Measurements & the stereo illusion

[jkeny]

Does every thread have to be dragged down to this old flame war topic & the snide remarks begin?

Can this question not be considered as an abstract discussion of possibilities & measurements without reverting to the usual bunkered attitudes or is there something threatening about discussing this in a rational, intelligent way?

 

[Phelonious Ponk]

Does every thread have to be dragged down to this old flame war topic & the snide remarks begin?

Can this question not be considered as an abstract discussion of possibilities & measurements without reverting to the usual bunkered attitudes or is there something threatening about discussing this in a rational, intelligent way?

Sure. Carry on.

Tim

 

[Groucho]

Does every thread have to be dragged down to this old flame war topic & the snide remarks begin?

Can this question not be considered as an abstract discussion of possibilities & measurements without reverting to the usual bunkered attitudes or is there something threatening about discussing this in a rational, intelligent way?

Fair enough, I'll not mention it again, except to say that instead of trying to track down an effect that may just be a phantom, approach it the other way round and listen while introducing phase shifts/delay etc. between the channels to directly observe the difference. It might give you a clue as to what you're looking for and how large the differences need to be.

 

[microstrip]

Can this question not be considered as an abstract discussion of possibilities & measurements without reverting to the usual bunkered attitudes or is there something threatening about discussing this in a rational, intelligent way?

S. Linkwitz addresses this subject in his site when addressing loudspeakers, not electronics or sources: A large contributor to Auditory Scene distortion is the loudspeaker. You can test your sound system's accuracy by how readily it reveals artificiality in recordings. See Accurate performance of a stereo sound system. Linear and nonlinear distortion in audio systems can be readily measured and expressed in numbers, though the degree to which these distortions are perceived is often not proportional to the numbers. Spatial distortion of the Auditory Scene can be perceived and described, but a method to measure the distortion and to express it in numbers has yet to be developed. I assume that spatial distortion has been a minor concern of recording engineers, because it is difficult to hear through the typical monitoring systems and because pan-potted stereo is such a convenient and popular recording method. Listeners to classical music expect to hear not only the direct sound from the instruments of the orchestra but also to have a sense of the venue's response. Consequently in classical music recording some attention is paid to rendering sources and space for a plausible Auditory Scene.
(my bold)

http://www.linkwitzlab.com/accurate%20stereo%20performance.htm"http://www.linkwitzlab.com/accurate%20stereo%20performance.htm .

 

[jkeny]

Fair enough, I'll not mention it again, except to say that instead of trying to track down an effect that may just be a phantom, approach it the other way round and listen while introducing phase shifts/delay etc. between the channels to directly observe the difference. It might give you a clue as to what you're looking for and how large the differences need to be.

Sure, I know that you are trying to be helpful, Of course I have thought of the possibility that it is not a "real" effect but I'm afraid mine & countless other ears disagree with that stipulation. I would prefer not to get into this warzone, however

The demo I linked to already proved that 0.22mS of timing difference between channels audibly shifts the sound so no need for your experiment. I just was asking the question if this is the lower limit of an audible shift in lateral imaging or does it go lower. My supplementary question was - if there is a fluctuating timing difference between channels, what is the lower limit of timing difference where this is audible - I suspect very much lower than 0.22mS.

 

[jkeny]

S. Linkwitz addresses this subject in his site when addressing loudspeakers, not electronics or sources: A large contributor to Auditory Scene distortion is the loudspeaker. You can test your sound system's accuracy by how readily it reveals artificiality in recordings. See Accurate performance of a stereo sound system. Linear and nonlinear distortion in audio systems can be readily measured and expressed in numbers, though the degree to which these distortions are perceived is often not proportional to the numbers. Spatial distortion of the Auditory Scene can be perceived and described, but a method to measure the distortion and to express it in numbers has yet to be developed. I assume that spatial distortion has been a minor concern of recording engineers, because it is difficult to hear through the typical monitoring systems and because pan-potted stereo is such a convenient and popular recording method. Listeners to classical music expect to hear not only the direct sound from the instruments of the orchestra but also to have a sense of the venue's response. Consequently in classical music recording some attention is paid to rendering sources and space for a plausible Auditory Scene.


http://www.linkwitzlab.com/accurate%20stereo%20performance.htm"http://www.linkwitzlab.com/accurate%20stereo%20performance.htm .

Thanks Micro, I have removed your bold & substituted my own. What to take from the emboldened text - is it to be interpreted that we perceive a greater disturbance than the numbers would suggest or the opposite?

 

[j_j]

To answer the question about changing sources. If the two sources are from the same digital bits, then any difference has to be ascribed to level, hardware, whatever. If you're talking about LP vs. CD, the distortions on LP are known to enhance the sensation of stereo soundstage and dynamic range, at the cost of going over the edge into annoying distortion if there's just a little bit too much.

So when you say "different sources" can you specify what you mean, the answer may be very different.

Not too many people use tape now days, but tape has another set of distortions that affect spatial sensation differently, sometimes positively, sometimes indifferently.

Rather than get into the "imagination" argument, I refer people to www.aes.org/sections/pnw/ppt.htm in particular the Heyser Lecture deck. There is also the recording of the local lecture I gave from the same material at http://www.youtube.com/watch?v=X_LN7qUiTpo

Watch the part around slide 14, at least. Notice, there is nothing to do with imagination, deception, delusion, etc, involved in the need for blind tests. It's simply how human beings process the data from their senses (which is many megabytes/second) into a few bytes/second that are the actual long-term memories and thoughts. The process is incredibly plastic, and after the basic sensory input, it's unlikely that anyone ever processes the same signal the same way twice.

 

[j_j]

The demo I linked to already proved that 0.22mS of timing difference between channels audibly shifts the sound so no need for your experiment. I just was asking the question if this is the lower limit of an audible shift in lateral imaging or does it go lower. My supplementary question was - if there is a fluctuating timing difference between channels, what is the lower limit of timing difference where this is audible - I suspect very much lower than 0.22mS.

As I already said, the numbers out there are 2, 5, and 10 microseconds. I'm confident in the 5 microsecond number myself, and that's under 220 microseconds by rather a lot, isn't it? On the other hand, fluctuating timing is going to reduce that sensitivity, not increase it.

 

[jkeny]

As I already said, the numbers out there are 2, 5, and 10 microseconds. I'm confident in the 5 microsecond number myself, and that's under 220 microseconds by rather a lot, isn't it? On the other hand, fluctuating timing is going to reduce that sensitivity, not increase it.

Yes, sorry, I missed that - thinking your were specifying milliseconds, not microseconds (I always got caught out by units in my science classes )

So, I guess my question comes down to - has anybody ever tested this level of timing in interchannel signals in audio reproduction?

 

[jkeny]

To answer the question about changing sources. If the two sources are from the same digital bits, then any difference has to be ascribed to level, hardware, whatever. If you're talking about LP vs. CD, the distortions on LP are known to enhance the sensation of stereo soundstage and dynamic range, at the cost of going over the edge into annoying distortion if there's just a little bit too much.

So when you say "different sources" can you specify what you mean, the answer may be very different.

Not too many people use tape now days, but tape has another set of distortions that affect spatial sensation differently, sometimes positively, sometimes indifferently.

Rather than get into the "imagination" argument, I refer people to www.aes.org/sections/pnw/ppt.htm in particular the Heyser Lecture deck. There is also the recording of the local lecture I gave from the same material at http://www.youtube.com/watch?v=X_LN7qUiTpo

Watch the part around slide 14, at least. Notice, there is nothing to do with imagination, deception, delusion, etc, involved in the need for blind tests. It's simply how human beings process the data from their senses (which is many megabytes/second) into a few bytes/second that are the actual long-term memories and thoughts. The process is incredibly plastic, and after the basic sensory input, it's unlikely that anyone ever processes the same signal the same way twice.

I'm talking about changing bit-perfect digital sources & not analogue sources.
Yes, an interesting lecture & I concur with a lot of what you say in it. Agreed the perception of hearing is plastic & malleable but everything isn't just expectation bias - particularly when you have no expectation or a negative expectation, yet you find your bias is actually overturned in real world listening.

 

[j_j]

Yes, sorry, I missed that - thinking your were specifying milliseconds, not microseconds (I always got caught out by units in my science classes )

So, I guess my question comes down to - has anybody ever tested this level of timing in interchannel signals in audio reproduction?

Uh, that's where the numbers come from, in fact, they come from signals generated artificially in a 16/48 system, too, showing once again that you can get very much sub-sample time resolution in PCM.

Now, we are talking about very precise systems and headphone listening, but that's what you do when you're trying to find a just noticeable difference.

 

[j_j]

If there is a testable difference with different sources fed from the same bits, there is either a hardware problem or a gain error, or something of that sort. I'm not going to get into arguments about expectation and unintended biases, because frankly there is zero, and I mean ZERO doubt about the accuracy of the science surrounding it. If you can't find it in a blind test, it is not due to the audio stimuli.

Now, very small gain errors may create audible effects, but you will have no idea that they are gain errors. Gain errors interchannel will, effectively, create the same effect as time errors in a loudspeaker setup, at least up to 800Hz or so, via interaural interference from a 2-channel system. So there are many possibilities that are not simply resulting from human perception. On the other hand, the only way to be sure is a blind test. No, you don't need to have any expectations at all in order to have inadvertent influence happen.

 

[jkeny]

Uh, that's where the numbers come from, in fact, they come from signals generated artificially in a 16/48 system, too, showing once again that you can get very much sub-sample time resolution in PCM.

Yes, the signals were generated with various interchannel time differences to do the test but did anybody take the analogue output signals & check that they hadn't slipped in timing i.e that they were exactly the same as the timing differences being fed in at the the source? Moreover, I don't expect that anybody checked if these interchannel timing differences remained constant even when the reproduction system was also processing other signals such as music - that wasn't what the test you mention was for but you see where I'm going? Do you have a link to the experiments/results?

Now, we are talking about very precise systems and headphone listening, but that's what you do when you're trying to find a just noticeable difference.

Sure

 

[jkeny]

If there is a testable difference with different sources fed from the same bits, there is either a hardware problem or a gain error, or something of that sort. I'm not going to get into arguments about expectation and unintended biases, because frankly there is zero, and I mean ZERO doubt about the accuracy of the science surrounding it. If you can't find it in a blind test, it is not due to the audio stimuli.

Now, very small gain errors may create audible effects, but you will have no idea that they are gain errors. Gain errors interchannel will, effectively, create the same effect as time errors in a loudspeaker setup, at least up to 800Hz or so, via interaural interference from a 2-channel system.

But the level of gain errors needs to be very much larger to create the same effect as a very small timing error

So there are many possibilities that are not simply resulting from human perception. On the other hand, the only way to be sure is a blind test. No, you don't need to have any expectations at all in order to have inadvertent influence happen.

Yes, a well controlled blind test like MUSHRA would be needed, I'd imagine, otherwise there are too many uncontrolled variables which will probably result in a null outcome

 

[j_j]

But the level of gain errors needs to be very much larger to create the same effect as a very small timing error

.2dB (applied to both channels exactly the same) will, with some material, in some cases, (I'm being very conservative there, by the way) create an impression of different quality in two versions of exactly the same signal, with and without the .2dB increase or decrease in level.

I'm not sure what level of gain error you're referring to.

 

[JackD201]

Has anybody plotted the summing and cancellations of a pair of loudspeakers acting in unison?

 

[j_j]

Has anybody plotted the summing and cancellations of a pair of loudspeakers acting in unison?

Depends under what circumstanes you are talking about, and with what time duration of signals. You didn't ask a simple question.

 

[JackD201]

Every loudspeaker measurement I have seen including polar pattern has been that of a single speaker. I've yet to see measurements of a stereo pair interacting. Since this thread is about sound staging and sound staging is a function of two sources, I'm wondering if anybody has actually mapped out the pressure levels, correlated and uncorrelated, in actual space. More to the point, is the "phantom" purely a psychoacoustic phenomena or is there actually a rise in pressure where the image apparently is located, and what does this contribute?

In practice, I use correlated pink noise when setting speaker positioning. When I get to the point where the noise is like a sphere between the speakers, everything pretty much falls into place. I used to use a mono track and "stack" instruments but the method I use now is more expedient as far as getting me in the ballpark leaving relatively minor adjustments for tonal balance.

I was just wondering why it appears nobody has made a 3D map of two speakers working together unless arrival time is the sole factor when it comes to localization. Somehow I don't think it is.

 

[jkeny]

.2dB (applied to both channels exactly the same) will, with some material, in some cases, (I'm being very conservative there, by the way) create an impression of different quality in two versions of exactly the same signal, with and without the .2dB increase or decrease in level.

I'm not sure what level of gain error you're referring to.

Well, I mentioned early on in this thread that Diffmaker might give a handle on ICLD & ICTD but I didn't know. I came across this interesting blog from CA titled "Why I never got that perfect null in difference testing part three"
http://www.computeraudiophile.com/b...rfect-null-difference-testing-part-three-258/

It does appear to answer my question of why I didn't get perfect nulls. Gain/sensitivity of DAC and ADC units vary a tiny bit over time. Enough to prevent complete nulls. Then even once you have adjusted for those tiny level shifts, very small timing shifts prevent quite getting the complete practically perfect null of having only low level noise show up.

It looks, however, like this may have been an anomaly of the method used whereby a phase shift can be introduced between D/A conversion & subsequent A/D recording unless some steps are taken to ensure the conversions are done in-phase.

 

[j_j]

I was just wondering why it appears nobody has made a 3D map of two speakers working together unless arrival time is the sole factor when it comes to localization. Somehow I don't think it is.

Arrival time is the primary cue for the ears. At low frequencies, level differences from two speakers interact with HRTF's to create a facsimile of ITD. At high frequencies, this breaks down badly, leading to a variety of effects in image movement, instability, etc.

But any such 3D map of radiation patterns would be absolutely specific to any given setup, and even you moving around the room will change the results.