 |
| Significant changes in the UI from the previous CTXMatrixBeta release. |
I want to remind 2-channel "immersive sound" lovers about the potential for using stereo crosstalk cancellation (XTC) / ambiophonics playback to enhance the 3D soundstage. You might have already heard of uBACCH in the audio magazines and on forums - why not try a similar effect for free (or through a small coffee donation)?!
You might recall, last year STC contributed a post on CTXMatrixBeta DSP for Windows. Well, it's great to see that he has continued to develop the software with updated user interface and abilities, this latest edition renamed CTXMatrix Lite, currently at version 1.3. It now includes a standalone processor which I'll spend some time discussing in this post, along with the 64-bit VST3 plugin as shown above that can be used in JRiver, Audirvāna and DAW software like Reaper.
Furthermore, there is now a VST3 plugin for Linux. Since I primarily use Windows for my music listening and processing, I'll stick with the Windows variant here with the hopes that some of this will be applicable. In my discussions with STC, he's aware of interest from Mac OSX users.
Perhaps the most important update in the last few months is that there is now improved support for standard ~60° stereo speaker angle (typically, the recommendation is 20-30° in ambiophonics). As a result, one should have a much better experience with the effect on most set-ups. In particular, hopefully more desktop computer music lovers will give this a listen without needed to move the speaker position and explore what you might be missing in your existing music library.
[If you run a miniPC streamer in the sound room like I do, CTXMatrix can easily be used like with HAF X-talk Shaper DSP discussed a couple years ago.]
IMO, the standalone processor is a nice addition to the package and will allow you to use whatever music player you wish by piping the audio into it through a virtual input/output loop. A free virtual loop driver to try is VB-CABLE. Download it, run the appropriate setup file (like "VBCABLE_Setup_x64.exe"):
Click on "Install Driver" and you should now see a device called "CABLE Input (VB-Audio Virtual Cable)" which you can point audio software like foobar2000 to use as the playback device:
Now run and configure the CTXMatrix Lite standalone app to accept the VB-Cable input, and configure output to whatever audio playback device or DAC you're using through the Settings window:
 |
| When you're sure you have no unpleasant audio feedback and ready to listen, unclick that "Mute audio input" box. For the other settings, I typically standardize my computer workstation at 48kHz these days; no need to futz around. For higher quality, try changing "Audio device type" to "Windows Audio (Exclusive Mode)", samplerate to 96kHz, and buffer size to 1920 samples (20ms). |
Alright, now make sure to take some time to run through the calibration procedure as discussed here (the procedure was also touched upon in my previous post). Given the numerous variables involved including audio hardware, set-up geometry, and individual characteristics (eg. one's unique HRTF), getting the calibration right is essential for accurate-sounding tonality and spatial cues. It's an important bit of initial investment.
.png) |
These calibration values work well for my ears on my computer workstation with Emotiva Airmotiv B1+ speakers, slightly toed-in. Speakers 55-60° angle from main seating position. |
For convenience, let me copy over a chart from STC's documentation showing the approximate expected calibration values based on angle between where you're listening and the two speakers:
You'll get a taste of the effect if you start with the "baseline" dB and sample delay (smp) values. Best to calibrate of course while using those numbers as guide.
Once calibrated, in foobar (or whichever player you're using), go enjoy some music with crosstalk cancellation turned on!
.jpg) |
| "Balance" knob is an improvement from the previous individual L/R trim controls. I like having the "Center" knob to anchor the central image a bit better on some albums that might sound too diffuse (default is at 0). Use the level meters to make sure there's no clipping especially if you boost volume with "XGain". |
In the same way as with foobar, I can point to the VB-Cable output as a Roon playback device:
%20-%202.png) |
| With VB-Audio Virtual Cable, you'll typically see 2 outputs installed and listed in Roon, choose whichever has been mapped to your CTXMatrix setup; in my case it was the second. For more see VB-Audio manual. |
 |
| Good to have the "Bypass" button available to turn off DSP easily since not everything will sound good with XTC engaged. Hey Roon, why don't you guys implement XTC natively in your Muse DSP? Among other obvious wish list items. |
I was shocked by how different, in a good way, "Angel [Mad Professor Remix]" on Mezzanine: The Remixes (2006) sounded with ambiophonics processing! Likewise, acoustic albums like "Venus" off John Coltrane's Interstellar Space (1967) sounded like I was listening in a much larger space than just sitting in front of my computer in the home office thanks to the expanded room reverb "air".
Sometimes, studio productions like Michael Jackson's "They Don't Care About Us" on HIStory: Past, Present and Future, Book I (1995) can sound too much like you're wearing headphones with the music localized inside the head which I'm trying to avoid using loudspeakers. Fun experience I suppose for those who like that kind of effect, but I prefer to turn off ambiophonics with these kinds of recordings. Speaking of MJ, have a listen to "Thriller". The footsteps at the start should sound well-defined like they're walking across from right to left behind you; this is how it sounds in a multichannel rig with Atmos mix.
There are some albums that can really benefit from the ambiophonics DSP. For example, you might already have some QSound encoded albums (eg. Sting's The Soul Cages, Pink Floyd's Pulse, Roger Waters' Amused To Death, Paula Abdul's Spellbound), binaural ones (eg. Yosi Horikawa's Wandering, Jean-Michel Jarre's Oxymore binaural mix, Macy Gray's Stripped), or those intended to be immersive but downmixed like the stereo version of Justin Gray's recent Grammy-winning Immersed (2025, stereo DR9, multichannel/Atmos DR13) - check out the track "Tapestry". Listening through calibrated CTXMatrix likely will get you significantly closer to the intended surround effect these albums were aiming for.
Something else fun to try is to take multichannel/Atmos music downmixed to 2-channels and then run that through crosstalk cancellation. For example, the other night, I had a listen to BTS' new album ARIRANG downmixed from multichannel (higher dynamic range DR12 vs. stereo DR5). While not exactly the same as with surround speakers, you should still be able to appreciate the effect of the rear channel content with sound as if coming from behind your head (like in the track "SWIM"). Another multichannel album that surprisingly blossomed in CTXMatrix is The Cranberries' Everybody Else Is Doing It, So Why Can't We? (1993), check out "Dreams".
Finally, beyond music player software that can be configured, if you want all Windows audio output to go through CTXMatrix DSP, you can set VB-Cable as your output device in the System → Sound settings:
 |
| I'm told that CTXMatrix's Mode A/B/C settings are variations to the cancellation steps. I don't hear a big enough difference to have a preference. |
As you can see, I'm enjoying Jon & Vangelis' "The Friends of Mr. Cairo" playing over YouTube in the Chrome browser, piped into CTXMatrix for the ambiophonics processing. Just as easily, I could be watching a movie with improved 360° soundtrack rendering.
Well done STC! Your DSP implementation of the RACE algorithm with ongoing improvements is a fitting homage to Ralph Glasgal's work bringing this to everyone. The plugin and standalone app has been stable over hours of listening and testing while doing work on my workstation computer. RAM usage not bad (~60MB max), and barely 0.1% CPU cycles at 48kHz during playback which increases to about 0.3% for higher quality "Windows Audio (Exclusive Mode)", 96kHz, 1920 samples buffer (AMD Ryzen 9 5950X machine).
Have fun music lovers and audiophiles! If you enjoy CTXMatrix Lite, go buy ST some coffee for his work. He might have saved you good money if you were thinking of purchasing the uBACCH plugin or even BACCH in general, as good as they might be and with advanced features.
--------------------
Let's end off with some new music.
As shown above playing in foobar, here's the recent track "The Mountain" (2026) by Gorillaz on the album of the same name:
The ending portion of the song (from 4:00 on) with the multiple overlapping voices saying "the mountain" sounds pretty cool with crosstalk cancellation turned on with each voice hanging in space with much greater width and depth.
As much as I love this album, my main complaint is that it's a sad, dynamically squashed DR5 stereo mix - even the 24/96 "hi-res" download is worthless when damaged like this! Given the complex layering of the production, it's a shame that the music is so unnecessarily robbed of breathing room, drowning out all the nuances. The louder portions are distorted, and low-level details get boosted to the point of just sounding gritty and fatiguing.
It's times like these when an audiophile needs to listen to the more dynamic multichannel/Atmos mix (DR10). 😏
 |
| Unfortunate dynamically compressed and distorted stereo mix above; HDTracks sells this as 24/96 "hi-res"! Do artists expect music lovers to buy badly damaged recordings? I'll happily take the superior-sounding lossy 768kbps Atmos downmix any day - it's like a different recording sonically. |
Happy Easter long weekend. Hope you're all enjoying some wonderful music!
Thank you, Archimago, for another thoughtful and detailed write-up on CTXMatrix Lite v1.3. I’m truly grateful for the time and effort you put into testing and sharing your impressions — and especially for the excellent step-by-step guidance on the standalone setup. That makes it so much more approachable for everyone. With your permission, I would like to add a link to this article in the CTXMatrix README.
ReplyDeleteOne aspect I particularly appreciate in today’s XTC discussions is the true benefit of crosstalk cancellation emerging most strongly in the Ambiodipole (narrow-angle) position. When speakers sit closer to the center line, they better suppress the monaural directional cues that a single ear would otherwise pick up. As the angle widens toward the conventional 60°, natural head-shadowing already attenuates much of the crosstalk, leaving primarily delay-based correction.
This thinking draws heavily from Ralph Glasgal’s foundational work on Ambiophonics, combined with my own experience: achieving effective cancellation at 60° requires actively suppressing those monaural cues. The most powerful approach involves modifying the original signal with an in-ear inverse measurement (HRTF-related correction). Since CTXMatrix deliberately avoids any EQ or frequency-domain processing to keep things clean and lightweight, its effectiveness gradually reduces as the speaker angle widens. A strong 3D effect can still appear even without full cancellation, but reaching the immersive, truly dimensional experience possible at 20–30° usually needs additional stereo-enhancer tools at wider angles.
Cont/
Cont/..
ReplyDeleteFrom my own journey, I stubbornly experimented with 60° setups for a long time before finally settling on the narrower 20–30° configuration — and the difference was night and day. I suspect many who start exploring 3D sound at 60° will eventually migrate to the narrower angle for the best results. Sadly, after nearly 80 years of conventional 60° stereo (and now with TVs dominating entertainment setups), very few high-end systems use the optimal narrow angle. In the last decade, I’ve yet to see a commercial XTC solution specifically optimized and demonstrated for a true 20° setup.
Interesting observation about the Michael Jackson tracks feeling a bit “headphone-like” or internalized. I will give them another listen later on my setup and report back. Meanwhile, I also noticed a BACCH user mentioning that the Time Out album didn’t feel entirely natural to them. As I was listening, nothing struck me as out of the ordinary at first, but it made me wonder whether XTC behaves differently at wider speaker angles. I’m unable to easily move my speakers to 60°, but I’ll check if the MJ tracks sound internalized at 20° or not.
This is a fascinating point, especially since I also observed that tilting my head upwards produced a noticeably wider soundstage. For example, the Buhina track (which is supposed to have a 270° stage with BACCH) behaved similarly with my 20° setup when I tilted my head up. I think this gives us more clues about the role that frequency response and the cancellation signal play in perceived soundstage. Always remember that XTC is not mere attenuation — it works with an inverse signal, so when the signals are not fully cancelled, it can produce some extreme effects. Just speculating on my part!
By the way, one quick tip for verifying calibration accuracy: the Chesky Channel Phase Test is excellent. It quickly reveals if cancellation is off — you’ll hear unwanted phasing or a shift to one side instead of a perfectly directionless/centered image when it should be neutral.
On a side note, I would strongly recommend users go with VB-Audio HiFi Cable (instead of the free VB-Cable) for high-resolution files, since the free version is limited to 48 kHz. That said, I’m really happy to see how well CTXMatrix performs even though it was originally designed for 384 kHz sampling rates. Also, have you had a chance to notice the latest v1.4 design yet?
Once again, thank you for your continued support of CTXMatrix and for helping bring Ambiophonics/XTC to a wider audience. Your reviews make a real difference for this niche!
Best regards,
ST
As I mentioned earlier, I gave “Heal the World” from the MJ Book 1 disc album a careful listen with CTXMatrix on my 20° Ambiodipole setup. When the girl started speaking, I actually had to double-check if I had accidentally turned on the rear Ambiophonics mode — her voice appeared so clearly positioned behind my head! Otherwise, the track felt nicely balanced overall. It sounded most natural to my ears when the center control level was increased by about 1 to 2 notches.
ReplyDeleteInterestingly, even with CTXMatrix completely off and using a conventional stereo setup at 20°, the girl’s voice already had a very strong forward presence. Perhaps it was an intentional “in your face” effect. Almost all great audiophile recordings have some sort of trick that made them different from others.
ST
Thanks for taking the time to discuss CTXMatrix and the ambiophonics technique in greater detail ST. The beautification process in 1.4 is looking really nice!
DeleteBy all means, feel free to reference any of the articles and discussions we've had on this blog. As far as I can tell, there's precious few articles about this topic so getting the word out there and encouraging audiophiles to try out the technique I hope is part of the fun of enjoying sound and music for us hobbyists 😊.
As you know, my main system these days is set up for multichannel/Atmos so I have the problem of being able to check out the sound of ideal ambiophonics with my speakers closer to 60° than 20°. Even so, the fact that I can enjoy the effect at 50-60° on my desktop setup is already great so long as I'm not aiming to be extremely picky!
Interesting comment about having your head tilted up for wider soundstage. The way I have my speakers set up on the desktop, the tweeters are about 2-3" below ear height and they're angled up so in effect that's like having my head tilted up if I had taller speakers. Maybe that actually helps the effect.
For Michael Jackson's tracks, yeah "Heal The World" sounds great over CTXMatrix! Nice spread of the synth at the start, kids voices playing, and the girl's voice located centrally and clearly, then MJ singing front-and-center with voice spreading wider when the chorus "Heal the world..." comes on. (This song came out in 1991 on Dangerous and made quite an impression on me back then as I was applying for medical school that year.)
Have a listen to "They Don't Care About Us" a couple of tracks down on HIStory. The repetitive marching percussion part is what sounds "inside the head" for me. Thankfully this doesn't happen too much and I suspect it's just the nature of that particular recording.
Keep up the great work!
Thank you Archi. Glad you like the beautification in version 1.4 😊
DeleteI haven’t heard the Michael Jackson tracks you mentioned yet — I’ve been having a bit of downtime lately, but I’ll definitely give “Heal The World” and “They Don’t Care About Us” a listen soon.
Regarding speaker angles, you’re right that your main Atmos setup has the front L/R at around 50-60°. Dolby recommendations commonly place the front left/right speakers between about 45° to 60° (some setups go as narrow as ~43°). At the narrower end (~43°), adding the plugin to the front L/R can give you a nice combination of strong XTC plus a mini DCH-like effect. However, for movies I usually don’t bother because the visual cues on screen tend to overwhelm the soundstage anyway.
ST
Nice work ST! I haven't tried your plugin yet, I already use uBACCH, but I will try at some point to make a comparison. I would like to comment on Archimago's note about listening "binaural" records on a desktop setup with cross-talk cancellation. In my experience, the result will greatly depend on what HRTFs were used for capturing/rendering the binaural record. The main reason is that when playing via XTC speakers their output gets filtered via your HRTF. Thus, if the binaural record was created using HRTF of another human, you will get double filtering which will affect the tonal balance greatly. It is safe (and is a great pleasure!) to listen to binaural recordings done with KU-100 head because KU-100 has relatively simple geometry, has no ear canal, and was tuned to be "loudspeaker" compatible—there are many recordings available on YouTube. But for example the binaural version from Yello's Touch BD sounds too elevated to me (that likely means the HRTF used for rendering was more human-like), while the stereo version sounds great.
ReplyDeleteThank you Mikhail for the kind words.
DeleteuBACCH is a VST plugin, and so is CTXmatrix. To test both side-by-side, you can simply copy the CTXmatrix.vst3 into the same folder where uBACCH is installed. In a DAW it becomes very easy to do quick A/B blind tests by switching between the two plugins. In JRiver you will need to disable one and enable the other (basically pressing the button twice), but it is still quite straightforward. BUt CTXMatrix strength is around angle of not more 40 degrees. Anything more than that you still get the benefit but no true 3D unless I use other effects to sweeten the soundstage.
Regarding binaural recordings and crosstalk cancellation (XTC), I agree it is a tricky subject. Binaural recordings are essentially a special form of stereo and are usually best experienced with in-ear monitors.
The Neumann KU-100 has proper head geometry and includes built-in EQ correction for the pinna (diffuse-field equalization). This was deliberately done so the recordings sound more natural and flat when played over loudspeakers. It is even possible to record with a Neumann-style setup without the pinna, which would be closer to classic Ambiophone recordings that use a head/sphere without pinna.
Chesky’s Binaural+ series is a true binaural recording but with an additional Choueiri pinna correction filter applied so it does not sound overly bright on loudspeakers. I’m not entirely sure why they chose the Choueiri filter instead of Neumann’s own correction at the beginning. If my memory serves me right, subsequent Chesky binaural recordings switched to using the Neumann filter.
The key point is that after proper pinna correction (whether Neumann’s diffuse-field or Choueiri’s), these recordings behave more like a spaced-mic arrangement with ~17 cm separation. They carry natural non-linear attenuation and time delays (similar to real interaural differences). This makes them excellent for XTC playback compared to most regular stereo recordings, which often rely only on panning or level differences. The result feels more natural to the brain and allows more accurate localization.
As far as I am concerned, both Neumann and Chesky binaural recordings are already pinna-corrected, so there should be no significant “double pinna filtering” issue when using XTC on speakers. I don’t have personal experience with the Yello Touch BD binaural version, so I can’t comment on why it sounded elevated to you.
For comparison, I have DPA, Sennheiser, and Roland binaural mics, none of which have pinna correction applied. Even when listening on in-ear monitors you can hear slight tonal differences, and the spatial cues can get a little distorted (for example, a sound at 36° azimuth / 20° elevation might localize as roughly 38°/18° depending on your own head size).
Looking forward to hearing your thoughts once you try CTXmatrix. Happy to discuss more if you have any questions!
Cheers!
ST
Thanks, ST! I haven't heard about the Choueri's compensation filter before. To my understanding, it might be similar to the correction curve calculated by Benjamin Bernschutz from his anechoic multi-point measurements of KU-100 (http://audiogroup.web.th-koeln.de/PUBLIKATIONEN/Bernschuetz_DAGA2013.pdf), which is more precise than the built-in correction of KU-100 made at Neumann from reverberant chamber measurements. But it is possible that the actual perceived differences are rather subtle, so maybe that's why Chesky switched away from it.
DeleteFor listening of personal binaural records on an XTC system, I would actually apply not "pinna correction" but rather equalization by the frontal field, measured using the same binaural microphones. This basically makes the response of a frontal source in the recording to be "flat," and the frequency responses of sounds from other directions become relative to the frontal source. In that case, when you are playing it back via XTC speakers placed in front of you, the acoustic path to your ears applies your natural frontal HRTF to the reproduced recording, and in theory you get the most natural tonality. David Griesinger has a lot of writings about that, see for example https://pub.dega-akustik.de/ICA2019/data/articles/000827.pdf
Reproduction of binaural records even via IEMs still wins from some personal equalization for achieving natural tonality and correct source placement.
If you wish to try to compare Yello's track in stereo vs. binaural, you can get it here: https://drive.google.com/drive/folders/1vOYDAWYg6ptqopdyGowbRflUwbWKt9-T?usp=sharing (for personal / educational use only!). Since they used Dolby Atmos, I believe their binaural rendering was be done via some "generic" (non-individualized) Atmos HRTF, which is not KU-100.
Thanks Mikhail for sharing the Yello files!
DeleteI think there might be some confusion caused by my use of the word “3D”. CTXMatrix is mainly meant to provide good, free crosstalk cancellation for two-speaker stereo playback — as Ralph Glasgal always wanted. I tried to make it as good as (or better than) any commercial plugin.
I’ll download the files and give them a listen. For a proper comparison though, I would need a good Atmos setup, which I don’t have (still running an old 7.1 HT system).
I learned a lot from you!
Regards,,
ST
A few thoughts:
ReplyDelete• XTC is the great new wave band from the UK that had such gems as “Dear God”, “The Mayor of Simpleton” and “Generals and Majors”. Cease and desist invoking their name for cross-talk cancellation.
• Any post processing meant to expand the stereo sound stage is going to be highly dependent upon the actual stereo mix and whatever tricks the engineer has used to expand the stereo image. Ideally, BAACH 3-D or whatever will be used in the mixing process so there’s no need for any other processing in the user’s end, but that’s rarely done.
• It’s so depressing to see that the practice of using multiband compression to wring the dynamics out of hi-res audio is still a thing. The only people buying 24/96 audio files are people who want more dynamics, not less. It’s like a Dolbyvision video release offering less much less dynamic range than the SDR version.
As Jerry Seinfeld always said, “I don’t get it… who are these people?”
This statement is partially true, but it overlooks an important distinction.
DeleteRecordings made with Ambiophone mics, a 17 cm sphere with Jecklin disk, or Chesky’s Binaural+ technique often sound more three-dimensional and natural than many studio productions where engineers manipulate phase, panning, and compensate for their own hearing limitations or room adaptation. In those cases, the spatial cues are already closer to real binaural listening, so playback-side XTC simply cleans them up rather than “expanding” artificially.
Even when engineers carefully place instruments (e.g., a trumpet at 40° left and a flute at 30° left), precise localization is highly individual. One mastering engineer might perceive them at 45°/35°, another at 38°/38°. If you don’t believe this, invite audiophile friends into a darkened room, make a click or clap sound, and ask everyone to point to the direction — you’ll see each person perceives it differently due to their unique pinna and head-related cues.
I cannot speak for BACCH. In my opinion, pure crosstalk cancellation works far better at 20° speaker placement than at 60°. The reasons are well documented in Ambiophonics work and in spatial hearing literature on monaural localization sensitivities (pinna cues are most accurate and sensitive around 30° off-center — coincidentally the classic stereo sweet spot). At 60°, strong monaural spectral cues from one ear anchor the images firmly to the physical speakers, so you need excessive delay/level/phase manipulation to override them. That’s why many widening effects (including some QSound-style processing) feel artificial.
It is also incorrect to call XTC an “effect.” It is a cleaning mechanism that removes interaural crosstalk after the sound has exited the speakers. The correction is highly individual, which is why CTXMatrix has those cumbersome calibration stages.
I have heard users complain about weird effects with hard-panned recordings like the West Coast Take Five when using some popular (and expensive) commercial XTC programs. I wish they would try CTXMatrix at 20° and share their opinion — in my experience it keeps the imaging coherent and musical rather than falling apart.
To better understand each other’s perspective: how many XTC programs have you tried, and can you give specific examples where you perceived weirdness or excessive stage that felt unnatural? That way we can actually address any misconceptions.
Looking forward to your thoughts
Hi Archi. Thanks a lot for the CTXMatrix Lite material.
ReplyDeleteI use it for headphones. The result is very decent. Yours sincerely, Vlad.
Hi VladHV! Could you tell us more about your use case? I'm intrigued. Normally headphones do not have cross-talk at all. Are you using the CTXMatrix to create it?
DeleteMe too.
Delete