-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
CTXMatrix - Free Ambiophonics Plugin
- Thread starter STC
- Start date
Keith_W
Master Contributor
Hello @STC. I tried the calibration again tonight. I suspected that most of the coloration I was hearing was due to the gain being set too high. So I changed my strategy - I set the gains low: -7.5 for left and right, and -3 for the crosstalk cancellation. I then adjusted the delays until the phantom image was at the occlusion point.
I also found that lower gain OR more delay: moves the phantom image laterally.
Step 1: Gain -7.5 Delay 0.02
Step 2: Gain -3 Delay 0.01
Step 3: Gain -7.5 Delay 0.02
Step 4: Gain -3 Delay 0.01
After the calibration, I listened again. MUCH better - but there is still noticeable colouration. Previously, I noted that the top end was shrill. Now there is too much bass, as if reverb has been added. I will try again with even less gain, but I suspect that I won't be able to dial a low enough delay to pull the phantom image to the occlusion point.
Anyway, I have some measurements for you. These are Natural vs. uBACCH vs. CTXMatrix. All measurements are of the left speaker. I didn't bother moving furniture out of the way to take the measurement, which is why things look a bit wonky.
Since you asked for pink noise, I compared the sweep with pink noise. They are nearly identical. Since I prefer looking at sweeps, I will only show sweeps from now on.
This is a LEFT sweep from 20Hz - 20kHz with both speakers active. Red is natural, green is uBACCH, and blue is CTXMatrix. You can see that uBACCH does not change the tonality very much. And this sweep confirms that my current settings with CTXMatrix produces way too much bass ... about 7dB louder than "natural", and it extends all the way up to 300Hz.
This is a LEFT sweep from 20Hz - 20kHz with the right speaker muted. You can see that the tonality of uBACCH and "Natural" is close to identical. CTXMatrix has more bass and less treble than either uBACCH or "Natural".
This is a LEFT sweep from 20Hz - 20kHz with the left speaker muted. In other words, the purpose of this sweep is to look at the crosstalk cancellation. uBACCH is in darker orange, and CTXMatrix is in yellow. In faint red is the "Natural" sweep with both speakers active for a SPL comparison.
I also found that lower gain OR more delay: moves the phantom image laterally.
Step 1: Gain -7.5 Delay 0.02
Step 2: Gain -3 Delay 0.01
Step 3: Gain -7.5 Delay 0.02
Step 4: Gain -3 Delay 0.01
After the calibration, I listened again. MUCH better - but there is still noticeable colouration. Previously, I noted that the top end was shrill. Now there is too much bass, as if reverb has been added. I will try again with even less gain, but I suspect that I won't be able to dial a low enough delay to pull the phantom image to the occlusion point.
Anyway, I have some measurements for you. These are Natural vs. uBACCH vs. CTXMatrix. All measurements are of the left speaker. I didn't bother moving furniture out of the way to take the measurement, which is why things look a bit wonky.
Since you asked for pink noise, I compared the sweep with pink noise. They are nearly identical. Since I prefer looking at sweeps, I will only show sweeps from now on.
This is a LEFT sweep from 20Hz - 20kHz with both speakers active. Red is natural, green is uBACCH, and blue is CTXMatrix. You can see that uBACCH does not change the tonality very much. And this sweep confirms that my current settings with CTXMatrix produces way too much bass ... about 7dB louder than "natural", and it extends all the way up to 300Hz.
This is a LEFT sweep from 20Hz - 20kHz with the right speaker muted. You can see that the tonality of uBACCH and "Natural" is close to identical. CTXMatrix has more bass and less treble than either uBACCH or "Natural".
This is a LEFT sweep from 20Hz - 20kHz with the left speaker muted. In other words, the purpose of this sweep is to look at the crosstalk cancellation. uBACCH is in darker orange, and CTXMatrix is in yellow. In faint red is the "Natural" sweep with both speakers active for a SPL comparison.
Last edited:
Keith_W
Master Contributor
I did some further processing of the graphs so that we can see more of what is happening.
The purpose of this graph is to look at equalization. To do this, I did Trace Arithmetic A/B. The black line is "Natural" divided by itself - it produces a flat line. Any deviation from the flat line is due to equalization. This is a sweep of the LEFT speaker with BOTH speakers playing. See for yourself the difference in equalization - both uBACCH and CTXMatrix are louder than "Natural", but uBACCH has equalized the bass downwards.
What I find interesting is that the loudness peaks produced by both uBACCH and CTXMatrix are the same. I don't have an explanation for that, maybe you do.
Anyway, let's look at individual measurements of the speakers.
This is LEFT speaker alone, with the right speaker muted (i.e. no crosstalk cancellation). You can see that uBACCH is almost flat, whilst CTXMatrix has noticeably more bass and less treble.
Now let's look at the crosstalk cancellation. This is a LEFT sweep with uBACCH. Blue is the left speaker with the right speaker muted, and yellow is the crosstalk cancellation from the right speaker (left speaker muted).
And this is CTXMatrix. It appears to me that uBACCH keeps the left signal the same and equalizes the crosstalk cancellation. CTXMatrix uses almost the same signal for crosstalk cancellation.
If you want to look at all the measurements, they are available in my Google Drive. Link here.
The purpose of this graph is to look at equalization. To do this, I did Trace Arithmetic A/B. The black line is "Natural" divided by itself - it produces a flat line. Any deviation from the flat line is due to equalization. This is a sweep of the LEFT speaker with BOTH speakers playing. See for yourself the difference in equalization - both uBACCH and CTXMatrix are louder than "Natural", but uBACCH has equalized the bass downwards.
What I find interesting is that the loudness peaks produced by both uBACCH and CTXMatrix are the same. I don't have an explanation for that, maybe you do.
Anyway, let's look at individual measurements of the speakers.
This is LEFT speaker alone, with the right speaker muted (i.e. no crosstalk cancellation). You can see that uBACCH is almost flat, whilst CTXMatrix has noticeably more bass and less treble.
Now let's look at the crosstalk cancellation. This is a LEFT sweep with uBACCH. Blue is the left speaker with the right speaker muted, and yellow is the crosstalk cancellation from the right speaker (left speaker muted).
And this is CTXMatrix. It appears to me that uBACCH keeps the left signal the same and equalizes the crosstalk cancellation. CTXMatrix uses almost the same signal for crosstalk cancellation.
If you want to look at all the measurements, they are available in my Google Drive. Link here.
Last edited:
- Thread Starter
- #24
Thank you so much Keith for the measurements. Now, I have a measurement to work with. Since you are using -7.5 dB gain then the value which I set -8dB in Beta is the best. But I understand that some users may not have large amplifier to compensate the 8dB loss . To address this I did EQ in Beta1.1 which I have not perfected yet. Your measurements now given me invaluable data to fine tune the plugin.
If you find the time please do the measurements for Beta which is what I am using now to built one with band cancellation like BACCH. My previous attempt did not sound as good as the plain Beta but now I have an idea how to address the issue.
Try to do measurements with same attenuation value and only adjust the delay to match the occlusion points. Also try Beta which is without any EQ. That measurements would be useful for the final version. I believe Beta should match much better to uBacch than Beta1.1.
You have no idea how much you have contributed for the development. Thank you once again.
Keith_W
Master Contributor
@STC of course you are welcome. What you are doing is awesome, and I would love to do anything to help.
But I don't think these measurements tell the full story. Like I said, if I use CTXMatrix with different delay settings, the colouration sounds different. My earlier attempts produced too much treble, and the current attempt produces too much bass. uBACCH sounds tonally correct no matter what half-span angle I apply, its only effect is with the soundstage. I know from experience from other free Ambisonics plugins I have tried in the past (that you and I discussed in emails a year ago) that wrong settings absolutely ruins the tonality. This suggests to me that uBACCH has some kind of dynamic equalization going on.
I do not understand enough about crosstalk cancellation to know why this might be the case. I think the way to investigate it is to redo the calibration with max delay / min gain and min delay / max gain. Would you be able to suggest a suitable test?
Also, I might be able to play with the matrix settings on my interface so that we can look at the output of the DAC instead of acoustic measurements. That might give us cleaner curves to look at. Would this be preferable to acoustic measurements?
But I don't think these measurements tell the full story. Like I said, if I use CTXMatrix with different delay settings, the colouration sounds different. My earlier attempts produced too much treble, and the current attempt produces too much bass. uBACCH sounds tonally correct no matter what half-span angle I apply, its only effect is with the soundstage. I know from experience from other free Ambisonics plugins I have tried in the past (that you and I discussed in emails a year ago) that wrong settings absolutely ruins the tonality. This suggests to me that uBACCH has some kind of dynamic equalization going on.
I do not understand enough about crosstalk cancellation to know why this might be the case. I think the way to investigate it is to redo the calibration with max delay / min gain and min delay / max gain. Would you be able to suggest a suitable test?
Also, I might be able to play with the matrix settings on my interface so that we can look at the output of the DAC instead of acoustic measurements. That might give us cleaner curves to look at. Would this be preferable to acoustic measurements?
- Thread Starter
- #26
@Keith_W , with different settings there is no proper cancellation. Without proper cancellation coloration sets in even when you can feel the spacious stage or 3D. They are not crosstalk cancellation but effects due to out of phase and delay which mastering engineer use all the time. Measurements wise you can see coloration but can you hear them? I know my first beta, CTXMatrixBeta have a build up of high frequencies peaking around 12000 Hz but I didn’t want to address them as half of us hardly could hear at the frequency.
You observation of elevated bass and lost HF is correct because in the other beta version CTXMatrixBeta1.1 ( not CTXMatrixBeta) which you did this measurement indeed was eq’ed at that region as I made that for you since you wouldn't move the speakers and to add the gain as the clippings is caused by frequencies at that region. It is still my position that you cannot suppress the one ear directional cue when sound comes from 30 degrees without canceling the natural pinna frequency shaping. And to do that you are no longer playing with the original sound but a pre EQ’ed sound. One member here created the impulse response for such approach. It was tonally so perfect but you need mics to measure the impulse response and need a convolution engine to do the crosstalk. Bacch approach is also similar as it uses impulse response.
OTOH, RACE uses the original signal without modification but since there is no pinna compensation the speakers position at about 20 degrees is critical. But with band cancellation maybe a more neutral measurements achievable. Will try that first with my system and then one for typical speakers position if successful. but all versions that’s used for 60 degrees speakers were not tested so I am really have to depend on users input to address those.
Thank you.
You observation of elevated bass and lost HF is correct because in the other beta version CTXMatrixBeta1.1 ( not CTXMatrixBeta) which you did this measurement indeed was eq’ed at that region as I made that for you since you wouldn't move the speakers and to add the gain as the clippings is caused by frequencies at that region. It is still my position that you cannot suppress the one ear directional cue when sound comes from 30 degrees without canceling the natural pinna frequency shaping. And to do that you are no longer playing with the original sound but a pre EQ’ed sound. One member here created the impulse response for such approach. It was tonally so perfect but you need mics to measure the impulse response and need a convolution engine to do the crosstalk. Bacch approach is also similar as it uses impulse response.
OTOH, RACE uses the original signal without modification but since there is no pinna compensation the speakers position at about 20 degrees is critical. But with band cancellation maybe a more neutral measurements achievable. Will try that first with my system and then one for typical speakers position if successful. but all versions that’s used for 60 degrees speakers were not tested so I am really have to depend on users input to address those.
Thank you.
- Thread Starter
- #27
Actually, all of them are crosstalk cancellation values. But looking at the values, they were not correctly adjusted. I know it is hard. Actually, I don want to put any values to be shown in the dialogue box so that users will have to rely on their hearing but that is so difficult even for me. I have to figure a way to do this reliably. Try playing at very low level during the occlusion stage to determine the exact spot. but since brain depends on both cues both values can vary. So you may get an exact occlusion point with -2.5dB/ 90 μs or -4dB/30 μs for 20 degrees setup. Or a combination of those numbers. Mathematically, the first should be the correct value but that depends on your system, ears and room. When you have conflicting level due to imperfect cancellation of different frequencies the delay takes precedence more than the attenuation.
Maybe, I should just drop the idea as this is getting too complicated to set up the stages without supervision.
Keith_W
Master Contributor
@STC this is what I think: the average user will adjust gain and hear the phantom image shift. He will adjust the delays, and hear the phantom image shift again. The instructions say to adjust delays and gains until the phantom image matches the occlusion point. But should you adjust delays? Adjust gains? Which should you prioritize? I found that reducing the gain as much as possible reduced the colouration, but that left me with not enough delay. So I think that the way forward is to find a way to do the calibration properly without affecting the tonality.
I was busy tonight, and I will also be busy tomorrow. I'll find some time to test the new beta and report back.
I was busy tonight, and I will also be busy tomorrow. I'll find some time to test the new beta and report back.
I've been playing with XTC for a long-long time, tried everything from a mattress to all kind of plugins.
Recently I've made a full sphere SOFA file of my personal HRTF and I thought I would simulate with a computer what are the theoretical limits of the different XTC methods. With the simulation I can get exact signals at the entrance of my ear canal without any room interaction, and evaluate how does this correspond to virtual and direct sources in different directions.
Every measurement below is using my personal HRIR simulation at the ear canal entrance and corrected with my diffuse field curve, so FLAT response IS NOT the desired response. It is more for comparison for different responses from different direction. I am using only the left ear, supposing that the HRIR is symmetrical, ( which is not), but it does not substantially affect the conclusions. I've processed the measurement signal with the XTC plugins then ran it through the SOFA convolution to simulate the summed response at the ear.
I compared the XTC solution playing L or R only signal with the perceived sound at the L ear with a real speaker at 60 deg, and playing mono signal in both channels with the sound of a real speaker at the center.
First the basic +-30 degree stereo triangle:
1. L 30deg: Left speaker heard at the left ear
2. R 30deg: Right speaker heard at the left ear
3. C 30deg: Center speaker heard at the left ear
4. L+R -6dB 30deg: Amplitude corrected virtual Center speaker heard at the left ear
Crosstalk cancellation from head shadowing is ~5db in the 600-8K range. in a room it is worse than that.
Here we can also see the famous BBC dip (green vs yellow) around 1.8k, because of the comb filtering from the 2 speakers producing the virtual center.
This is what is happening with a real Center speaker at the left ear when I turn my head left or right:
Basically the smooth change in FR tells us a little horizontal shift in the source position, No dramatic change especially in the position of the HF dips.
In case of a virtual Center speaker (L and R playing the same signal):
The BBC dip remains but shifts, below and above some considerable changes are coming in FR an in the HF both the position and volume of the dips are changing. These are causing tonality changes and localization confusion. It is getting messy, not something you can EQ out. And this is with keeping the center position, just turning your head, moving to the left or right is much worse.
Let's see the crosstalk cancellations. All the measurement below are made with the simulated speakers moved to the +- 10 deg position.
First the simplest one, a barrier between the speakers. For this measurement I simulated the barrier as progressively increasing damping above 200 Hz up to 15 dB at 20kHz.
Considerable damping between the ears, but how does this compare to a real speaker at +60, basically to a double wide soundstage compared to the stereo triangle. Crosstalk cancellation is usually promising wider soundstage.
Actually not that bad. The closer ear is not loud enough, because at 60 deg the left speaker would have a straight shot at the ear, while the real speaker at +10 degree is a little shadowed by the head. The other ear gets somewhat similar signal what would come from a real speaker on the other side at the +60 degree position. The HF peaks and valleys are different which means the virtual images created with the barrier are not too resilient to head rotation. This is the effect Mr. Glasgal wanted to overwhelm with the pan-ambio setup, where he put an other XTC speaker pair in the back to confuse the HF pinna cues. Center images have no problem because the speakers are basically in the center and playing non-processed signal.
Lets see Bacch (Bacch plugin, 5degree half-span angle, 0 center boost:
It shows the Bacch philosophy, pushing the XTC to as low as possible. It creates >15dB XTC in the 200-5000 Hz range. But it also shows what I always felt, the Bacch's lack of impact in the midbass. The separation effect is great, but the sound is very lightweight.
Compared to a real speaker at 60deg, it has the right amount of cancellation above 1k, but below 1kHz it is much more what we can ever hear in real life. But pumping that much energy in to XTC that low causes a cancellation level drop. I think they hope that the off axis response of the stereo dipole together with the room will boost this range to the proper level, but I've measured the same hole in my room with real speakers and microphones.
Same funny business is going on with the center image. Also, it is a bit too hot in the 1-4k range, otherwise would be ok.
Neutron Player:
I can't test the CTXMatrix plugin, as I am running on a Mac, but the Neutron is also a straightforward RACE implementation.
Effective XTC range 400-4k, ~10dB.
General tonality is not bad. Little less separation at the ear than the real speaker at 60 deg, but not horrible. Rolls off above 4k too on the speaker side, but too high on the other side, so in a real room it might not feel lacking in HF.
What is good is the center image, it tracks the real center speaker very well up 3k. Above that the important dips areas 8 and 10K are the right position, but the level is wrong. This is one of the reason why race was originally bandwidth limited the XTC effect to 400-4000Hz. This simulation Is full range, the peaks and dips in the HF can be reduced with that limiting, or the pan-ambio setup mentioned at the barrier setup.
Level diffs above 3k between L and C image are in the same direction, you can try to correct it using EQ.
Worth mentioning that Prof Angelo Farina's method of measuring in ear HRTF for the +-10 deg speaker position and calculating an amplitude limited inverse full stereo filter gives in trend very similar curves to the RACE.
Finally XtalkShaper:
Lots of separation in the 400-4000 range, but a bit jagged. But also there is a huge latency from the plugin.
.
It exceeds what we really hear.
Center is good, but HF rolloff starts early. There is a similar drop in the L only signal in the 4-8k range, so general EQ boost in this range can help with tonality.
All this seems a bit of gloom and doom at the first sight, but have a look at the next graph. It is Neutron with the head turned left and right, but the other XTCs behave similarly:
Smooth and gradual changes on the speaker side, preserved HF dips and still good XTC results on the other side. So if you can get good tonality with the XTC and EQ, it will retain good imaging and consistent tonality even with head turns and side movements too. Not like with the stereo triangle.
So that's it for today. Make your own conclusions. Mine was that XTC works as a part of the solution, which is to simulate a venue with multichannel convolution, and place the direct sound in this sound field with XTC (ambiophonics). Tonality comes from the (simulated) room as much as from the direct signal.
Recently I've made a full sphere SOFA file of my personal HRTF and I thought I would simulate with a computer what are the theoretical limits of the different XTC methods. With the simulation I can get exact signals at the entrance of my ear canal without any room interaction, and evaluate how does this correspond to virtual and direct sources in different directions.
Every measurement below is using my personal HRIR simulation at the ear canal entrance and corrected with my diffuse field curve, so FLAT response IS NOT the desired response. It is more for comparison for different responses from different direction. I am using only the left ear, supposing that the HRIR is symmetrical, ( which is not), but it does not substantially affect the conclusions. I've processed the measurement signal with the XTC plugins then ran it through the SOFA convolution to simulate the summed response at the ear.
I compared the XTC solution playing L or R only signal with the perceived sound at the L ear with a real speaker at 60 deg, and playing mono signal in both channels with the sound of a real speaker at the center.
First the basic +-30 degree stereo triangle:
1. L 30deg: Left speaker heard at the left ear
2. R 30deg: Right speaker heard at the left ear
3. C 30deg: Center speaker heard at the left ear
4. L+R -6dB 30deg: Amplitude corrected virtual Center speaker heard at the left ear
Crosstalk cancellation from head shadowing is ~5db in the 600-8K range. in a room it is worse than that.
Here we can also see the famous BBC dip (green vs yellow) around 1.8k, because of the comb filtering from the 2 speakers producing the virtual center.
This is what is happening with a real Center speaker at the left ear when I turn my head left or right:
Basically the smooth change in FR tells us a little horizontal shift in the source position, No dramatic change especially in the position of the HF dips.
In case of a virtual Center speaker (L and R playing the same signal):
The BBC dip remains but shifts, below and above some considerable changes are coming in FR an in the HF both the position and volume of the dips are changing. These are causing tonality changes and localization confusion. It is getting messy, not something you can EQ out. And this is with keeping the center position, just turning your head, moving to the left or right is much worse.
Let's see the crosstalk cancellations. All the measurement below are made with the simulated speakers moved to the +- 10 deg position.
First the simplest one, a barrier between the speakers. For this measurement I simulated the barrier as progressively increasing damping above 200 Hz up to 15 dB at 20kHz.
Considerable damping between the ears, but how does this compare to a real speaker at +60, basically to a double wide soundstage compared to the stereo triangle. Crosstalk cancellation is usually promising wider soundstage.
Actually not that bad. The closer ear is not loud enough, because at 60 deg the left speaker would have a straight shot at the ear, while the real speaker at +10 degree is a little shadowed by the head. The other ear gets somewhat similar signal what would come from a real speaker on the other side at the +60 degree position. The HF peaks and valleys are different which means the virtual images created with the barrier are not too resilient to head rotation. This is the effect Mr. Glasgal wanted to overwhelm with the pan-ambio setup, where he put an other XTC speaker pair in the back to confuse the HF pinna cues. Center images have no problem because the speakers are basically in the center and playing non-processed signal.
Lets see Bacch (Bacch plugin, 5degree half-span angle, 0 center boost:
It shows the Bacch philosophy, pushing the XTC to as low as possible. It creates >15dB XTC in the 200-5000 Hz range. But it also shows what I always felt, the Bacch's lack of impact in the midbass. The separation effect is great, but the sound is very lightweight.
Compared to a real speaker at 60deg, it has the right amount of cancellation above 1k, but below 1kHz it is much more what we can ever hear in real life. But pumping that much energy in to XTC that low causes a cancellation level drop. I think they hope that the off axis response of the stereo dipole together with the room will boost this range to the proper level, but I've measured the same hole in my room with real speakers and microphones.
Same funny business is going on with the center image. Also, it is a bit too hot in the 1-4k range, otherwise would be ok.
Neutron Player:
I can't test the CTXMatrix plugin, as I am running on a Mac, but the Neutron is also a straightforward RACE implementation.
Effective XTC range 400-4k, ~10dB.
General tonality is not bad. Little less separation at the ear than the real speaker at 60 deg, but not horrible. Rolls off above 4k too on the speaker side, but too high on the other side, so in a real room it might not feel lacking in HF.
What is good is the center image, it tracks the real center speaker very well up 3k. Above that the important dips areas 8 and 10K are the right position, but the level is wrong. This is one of the reason why race was originally bandwidth limited the XTC effect to 400-4000Hz. This simulation Is full range, the peaks and dips in the HF can be reduced with that limiting, or the pan-ambio setup mentioned at the barrier setup.
Level diffs above 3k between L and C image are in the same direction, you can try to correct it using EQ.
Worth mentioning that Prof Angelo Farina's method of measuring in ear HRTF for the +-10 deg speaker position and calculating an amplitude limited inverse full stereo filter gives in trend very similar curves to the RACE.
Finally XtalkShaper:
Lots of separation in the 400-4000 range, but a bit jagged. But also there is a huge latency from the plugin.
It exceeds what we really hear.
Center is good, but HF rolloff starts early. There is a similar drop in the L only signal in the 4-8k range, so general EQ boost in this range can help with tonality.
All this seems a bit of gloom and doom at the first sight, but have a look at the next graph. It is Neutron with the head turned left and right, but the other XTCs behave similarly:
Smooth and gradual changes on the speaker side, preserved HF dips and still good XTC results on the other side. So if you can get good tonality with the XTC and EQ, it will retain good imaging and consistent tonality even with head turns and side movements too. Not like with the stereo triangle.
So that's it for today. Make your own conclusions. Mine was that XTC works as a part of the solution, which is to simulate a venue with multichannel convolution, and place the direct sound in this sound field with XTC (ambiophonics). Tonality comes from the (simulated) room as much as from the direct signal.
Attachments
Last edited:
- Thread Starter
- #31
Thank you so much for the measurements; they will be useful for fine-tuning the plugin, especially the physical barrier measurements. I don’t recall doing this with DPA mics, but I think my measurements with the Roland mic were not as good as yours with the physical barrier.
I was supposed to respond much earlier but got tied up. As far as I remember, PanAmbio was designed specifically for the 4.0 format. Ralph couldn’t get anyone to research its role in stereo, although both he and I observed that visitors preferred it with the rear Ambio on, even for two-channel stereo. This likely has to do with added envelopment, which becomes redundant when using DCH. I’d appreciate any reference where Ralph stated it was meant to smooth pinna cues.
@Keith_W , I need time to think which should be done first. My observation can be tainted as I have been doing this for a long and may have developed some preference over the method. Just hoping to find some fresh ears who could spend few hours to be sure whether delay or attenaution should be first. Difficult to find anyone interested. Maybe, I should say cable blind test to have more people intersted.
- Joined
- May 12, 2019
- Messages
- 231
- Likes
- 104
Hi @STC! I just had the displeasure of reading through the 2017-18 ambio thread on SBAF, and aside form the sad reminder of what "moderator" behavior was like on there, one technical tidbit also stuck with me: you said XTC wouldn't work with soundbars. Why not? AFAIK they have two speakers in them, and the distance between them should be right in the recommended range you're always talking about. So what would prevent them from performing right for ambiophonics?
Keith_W
Master Contributor
- Thread Starter
- #35
HI,
I don’t recall much now. I checked, and it seems the forum was edited. One moderator posted a crude drawing of “——-“ to mock Ambiophonics, which made me stop visiting there. It is common for me to get banned or suspended when questioning authority—even my mentor, who established the Ambiophonics Institute, was not welcome in some forums and you can see his name in almost in all forums and he loves to engage with academicians.
Thank you for mentioning the moderator issue—it took seven years for someone to bring it up. I am used to it by now, especially when some tell me to “go back and learn basic math” before challenging established figures. Forums like that are hard to engage in when people side with big names acting like authorities. But with the magic wand to edit for some it is often misrepresent the whole thread and my context.
Regarding soundbars, you are correct—I may have said that it wouldn’t work. For XTC, they can work, but only if you bypass their internal DSP and treat them as two speakers arranged in an Ambiodipole setup, with close positioning as I recommend. As shown in my YouTube video, there is no noticeable coloration with XTC, and no fancy EQ is needed. While signal measurements may show extreme coloration, the real question is whether you can hear it. My plugin uses simple delay and attenuation, not complex algorithms claiming special features.
I believe my plugin speaks for itself in achieving effective crosstalk cancellation. (The funny part is the YouTube video above is partially muted by YT. I think the muted the part actual stereo sound playback recorded with the phone but didn’t mute the XTC portion. LOoks like YouTube algorithm use frequency spectrum to compare original recordings and obviously XTC will measure different.
Have you tried any soundbar setups for XTC?
ST
- Thread Starter
- #36
Sorry Keith forgot to reply just now. I appreciate your feedback as I do not know how it will behave with 60 degrees. I think there is some difference when I limit the delay for narrow angle setup and wider angle. BUt is is hard to be sure with the $10 computer speakers that I use to develope this. So far I only test them with my 20 degrees angle of my main system. The display values are not correct as I suspect something is not exactly reflected in the values. It is working internally but not sure why the values are shown differently. HOpe to address them in the new version.
- Joined
- May 12, 2019
- Messages
- 231
- Likes
- 104
Nuuu, I haven't tried any soundbar for anything, nor XTC on anything (except maybe proprietary interpretations of XTC in whatever "spatiality" or "3D" effects I may have randomly come across so far).
I just found out about XTC/ambio as its own thing with its own supporting research 2 days ago, and I'm just trying to get the concepts cleared up, see what the prerequisites and constraints might be, and what results could be expected, going toward conclusions on what experiments might be worth trying for myself.
Speaker positioning is a big issue if you want the largest number of people to give this a shot, i.e. people who don't have a "man cave" listening room where the speakers can take center-stage but rather tend to consume media in a living room where a TV takes center stage. Showing that this works with soundbars I think is the breakthrough you need to get Average Joe in on this.
- Thread Starter
- #38
Optimal stereo is for one man only. Yes, it is kind of anti social. Crosstalk cancellation is well researched. I have archived most of the research papers. MIller’s website also have much more. YOu can get some idea about Ambiophonics here at Soundstage.
Last edited:
Keith_W
Master Contributor
Hello STC. I tried the new beta tonight. Here are some comments.
1. There is a problem with your webpage. It looks like this.
2. The README file contains the same instructions as the previous betas except for the addition of Stage 5.
3. From my previous experience, I decided to turn the delay down to 0 and adjust attenuation to the phantom image. I came up with the values of:
Stage 1: Gain -8 Delay 0
Stage 2: Gain -2 Delay 0
Stage 3: Gain -8 Delay 0
Stage 4: Gain -2 Delay 0
These values were obtained by replicating my experimental values for the left channel to the right. In one of my earlier attempts, I tried to subjectively tune both channels independently. The result was that the soundstage was shifted noticeably to the right.
4. After completing all the stages, I listened and noted that it was ALMOST the same tonality as with CTX bypassed but much lower in SPL. (by "bypass", I meant deactivating the VST in JRiver, NOT pressing the "bypass" tickbox in the VST). After I turned up the output gain, the sound took on a "processed" quality ... it's hard to describe. The soundstage seemed to flatten vertically and the treble became more prominent. Even turning down the Treble trim didn't fix it. That's when I remembered your warning about digital clipping, so I turned the output gain back down. That restored the tonality.
After fiddling with the tone controls a bit, I managed to get the tonality pretty close to the original.
I don't know if it's something to do with the mood I am in tonight, but I do not hear ANY soundstage widening effect. In fact it sounded "smaller" than with CTXMatrix off, no doubt due to the SPL loss. I did a sanity check by listening to uBACCH. There was a slight soundstage widening effect. I'll try again tomorrow night.
1. There is a problem with your webpage. It looks like this.
2. The README file contains the same instructions as the previous betas except for the addition of Stage 5.
3. From my previous experience, I decided to turn the delay down to 0 and adjust attenuation to the phantom image. I came up with the values of:
Stage 1: Gain -8 Delay 0
Stage 2: Gain -2 Delay 0
Stage 3: Gain -8 Delay 0
Stage 4: Gain -2 Delay 0
These values were obtained by replicating my experimental values for the left channel to the right. In one of my earlier attempts, I tried to subjectively tune both channels independently. The result was that the soundstage was shifted noticeably to the right.
4. After completing all the stages, I listened and noted that it was ALMOST the same tonality as with CTX bypassed but much lower in SPL. (by "bypass", I meant deactivating the VST in JRiver, NOT pressing the "bypass" tickbox in the VST). After I turned up the output gain, the sound took on a "processed" quality ... it's hard to describe. The soundstage seemed to flatten vertically and the treble became more prominent. Even turning down the Treble trim didn't fix it. That's when I remembered your warning about digital clipping, so I turned the output gain back down. That restored the tonality.
After fiddling with the tone controls a bit, I managed to get the tonality pretty close to the original.
I don't know if it's something to do with the mood I am in tonight, but I do not hear ANY soundstage widening effect. In fact it sounded "smaller" than with CTXMatrix off, no doubt due to the SPL loss. I did a sanity check by listening to uBACCH. There was a slight soundstage widening effect. I'll try again tomorrow night.
- Thread Starter
- #40
Thanks for the detailed beta feedback.
1. Webpage Issue: When I composed the README, it displayed correctly on my PC, but I apologize for not checking on phone. My fault—the overlapping is fixed in next version.
2. Beta Release: Sorry, I can’t explain the delay issue in more detail right now. DEEPAVALI PREPARATIONS have delayed me. I don’t think it’s correct to release a new beta just a few days apart.
3. Ambiophonics Standard: I must apologize—effective crosstalk cancellation without fancy stereo widening effects only works optimally at ~20° speaker angle. Your 60° setup is challenging per physics.
4. HF Coloration: I’ve received enough data to realize younger listeners may perceive subtle HF coloration. Beta 17 addresses this with EQ—response is now almost identical to flat, although I believe flat response is incorrect for 20° geometry.
5. Delay Adjustment: You MUST adjust the delays—they represent your HRTF timing. I may have mistyped “attenuation” instead of “delay” in instructions. My error.
6. My Testing: On my main system, I prefer this VST over years of Reaper implementation. I’m confident it’s fully transparent.
7. Beta 17 Final: Will deliver flat response. This will be my last beta—then I focus on the final product.
Thank you for the support. I expected more feedback from professional listeners, but it’s been a one-man journey. Special thanks to Keith and fcserai—you’ve provided crucial data for these improvements.
Looking forward to your retest with correct delays!
Similar threads
