BACCH and other XTC

[LIΟN]

Thanks to @Lion who asked for actual measurements of attenuation of crosstalk cancellation of RACE measurements, here are the graphs. The complete graphs will be attached in a video as they are too many.

The best measurement is 26dB difference around a single point of 1277kHz.

View attachment 386959View attachment 386960View attachment 386961

I think the shape of the graph is more intuitive. Good.
Can I also see a change around 6000-8000hz? I've seen people using Bacch talk a lot about height awareness or expansion and I think it's due to a variation of 6000-8000hz (whether it's intended or unintended motion).
And as far as I know, Race has some restrictions at 6000hz.

Btw, his Race graph was beautiful, and the recorded tom's diner was also beautiful, like the motion of the beautiful graph.
Sound that is most true to the foundation of XTC

 

[STC]

I think the shape of the graph is more intuitive. Good.
Can I also see a change around 6000-8000hz? I've seen people using Bacch talk a lot about height awareness or expansion and I think it's due to a variation of 6000-8000hz (whether it's intended or unintended motion).
And as far as I know, Race has some restrictions at 6000hz.

Btw, his Race graph was beautiful, and the recorded tom's diner was also beautiful, like the motion of the beautiful graph.
Sound that is most true to the foundation of XTC

Initially, RACE added HP and LP filter as cancellation above 6000Hz (or was it 5000Hz) is counter productive given the length of the sound wave is extremely short so precise cancellation is not possible. I don’t think we can find exactly two speakers that are exactly the same. In theory, if we sit motionless ( deep freezing would be helpful here) and two identical speaker with exact sound waves used then we can have perfect cancellation with inversion of the signal.

You can see the attenuation is almost getting the same as without crosstalk here in the below. This is the reason why XTC can sound honky if not EQed. Others, simply limit the recursive numbers so that the higher frequencies not increased.

 

[STC]

I've seen people using Bacch talk a lot about height awareness or expansion and I think it's due to a variation of 6000-8000hz (whether it's intended or unintended motion).

Really? As far as I know stereo got no height info. The sense of perceived height is due to manipulation of HF. But like LEDR test tone we can always use DSP to create false sense of height with stereo. RACE is just about crosstalk cancellation of stereo playback.

 

[STC]

The complete graphs will be attached in a video as they are too many.

 

[LIΟN]

Really? As far as I know stereo got no height info. The sense of perceived height is due to manipulation of HF.

Maybe you saw it in another XTC bacch thread too?
I remember seeing a lot of times in some users' reviews saying, "The stage got bigger. The space got bigger." And in some cases, the height got better or higher. <--- I wrote it down in case you misunderstood it, but it's not my opinion, it's the words of some people I've seen

I also noticed it when I did XTC myself. The frequency actually changed. Maybe it wasn't the intended effect.
So whether it's good or not, I think it's true that it worked to make those people feel some effect differently from the original purpose of XTC (and this might also be true of the opinion that another bacch adds something info)

 

[LIΟN]

RACE added HP and LP filter as cancellation above 6000Hz (or was it 5000Hz

Good Info. I also saw XtalkShaper.
I couldn't confirm what filter configuration it consists of.
And what you've seen in many papers, including Bacch
What keeps changing and being mentioned was that the kind of "filter" that is applied and produced as both channels are mixed should be weighted.
In some papers, it is applied separately by dividing it into a kind of multi-band, and in the case of bacch, weights are applied by frequency depending on the parameters.
(Filters go in harder at some frequencies and weaker at others).
In the end, it seems that everyone's interest is to increase performance while minimizing the coloration of the XTC itself, what kind of filter to use and how much strength to mix it with.
So I think Race's cutoff frequency selection is very right.

 

[STC]

I also noticed it when I did XTC myself. The frequency actually changed. Maybe it wasn't the intended effect.
So whether it's good or not, I think it's true that it worked to make those people feel some effect differently from the original purpose of XTC (and this might also be true of the opinion that another bacch adds something info)

The answer rather simple. Be it recursive or IR based cancellation, what you do you add delayed and attenuated signal to the original and an inverted signal also attenuated and delayed to the other side.

So changes in frequency response is something to be expected. What’s important is how much of it audible enough and can it be corrected with EQ?

 

[LIΟN]

What’s important is how much of it audible enough and can it be corrected with EQ?

Nope. I'm almost sure EQ can't fix that. That's the biggest problem. (If it is caused by EQ, it can be corrected with EQ, but since it is a mixture of information from both sides, not EQ, it is very mission-impossible to correct it with EQ, considering both before and after mixing) :: We should also consider how both ears will accept it. Why are our ears different? Why on earth! lol
What I'm really sure is that the strongest and most effective way is to find the "cause" that would cause problems before and after XTC and make changes to it.
Of course, we can do some tonal balance matching with EQ, but it's complicated.
I think I'm happy to get rid of those variables as much as possible. ...

Btw, It's amazing that you made a graph like that with Excel. It's really intuitive.

 

[STC]

Nope. I'm almost sure EQ can't fix that. That's the biggest problem. (If it is caused by EQ, it can be corrected with EQ, but since it is a mixture of information from both sides, not EQ, it is very mission-impossible to correct it with EQ, considering both before and after mixing) :: We should also consider how both ears will accept it. Why are our ears different? Why on earth! lol
What I'm really sure is that the strongest and most effective way is to find the "cause" that would cause problems before and after XTC and make changes to it.
Of course, we can do some tonal balance matching with EQ, but it's complicated.
I think I'm happy to get rid of those variables as much as possible. ...

Btw, It's amazing that you made a graph like that with Excel. It's really intuitive.

You cannot correct with EQ and I agree with. But what’s matter is what you hear. Audiophiles will never agree to this but as long it sound good and better than plain stereo I am all for it.

 

[STC]

Experimenting with IR cancellation with RACE surprisingly put the rear image of binaural recordings well behind the head. Indeed it is much better than RACE alone implemention. I think you can get better cancellation if you do ear measurements and modify the IR.

 

[Tim Link]

Listening to Xtalk Shaper I'm hearing a nice widened soundstage, but more importantly to me the phantom center is great! That's never the case to my ears with a normal stereo triangle.

 

[Keith_W]

Listening to Xtalk Shaper I'm hearing a nice widened soundstage, but more importantly to me the phantom center is great! That's never the case to my ears with a normal stereo triangle.

Did you have a chance to compare it to uBACCH? It is available as a free trial. A comparison would be very helpful!

 

[Tim Link]

I tired uBACCH but it was having some problem causing distortion. I never got it figured out and now the trial period is expired. Xtalk Shaper is on demo mode which seems to just cut the sound out temporarily on a regular basis. I prefer that because it gives me a while to demo it, make sure it works right. My only complaint is that it seems to add about 300ms latency. Won't be able to use it for movies or games.

I've got a crosstalk barrier setup at work and I sitll think it takes the cake for bringing out details in a recording, especially ambient cues, but it's hard to tell for sure because that room is so different than what I've got here. Also I'm using little bookshelf speakers there and big horns here. The room reflections at work are completely under control, speakers and my listening position are far from any wall, plus the barrier prevents sound from the left speaker from bouncing off the right wall and vice versa. That kind of crosstalk is also fairly destructive.

Xtalk Shaper is making everything sound very solid and clear. Nothing ghostly or weird sounding about it. Regular stereo sounds weirder to me by a significant margin. Testing it with left to right panning is giving me a near full 180 soundfield. Most recordings don't sound so extreme though. Tonally the overall impression is a big improvement, which has not been my experience with past recursive crosstalk programs. They've definitely gotten better, or this is just happens to gell with my system.

 

[Tim Link]

The answer rather simple. Be it recursive or IR based cancellation, what you do you add delayed and attenuated signal to the original and an inverted signal also attenuated and delayed to the other side.

So changes in frequency response is something to be expected. What’s important is how much of it audible enough and can it be corrected with EQ?

Changes in frequency response should be expected and desired. Crosstalk screws up the response progressively worse as things are panned toward the phantom center. That should change for the better with XTC. I'm clearly hearing it work with Xtalk Shaper, and if that's all it accomplished I'd be pretty happy! One thing that helps the phantom center is moving the speakers much closer together, which happens to work best with DSP XTC and physical barriers too. Because of that, it's hard to hear as much of a difference in the phantom center when comparing bypassed mode because the problem has already been largely corrected. So, once the speakers are spaced closely and the XTC is working well, it's fair to say that the phantom center tone shouldn't change much. The soundstage will just get wider and the center will sound more natural within that soundstage.

 

[Stefano_Milano]

Changes in frequency response should be expected and desired. Crosstalk screws up the response progressively worse as things are panned toward the phantom center. That should change for the better with XTC. I'm clearly hearing it work with Xtalk Shaper, and if that's all it accomplished I'd be pretty happy! One thing that helps the phantom center is moving the speakers much closer together, which happens to work best with DSP XTC and physical barriers too. Because of that, it's hard to hear as much of a difference in the phantom center when comparing bypassed mode because the problem has already been largely corrected. So, once the speakers are spaced closely and the XTC is working well, it's fair to say that the phantom center tone shouldn't change much. The soundstage will just get wider and the center will sound more natural within that soundstage.

I also confirm. I have been using Xtalk Shaper for months and I am very satisfied. I do not notice any problems with distortion or timbres.

 

[Tim Link]

I also confirm. I have been using Xtalk Shaper for months and I am very satisfied. I do not notice any problems with distortion or timbres.

This plug-in sounds good to me but I'm not going to be able to use it because I game through my system and the latency is too high for that. Overall I'm finding that my wide speaker spacing is providing a good phantom center, with very good tonal balance across the sound field as tested by panning left ot right. Xtalk actually was rolling off the highs towards the edge of the soundfield. That might have been because I didn't have it fully dialed in. Anyways, I liked it a lot.

 

[boxerfan88]

Can u guys recommend some tracks to test XTC?

Thus far these two seem to show the effects quite significantly:
* New Rules - Dua Lipa
* Lovesong - Adele

 

[Gwreck]

Kraftwerk especially Tour de France

 

[doorofnight]

Kraftwerk especially Tour de France

And all songs in this list:

Open Qobuz

open.qobuz.com

Or this one:

 

[fcserei]

Hi
I am new to this forum.
I am using Ambiophonics/DCH with some breaks for almost 20 years, since my first demo and consultations with Mr. Glasgal. He ruined stereo for me forever.

Currently I use small Martin Logan electrostats and 2 12" subs in panambio setup at 15 degree front and around 20 back, with about 3' from speakers to head distance.
I have a small room, not much room treatment. I usually listen alone in my comfy chair, sweet spot size is not an issue. It is big enough and not sensitive for head movements.

I've tried all kind of XTC algorithms (and the mattress) over the years and stayed with the full range RACE at the end. The XTLALKShaper and BACCH has all have weird ringing to my ear. ( see the 3 impulse responses for RACE, XTS and BACCH for the same 15 degree speaker config below. Note the different scales, the RACE is sample level , the others several ten milliseconds with lots of pre and post ringing.)
I have no problem with tonality change for RACE. Adjusting the delay parameter to the speaker setup is the most important to reduce artifacts. I've ended up in the 30uS range. Panning the same mono source to mono, XTC center and XTC sides preserves the tonality IF you turn your head towards the source. Of course tonality changes if your head is fixed and the source moves. As Mr. Glasgal said, it should be only one pinna response in the whole record/reproduction chain and it better be yours.
Also, the rear XTC speaker pair have a big effect on increasing realism ( fooling your ears), reducing effects of head movement, and as you will see later, creating ambience.

On the Domestic Concert Hall side, it is a very important part of the realistic reproduction. 2 channel recordings can not preserve the correct direct/reflected sound ratio of a real acoustic event, because if you try to record the full 360deg sound field and downmix it with the proper power ratio to the front stage, the record will sound like it was made in a bathroom.
The same way if your listening room is very dead, most record will sound very dry, because you only get the reduced level reflections only from the front and, way down from the proper direct/reflected ratio. Most regular listening rooms have 0.8-1.5 reverb time which will reinforce the record's lacking ambience, but with usually the wrong tonality, direction and power. A nice listening room can be similar acoustic environment to a packed small club. No wonder the preferred audiophile demo genre is jazz or girl/boy with a guitar in a small venue.
Over the years I've tried many multi speaker simulations. The usually work unbelievably well to transpose you to a real event, but two downsides. One: complexity with the extra channels, two: you need lots of decorrelated channels or your ear will start to pick up the ambience speaker locations and they will not blend in well.
Nowadays I use only the 4 main speakers and mix in convolver generated front and rear ambience before the XTC. This is for me the best compromise between simplicity and realism. Attached the schematics of my signal processing in Element on Mac.

As for testing XTC, I am classical music fan, but try the Doors at the Hollywood Bowl on YouTube. At the 9 min mark when the music dies and the audience starts to shout in several directions, it is scary "beeing there" with XTC and a generated big venue ambience.