Something I never understood-Vertical Imaging

A perceived varying depth of image with varying source is physically possible, but I don't believe vertical image variation is possible except as an artifact of non vertically symmetric speaker designs or room acoustics. i.e., there is no physical way to encode vertical information into a L/R signal.


Depth cues are often handled at the mixing level with various amounts of reverb. Psychoacoustically, longer reverb times on, for example, a vocal track tend to make the vocalist sound further back on the "stage," while a dry mix (no reverb) sounds very forward, close, in-your-face, etc. From my admittedly limited knowledge of mixing techniques, instruments can be placed in both horizontal and depth directions using pan and reverb. There might also be a way to use phase to give depth cues... L/R placement can be handled by intensity shifting or phase shifing, with intensity shifting being the most common for mixing pans. I'm not a recording engineer, so I'm not sure what the phase shifting does to depth cues, or what the downsides are to using it.

binaural recording hints at the answer, I think.


The human with an ear perceives the vertical sound location by subtle phasing and frequency response cues caused by the interaction of the ear itself with the sound. In other words sound sources from high up "see" a different ear shape than sounds that come from low. So they get subtly changed.


Having sounds "forward" versus "Back" is a very well understood thing that can be done with phasing. Take a loook at Dolby heaphone technology. And there are various systems that can image rear channels behind the listener in a 2-speaker system. So there shouldn't be any question that you can image the forward/back dimension with just 2 channels (of course the success of various systems is questionable, but the fact that they make an impact isn't).


Anyway, where I'm going is that the better a speaker preserves subtle phasing relationships and generally the better it reproduces the signal waveform, the better these subtle phase and frequency response cues will be preserved, so the more vertical depth imaging you will notice.


-Tom

Quote:
Driver directivity/position and the room do interact to produce frequency dependant image shifts, althouh there are other mechanisms at work.


I get radically different vertical placement on standup bass, electric bass, and drums depending on the recording although the frequencies are essentially identical. Some have them close to the floor where my left/right sub drivers are, some have them about 3' off the ground. Maybe I'm missing higher order harmonics on some of the recordings, although where that's the case the instruments sound wrong.


Some recordings have the vocals noticeably above my tweeters which are 42" off the ground.


Perhaps the pickup pattern of some microphones has a frequency response which looks like our head related transfer functions (HRTFs). Or maybe you key off timing differences/constructive+destructive interferance from the floor and ceiling reflections present in the recordings.


There could also be other psycho-acoustics at work - you expect kick drums to be on the floor. The percussion section is always behind the rest of an orchestra. The realism of the reproduction could influence when your brain starts fooling you.

I don't know what causes it, but I am POSITIVE that some of your typical garden variety LPs and CDs have very specific vertical imaging.


The example I can put my fingers on immediately is "59th Street Bridge Song" on Simon & Garfunkel's Parsley, Sage, Rosemary and Thyme. On every system I've heard, Garfunkel is heard distinctly "up" and slightly right of center. I've heard both the CD and the LP on a number of systems, as this is one of the ones I used as a reference. My speakers are Martin Logan Odysseys, which play everything over 250 Hz through a single electrostatic panel, so on my system it's not due to time alignment (or lack thereof) in the drivers.


I've heard the same image on Avalon Opus (perhaps the Opus Ceramique), and also on a pair of AvantGarde Duos, so I know it's there. I can't figure out how this could be happening, but at this point I know I'm not imagining it.

Drew's explanation makes the most sense: "Perhaps the pickup pattern of some microphones has a frequency response which looks like our head related transfer functions (HRTFs)."


Without a binaural dummy head, there isn't any way I'm aware of to encode that vertical information intentionally into a L/R recording. I'm sure processing will be availble at some point that tackles this problem with some success (they have been working on it for years), but the fundamental problem is the differences between ear shapes in humans. A binaural recording played back through ear-canal buds might produce scary results, but that's almost impossible (if not entirely impossible) to approach with conventional loudspeakers.


Now, as Drew suggested, it is certainly possible that through some quirk of the recording, processing, speaker, or room interaction, the sound that reaches you will, after the transfer by your ears, contain distortions that closely resemble vertical cue transfer functions.


The important point, IMO, is that there isn't a way to take advantage of that and produce a repeatable result. Interestingly, it could also very well be that what you perceive as vertical imaging on one track I or someone else may not, and vise versa, due to the physical differences in our ear shapes.

Here's one of my recent experiences, and also a theory.


I was listening very carefully to Puccini's opera Turandot - specifically, the London/Decca CD with Pararotti, Sutherland, Caballe et al. I was comparing my standalone Classe CDP-.5 connected via balanced analog interconnects to the same unit used as a transport feeding a Classe SSP-75's internal DACs. The power amp is a Classe CA-200 with ML Odyssey speakers.


I wrote in my notes that especially the SSP seemed to get the image wrong, and in fact it seemed to be backward with the orchestra in front of the singers. (The CDP's internal DACs yielded a much "flatter" image front-to-back.) And the vertical image seemed goofed up, too, since the singers appeared to be higher vertically than the instruments traditionally found at the back of the orchestra (e.g, double bass, precussion).


The penny dropped on the famous "Non piaget"*, when I noticed that Liu's (Caballe, I think) sonic image was distinctly lower than that of Calaf (Pavarotti). Of course! Liu is begging. I've never seen the opera myself, but it makes sense that she would have a lower vertical image position on this track, because she would be on her knees. Then it hit me that this is an opera, dummy, not a choral performance - so the orchestra is supposed to be in the pit, DOWN in the front, and the singers are UP on stage, behind the orchestra.


The opera regulars are probably screaming "DOH!!!" but I'm new to opera. I had fixed in my mind a picture of a performance of an oratorio such as Handel's Messiah, in which the chorus is usually arrayed behind the orchestra with the soloists in front.


In the end, it was the vertical image information that gave me the clues necessary to understand the recording despite an almost certainly erroneous initial mental picture. And moreover, a pretty decent CDP wasn't sufficient to tell - it took the better DACs in the SSP75 to see it at all.


A week later I was able to listen to the same performance on LP, using a vastly more expensive system [SME 20 / Graham 2.2 / Koetsu Rosewood? / Aesthetix IO / Aesthetix Callisto / Wavestream Kinetics / Avalon Eidolon Diamonds]. It was very easy to identify the image, although I wonder how much of that is due to the clearly superior capabilities of that system, how much is due to the LP vs CD encoding, and how much is due simply to the fact that now I had recognized what the reality almost certainly was. My guess is that all three are in play.


I'm pretty certain that the imaging is both correct and reproducible in all dimensions in this case, although I couldn't prove it unless I see a video or photograph of that performance.


-----


After thinking about it some more, I think I have a theory. Although we're hearing "only" two channels (the above is definitely a stereo recording), they carry more than two point sources worth of information. In this case, I think what is happening is that the various reflections from the hall were accurately captured by the mikes. Perhaps there is some way in which we correlate reflections arriving at different times in each ear with vertical (and other) position information. However we psychoacoustically reassemble them, the better systems were in this case clearly able to convey the image more precisely as it was recorded.


* apologies for the massacre of Italian spelling!

Quote:
Unfortunately, nobody in a studio even remotely pays attention to this sort of thing for vertical imaging... also keep in mind that studio recordings are mono tracks placed in a stereo field and in reality contain absolutely no stereo information whatsoever ( a few engineers will fly stereo room mics for drums and things but it is uncommon)... all studio imaging is is a mono track that is sent more to 1 channel versus the other creating an illusion of placement in the stereo field.


However, with most live classical recordings where there are two seperate microphones capturing 95% of the left v. right information (different phase and all), there is a real attempt at capturing a true stereo image (some techniques are more effective than others).


As far as the post above about the opera... part of the reason for what you are hearing is the common use of boundry layer microphones when capturing opera which do a very good job at capturing singers on a stage without having to have the microphones right in front of them.