Re: [Sursound] Transient time differences
On 04/02/2012 11:51 PM, Eric Benjamin wrote: I can't answer the question precisely without either doing an experiment or by doing many hours of calculations. a good online tool that does many of the calculations relevant to ITD and ILD reproduction is available from helmut wittek at http://hauptmikrofon.de/ima2-folder/ImageAssistant2.html i find it an invaluable tool in trying to understand what different miking techniques are trying to accomplish. -- Jörn Nettingsmeier Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487 Meister für Veranstaltungstechnik (Bühne/Studio) Tonmeister VDT http://stackingdwarves.net ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Transient time differences
On 04/02/2012 08:37 PM, Eric Benjamin wrote: I believe that the "glockenspiel effect" that you describe arises because the localization cues experienced by the listener are different for ITDs than for ILDs. Because we primarily rely on ITDs at low frequencies and ILDs at high frequencies, if the reproduction system doesn't handle them in the same way then the listener experiences a disparity. This happens in both Blumlein stereo and in Ambisonics. at the risk of eternal damnation: if you want to fix this in stereo, you can. the solution is called ORTF, NOS, or any other slight variation thereof :-D incidentally, higher-order panned sources are also way more stable with respect to timbre. and now: "duck, and cover :)" -- Jörn Nettingsmeier Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487 Meister für Veranstaltungstechnik (Bühne/Studio) Tonmeister VDT http://stackingdwarves.net ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
[Sursound] Transient time differences
--- On Mon, 4/2/12, Eric Benjamin wrote: > Don't take this to mean that I don't like ORTF recordings. > I do like them. The > best stereo recording that I have ever made was an ORTF > recording. But then, I'm > not a very good recording engineer. I think that one of > the reasons that I like > ORTF is that it introduces an artificial spaciousness which > may compensate for > the spaciousness that is lost in stereo reproduction. I think you might find that this lost sense of spaciousness (IACC) is attributable to 60-degree stereophony. Three-speaker stereophony (Trifield decoded) with the left/right loudspeakers subtending a 90 degree arc does not suffer from a lack of spaciousness thus obviating the need to create artificial spaciousness a la spaced-omnis in order to compensate for 60-degree stereophony. ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
[Sursound] Transient time differences
--- On Mon, 4/2/12, Eric Benjamin wrote: > In subsequent thinking about his question it occurs to me > that the plausibility, > not of the signals in the recording but of acoustic signals > that enter the > listener's ears, is an important indicator of whether the > listener finds the > reproduction to be realistic or not. If our ears receive a > large number of cues > that are wrong, or at least implausible, then the > reproduction is unrealistic. I would hasten to add visual cues as well. Seeing a small listening room and observing loudspeakers interferes with the creation of the illusion. Listening in a pitch black room (no light whatsoever!), as silly as it may seem, is imperative to create the suspension of disbelief. Try it! ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Transient time differences
Well, don't get the idea that I do not like Blumlein. My once(actually twice as it happened) in a lifetime chance to record major orchestras with Kavi Alexander in charge, we did use Blumlein. And ORTF sounds a little colored as to timbre to me (we have some recordings made with identical mike placements but the two methods, though this is hardly a theoretical test since the mikes themselves are of necessity different). But I have always been told as "conventional wisdom" that ORTF gets the time of arrival right in a way that Blumlein does not. And I always regarded the choice as a tradeoff--absolutely spot on timbre with Blumlein versus correct space with ORTF. But maybe this is wrong! If things work up to 6k, that would surely alter the viewpoint. I await this with interest. I would also be interested to see the calculations that would explain why ORTF comes out a bit colored, at least to my ears. Robert On Mon, 2 Apr 2012, Eric Benjamin wrote: I can't answer the question precisely without either doing an experiment or by doing many hours of calculations. But one thing to consider is that the Blumlein recording won't fail to produce correct ITDs at frequencies above 700 Hz. At least, not theoretically. I can demonstrate by calculations using a good head model that the ITDs continue to be correct up to some relatively high frequency, say up to 6 kHz. I have the calculations already done to demonstrate that is works for an Ambisonic system. But one of the differences between theory and practice is that the listener won't necessarily be exactly at the sweet spot. That is to say, if the listener shifts 10 cm to the left then he has undone the 'correct' time differentials provided by the ORTF system. Don't take this to mean that I don't like ORTF recordings. I do like them. The best stereo recording that I have ever made was an ORTF recording. But then, I'm not a very good recording engineer. I think that one of the reasons that I like ORTF is that it introduces an artificial spaciousness which may compensate for the spaciousness that is lost in stereo reproduction. I will do some calculations on ORTF stereo so that I can understand it better. - Original Message From: Robert Greene To: Surround Sound discussion group Sent: Mon, April 2, 2012 1:44:28 PM Subject: Re: [Sursound] Transient time differences Thanks for the information. But here is my question in more precise form: Suppose you do a recording with ORTF(which of course has its own set of problems). Suppose you record a source that is say 15 degrees left of center. and that the source is a pistol shot(an impulse). Now the impulse will arrive at the left mike before it arrives at the right mike. The time difference is the same more or less as it would be for a dummy head recording since the distance between the mikes is the same (more or less) as the distance between the listeners ears(or the dummy head ears). On playback, the impulse will also arrive at the left ear the same amount of time before the right ear-- as far as the high frequencies are concerned. Namely they are heavily shadowed by the head so that the arrival at the left ear first is blocked from the right ear, and the right ear hears only the right speaker. This is in the highs. Below around 700 Hz, Blumlein would have put the phase shifts right and that part of time would be there. But the head shadowed part , the high frequency part, is right for ORTF but wrong for Blumlein-- the direct arrival of the high frequency part of the impulse as recorded in Blumlein is simultaneous in the two speakers(as is everything) but there is no reconstruction via head effect because the head effect is essentially total shadowing. Of course there is some head shadowing in the real world, too. But 15 degrees off to one side is not enough to block the highs so completely as 45 degrees (or 30 degrees). So there is some range of angles where the timing is off because of the greater shadowing(almost complete) from the wide speaker separation is not representing the real situation. Is this wrong? This is not my private theory. I think Blumlein was aware of this, and I known other people have mentioned it. Maybe is does not really matter, but it seems real enough. (I believe it is known that time delays in the high frequencies play a role in location) Robert ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Transient time differences
I can't answer the question precisely without either doing an experiment or by doing many hours of calculations. But one thing to consider is that the Blumlein recording won't fail to produce correct ITDs at frequencies above 700 Hz. At least, not theoretically. I can demonstrate by calculations using a good head model that the ITDs continue to be correct up to some relatively high frequency, say up to 6 kHz. I have the calculations already done to demonstrate that is works for an Ambisonic system. But one of the differences between theory and practice is that the listener won't necessarily be exactly at the sweet spot. That is to say, if the listener shifts 10 cm to the left then he has undone the 'correct' time differentials provided by the ORTF system. Don't take this to mean that I don't like ORTF recordings. I do like them. The best stereo recording that I have ever made was an ORTF recording. But then, I'm not a very good recording engineer. I think that one of the reasons that I like ORTF is that it introduces an artificial spaciousness which may compensate for the spaciousness that is lost in stereo reproduction. I will do some calculations on ORTF stereo so that I can understand it better. - Original Message From: Robert Greene To: Surround Sound discussion group Sent: Mon, April 2, 2012 1:44:28 PM Subject: Re: [Sursound] Transient time differences Thanks for the information. But here is my question in more precise form: Suppose you do a recording with ORTF(which of course has its own set of problems). Suppose you record a source that is say 15 degrees left of center. and that the source is a pistol shot(an impulse). Now the impulse will arrive at the left mike before it arrives at the right mike. The time difference is the same more or less as it would be for a dummy head recording since the distance between the mikes is the same (more or less) as the distance between the listeners ears(or the dummy head ears). On playback, the impulse will also arrive at the left ear the same amount of time before the right ear-- as far as the high frequencies are concerned. Namely they are heavily shadowed by the head so that the arrival at the left ear first is blocked from the right ear, and the right ear hears only the right speaker. This is in the highs. Below around 700 Hz, Blumlein would have put the phase shifts right and that part of time would be there. But the head shadowed part , the high frequency part, is right for ORTF but wrong for Blumlein-- the direct arrival of the high frequency part of the impulse as recorded in Blumlein is simultaneous in the two speakers(as is everything) but there is no reconstruction via head effect because the head effect is essentially total shadowing. Of course there is some head shadowing in the real world, too. But 15 degrees off to one side is not enough to block the highs so completely as 45 degrees (or 30 degrees). So there is some range of angles where the timing is off because of the greater shadowing(almost complete) from the wide speaker separation is not representing the real situation. Is this wrong? This is not my private theory. I think Blumlein was aware of this, and I known other people have mentioned it. Maybe is does not really matter, but it seems real enough. (I believe it is known that time delays in the high frequencies play a role in location) Robert ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Transient time differences
Thanks for the information. But here is my question in more precise form: Suppose you do a recording with ORTF(which of course has its own set of problems). Suppose you record a source that is say 15 degrees left of center. and that the source is a pistol shot(an impulse). Now the impulse will arrive at the left mike before it arrives at the right mike. The time difference is the same more or less as it would be for a dummy head recording since the distance between the mikes is the same (more or less) as the distance between the listeners ears(or the dummy head ears). On playback, the impulse will also arrive at the left ear the same amount of time before the right ear-- as far as the high frequencies are concerned. Namely they are heavily shadowed by the head so that the arrival at the left ear first is blocked from the right ear, and the right ear hears only the right speaker. This is in the highs. Below around 700 Hz, Blumlein would have put the phase shifts right and that part of time would be there. But the head shadowed part , the high frequency part, is right for ORTF but wrong for Blumlein-- the direct arrival of the high frequency part of the impulse as recorded in Blumlein is simultaneous in the two speakers(as is everything) but there is no reconstruction via head effect because the head effect is essentially total shadowing. Of course there is some head shadowing in the real world, too. But 15 degrees off to one side is not enough to block the highs so completely as 45 degrees (or 30 degrees). So there is some range of angles where the timing is off because of the greater shadowing(almost complete) from the wide speaker separation is not representing the real situation. Is this wrong? This is not my private theory. I think Blumlein was aware of this, and I known other people have mentioned it. Maybe is does not really matter, but it seems real enough. (I believe it is known that time delays in the high frequencies play a role in location) Robert ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
[Sursound] Transient time differences
Since this has moved away from the point of the original query and to a technical issue, I'm starting a new thread. Please forgive me if I get a bit pedantic here. I'm still concerned that you mean something different by "transient time of arrival differences" than I do by ITDs. I'll repeat my assertion that time of arrival differences only exist at the point of a listener hearing the sound. If there is a violin playing on stage, slightly to the right of center, the sound generated by the violin arrives at the point occupied by the listener all at the same time. The listener locates the violin largely by the fact that his ears are at two different places, and those places have different times of arrival. What we are concerned with is the sounds (ear signals) that the listener hears when he is listening to the reproduced sound. Stereo, and now I need to state explicitly that I mean ordinary intensity stereo as embodied in Blumlein recording, is a sort of coding and decoding system. The microphone array codes the direction of sound arrival by different intensities in the two channels. Those two channels are mapped to two spaced loudspeakers and it is the way in which those sounds are combined in the reproduction venue that results in the listener's perception of location. I believe that the "glockenspiel effect" that you describe arises because the localization cues experienced by the listener are different for ITDs than for ILDs. Because we primarily rely on ITDs at low frequencies and ILDs at high frequencies, if the reproduction system doesn't handle them in the same way then the listener experiences a disparity. This happens in both Blumlein stereo and in Ambisonics. It's just not possible to get the ILDs exactly correct, at least not with loudspeakers spaced by 60 degrees or 90 degrees. There simply aren't enough degrees of freedom, not by a couple of orders of magnitude. So the best that we can hope for is that the ILDs will be correct on average. That they won't be worse in some directions than in others. Think about what happens in Blumlein recording and reproduction. If the sound source is on the axis of of one of the microphones then it appears exclusively in that channel. It then appears exclusively in the associated loudspeaker. It sounds as though it were coming from that loudspeaker because it IS coming from that loudspeaker. It the sound source is located exactly between two microphones then the signal appears in both channels and is reproduced equally by both loudspeakers. The signals that arise at the listener's ears have substantial comb filtering due to the acoustic crosstalk. Crosstalk that wouldn't have been there if the listener were present in the recording venue but is present for the listener in the reproduction venue. The reproduced sound is more diffuse for the center image than for sounds combing from the directions of the two microphone capsules. The difference between Blumlein and Ambisonic reproduction of the sound (here I am referring primarily to first order) is that both the low- and high-frequency cues are much more even with respect to direction. But clearly there is a lot of room for improvement. I can demonstrate that all of this works the way that I described it by doing an experiment. We can go to the concert hall and make a series of recordings. Perhaps we might want to do three recordings. One with a dummy head (we'll use my head. It's available...), one with a Blumlein array, and one with a soundfield microphone. We will then analyze the recorded signals from the dummy head which will give us both ITDs and ILDs as experienced by the listener (me). We then reproduce the Blumlein and Soundfield microphone recordings using an array of four loudspeakers. That will gives us a 90 degree stereo pair which is, as I understand it, how you like to listen. We then record the acoustic signals going into my ears when I'm in the sweet spot and analyze those. The result that I expect is that the low-frequency ITDs will be correct and the high-frequency ILDs will not be correct. And they will be more confused for the Blumlein recording than for the Soundfield recording. Now I've actually done this for the case of Soundfield recording, although not as carefully or as thoroughly as I would like. I intend to do more such experiments. In particular I intend to explore what I call the plausibility of the recording. Several years ago I was asked by Svein Berge how to tell if a particular Soundfield recording was good or not. It took me a bit to realize that he was asking, not if it was aesthetically pleasing, but whether or not is was correct; that is, could those signals in the recorded file actually represent a real sound field. And as it turned out, in some of those cases they couldn't. Or at least they were highly implausible. In subsequent thinking about his ques
