A article on BACCH4Mac that answers a lot of questions.

[STC]

My system is well optimized and my statement was about of an exaggeration of the difference between recording types with BACCH. Early stereo recording with hard panned instruments will often place to instruments at 3 o’clock and 9 o’clock or even a bit further back. If one does not like the xtc they can dial down the XTC or shut it off altogether. I am getting over 20db of XTC with my BACCH-sp which is an indicator of optimal settings and speaker type and placement.

Hard-panned sound simply means the signal is sent fully to one channel only—so 100% left or 100% right, coming from just one speaker in playback.

Without any crosstalk cancellation, we localize that sound pretty much right at the speaker (or very close to it) because both ears still get the full signal plus the natural binaural cues, HRTF, visual references if eyes are open, and all our years of two-ear learning. It stays anchored near the physical speaker position.

When you apply proper interaural crosstalk cancellation—meaning only canceling the unwanted leakage to the opposite ear, nothing extra—the opposite ear gets almost none of the signal. Localization then shifts to mostly one ear (monaural-like for that channel), and the perceived direction moves slightly outward from the actual speaker because the inward-pulling crosstalk is gone.

There’s no reason for a hard-panned sound to suddenly jump to 3 o’clock or 9 o’clock (extreme positions) in a standard 60-degree speaker setup. That would only happen with over-processing or added artificial widening beyond pure cancellation. If it’s truly just correcting the crosstalk error, the sound emerges slightly outside the speaker edges at most—not way out wide or behind unless your speakers are spread to something crazy like 140 degrees. having said that, some short frequencies can indeed make it like sounding at the extreme. I think we discussed about this before where you gave me some tracks to test?

Try the blindfold test with one ear plugged: localize a phone or speaker playing a tone. You’ll notice it pulls further outward than the real position. That’s the monaural shift in play—exactly what clean XTC does to hard-panned material without inventing extra space.

So yes, if your hard-panned stuff is landing way at 3/9 or beyond on a 60-degree triangle, something’s pushing the processing too far past basic cancellation. Maybe it need to be calibrated again.

Mind sharing how the 20dB cancellation is measured? I recall BACCH’s display is simulated based on the filter itself rather than an actual real-time ear measurement

 

[Gwreck]

Hard-panned sound simply means the signal is sent fully to one channel only—so 100% left or 100% right, coming from just one speaker in playback.

Without any crosstalk cancellation, we localize that sound pretty much right at the speaker (or very close to it) because both ears still get the full signal plus the natural binaural cues, HRTF, visual references if eyes are open, and all our years of two-ear learning. It stays anchored near the physical speaker position.

When you apply proper interaural crosstalk cancellation—meaning only canceling the unwanted leakage to the opposite ear, nothing extra—the opposite ear gets almost none of the signal. Localization then shifts to mostly one ear (monaural-like for that channel), and the perceived direction moves slightly outward from the actual speaker because the inward-pulling crosstalk is gone.

There’s no reason for a hard-panned sound to suddenly jump to 3 o’clock or 9 o’clock (extreme positions) in a standard 60-degree speaker setup. That would only happen with over-processing or added artificial widening beyond pure cancellation. If it’s truly just correcting the crosstalk error, the sound emerges slightly outside the speaker edges at most—not way out wide or behind unless your speakers are spread to something crazy like 140 degrees. having said that, some short frequencies can indeed make it like sounding at the extreme. I think we discussed about this before where you gave me some tracks to test?

Try the blindfold test with one ear plugged: localize a phone or speaker playing a tone. You’ll notice it pulls further outward than the real position. That’s the monaural shift in play—exactly what clean XTC does to hard-panned material without inventing extra space.

So yes, if your hard-panned stuff is landing way at 3/9 or beyond on a 60-degree triangle, something’s pushing the processing too far past basic cancellation. Maybe it need to be calibrated again.

Mind sharing how the 20dB cancellation is measured? I recall BACCH’s display is simulated based on the filter itself rather than an actual real-time ear measuremen

With ORC the filters are remeasured. And there are graphs produced showing the optimal measurement and the measured correction for frequency response and the xtc will change with the remeasurement. If there is some alternative method to measure XTC outside of the BACCH process I am not sure. It would be interesting to see what would happen if some industrious BACCH user found a way to measure the XTC to verify the BACCH theoretical number.

My experience with BACCH and my system is that when the measurements are not accurate for various reasons the XTC becomes unstable or collapses back to plain stereo. I’m not sure what a person would do to improve my settings or measurements. If there were some setting to change that would improve BACCH for me I would’ve glad to know. When having my Sp set up Dr. Choueiri literally giggled a bit when he saw how good my readings were. (XTC, frequency response, impulse response).

 

[mieswall]

BACCH is fundamentally a crosstalk cancellation (XTC) system, just like other XTC solutions out there. Not "just like". This one really works (among other reasons, because it not only deals with the source like all others, also does it with the medium where it performs: your room); then it doesn't alter spectral content, like most others (for mathematical reasons explained in BACCH papers). All of them work primarily through phase manipulation to cancel the unwanted signal that crosses over from one speaker to the opposite ear. I'm not sure the issue is that simple. A recursive algorithm is required, cancelling the information added in the opposite channel, all done in real time. A second key point is that xtc is calculated with data points measured at your ears, considering your anatomy and, more important imho, the acoustic conditions of your listening environment and capabilities of your system. AFAIK, no other XTC solution is able to do that, and it's a key aspect for BAACH to be as effective as it is.

The thing is, you can push phase manipulation so far that the stereo image gets shifted dramatically—even out to ±90° or beyond. At that point, though, you’re no longer doing proper interaural crosstalk cancellation; you’re basically creating an artificial widening effect instead. If so, that's a problem with the recording, not the xtc process. That source will also sound unnatural in normal stereo, as we all know from those early stereo recordings. Anyway, you can control the amount of XTC desired in those cases.
But even when that happens, the result is so much realistic than just the mere positioning of instruments in space; it is the way each of those instruments is clearly identified in complete separation of others, instead of the "mass" of sound that normal stereo achieves in comparison; and the retrieval of the acoustic conditions of the recording venue, as the interaural difference in reflections and reverberation is a key element in that definitions of the "space" captured in the recording.

This extreme image shift often happens when the cancellation parameters (delay, attenuation, filter curve) aren’t dialed in correctly for the specific listening setup. It’s a very common reason why people try XTC, don’t hear the expected natural improvement, and conclude it “doesn’t work” or sounds weird. No, this extreme shift happens when the recording has those weird effects. It is not "created" by XTC.

In a well-calibrated XTC system, hard-panned sounds (full left or full right in the recording) should still appear only a few degrees outside the physical speaker positions—not floating way out wide or behind you. When they do stretch unnaturally far, it’s usually a sign the settings are off. I don't know how you infer those "laws". It is perfectly possible that the mix wants you to be immersed in the music as much as possible. Again, it is not an issue of XTC, but of the recording. If you have the fortune to achieve that extreme shit always, with every recording (which is impossible, as tXTC doesn´t create content, it works with content as input and results depend on the quality of that content), and then believe it is just too much, simply set permanently the xtc slider at, say, 40%, 50% or any other amount pleasant to you.

And just to be clear, this isn’t unique to BACCH. Pretty much every XTC implementation can run into the same issue if the cancellation isn’t properly matched. Then again, all this theorical discussion by people judging a technology they don't have even heard. BACCH has many particularities that set it apart from other known XTC implementations, and Choueri has discussed the matter in depth in published papers and conferences.

STC: Comments to your post in red above.

 

[STC]

STC: Comments to your post in red above.

I agree that strong crosstalk cancellation can extend the soundstage to extremes in some cases, and the FAQ does mention how certain studio-mixed recordings with extreme spatial panning can produce very wide or near-ear images, especially in dead rooms.

That said, true hard panning means the signal is fully in one channel only—the opposite channel is silent. In that situation, without opposing cues from the other speaker to create large ITDs or ILDs, localization relies primarily on monaural pinna cues, which typically shift the image slightly outward from the physical speaker position. The amount depends heavily on frequency content, duration, and individual anatomy—some short high-frequency sounds can indeed feel like they’re at or beyond 3/9 o’clock, but it’s not a guaranteed or consistent extreme to 90°.

The FAQ ties ‘extreme spatial panning’ specifically to hard-panned sources in the absence of reflections, so we’re dealing with one-sided panning rather than material that has deliberate content from both channels to generate exaggerated ITDs/ILDs.

The cleanest way to explore it is probably the mattress test Choueiri describes: pan a vocal or instrument track fully to one side, play it through a single speaker with a partition blocking the other, and see where listeners localize it. That removes digital processing variables and shows what pure physical crosstalk elimination does on a hard-panned source.

Curious if anyone has tried that exact setup—what did the localization feel like?

 

[goat76]

The cleanest way to explore it is probably the mattress test Choueiri describes: pan a vocal or instrument track fully to one side, play it through a single speaker with a partition blocking the other, and see where listeners localize it. That removes digital processing variables and shows what pure physical crosstalk elimination does on a hard-panned source.

Curious if anyone has tried that exact setup—what did the localization feel like?

Yes, but you don't even need to use a mattress for this test; you just have to block the opposite ear with a finger to eliminate the crosstalk almost completely.

With a hard-panned sound coming from the left speaker, hearing it with both ears, the sound will be heard from the exact position of the left speaker.
When blocking the right ear with a finger, which will give you an almost 100% crosstalk cancellation, it will make the hard-panned sound appear to be coming from a position, maybe a couple of feet outside to the left of the left speaker, but not any more extreme than that.

As in the above example. When blocking the opposite ear with the finger (which again will give close to a 100% XTC), and as that will still not position the hard-panned sound object to any extreme wide position not much further than a couple of feet outside the speaker, you can be quite certain that if your XTC program makes that same sound object appear to be coming much further out to the side than that, there is obviously a "phase trick" going on and the XTC program is introducing an artificial widening effect to the sound object.


The problem I experienced when trying uBACCH wasn't actually the widening effect of hard-panned sound objects, as that could sound pretty impressive. The main problem was this strange feeling of being almost "deaf" in the opposite ear to the hard-panned sound, which is similar to hearing it in a pair of headphones. The hard-panned sound object becomes strangely isolated to just one of the ears, and such an effect will never occur with real-life sounds in any circumstances. Hard-panned sound objects simply need crosstalk to sound as they are intended to sound. Otherwise, they will sound unnatural.

 

[STC]

Yes, but you don't even need to use a mattress for this test; you just have to block the opposite ear with a finger to eliminate the crosstalk almost completely.

With a hard-panned sound coming from the left speaker, hearing it with both ears, the sound will be heard from the exact position of the left speaker.
When blocking the right ear with a finger, which will give you an almost 100% crosstalk cancellation, it will make the hard-panned sound appear to be coming from a position, maybe a couple of feet outside to the left of the left speaker, but not any more extreme than that.

As in the above example. When blocking the opposite ear with the finger (which again will give close to a 100% XTC), and as that will still not position the hard-panned sound object to any extreme wide position not much further than a couple of feet outside the speaker, you can be quite certain that if your XTC program makes that same sound object appear to be coming much further out to the side than that, there is obviously a "phase trick" going on and the XTC program is introducing an artificial widening effect to the sound object.


The problem I experienced when trying uBACCH wasn't actually the widening effect of hard-panned sound objects, as that could sound pretty impressive. The main problem was this strange feeling of being almost "deaf" in the opposite ear to the hard-panned sound, which is similar to hearing it in a pair of headphones. The hard-panned sound object becomes strangely isolated to just one of the ears, and such an effect will never occur with real-life sounds in any circumstances. Hard-panned sound objects simply need crosstalk to sound as they are intended to sound. Otherwise, they will sound unnatural.

The finger-in-ear test ( I call it the occlusion method) is a solid quick way to mimic near-complete crosstalk cancellation on a single hard-panned source, and your experience with it—slight outward shift, maybe a couple feet beyond the speaker but nothing extreme—lines up with what a lot of people observe in physical barrier setups too. BUt you need the mattress to double confirm to listen to stereo to understand some extreme position users reporting.

That “deaf in the opposite ear” isolation you felt with is a known thing when cancellation gets pushed high. It basically makes hard-panned signals feel headphone-like but unnatural since real acoustic sounds never fully separate the ears like that. Bauer highlighted this back in the 60s: the lack of crossfeed exaggerates separation and can make things over-wide or detached.

So perhaps with high attenuation, the separation becomes so clean that it starts needing a bit of intentional crosstalk dialed back in to restore the natural acoustic blending our ears expect.

The core purpose of crosstalk cancellation is to deliver the encoded ITD and ILD as precisely as possible—speakers suffer from inherent bleed that blurs those cues, while headphones deliver perfect isolation but lack the real-world crosstalk that makes sound feel natural in space due to lack of pinna frequency shaping. The trick is finding the right amount of cancellation that gets us closer to live listening without tipping into unnatural isolation. Too much can create that headphone-y detachment; too little leaves the blur.

But this is a whole different topic and I just want to address the reason of extreme positioning which seems a problem almost with all types of XTC. It is actually the calibration issue but users without understanding how it should sound often make mistakes with the calibration

 

[Gwreck]

I have 2 sets of speakers in my listening room(large basement family room).
I am adding the measurement plots from both my ESL speakers and My Ascend ELX towers in the same room and almost colocated. The systems run independently. My Sanders ESL’s use the BACCH SP as the preamp and have 2 Sanders Magtech amps as the speakers are bi-amped. The ELX are part of my home theater system and run with my Denon A1h with Audessey correction. The ELX towers are running also with 3 monolith THX15 subs and minidsp with MSO. My room is basically a 40x30ft basement family room and my system along the longest wall.



I have added plots comparing my ESL’s and ELX’s. You can see the difference in the amount of XTC on one set of plot between the different speakers. I also have measured impulse response plots between the speakers. You can clearly see the difference in impulse response and XTC between the different speakers in the same room. The ESL’s have much higher XTC and cleaner impulse response. The measurements are all notable not flat above 1000hz due to the effect of ones torso and higher due to the pinna.

 

[phoenixdogfan]

Theoretica is now offering the capability in its Baach4Mac package for the creation of 12 channel virtual Atmos layouts (7.1.4) over a single pair of stereo speakers.
The module is called Stratos, costs $3000 and is only available in its Audiophile and above version of the package which, in and of itself, costs $4980 and only runs on Apple Mac machines.

Moreover, the Stratos module has its limitations, as witnessed by Theoretica's explanation on the website. Quote:

"BACCH-Stratos does not decode Dolby Atmos® content. Instead, BACCH-Stratos accepts up to twelve discrete PCM audio channels that have already been decoded on the Mac by a Dolby-licensed playback application. Supported sources include Apple Music® and Apple TV®, as well as the Dolby Reference Player, which can play lossless Dolby Atmos audio and video files.'

So basically what it's saying here is you can use Apple sources for Apple Dolby Atmos content which Apple Macs are licensed to decode or you can buy the Dolby-Licensed playback app which will decode any Atmos bitstream files you can find and load onto an Apple machine, but not streaming sources such as Netflix, not ever.

The creation of those Dolby files is a Herculean task and was covered in a thread here which was, in turn, coverage of a thread on Audiophile Style detailing everything required to do that. The consensus was clearly that the juice was just not worth the squeeze.

So basically, that Stratos module is a great idea but it's $3 k on top of $5k to listen to Apple Atmos content over speakers. So, like everything Theoretica, a great idea but priced in the 'Stratos' sphere and having a highly limited use case.

 

[Gwreck]

Theoretica is now offering the capability in its Baach4Mac package for the creation of 12 channel virtual Atmos layouts (7.1.4) over a single pair of stereo speakers.
The module is called Stratos, costs $3000 and is only available in its Audiophile and above version of the package which, in and of itself, costs $4980 and only runs on Apple Mac machines.

Moreover, the Stratos module has its limitations, as witnessed by Theoretica's explanation on the website. Quote:

"BACCH-Stratos does not decode Dolby Atmos® content. Instead, BACCH-Stratos accepts up to twelve discrete PCM audio channels that have already been decoded on the Mac by a Dolby-licensed playback application. Supported sources include Apple Music® and Apple TV®, as well as the Dolby Reference Player, which can play lossless Dolby Atmos audio and video files.'

So basically what it's saying here is you can use Apple sources for Apple Dolby Atmos content which Apple Macs are licensed to decode or you can buy the Dolby-Licensed playback app which will decode any Atmos bitstream files you can find and load onto an Apple machine, but not streaming sources such as Netflix, not ever.

The creation of those Dolby files is a Herculean task and was covered in a thread here which was, in turn, coverage of a thread on Audiophile Style detailing everything required to do that. The consensus was clearly that the juice was just not worth the squeeze.

So basically, that Stratos module is a great idea but it's $3 k on top of $5k to listen to Apple Atmos content over speakers. So, like everything Theoretica, a great idea but priced in the 'Stratos' sphere and having a highly limited use case.

The BACCH-sp will be getting stratos in June and having Apple Music for a source on the SP is not straightforward. I end up using ROON as my source and stream from tidal with my SP. I assume for stratos to work with the SP it will
Need to accept more than Apple Music. Also technically with b4m pro version can already do multichannel mixes. I had already tried 8 channel (7.1) content with my UCX II analog input. It worked well for the spatial aspect but since I don’t have my own HRTF the tonality was off. So there are other ways put stratos makes it simpler and better sounding. I hope with the SP possibly surround discs (or other surround sources) can be used.