Dr. Edgar Choueiri explains BACCH

[Davide]

These days I finally managed to get my system back online.
As I said initially, as an audio enthusiast I am quite curious to hear the effect of an XTC on my system.
So I was about to download HAF X-Talk, which I think is still reasonably priced. However, I realized that the uBACCH demo does not require any more effort to set it up, I felt that it could actually be a better test for experimenting with an XTC, being an algorithm theoretically without tonal/dynamic penalties.
So I put it into the system, which fortunately in my case is already set up to insert VST plugins, and therefore I don't have to purposely introduce further complication to the system (which is relevant for many).
According to what the manufacturer says, my DALI speakers have wide dispersion and my listening environment is very reflective (RT60 which reaches 700ms on 3kHz), furthermore it is asymmetrical. I necessarily use a DRC (Dirac or Audiolense) to correct problems below Schroeder, while above I have to limit it because the correction reduces its effectiveness (in area) and reduces the soundstage, which in fact normally benefits from early reflections.
By inserting uBACCH in this configuration, adjusted with the actual angle of the speakers, at first glance I noticed a slightly different spatial effect that depended from song to song. I couldn't say exactly in what terms, but it wasn't particularly better.
At some point I reasoned that room reflections are taking place in the XTC band, so I decided to toe-in the speakers and try using Audiolense to correct up to 6kHz with a rather wide FDW, in order to improve direct/indirect sound ratio in the listening position.
Then I adjusted uBACCH with a limited pink noise signal on the band affected by XTC, in order to obtain the most lateral sound possible by ear (as suggested on their site).
The effect obtained in this way was decidedly different. I experienced a much more defined and "3D" soundstage. Really nicer than the system without XTC, at least in that point.
This confirmed to me the dependence on the impulse response for XTC to be effective.
Unfortunately I don't have a binaural microphone so I can't see what's happening...
There is a further consideration: The fact that the soundstage obtained with only DRC up to Schroeder is narrower than that with XTC and DRC beyond Schroeder, and at the same time wider than that with only DRC beyond Schroeder, could mean that the reflections normally present compensate for the negative effect of crosstalk in recreating the soundstage.
By virtue of this, to compare the results with a more significant criterion, I created two projects in Reaper (which allows you to switch them on the fly). One with DRC up to Schroeder and the other with DRC up to 6kHz plus XTC.
From this comparison I can confirm that XTC and extended DRC gave me the best impression of spatiality.
Now, there is no point in thinking again about what is more correct because we agreed there is no single answer to the question. It's more a matter of personal paradigm.
The fact remains that I had not previously considered the possibility of using XTC in a complementary way to a DRC beyond Schroeder. From this perspective, I recognize that the variability of the result is lower than that theoretically expected with XTC alone, and that in any case the result is subjectively better.
Almost to the point of reconsidering my opinion on the cost, in view of the fact that for this amount it is difficult to obtain a better subjective improvement.
It would be nice to have more concrete objective elements to support this consideration, especially to understand the coexistence relationship between xtc and reflection with wide dispersion speakers and highly reflective room.
Being passionate about audio at heart, I think I will continue to investigate.

UPDATE:
After days of subjective test I have the idea that uBACCH can be, in my specific case, an appropriate complement to the correction of the speakers in terms of spatiality. With DRC I have never been able to improve in this sense, probably for the benefit of the reflections, in fact I limit it to Schroeder and planned to treat room. uBACCH improves the soundstage in a clear and broad way in practically all cases (I listen mainstream music, presumably with artificial spatial cue), although it does not provide that 360-degree audio effect of headphones with binaural coding and it works on about 20 cm per side in my case (the biggest disadvantage for the price to me).
I also checked with measurement outside the sweet spot and there is no FR alteration with it, so no additional penalties.
As a correct spatial reference/target cannot be conceived in most cases, for that relatively limited point where the IR is more optimized by the DRC and uBACCH is effective, it seems to me the latter could represent an advantageous investment for the mid band compared to the treatment of the room (which is relatively uncertain and unsightly).
I would be curious to understand if the flat directivity index to have a spectrum of reflections consistent with direct sound is still an important thing or not, and if the presence of reflections mitigates the extreme panning where present in the audio track.
Unfortunately, with other XTC I just tried I couldn't get the same good soundstage... I suppose that this is due to the peculiar technology that avoid tonal penalties and works at more traditional speaker angles, plus the feature to adjust the center channel gain.
With these criterion I talked a bit with my wife and after a practical reflection we agreed that this could be a better and cheaper compromise than the treatment of the room I was thinking about (for her immense joy). Especially because we are renting and in my house I am the only one who cares about the audio at the moment, so saving money for a single sweet spot makes sense.
So, fresh from this epiphany, not denying substantial subjective improvements and seeing the current discount I decided, a little impulsively, to pull the trigger, contravening the most objectivist part of me.
Objectively I was skeptical for the price due to the various reasons said (IR dependence, artificial audio tracks, single person limitation), but this convenience and subjectively good result unexpectedly prevail, in fact this makes me feel comfortable enough.
I believe this experience shows that the value of the technology cannot be found entirely on a theoretical and general level. It can be practical and user specific, as well as requesting proper setup. And also it is not very immediate to realize, especially for people adapted to the most canonical audio science.
Therefore, debating at theoretical/general level, however correct and argued, it is not particularly useful, as already mentioned. But overcoming the objective barrier is not immediate when something has an additional relatively high cost. I'm a little sorry but I suppose the usefulness of this discussion is in this regard.
However, this makes me curious to experience the Audiophile version to understand for the price how far can it improve in the dynamic sweet spot, and to remain ok outside and/or for others.
ASR should be useful in this investigation, but given the results so far, I took someone's suggestion and I tried to ask the company a few questions about.
As an audio enthusiast and basically fascinated by this refined technology, I would be curious to try it on my system, especially given this experience with Ubacch.
I will try to understand if I can risk losing those $200 for the trial.
In the meantime, I want to try experimenting with narrower directivity speakers (Klipsch) to see what subjectively you can get from something cheap, supported by BACCH technology.

 

[LIΟN]

At some point I reasoned that room reflections are taking place in the XTC band, so I decided to try using Audiolense to correct up to 6kHz with a rather wide FDW, in order to make the impulse response cleaner, albeit in a limited area based on frequency.
Then I adjusted uBACCH with a limited pink noise signal on the band affected by XTC, in order to obtain the most lateral sound possible by ear (as suggested on their site).
The effect obtained in this way was decidedly different. I experienced a much more defined and "3D" soundstage. Really nicer than the system without XTC, at least at that point.
This confirmed to me the importance of the impulse response for XTC to be effective.

You are right, frequency range is roughly 200-6000hz, which means that the gating response is roughly 5ms.
A sub-5ms response means that the response should be as pure as possible (without any Negative comb filtering).
Some impulse compensation (with FIR filters or Dirac, DRC, etc.) can match the impulse somewhat, but obviously it won't restore the altered response.
So the more uncontrolled (and therefore more variable) such a hyperbasic is, the more unpredictable it can be in XTC, where impulses from both sides are mixed.
(e.g., it should be similar to the original tonal balance, but it's not, or the left and right stages feel different when they should be, etc.)

 

[mieswall]

Hagen Quartett, Mozart K. 546, Fugue
https://open.qobuz.com/track/4254603

Here, the strings are extremely well focused and staged in their customary places both in width and depth. This is from a multi-CD set of all Mozart string quartets and contains recordings made at different times and different venues. The ones done at the Schloss Mondsee Festsaal, including this one, are the best ones in terms of pinpoint imaging. In this fugue the instruments enter one by one, which makes it easy to gauge their locations.

G R E A T !! the (aural) world doesn't have to be flat. And ORC here, at last in my setup, behaves particularly well (not always the case).
That's almost exactly how the real thing sounds in the front seats of a good, medium sized concert hall.

 

[doorofnight]

It only works if the recording is binaural.

[Edit]
To add some clarification to my response above a bit more, when I said "it only works" I mean it works as an accurate spatial reproduction of the sound. With non-binaural recordings, XTC will most often give a pleasant "enhancement" to the spatial feel of the reproduced sound over standard 2 channel stereo. But it is not going to be an accurate reproduction.

I agree about the more pleasant effect with normal (stereo) recordings, but I am not sure if I agree with the (not so) accurate reproduction due to Bacch. I've been listening to music in a few setups with mainly direct sound (in highly acoustically adapted rooms) without and with Bacch. I believe that in those rooms reproduction is quiet accurate, at least more accurate than in not treated rooms.

My room is acoustically treated (ceiling, walls, floor) but not that far-reaching as those rooms that I've been in, but I use Bacch for some time now. To my ears, the sound in my room comes closer (in terms of the 3D sound bubble) to the sound in highly acoustically adapted rooms now with Bacch. And yet, there is a difference. In those extreme treated rooms with direct sound the sound was more close to the headphone-effect. With Bacch it comes closer, but without the oppressive effect of a headphone. It's more open. I love the 3D bubble Bacch creates with normal recordings. It also enhances Q-sound effects in my setup.

I am not sure which situation (the headphone effect of mainly direct sound in higly damped rooms) or my situation (stereo with Bacch in a reasonable treated room) is more accurate, but there are more similarities than differences. So I guess, the Bacch effect in 2 channel stereo could give a more accurate reproduction than without.

 

[Sometimes I'm right]

In my experience, XTC processing is the biggest upgrade you can do for your box speaker stereo sound system.

For basic XTC software, this browser plugin is good start. https://magic.audio/web-extension-addon
For basic XTC hardware, MV Silicon's BP1064a2 DSP chip is an option. (As found in most Arylic products.)

XTC not only works for box speakers, it can improve soundbars also.

Interestingly, with open baffle speakers both delayed and out of phase signals radiate from the back side of the drivers by design.

 

[Gwreck]

In my experience, XTC processing is the biggest upgrade you can do for your box speaker stereo sound system.

For basic XTC software, this browser plugin is good start. https://magic.audio/web-extension-addon
For basic XTC hardware, MV Silicon's BP1064a2 DSP chip is an option. (As found in most Arylic products.)

XTC not only works for box speakers, it can improve soundbars also.

Interestingly, with open baffle speakers both delayed and out of phase signals radiate from the back side of the drivers by design.

Open baffle I suspect work well for XTC due to their dispersion pattern.

 

[AudioJester]

Open baffle I suspect work well for XTC due to their dispersion pattern.

Indeed. Gives the impression you have discrete channels next to and behind you

 

[Gwreck]

Indeed. Gives the impression you have discrete channels next to and behind you

This channel has binaural mixes of quad material and it’s pretty awesome with BACCH despite not lossless.

 

[jimbill]

What have you found to be the best camera for headtracking? Is there a good one with night vision?

 

[sweetsounds]

What have you found to be the best camera for headtracking? Is there a good one with night vision?

Bought an ELP infrared USB camera on Amazon.
First one died after a year, Amazon replaced it.
The 3.5mm lens is good if you place the camera close to you.

I found the model with an 8mm lens, which is good for a seating distance of 2.5m.

Nice thing is it switches between normal image sensing and illuminated infrared light in the dark.

Very happy.

https://www.aliexpress.com/item/32997227451.html

https://www.aliexpress.com/item/1005007675380320.html

 

[LIΟN]

I’m joining this thread again after a long time. Wishing everyone a happy and prosperous New Year in 2025!
This thread is related to Bacch, but since it’s the thread on ASR with the largest number of people interested in XTC, I’m leaving a post here. (To be precise, I also made a brief post in Tim Link’s personal XTC thread, but most XTC users are here.)

I’d like to hear the opinions of those who have experience with XTC regarding the points below.
This isn’t anything too profound—it’s just my experiment.
There are a few graphs and some text, so you might have to scroll down quite a bit. I apologize in advance for that.

In most reviews of XTC, the phrase "the speakers disappear" is quite common.
I also agree with this, but I was wondering if this was the intended outcome or if it was due to additional effects.

Starting briefly with the physical barrier, its purpose is to prevent the opposite ear (in this case, the right ear) from hearing the sound when the left speaker is playing.
It functions similarly to IEMs or headphones, delivering sound to only one ear.
However, in my opinion, this barrier blocks most of the ILD/ITD that naturally occurs when listening to speakers in real-life situations.
(If we assume that stereo speaker listening has its own identity, then that identity is hindered and lost due to the barrier.)

Therefore, array-based methods and basic recursive cross-iteration DSP XTC processing approaches ensure that both ears can hear the sound. (This applies regardless of whether it is universal or personalized.)
The cancellation signal leaves traces in the opposite ear, and it repeatedly reduces them to eliminate those traces.
But that’s unavoidable. In reality, we always hear with both ears, and it’s impossible to play a desired filter from one speaker and have it reach only one ear accurately.
However, since I can freely manipulate or synthesize each channel using BRIR impulses, I decided to give this a try.
(Originally, I also performed XTC using the traditional Bacch or Race methods.)

So, I recently tested this using the files of another individual I was assisting with calibration.

I extracted only the crosstalk corresponding to each path from his binaural room response.

The yellow line represents the right ear response from the left speaker (L_RightEar), while the blue line shows L_RightEar after cancellation.
And now, we need to check the combined response.
(The left ear and right ear responses from the left speaker.)

The Red graph represents the original combined response, while the Green graph shows the combined response after cancellation.
Although it’s not a perfect 30-degree stereo setup, the dip patterns resemble the typical stereo dip mentioned in Dr. Toole's book.
However, in the Green graph, you can see that these dips have completely disappeared.

Since this file is not mine, I couldn’t make an exact judgment, but even when listening to pink noise, the tonal balance differences were clearly noticeable.
I sent this to the owner of the file and waited a few days for their response.

I told him that when I sent the file, I approached it with the thought, “Could this feel like having a transparent head?” and decided to give it a try.

He then mentioned that when listening to his own BRIR, he could sense that his head was no longer blocking the sound, and it felt as though his head had become transparent. While the stereo image was maintained perfectly in front of him, at the same time, there was a sensation of sound passing through his brain—not his ears, but his brain.

I also recommended some music tracks and binaural tracksI also recommended tracks like a binaural car track where a car approaches from the front, passes by, and fades behind, or typical ASMR binaural tracks.) to him, and he was mostly satisfied and repeatedly emphasized his experience. He described it as feeling like the best aspects of speakers and headphones were combined, with contradictory sounds occurring simultaneously. His brain perceived the differences in distance and direction between these sounds, and he repeatedly stressed that this seemed paradoxical and almost incomprehensible.

And I asked him again.
I inquired whether there were any differences in the panning elements or perceived staging.
However, he mentioned that there wasn’t much change in the imaging.
The reason I asked this is that the graph I attached represents something impossible in reality—only the personalized crosstalk from the opposite ear path was precisely removed, and I didn’t add anything else.
I had anticipated that he might have impressions of specific panning elements, similar to the way people describe their experiences in reviews of general DSP XTC processing. However, that was not the case at all.
This led me to wonder if changes in certain panning elements or staging might be additional effects resulting from the repeated mixing of information between the two channels.
(Just to clarify, I appreciate and respect all XTC software, hardware, physical barriers, and theoretical approaches. I enjoy learning from them and have no negative thoughts about them whatsoever.)

You might say, “Isn’t it meaningless to discuss this since it’s something impossible in reality?” However, I’m curious to hear your opinions on these results.
I hope I’m not hijacking the OP’s thread. My apologies, and thank you for reading my post.