Localisation cues in recordings

[jkeny]

A lot of people know that my focus on analysing audio for a while now has been in the area of timing. There are a number of reasons for this focus but one of the main ones being that sound stage is a noticeably better focused in the better sounding systems. I attribute this to better reproduced timing in the reproduced sound field.

The foundation for this sound stage premise is based on psychoacoustic underpinnings - in nature, we localise sounds based on the difference in the timing of the signal reaching the two ears (interaural time difference - ITD) & also on the difference of the intensity of the sound between the ears. Here's an interesting, hands-on demo of just this effect with ITD & IID http://auditoryneuroscience.com/topics/ ... ty-trading
Note: the lowest values used in this demo are ITD of 0.22mS & IID of 3dB - very much bigger than the just noticeable hearing limits for ITD of 2us & IID of 0.05dB

Now the question that always bugged me was that an awful lot of recordings are not live recordings - they are products of a studio engineer - the question being do these engineers observe psychoacoustic rules when creating a sound stage on the recording? Not being intimate with the process I just didn't know - I suspected that just volume differences were used to create this sound stage.

The question then becomes how we deal with this anomaly when it occurs in playback i.e the timing & volume aren't in accordance with natural localisation of sounds. It seems we accommodate to this in our listening?

I came across this recently when doing some further reading of Jneutron's comments on group delay (timing differences across frequencies) in speaker cables. He stated this which makes some sense http://www.audiocircle.com/index.php?to ... #msg479123

Unfortunately, the bulk of the music heard today is of studio origion. All the artists don't even have to be on the same continent.

Given say, 24 individual tracks that are mixed down to two, ITD loses. The mixdown engineer turns a knob to position a track in space.

Music content which has been "panned" for position, is a forced response media. In other words, we humans are forced to interpret the inaccurate localization parameters to determine a position in space.

Unforced, or natural localization, uses both intensity difference as well as time difference. If you get a chance, try altering one or the other, and listen to how your mind reacts. If you reduce one ear's intensity (say with a towel, or a plugged ear due to water), note that you can still determine the source location fairly well...this is because you are sensitive to the time difference. In fact, as you lower one channel intensity, you can "see" that the source direction is "fighting" the intensity error, trying to remain centered. This is ITD at work.

Conversely, if you have a multiplexed dac feeding two channels into headphones, and the S/H system is not syncronous, you get an 11 uSec interchannel delay (typical of really cheap sound cards). With a built in 11 uSec delay, there is nothing you can do with the balance control to get the image field correct... It is really weird.

Nobody measures the ITD information which is presented to our ears to determine if the system has altered it. Since half the information we use to distinguish direction is ignored in the measurements, how then do we determine the accuracy of the delivery system?

Human ITD sensitivity is in the 2 to 5 uSec range. Very small number, eh?

Cheers, John

[jkeny]

And also this particular throwaway comment (which I have bolded) also makes you think http://www.audiocircle.com/index.php?to ... #msg478755

To date, all the two channel musical product you buy (with some exceptions of course), are mixed down using pan pots which utterly destroy the localization parametric known as Inter-aural Time Delay, ITD. This is done primarily to keep the two channel product backwards compatable with mono. MONO, believe it or not..sheesh. If ITD were introduced, or even maintained, the mono result would be heavily combed and rather distorted.

To properly measure how amps, cables, whatever, impact a two channel visualized soundstage, what is required is a correlated two channel spectral analysis. To date, everybody is content with ONLY single channel FFT magnitude analysis. As a product rep from B and K stated (actually, I got him to admit it on a forum... :oops:), there are an infinite number of signals that can have the same FFT magnitude spectrum. So even though two independent signals look good by spectra, that doesn't necessarily mean that they sound the same when analyzed by the human brain in a stereophonic application. JND studies have been done along these lines for single channel only, but nobody has done correlation studies interchannel with complex signals (like music), nor even I/O studies with magnitude/phase correlation.

Edit: btw, while dbt's are rather sensitive instruments, how many are performed using a source program which has been tailored to a speaker system that is a specific distance from the subject and a specific angle with a specific horizontal dispersion characteristic? The human brain is wondrously adaptable, that is why we "see" a sound image off center even though the right-left delay in the source program may be zero.

So DBT's may well be compromised because we use source material with unnatural localisation cues. This effectively limits identifying some of the audible differences related to sound stage (& this is where I hear improvements in better systems). So let's keep an open-mind about DBTs & the source material used for audio testing.

However, (playing my own devil's advocate here) could this explain why when sighted & we know what we are listening to, we can differentiate between 2 system i.e somehow compensate for this unnaturalness but unsighted we are left without this mechanism? I'm not convinced by this but I remain open-minded.

[jkeny]

What also prompts me in starting this thread was my (& others) experience recently in Tony's get-together.

When we were trying to differentiate between the two PC configurations (Tony's & Nige) I could immediately hear it. I heard a harshness in Tony's that wasn't in Nige's. But when the tweeters were retired I could not. I believe others came to this same conclusion also. Question is - was this just a difference in the noise level/HF content/timing? I don't know?

Another issue that I heard was when we were comparing Jplay Vs Vinyl using the same Oscar Peterson track - I felt that Jplay was curtailing the ending of or tail off of the piano notes - vinyl was more lifelike (I much preferred it). Interestingly, MQN did not do this & in fact sounded better than vinyl, to my ears.

[jkeny]

For those who have experience with or know about professional recording techniques, maybe they could confirm or otherwise, the view expressed above - that volume panning is used rather than timing differences to create sound stage on a recording?

What about acoustic & classical recordings - are these using more natural techniques?

Any insight into these matters?

[Diapason]

All very interesting, JK. Some classical recordings are indeed more naturally recorded, with minimal mics and space being captured naturally. Not all, of course...

Purely anecdotally, I remember when Nordost demoed the new Odin cables in Blackrock a few years ago. As always with cable dems I wasn't too sure that differences were all that apparent, but when somebody (Ciaran?) asked for a classical track, differences were far more apparent. Yes, we may have imagined it, usual caveats, but I thought it was about the most convincing cable dem I'd ever heard. I also find spatial cues in classical music to be one of the best ways of assessing different valves in my amps at home.

Anyway, can't contribute technical content, but I definitely share your views on the importance.

[nige2000]

Really interesting stuff there john

i havent studied this stuff and i could be well of the mark in my 5 min of thought

if you think about our perception through hearing
we've only got two ears and with them we have an idea of position of the source of sound whether it be in front, behind, left, right, above, below close or far away
i dont believe that our ears are directional, maybe we only differentiate direction because the exact same sound sounds slightly different to our ears depending on where the source is

our ears are probably more sensitive than we think
maybe if is all figured out we could have surround sound with just one speaker

[Fran]

Yes, very interesting John - and there is probably an interesting paper in it I think.

Probably tough to test though without the appropriate recordings.

Did some of the lads from the listening suite work in production/mixing? Might be worth asking them. Or even a straight email to a well known studio person (engineer?) to explain the thinking and ask them?

Fran

[jaybee]

I wonder would you find a correlation between "live room" studio recordings ( ie minimal engineer generated localisation) and those recordings considered "audiophile"

[jkeny]

It's well known how our ears determine the direction of sound - it's by ITD & IID
What hasn't been done is to measure these aspects in audio playback.
Testing this for audibility, as Fran says, is difficult unless we can guarantee the recordings have been done with these ITD & IID elements preserved.
I'm working on other ways of doing this however & may produce some measurements eventually.

However, it is worth bearing in mind
- without this sort of correct source material we are hampering our listening
- DBTs are also hampered by such incorrect source material
- this ITD & IID may well be some of the missing measurements in audio that correlate to what we hear

[jkeny]

BTW, guys when we compared Jplay to Vinyl, were we using the supertweeters? I don't think we were?