherent limitations in creating a realistic soundscape. The Ambiophonics Institute was a pioneering organization that promoted and supported the research and development of ambiophonics and related technologies.

A picture of Ralph Glasgal's sound room circa 2019.

The Ambiophonics Institute was also where Edgar Choueiri of BACCH was involved, working alongside Ralph to improve the crosstalk cancellation. Ralph’s RACE algorithm is incorporated in many other products. BACCH evolved using a different approach called Band-Assembled Crosstalk Cancellation Hierarchy through impulse response convolution.

[Ed: we can see convolution techniques used in various papers over the years such as this one, or this one. Here's an article from 2010 showing Choueiri's "ambiopole" speaker arrangement discussed below, and one of Choueiri's early videos more than a decade ago about the technique and the Princeton lab. Repository of publications here.]

RACE uses the original signal and inverts it and reproduces it delayed via the opposite channel speaker, so that the secondary interaural crosstalk from the first speaker and the inverted signal arrive at the same time at the unintended ear, thus reducing crosstalk level. This process is repeated until the signals reach above 60dB attenuation. This can be achieved by anyone with a DAW using delay and an effect such as "Ping-Pong" plugin that allows adjustments in microseconds. [Ed: for those using Reaper as a DAW, check out this discussion.]

The original RACE block diagram. Notice processing within the 250Hz-5kHz band, recursive 60-100us delay with inversion for each pass. Bass <250Hz and higher frequencies >5kHz are typically unaffected by the DSP.

The theory of cancellation is well documented by Ralph Glasgal and Robert Miller's Filmaker website who also worked with Glasgal in the early stages. Since early 2023, Glasgal (92 years old in 2023) has closed the Ambiophonics Institute but the archival contents are still available here. Check out Glasgal's Ambiophonics: Beyond Surround Sound to Virtual Sonic Reality book for more (free PDF here). Also, see this 2019 article with a visit to Glasgal including some fascinating background on his life.

BACCH is said to use different levels of cancellation based on frequencies as head shadow attenuates higher frequencies more than the lower ones. BACCH's approach is similar to the RACE "ping-pong" effect but implemented in a convolver using impulse responses for the contralateral signal attenuation.

In the end, what’s the audible difference between BACCH and RACE based products?

Unless you have access to the various products implementing XTC from SoundPimp (demos here), uBACCH, Neutron MP, etc. you are at the mercy of opinions and subjective claims; as usual, it's best to experiment and try for yourself.

IMO, the best cancellation can be achieved without spending $$$, such as with a DAW or music player with DSP. The trick is how to get the tonal accuracy. Audiophiles can make bees fly around the head with free XTC software and some know-how. The ability to achieve this using standard 2-channel audio content is something that should not be expensive but one will need to make an effort to understand how this works so you can implement it well in your own room and own system.

Part II: Benefits and the ambiophonic set-up (by STC & Archimago)

In this next section, let's talk about what we expect to hear from a properly set-up XTC/ambiophonic system.

STC's system: Sound Lab electrostatic speakers, powered by Crown XLS amplifiers. Custom implementation of RACE.

Notice that the speakers in an ambiophonic system are typically placed not in something like a 60° (ITU-R BS.775-3) equilateral triangle arrangement. Rather, they're much closer at around 20° apart in front of the listener. 

We can see this kind of closely-spaced arrangement in the papers by Glasgal like this one showing the most basic "ambiopole speaker pair" spaced 20-30° apart fed with standard 2-channel input through the XTC processor:

Notice that there's also the potential to use captured concert hall impulse responses to re-create ambiance using various surround speakers, spaced every 15-30° for example. (You can see some of this work capturing concert hall impulse responses in presentations like this from Farina.)

As you can see in the paper, there are many other layouts and processing options that can be considered such as adding a rear ambiopole speaker pair (see "360° Localization via 4.x RACE Processing" AES 2007) which can be used to convey multichannel like quad or 5.1 content. Glasgal also discussed the use of height channels in what he called "periambiophonics" with captured "upper hemisphere" soundscapes.

While I will stick with discussing just the 2 speakers up front, STC indicated that he also has some small Panasonic "rear ambiopole" speakers, fed a similar signal as the front, to expand the 360° panorama. Also, he has a stack of amplifiers powering his surround speakers when playing music in his "virtual concert hall".

Stack of Line5 A960 amps...

You might be wondering why the 20-30° closely spaced front speaker layout. This is for psychoacoustic reasons since human frontal localization via the pinna is poor compared to the sides (the pinnae as direction-finding mechanisms are most sensitive to sounds coming from 30° off-center). With both speakers closer up front, the crosstalk correction mechanism can assume minimal pinna coloration and relative independence from the listener's head size and shape. Furthermore, this speaker arrangement could inherently reduce side room reflections, improve center imaging, and reduce comb filtering. 

Here's STC's take on some of these elements:

1. Reflections - Glasgal was aware that RACE XTC reduces the impact of sonic reflections and he shared his thoughts about it within the archives. RACE is a recursive signal of in-phase and out-of-phase effects separated by ~80μs so the reflections can get cancelled by itself. BACCH is also recursive since that's what the convolution engine does.

2. Comb filtering - comb filtering will always exist when two sources produce the same signal. In stereo you are always hearing 220μs of comb filtering and no one complains. In ambiophonics, the cancellation is supposed to reduce or eliminated the comb filtering as the ears only hear a reduced level from the other side. Of course, the comb filtering is already happening from the two signals traveling from the speakers to a certain extent. I have done many listening tests with music students, non-audiophile music lovers and musicians on my system, all of them preferred the XTC version as the sound is more detailed and refined.

3. Centre Image - This tends to be excellent with ambiophonics. My own XTC will be slightly different from a DSP like AmbiophonicDSP [Ed: we'll talk about this below] but listeners (music students) have preferred the XTC version as it sounds more natural compared to typical stereo speaker placement. The question is whether it's the same or better than mono vocals or center channel solo vocals. Listening tests need to be done and since my children are grown up now, I do not have virgin ears to explore this further!

As usual, the results can vary between systems. For example, my setup is literally 1.5 x 1.8m of radiating electrostatics surface listening from about 2.4m away which could be very different from what you use. I am very familiar with AmbiophonicDSP and despite giving all the parameters of my listening system, some users cannot seem to achieve the intended sound quality with their systems. One guy who was a recording engineer was shocked at how well AmbiophonicDSP worked on my system. At that time I was using Harbeth speakers, but he could not get it done correctly in his studio. So it can vary! Just too many variables and sorry to say this, but I am beginning to suspect not all audio engineers are equally made. Some cannot even seem to get the speaker angles correct. 😂

[Ed: Indeed on that last point! ;-) Sadly, some audio engineers apparently can't even tell that their DR4 dynamically compressed mastering sounds like crap...]

Also see "What to expect to hear with Ambiophonics" by Robin Miller.

Another perspective on crosstalk cancellation from Floyd Toole; the thread is interesting as it proceeds into discussions of BACCH and back-and-forth with Choueiri.

A comment about listening to audiophile vocal recordings with speakers spaced closer from Glasgal.

Part III: Try XTC for yourself (by Archimago)

Thanks STC for the information above! Great stuff and I hope the background on crosstalk cancellation (XTC), ambiophonics, and BACCH intrigues you, dear audiophiles. I know the terminology might be quite alien to many readers although I think the concepts should not be difficult to grasp.

In Part III then, let's get into it and see if we can try this for ourselves. First, it's important to space out the speakers so they're close to that 20° angle to the listener. For me, in the living room, I have some small Tannoy Mercury mX2 speakers that I can move around temporarily like this:

As you can imagine, normally these speakers would be on either side of the 65" TV which is connected to my gaming computer. The speakers are powered by a Denon AVR-3802 receiver from decades back which still sounds excellent. You can see on the screen that I have JRiver 64-bit running with the "Universal BACCH Filter" (uBACCH) plug-in window open. The software I'll be discussing are all Steinberg VST plugins which can be used in digital audio workstations (DAW) or playback software supporting this standard like JRiver.

First, let's go to the download/purchase link for BACCH Labs' uBACCH where you can find a 14-day trial of the VST plug-in for Mac or Windows. Install the file. Afterwards, on my Windows machine, using JRiver's DSP Studio, I click on "Manage Plug-ins..." and find the uBACCH directory where the .vst3 file lives (in my case, "C:\Program Files\Common Files\VST3\BACCH Laboratories"):

Now open up the uBACCH window and you can play with the very simple "u-BACCH Angle" setting to find the best value for your set-up:

My speakers are 20° apart (+/-10° off center) so putting that angle for the setting turns out quite nicely as expected. Notice the uBACCH filter also has the "Tuning" button which you can click on and that will play some pink noise so you can fine-tune the effect. Unfortunately, there's a bug somewhere and JRiver needs to be playing music in the background in order to hear the pink noise sound. Here's what it looks like to have the L channel playing the pink noise while I slide the angle width adjuster around:

What you want to hear is the lateral R/L sounds emanating from just beyond that corresponding ipsilateral ear, and as little sound (crosstalk) as possible heard by the other ear. BACCH Labs has a video discussing this.

To test out how well the plug-in works, have a listen to some binaural content. One of the tracks to check out is "Edgar (Choueiri) The Barber" found on Chesky's Sensational, Fantastic, and Simply Amazing Binaural Sound Show! (Binaural+ recording**). This binaural track should sound "3D". You should be able to hear the spatial movement of the scissors corresponding with the narration, and if it's working well, you should even hear the scissors moving behind your ears during those portions of the demo.

So that I can listen to music from a variety of sources (including Roon), JRiver has a handy WDM driver you can turn on so all your computer audio can be piped into the output device through the DSP:

24/48 sounds great already. Feel free to process at 24/96 if you prefer. As usual, for DSP precision, 24+ bits are beneficial. Samplerate is not nearly as important.

You can set Windows' default audio output to the "JRiver Media Center" device as I did above. With this, I was able to achieve a similar experience playing the Virtual Barber Shop off YouTube hearing a wide spatial field, reasonably good tonality, and portions where the voices and sounds seeming to be coming from behind my head; it's an older recording so resolution isn't the best.

Overall, I found uBACCH worked quite well and sounded good. Do not confuse BACCH Labs who sell uBACCH (US$399) with Theoretica Applied Physics and their BACCH products currently only for MacOS along with more expensive standalone versions.

US$400 for a software audio plug-in isn't cheap though, so let's see if we can find some free/inexpensive options.

Here's a free one, check out ambio.one by Weldroid, released initially in 2010. The links on the page might not work to download from any more, so try this link. Note that it's a 32-bit plug-in so it'll only run on the 32-bit version of JRiver.

While this one did expand the width of the soundstage, it definitely wasn't as good as uBACCH for the "behind the head" effect or impression of being spatially enveloped when listening to my music. Feel free to play with the control knobs to see if you can fine-tune the results to your system. Hey, it's free! ;-)

Finally, let me direct you to this 32-bit VST plug-in that I fear unfortunately might not be available to purchase anymore - AmbiophonicDSP (only US$10 last published price, released initially in late 2009 - link provided for info even though I think no longer sold).

It works very well (as mentioned above by STC) and I quite like the "Jazz / acoustic" preset in the drop-down menu. STC suggested increasing the "Space" knob to further expand the spatial envelopment. I found that going up to 3-4 was good but beyond that I started noticing some mild distortion with sustained decays in my system, Also, push the "Zentrum" knob higher if you want to strengthen the central image, especially if voices sound too recessed. YMMV.

[Note from STC: I was using JRiver and AmbiophonicDSP but when I upgraded to JRiver 25, I had issues with it and also the virtual concert hall using surround speakers (or "domestic concert hall" - DCH). I did implement DCH with JRiver and there is a complete guide in the old Ambio website. The best part with JRiver was I could use various inexpensive USB DACs and implement DCH using their Zones. Unfortunately, after the upgrade to version 25 the sync wasn’t possible. JRiver was unable to resolve it and thankfully forced me to finally learn to use a DAW to created my own crosstalk cancellation procedure!]

Clearly the DSP widened soundstage, achieved good clarity, I did not hear any distracting coloration (unless parameters moved to extremes), and that "surround" behind-the-head effect was at least as good as uBACCH in my simple Tannoy set-up. Make sure to have a look at the user guide for more details and recommendations on the values of those settings; for example, depending on the listener-speaker angle, set the delay closer to 250µs as you get towards 60° if you're not able to pull speakers together. 

Beyond just binaural samples like the virtual haircuts, when listening to music, it's easy to hear the pleasant, natural, expansion of the soundstage. For example with classical music, it can sound as if one is sitting and watching the orchestra unfold in front with the various instrument sections - I really enjoyed Reference Recordings' recent Tchaikovsky: Symphony No. 5; Schulhoff: Five Pieces for String Quartet (2023, DR12). At other times it can be startling to hear sound effects in some mixes I didn't pay as much attention to before. For example the rain and thunder at the start of aha's "Crying In The Rain" (off East of the Sun West of the Moon) is rendered with a jarring presence of the raindrops falling around me, nice expansion of the soundscape with the thunder claps apparently above and behind my head.

Vocals like Adele on "Make You Feel My Love" off 19 sounds well defined in space as a tighter center image, the voice seeming closer and more forward than I'm typically used to. The tonality of typical audiophile female vocals like Lori Lieberman's Truly (2022, DR11) and the piano solo "Sonata in Darkness" (from The Batman) sounded pleasing as well. Also, check out some of the QSound encoded CDs out there - Madonna's Immaculate Collection, Roger Waters' Amused To Death, Sting's Soul Cages, Pink Floyd's Pulse, even Def Leppard's Yeah!. They really "pop" as intended.

You might also want to have a listen to The Cowboy Junkies' Trinity Session (1988, DR12) which was recorded with a single Calrec ambisonic omnidirectional microphone (more on the album here). In fact, have a look at the list of ambisonic recordings out there in this FAQ which includes Holly Cole's Girl Talk (1990, DR14). It can be quite surprising the amount of spatial depth and details captured in these. (Whether you're pro or against using ambisonic microphones, here's an interesting discussion.)

Although I have not tried it extensively yet, for those using foobar2000, there's foo_dsp_ambio plugin released in late 2011. It's no longer under development, here's an alternate download link. Good alternative if you don't want/have VST processing; I'll probably use this most often on my desktop machine.

Within the reasonable settings I used, I didn't detect unpleasant tonal distortion or "phasey" anomalies across the audio spectrum using these DSP plugins which I've heard some listeners complain of. Make sure your processing/playback is at least 24-bits and the parameters are optimized for your system.

Enjoying the sound of crosstalk cancellation? 😯

Summary...

It has been fascinating jumping into the world of crosstalk cancellation and ambiophonics these last few weeks while chatting with STC about it. Appreciate the collaboration, STC!

There have been all kinds of products out there already featuring stereo width expansion over the decades. Most of these play with phase to artificially widen the stage rather than proper crosstalk cancellation (XTC) to remove the unintended sound reaching the opposite ear. These days, in the 21st Century, with the ubiquitous computational power of modern computers where even a Raspberry Pi should have no problem implementing 2-channel RACE in hi-res, we can do XTC in DSP with good results like the VST plugins above so long as the speakers are arranged appropriately and we tune the settings to match.

I've heard criticisms around using crosstalk cancellation as being unnatural because in the studio, the engineers and artists typically are not recording and mixing while employing XTC. I think this is a true and fair statement. The sound achieved is quite likely not "as the artist intended" or at least "as the artist heard" in the studio. Having said this, I'm not sure I actually care much about what the "artist intended" these days especially with pop studio productions. When it comes to rendering soundstage, I care more about whether the reproduction sounds interesting, entertaining and "cool"! Employing XTC, Madonna's "Ray of Light" is almost psychedelic with AmbiophonicDSP and I've not heard New Order's "Blue Monday 1988" sound so expansive before. I'm pretty sure I don't care what Madonna or New Order "intended" all those years ago in the studio. 🤔

Due to the front speakers being close together, they act as a common/center channel if the listener is further off to the sides. The sweet-spot for hearing the enveloping "surround" effect is reserved for those sitting centrally and narrower than typical ~60° stereo arrangement so it won't be as satisfactory if you have 3 or 4 people sitting side-by-side. I suppose one could arrange a few rows of chairs behind each other to have a few listeners within the central axis, but that would be a little odd in most domestic listening rooms - my wife and kids will have none of that.

Alas while I quite enjoyed the sound, as a pragmatist, unless I have an ambiophonics-specific set-up in one of my rooms, I would not be able to keep my speakers in that exact 20° orientation since all my sound systems here at home feature a screen and the speakers typically positioned on either side of the TV/monitor. Furthermore, since I've been listening to quite a bit of multichannel content these days, a standard 5.1 or 7.1 "bed" speaker layout with or without height channels would still work best.

[Note from STC: For home theater, transparent acoustic screens can be used to visually hide speaker placement. However, when vision is involved, auditory cues take a secondary role. There are research papers (like this) exploring changes in sound quality perception when accompanied by video. Furthermore, most of the positional information is predetermined by the eyes and the surround sound reinforces that localization.]

For those Tannoy speakers in my main living room, one solution I can do is to keep the speakers on either side of my TV but sit further back to listen. With the speakers 5.25' apart, sitting on the sofa 13' away will give me +/-11.4° or 23° between speakers and listener which is within the 20-30° layout. When I do this and activate the DSP, the sound is not as clear, the center image not as stable, nor is the "surround" effect as convincing as sitting closer due to the suboptimal layout and room acoustics, but I still experience a good amount of soundstage envelopment. A reasonable compromise, I think!

It's pretty cool watching a movie with XTC to expand the auditory "surround" experience using 2 speakers:

Excellent sounding "Diva Plavalaguna" scene (start at 1:27) from The Fifth Element.

For completeness, I want to mention that XTC has also been implemented in devices like the miniDSP products. Polk has their Legend L800 speakers (which I had a listen to at RMAF2019). On a web browser, Aria3D is also based on the RACE algorithm; I see the page is still up but I'm having difficulties accessing the demos on Chrome these days, you might have better luck. The same founder has another project called Magic Audio which works.

Of course check out the BACCH systems which have been promoted to audiophiles over the years. The recent TAS review of the BACCH-SP provides a good description. Starting at US$20k for a system like that without DAC functionality clearly represents quite a significant investment, but it does allow for further listener customizations like measurement of HRTF, head-tracking, and special headphone filters. BACCH4Mac is more affordable starting at about $1,000 for what looks like the generic DSP playback only. I wonder what is the difference between something like a Mac Mini running BACCH4Mac Intro and uBACCH VST playing in JRiver given the price difference.

[By the way, with "universal" BACCH VST plug-in, I was curious if I could measure the difference between different settings like 20° vs. 60°. Yup, we can capture an impulse response, overlap and show the change:

I suppose someone with BACCH4Mac Intro could capture the impulse response on that with the same angle settings and compare to see if it looks like uBACCH VST above.]

As usual, I'll leave it to you to experiment, audiophiles. I hope this article provided a reasonably comprehensive launch pad to help understand the technique and points you to some interesting audio experiences if you have not tried this before.

You might be very surprised by what you hear in your existing 2-channel audio library!

Further reading: A Psychoacoustic Rationale for the Subjective Evaluation of Stereophonic Sound Systems by Reviewers and Audiophiles (Ralph Glasgal)

Also check out Home Audio Fidelity's X-talk Shaper as another great option for DSP processing.

--------------------

** I want to add an extra comment here about using binaural recordings and listening with the ambiophonic set-up and processing.

Strictly speaking, even though they can sound very good with ILD and ITD information, binaural recordings are not the most proper content for best reproduction through the ambiophonic system. Binaural recordings are done typically with a dummy head + pinnae which are supposed to represent some kind of "average" human HRTF.

Neumann KU 100 Dummy Head binaural mic ~$9.5k.

With ambiophonic listening, we would be using our own ears/pinnae, so we actually do not need those pseudo-ears imparting pinna frequency shaping in the recording. This is why if we look at Glasgal's 2003 paper, he showed a prototype "ambiophone" microphone that looks like this:

Notice that the mics would be inserted where the holes are on either side of that central "head". The top in particular is tapered to reduce capturing ceiling reflections. A pair of these can be used to capture front and rear hemispheres. See Glasgal's comment here. More on the ambiophone from AES 2001.

This means if we listen to a binaural recording with the HRTF of a person or dummy head including pinna coloration, the tonality sounds bright with speakers and thru ambiophonics. You can hear this on pure binaural recordings like Stax's The Space Sound CD, the binaural mix of Jean-Michel Jarre's Oxymore, Lou Reed's Street Hassle among others over your speakers.

Interestingly, Chesky's "Binaural+" Series recordings (of which "Edgar The Barber" is one) are supposedly made for both headphone and speaker listening. It looks like an EQ has been applied on these recordings as discussed in this 2013 Sound & Vision Chesky interview indicating the change (thanks STC for the link):

S+V: How do you process the recording so it plays back on speakers?

DC: We add a diffuse EQ filter so that you can play it back on speakers. Without that filter, the sound will rip your head off, it'll be ridiculously bright. We do it in Sonic Solutions. We have a filter algorithm that was supplied by Edgar [Choueiri]. This makes it theoretically flat again.

--------------------

To end, with all that's going on in the early fall season at work and in the family, it has been hard finding time for the audio hobby. ;-) I consider this a good thing since it also means, ironically, that I'm enjoying the music more even though in shorter spurts! More just listening to the music and thinking about things than having the time to actually write stuff.

For other interesting reading material, check out Whit Turner's JWTAudio blog. It's always good to look around and see what others have been working on!

Have a great October, audiophiles. Hope you're enjoying the music.

Oh yeah. Happy Canadian Thanksgiving 2023, eh?!