Measurements & the stereo illusion

[JackD201]

Yes it would but why not a baseline as is done with stereo optics? The same could be said for mono (specific to any given set up) since we don't have our own anechoic chambers. There must be some utility to such mapping as far as placement/focus recommendations goes at least as starting points. As it stands most have to move around large speakers by trial and error and to make things worse, they are likely to be spiked at some point in the process.

The question is, if there is a 3dB sum in an area of space between the two sources (easily measurable as is any area of a room), how much of this is heard from the listening position? Surely even a bit of this but how can one tell when you can't isolate it from the direct sound?

 

[jkeny]

There is a nice test which directly relates to timing issue differences between channels - it's the IQtest from Jim LeSurf http://www.audiomisc.co.uk/Linux/Sound3/TimeForChange.html
"The IQ-Test is a method for measuring the rate of replay of a stereo (or multichannel) digital system. In essence it is a form of ‘wow and flutter’ measurement. The approach is based on general FM and comms theory."

In which he generates an input file for playback & the results analysed. This input file takes advantage of " IQ modulation and demodulation (as widely used in modern communications) provides an easier approach. This is based on having two waveforms transmitted in parallel. These share the same frequency and amplitude. But they differ in phase by ninety degrees. We then say one of these is the ‘In Phase’ component (I) and the other is the ‘Quadrature’ component (Q) of this composite IQ waveform. This approach has a number of convenient mathematical features that make determining the phase much simpler."

So the input file has a mathematically defined phase relationship between interchannel sample pairs. Any change in this phase represents a timing difference between the pairs created by the electronic reproduction system. The full mathematical details are here http://www.audiomisc.co.uk/Linux/Sound3/TheIQTest.html

A stark graphical difference of two bit-perfect sources is shown here, which clearly demonstrates that timing differences can exist when different bit-perfect sources are used as input


What can be seen in this picture is a comparison of the USB input of the DAC Magic Vs using a Halide Bridge USB receiver feeding SPDIF into the DAC Magic. Clearly there are timing differences shown.

A more subtle timing differences can be seen in the two graphs of the Arcam rDAC & Halide Bridge




The spike at 0.1Hz is the obvious difference between graphs but there are others.

Whether these differences account for the audible differences between rDac & Halide Bridge is of interest. The more significant differences between DAC Magic & Halide bridge would seem to qualify.

 

[j_j]

Yes it would but why not a baseline as is done with stereo optics?

Well, because it doesn't work like stereo optics works, for starters, and then stereo optics do not have to contend with reflections (early and otherwise) in the room, either. Stereo optics are one image applied to each eye. Each speaker has a path to both ears, just for starters, quite aside from any room issues.

The question is, if there is a 3dB sum in an area of space between the two sources (easily measurable as is any area of a room), how much of this is heard from the listening position? Surely even a bit of this but how can one tell when you can't isolate it from the direct sound?

Are you referring to summing of energy, as opposed to summing of amplitude, or what, here? I can't make out what you're asking.

Some speakers do publish 2D polar patterns, but they are few and far between, sadly.

More interesting would be publishing both direct (0,0 direction) and total energy magnitude response, which some speakers do publish, as well, but there's still a lot of missing information.

 

[JackD201]

I'm not trying to be a wise guy JJ. Honest. Using one speaker, that one sound event has a path to both ears. Two speakers, plug one ear, two events have a path to the one open ear. In an anechoic chamber, there are no reflections. I'm referring to summing of amplitude.

Yes, instinctively I feel there is still a lot missing. Sadly I do not know what is. That is the spirit inwhich I'm asking all these crazy questions.

 

[j_j]

I'm not trying to be a wise guy JJ. Honest. Using one speaker, that one sound event has a path to both ears. Two speakers, plug one ear, two events have a path to the one open ear. In an anechoic chamber, there are no reflections. I'm referring to summing of amplitude.

Yes, instinctively I feel there is still a lot missing. Sadly I do not know what is. That is the spirit inwhich I'm asking all these crazy questions.

I'm trying to figure out what you're asking, really. When you sum amplitude of two identical signals, you get 6dB. When you sum two incoherent signals with the same spectrum, you get 3dB.

Yes, there is a lot missing, it's hard to add unless one measures the (*&(*&*( out of a single specific room.

 

[JackD201]

Is the 3dB difference due to difference in phase/direction? I barely passed physics. I regret this because I just wasn't interested at the time. I am now.

Sometimes I just try to imagine the signal carrying say a tom strike as it gets distributed to a multi-driver speaker and how the drivers are behaving in time. I mean, how the bulk of the acoustic energy comes out in the mids and tweeters but as this dies off the woofers carry on the decay. It's a bit mind boggling how any of this works at all. Then we talk about sound staging and now we're talking about two of these speakers set in an environment that can pretty much be described as random at best. Yet it works. Not as well as having many discreet channel but it does work and arrival time isn't all of it even if it is the bulk of it. I guess, going back to the original question, I'm asking what explorations the scientific and engineering community have been doing with regards to these other factors. We know that pairs of speakers with differing polar patterns "stage" differently because of the way they excite the air in the room differently. What ties them all together? What are the bare essentials, the baselines? Harman says smooth off axis response is one factor that their subjects prefer but again these speakers are tested singly and not in pairs. If anything, I'm just wondering why products sold as pairs, meant to be used in pairs, are tested singly. I can imagine it seriously complicates things introducing so many variables but I'm pretty sure if the great minds put their backs into it, some protocol could be devised. The disappointing thing would be if they just said "what for?", "just for the heck of it" would be good enough reason for me. If I had the know how and the resources to do it. I'd do it myself. Obviously I have neither.

 

[j_j]

Is the 3dB difference due to difference in phase/direction? I barely passed physics. I regret this because I just wasn't interested at the time. I am now.

Nobody can know everything. If you look up linear system theory, signal addition in amplitude vs. in power, cross-correlation, that kind of things, you can get an idea of what's going on.

Maybe it's time to write a blog post on why polar pattern matters.

 

[JackD201]

Thanks JJ that would be nice of you.

 

[jkeny]

Thanks JJ that would be nice of you.

I guess the whole ICTD issue is ignored then?

 

[JackD201]

No I don't think so John. It's just another part of the puzzle. One in the electrical and the other in the acoustic domain.

 

[jkeny]

No I don't think so John. It's just another part of the puzzle. One in the electrical and the other in the acoustic domain.

Sure, but it seemed to be ignored as I posted actual measurements which showed inter-channel timing differences between bit-perfect sources - something which directly contradicts the position taken by many. To me it raises a whole series of questions regarding the playback system & the assumptions underlying it's operation. It is interesting that nobody has the inquisitiveness to ask these questions.

 

[JackD201]

I'd have asked John but honestly, at this point in time, I can't make heads or tails of the measurements and what their implications might or are supposed to be. I just don't know the context.

 

[jkeny]

I'd have asked John but honestly, at this point in time, I can't make heads or tails of the measurements and what their implications might or are supposed to be. I just don't know the context.

Sure, I know it's a bit of a read into the technique but in simple terms what he's measuring is the "wow & flutter" of digital playback in parts per billion & parts per million i.e very sensitive measures. He does this by generating a file which has two different signals generated on each channel of a stereo signal. These two channels have a very precise mathematically defined phase relationship between each sample pair (one channel is a sine wave & one channel a cosine wave) which can be derived mathematically.

He plays this generated file back through a playback system & records the output which he then analyses. The variation in phase between any two sample pairs can be then analysed & any variation from the theoretical phase shows a timing difference which arose in the playback system. This can be done for one sample or a any range of samples. His analysis program derives the timing differences across both LF & HF frequencies.

I guess it's somewhat like a jitter measurement but more generally applicable to any digital playback system & more capable of revealing both short term (flutter or jitter) & longer term (wow or drift) discrepancies. Jitter measurements focus on a specific signal & are short term measurements - this input signal varies across the full spectrum & seems to be able to analyse timing issues across any duration.

In the first plot it can be seen that playing the file through the DAC Magic's own USB input results in a jump in timing every 10 secs, compared to the playing back of the same input file through an external Halide bridge USB to SPDIF receiver & then into the DAC Magic's SPDIF input which doesn't show this jump. Substituting an rDac USB to SPDIF receiver instead of the Halide bridge shows a similar pattern to the Halide but not exactly the same.

I am interested in the veracity of the results as it pertains to my initial question - do better reproduction systems (this includes the PC) have better timing characteristics at these low levels?

 

[jkeny]

perhaps i should add that when my technical audiophile friends and i did null tests, with music, in the early eighties, while with one song the gear would null like 40db or so, the next song it would null a bit more or sometime quite a bit less, like 8 db or so, so, it showed how the particular frequenices of the song were affected by our solid state feedback amplifiers at the time.

But the song still sounded good, each song, cause really now...

.................there is no reference in stereo recordings......only direct comparison to another flawed unit.....

This test seems to provide an absolute reference by using a file which has specific phase differences across each sample pair

 

[jkeny]

as i say in my byline, i pursue detail/tone over soundstage. i appreciate soundstage, but plain old stereo, has so many issue with balance (in all manners including time and phase and group delay) between channels, in the putting together of the recording, the storage medium, the playback electronics, the speaker balance, the room balance, your head locked in a vice, your ears balance, your ear/brain interface and expctations........welcome to the real world of infiinite variables with at best just getting the least damage from the entire signal chain..

i guess i would add that while these changes to timing we are discussing here in this thread happen, and differently with different gear, we can still enjoy the stereo effect, so while it is electrically "affected" by these issues, those who understand that a stereo recording can not duplicate a live event, dont seem to be too worked up by this issue as we listen from song to song....ie if its a problem, it may or may not be the reason two different electronic components (forget transducers) sound "different" but dont sound "bad".

Agreed Tom, that there are many variables when changing from one system to another, one room to another & yet we still seem to be able to enjoy the reproduction (within reasonable parameters of room reflections, etc.)

But when changing just one parameter in an existing system (forget transducers for the moment) & this changes the sound stage, I ask why. My thoughts here are just an attempt to delve into the possible whys (based on the little I know about psychoacoustics) & also to ask if it has been examined before in any detail.

 

[jkeny]

I should say that his recording of the playback is directly from the analogue output of the DAC Magic i.e before it goes through an amplifier or speaker. So these variables are eliminated but I have no doubt that the system could be used to record the output from an amplifier or speaker for that matter - I just don't know if the results would be of value when recording from speaker outputs.

 

[jkeny]

i am agreed that we can track phase all the way to our ears if need be. It is a good test at my first look at it. stereo is about timing and amplitude..change either one and all others being equal, and the change big enough, the soundstage illusion changes. like everything else, the question is "how much" of a change is audible or perhaps disagreeable. Thanks for bringing the test method to our attention. it is more specific than a null test for sure as far as what it is desinged to look at (phase).

i dont think any commercial manufacturer quotes their phase deviation across the audio frequency band. and never will?

Yes, the just audible phase differences that matter was a question I asked JJ & his reply was 2, 5 or 10 microseconds; 5 being the number he would stand over.
So the question becomes - does/can a playback system vary it's phase differences by 5 microseconds or more. Note, I don't mean a fixed 5 microseconds but a varying phase difference during replay. My next question is what effect such a variation in timing might have to the listener? Now you may say this happens all the time anyway, by our head movements during listening but is that correct or do we have an inbuilt mechanism for adjusting our head movement which compensates for this? Whereas when such small apparent movement is externally happening, we perceive it as a vagueness in the position of the location of an object in the sound field.

Here's an interesting video of just such a use of head movement in fox hunting a prey under the snow
http://auditoryneuroscience.com/sites/default/files/foxInSnow.mp4

Edit: Isn't nature something to behold??

Edit2: Ah, just found out - this head movement is called "Purposive Motion" so we map our head movement to our individual HRTF to locate the position of objects in the sound field. Therefore we don't need our heads to be locked in a vice to recognise/evaluate sound stage & it also goes some way to explaining why headphones do not cut-it for evaluation of sound stage differences in audio

 

[Robh3606]

So the question becomes - does/can a playback system vary it's phase differences by 5 microseconds or more. Note, I don't mean a fixed 5 microseconds but a varying phase difference during replay. My next question is what effect such a variation in timing might have to the listener?

Hello jkeny

I quoted JJ's post below. Seems to me it is implied that this testing is done in rather contrived conditions. Forget about head movement what about the delays in the sound field due to reflections and multiple paths to the ear. How could you hear the timing fluxuations? Seems to me the conditions we normally listen under are not the best conditions for hearing small differences in time of phase. If anything wouldn't they tend to mask them?

Which could be why you need special conditions to test for them and hear them?

The threshold of hearing for ITD is reported to be 2, 5, or 10 microseconds, depending on who you believe, and what kind of utterly artificial, carefully designed stimuli one uses in order to make the test as easy and sensitive as possible.

Rob

 

[jkeny]

I mentioned a simple thought experiment before (which could also be done as a real experiment) to provide a rough analogy to what might be the perceived effect of such timing differences in an audio system.

Say we mounted the speaker boxes on platforms that could be independently moved by some mechanical device. What would the perceived difference be if those speakers were individually vibrated back & forth in relation to the listener by small amounts. Remember, I'm not talking about both speakers moving in unison by X millimeters but independently (as this is what I believe might be happening in the electronics) How much vibration movement would be audible as a more diffuse sound stage?

So if we apply some maths (& correct me if I'm wrong) - the speed of sound in air is about is 343metres/s therefore in 1 microseconds it travels 0.3cm. Now let's take the just audible timing differences to be 10 microseconds, therefore if the speaker boxes vibrated by a distance of 3cms would this be noticeable in the sound stage? But if we consider that the speakers are moving independently then a 3cm vibration would result, at certain times, in the speakers being out of alignment by 6cms i.e the differential between the sound from L speaker & right speaker reaching our ears would be 20 microseconds

Ok, so reflections, etc are also part of this scenario but they would also be effected by the speaker vibrations & come at varying intervals. Rather than this masking the effect, I imagine it would heighten the diffusiveness of the perceived soundfield - psychoacoustics & the workings of exactly how we localise sound come into play.

Again, I'm simply trying to tease out the possible issues.

 

[jkeny]

Hello jkeny

I quoted JJ's post below. Seems to me it is implied that this testing is done in rather contrived conditions. Forget about head movement what about the delays in the sound field due to reflections and multiple paths to the ear. How could you hear the timing fluxuations? Seems to me the conditions we normally listen under are not the best conditions for hearing small differences in time of phase. If anything wouldn't they tend to mask them?

We seem to be able to do the job of hearing the illusion of sound stage from stereo given reasonable reproduction & room. So if your premise was correct then we would not be able to do this except in very limited, controlled circumstances, yet we regularly hear sound stage produced from our replay systems despite less than optimal condition, no?

I don't know the test setup JJ is talking about or what/how it was tested but my experiment could be of interest?