Intensity panning (done on most recordings) only creates L-R level differences and hence the higher frequencies will be more or less correctly positioned, but the lower frequencies will remain closer to the center. Localization will only be correct for extreme Left or Right.

Hmmm, I think he means that even the mastertape has flaws regarding stereo imaging that can be improved.

But.. I don't understand how 'increased crosstalk at higher frequencies' solves the problem at lower frequencies. As far as I know this crosstalk process does not introduce interaural time differences in the lower spectrum in any way, or am I wrong?

Just brainstorming.

Summary of localization experiments:

A. At low frequencies rooms tend to decrease the separation of images, compared to the same loudspeakers in free field, effectively pulling acoustic images toward the center of the room. The L-R component of the signal to the loudspeakers is less effective than it should be in producing localization -- it merely broadens the image and gives leaves a vague sense of spaciousness.

B. At high frequencies rooms affect separation less than at low frequencies, but with some placements they tend to increase separation, smearing the higher formants and consonants of voices outward toward the loudspeakers. Rooms do the opposite of what recording engineers would like. They decrease the LF separation and sometimes increase HF separation. FOr coincident or pan potted recordings you would like them to increase separation at LF (for spaciousness) and reduce it (or leave it alone) at HF.

David Griesinger is a physicist who works in the field of sound and music. Starting in his undergraduate years at Harvard he worked as a recording engineer, through which he learned of the tremendous importance of room acoustics in recording technique. After finishing his PhD in Physics (the Mösbauer Effect in Zinc 67) he developed one of the first digital reverberation devices. The product eventually became the Lexicon 224 reverberator. Since then David has been the principle scientist at Lexicon, and is chiefly responsible for the algorithm design that goes into their reverberation and surround sound products. He also has conducted research into the perception and measurement of the acoustical properties of concert halls and opera houses, and is the designer of the LARES reverberation enhancement system

Sorry if I appeared grumpy in my first email..... I just didn't like being accused of being "full of shit!"

Doesn't 'crosstalk' basically have the effect of increasing the correlation between both channels? In that sense it is logical that it results in less smearing/blurring, but the question is: is it really an improvement?

Maybe there is more to this, therefore I'm interested in your paper.

This is interesting. Anyone capable of writing a simple filter for this? I'm not well-versed enough.

There is also a snicker that the products are no longer in production.

Quote from: Richard Brice on 2008-10-23 11:06:10

There is also a snicker that the products are no longer in production.

I hope you don't think I pointed this out to snigger - it was to point out that your page couldn't simply be typical marketing BS, because it is not selling anything.

Welcome to HA.

Cheers,
David.

Quote from: Jebus on 2008-10-24 19:42:18

This is interesting. Anyone capable of writing a simple filter for this? I'm not well-versed enough.

If you're referring to my note, you can calculate M and S, yes?

Then adjust them so that the energies are equal.

Reconstruct, and adjust gain to not clip.

Now, measure your interchannel correlation.  SURPRISE!

Quote from: Woodinville on 2008-10-24 22:08:10

Quote from: Jebus on 2008-10-24 19:42:18

This is interesting. Anyone capable of writing a simple filter for this? I'm not well-versed enough.

If you're referring to my note, you can calculate M and S, yes?

Then adjust them so that the energies are equal.

Reconstruct, and adjust gain to not clip.

Now, measure your interchannel correlation.  SURPRISE!

Just remember that the correlation is increased, but only in the HF range: in other words, in the sum/difference domain, the gain of the difference channel has to vary with frequency.

Here is a plot (from Tina circuit analysis) of the Francinstein filter with a source with 15dB difference in inter-channel input. Imagine this as a source panned hard left. The red trace shows the left-channel output: the green trace shows the right.



As you can see, the difference is not affected at LF, there's a transition point at about 700Hz with the channel difference being degenerated to about 8dB at HF.

If you look at..



you can see this is about the appropriate level of correction required to map the wider HF image back onto the narrower LF image.

Best wishes

Richard

So, basically you're converting L/R to M/S, apply a filter on S that damps frequencies above 700 Hz by about 3.5 dB, and convert it back to L/R -- or some equivalent process. Right?

Cheers,
SG

Isn't the problem here that, if a source is panned hard right (for example), then it really should come from only the right speaker. Phantom source = actual source in this example.

Phantom sources in between the speakers may need/benefit from this processing, but sources at the speaker can only be degraded by this processing.

I suppose you could argue that in the kinds of recordings that would benefit most (e.g. signals from real microphone arrays, not pan/pot and mainly LCR stereo), wouldn't have signals hard left or right.

Cheers,
David.