Header image  
by Ralph Glasgal
line decor
Home Tutorials Tech
Kudos and
Demos Bio Free Ambio
Glossary The Home
Concert Hall
Rec Engineers
FAQ/Forum Links Contact us
line decor

Myths about Ambiophonics Exposed
   by Ralph Glasgal

Basically RACE (Recursive Ambiophonic Crosstalk Eliminator) BACCH (Theoretica), Ambidio, Amtra, NeutronMP etc. are similar recursive crosstalk cancelling methods. RACE is meant to be for music and movies where extreme values of crosstalk cancellation are not needed. In concert halls you are lucky to have 5dB of interaural level difference. But in the time domain you can easily find the full 700 microseconds ITD that a human head can produce for a sound at the far side in everyday hearing on recorded media. RACE can easily achieve 10 dB and the 700 ┬Ásec ITD so it is quite adequate for any standard home media you prefer. If you want a bee buzzing in your ear from special recordings then something like BACCH is for you. But BACCH does not do full 360 degree surround in the horizontal plane so it is mostly a frontal only paradigm. But I assume in time they will go to more speakers just as in Ambiophonics. RACE works easily without HRTF measurements or head tracking. It is also inexpensive to use two RACE programs at once with four speakers to achieve a full circle of direct sound in the horizontal plane from 5.1/7.1 home theater or newer home theater formats like Dolby Atmos, and Auro3d. Ambiophonics also has flatter frequency response than any audiophile caliber stereophonic loudspeaker system.

Ambiophonics like stereophonics is free for anybody to use. Other methodologies can cost a lot. In an effort to compete with Ambiophonics, a claim is often made about poor frequency response. That is false news. First, just think, if the crosstalk cancellation is perfect, then the left ear hears only the signal from the left speaker which must therefore be flat and the right ear hears only the right speaker so that is also by definition flat if the speaker response is flat. If the crosstalk cancellation is less than perfect, then you just have some ordinary stereo left over mixed in with the normally much louder non-stereo signals at the ears. The audio engineering fraternity often makes a fundamental mistake in confusing what is going into a speaker with what is heard at an ear at the listening position. Yes in RACE a given speaker is outputting say the flat left signal followed by a whole bunch of delayed, attenuated and polarity reversed but still flat samples of both input signals. All these samples are flat since they are not filtered in any way. But if you use an averaging RMS frequency response meter or integrating RTA to measure this train you will not see flat response but a response that keeps changing with time and the music content. This is like trying to measure the frequency response at a seat in a concert hall. The integrated or short term average response will be far from RMS flat and will vary with where the sound from the stage originates and what instrument is playing. In contrast, measuring ordinary 60 degree loudspeaker stereo signals at an ear rather than at a speaker, will show really poor frequency response due to crosstalk induced combing. That is one of the things Ambio is designed to fix, that is flat instantaneous response at the ear rather than anywhere else.

Now these delayed samples, that are reaching the ear to cancel the crosstalk, mix everywhere in the room once you get a foot or so from the speakers, so the response even away from the listening area is flat and you hear both channels clearly, even if far off center. So even room reflections will not be colored since these reflections are a mixture of all these samples with even more delays and attenuations, thus averaging out to be flat just as in a concert hall. What it really amounts to is that the direct signals are accompanied by a rich bunch of early reflections again just like in a real space. If you use two rear speakers, as in Envelophonics, you can actually produce and exploit this effect directly and generate envelopment.

So at least one vendor, wanting the response at the loudspeaker to be flat in the traditional integrated sense, splits the frequency band into three or more regions and applies RACE to each of these with more ideal values of attenuation and delay. They then also apply frequency response filters so that the RMS response at the speaker terminals is flatter than just a plain multi band version of RACE would be. Of course, if such a signal is flat at the speaker terminals then it cannot be flat at an ear or do all the cancellation properly. So to correct for this one measures the response and the cancellation at the ears and introduces another filter to correct for this error. Of course now the speaker signal is not flat, but that is as we now know inaudible. However, unless you are an expert, when you measure anything at the entrance of the ear canal you usually also end up with an unwanted pinna response function, so that is a problem that Ambiophonics methods avoid.

Indeed dummy head recordings, played using HRTF, speaker measurements, and multiband crosstalk optimization are a real accomplishment. This way one can get up to 20 dB of level difference between the ears at all the mid frequencies and this means you can actually have someone whisper in your ear. You could add all this to RACE or similar products and only play special recordings, but that is not what Ambiophonics is meant to do. RACE is now being used in tablets, laptops, four speaker surround systems, as well as audiophile caliber systems where its two easily adjusted parameters ensure its universal compatibility with existing media, computers, phones, home theaters, etc. This makes sense since there are almost no media sources like music, movies, or video with ILD levels even up to 10 dB. For some games and virtual reality you do need to have care to recover all the localization data you can. The advantage of RACE in these applications is that it can deliver a full circle of direct sound in the horizontal plane from speakers rather than earphones.

The brain does not use the interaural cues of delay and level much above 1500 Hz where it switches to pinna direction finding. Since crosstalk in 2.0 speaker systems is only detrimental to ITD and ILD, you really don't need crosstalk cancellation at higher frequencies. Thus RACE products often include a control to limit their function to the lower frequencies if you wish to use it. So you can have a mixture of XTC and stereo to whatever degree you wish. I have found no evidence that using this control is seriously audible but a listening panel is needed to determine this. But if you have flat RMS frequency response at the speaker input for the higher frequencies, then with RACE bypassed and with plain stereo at the higher frequencies you now have the return of the stereo combing problem, so that the response at the ear is again definitely not flat. This higher frequency combing mimics pinna direction finding patterns so it is not a good idea to have stereo at any frequency below 10 kHz. However, the fact that RACE allows the speakers to be much closer together means that this non-RACE HF stereo combing begins more than one or two octaves above the usual 60┬░ speaker arrangement so that is an improvement.

The operation of RACE, at the higher frequencies, where it is still cancelling crosstalk that does not affect ITD or ILD, and is now becoming less accurate with such short wavelengths, just produces a random shower of early reflections that have every phase angle and absolute polarity imaginable. Whether this is audible needs more research but you have similar showers in concert halls from nearby heads and seats. Using two RACE enabled speakers in the rear ensures that this shower is even more diffuse and thus inaudible to the pinna direction finding mechanism. The unavoidable room reflections of the direct sound and the delayed cross talk cancelling samples, seem like early reflections to the brain since there are so many more delayed replicas involved. Like all room reflections these can add and subtract in many unpredictable ways. But the brain bases it opinion of frequency response on the first arrival direct sound, not the later room or hall ambience.

The novel concept being advanced in some AES circles that ILDs of more than say 10 dB are truly necessary for spaciousness or anything similar except things like whispering in the ear, is false and not substantiated by listening panel tests. For normal music, games, movies, home theater, etc. Ambiophonics does the job at low cost, minimal complexity, and easily delivers a full circle of direct sound in any room. The listening tests at NYU confirm the superiority of Ambio over normal stereo and illustrate its flat frequency response. (You can read the Ambidio thesis and the shorter AES paper if you go to www.ambiophonics.org and download it.)

The Yarra 3D and other soundbar schemes are great and much better than the usual TV speakers. But they are not really meant to do what a normal binaural loudspeaker system or a home theater system like Dolby Atmos or Panambiophonics can do. Like stereophonics, there are many ways to implement Ambiophonics and it is possible to have a soundbar type of horizontal speaker array producing a full width front stage. Basically the Yarra idea that you can make a speaker have a sonic beam so focused that it will only stimulate one ear is nonsense. Fine tuning it with HRTF filters cannot help. However, like using a physical barrier between two speakers, this design is still better than standard stereo even if these are not audiophile grade speakers. I would say this is a product for listeners who are easily satisfied with partial localization. Also despite some Yarra claims for surround, or 3D the curves do not show any localization from the rear for 5.1 sources, which is a weakness of most soundbars except when they luck out using a side wall reflection or a pinna response that just happens to match a listener's.

If you are listening to standard LPs, CDs, high rez downloads, or 5.1 SACDs, these almost never have level differences between the channels of more than 5 to 10 dB, and much less for vocalists with a guitar. So what does it mean if they are played back using a system that has the capacity to deliver 15 to 20 dB ILDs. It means nothing audible, of course, unless you believe you can hear the difference between 48/24 and 96/24 for say a signal at 300 Hz. Also ITD is a lower frequency localization cue and so RACE delivers that with ease, again, only if it is on the recording.

There is a problem with both RACE, and all similar processes, that is avoidable. That is if you cancel crosstalk that is not there. For instance if you tell RACE your speakers are at 45 degrees and set things accordingly, but the speakers are really just 10 degrees apart, you will over cancel and that means introducing reverse polarity signals into the speakers that now have no job to do when they reach an ear at the listening position. This is sort of like listening to stereo with one speaker reversed in polarity. The loss of low bass is a clue that you have made this error. That can happen with BACCH for example. if you move the speakers after you have made all the mic measurements. RACE can just be adjusted with two controls anytime you get new speakers or move them a whole lot. You can also use the controls to season the localization to suit a recording or your taste.

Line array speakers like electrostatics image well with RACE but so do oval shaped radiators like the MBL. A speaker that does radiate like a laser beam would, in theory, be great. The Soundmatters speaker bars have two tiny speakers four inches apart and they image like crazy, but you need a subwoofer of course. They indeed make a fantastic demo, when you hold this little thing in front of you and hear sounds extending to 90 degrees on each side of you.

With four speaker Ambio, once you have a great wide front stage, and the rear pair going, the listening room ambience will seem much less audible. It is easier to keep the speakers relatively close if the angle is narrow so the room will be even less audible in the near field. Actually with the rear pair going, the longer room delays do add a few later reflections at lower levels that the brain likes.

Room treatment will not do anything for crosstalk elimination. XTC involves delays on the order of tens of microseconds. Room reflections are tens of milliseconds and arrive too late to have anything to do with localization or crosstalk. They do influence your sense of space in a concert hall or room but that has nothing to do with stereo loudspeaker crosstalk even if many writers do get frontal localization and general spaciousness confused a lot.

Do you think there is a problem with indirect reflections on imaging in a concert hall? The only problem with stereo or Ambio spaciousness (not imaging or localization) in homes is the lack of proper room reflections not their presence. In a stereo, Dolby, Auro, Theoretica, 5.1, etc. system all the room reflections from the two front speakers arrive at the ears with the same directional parameters no matter where the front image is located. In normal hearing, each frontal source produces a quite different set of room reflections with a great variety of ILDs and ITDs that, again, vary with each frontal source position. In all the frontal only speaker systems there is only one ILD and ITD reflection value set from each speaker no matter what instrument is playing or from what position. The loudness of a room reflection may change but its directional cues remain fixed. The brain finds this abnormal and says you must be listening to something like a normal stereo recording despite XTC. Ambiophonics, of course, easily avoids this defect by using two inexpensive speakers behind the listening area. Many vendors think they need to do room treatment but Ambiophonics users and product creators can avoid this.

How can you tell if you have an Ambiophonic system working properly? If you put just a left signal into the system that implies a very large value of level difference, well over say 60 dB. In normal hearing such a signal would seem to be at the 90 degree side. So when you apply RACE to say a left only signal, if the crosstalk is properly cancelled, you should hear the sound very very far to the side. It may not be quite at 90 degrees because of a pinna direction finding error. You need the rear speakers to get the last few degrees, but if the sound is close to the side you know RACE is working well indeed.

In other XTC product demos I have heard the stage is not nearly this wide in this test because the cancellation is compromised to get flat speaker response or the speaker angle used is too wide to avoid pinna direction finding distortion. Remember that a left only sound in real life goes around, under, and over the head to reach the other ear so the ILD is likely on the order of 10 dB or less, unless the sound source is really close to an ear, like in the haircutting demos, the bee buzzing, and the whispering in the ear recordings. The ITD should of course be about 700 microseconds and it can never be more in normal hearing since that is the maximum possible delay between the ears. So for a left only signal, if the XTC is good the signal at the left ear should reach the right ear in 700 microseconds as in normal hearing rather than in 220 microseconds as in the stereo triangle. So this far side localization should be easy to detect if RACE parameters are set properly and the speakers are equidistant from the listeners and have about equal loudness.