I don't believe cables make a difference, any input?

Status
Not open for further replies.
GlidingDutchman said:
Bistening to sinewaves have absolutely no purpose in assesing the sonic signature of audio gear.
Well, if one were to reduce test signals to steady state sines, there is something to you statement (still not very much, though). But, there are myriads of proper test signals (some even very similar or even derived from "real" music signal), and the primary requirement for a test signal is that it tries to isolate some phenomenon, decreasing the observer's "signal-to-noise ratio". For example I have created a series of complex test signals, trains of sine bursts with very specific characteristics, which allows the user to judge/verify/optimize the bass response/"character" of his system (speaker+room, predominantly) in greatest detail, without the need to measure something (the ear is the detector), since most people don't have measurement gear and don't wont to bother with that anyway (for good reasons). This signal set is already quite popular in Germany (it's a CD kit, that is, the test signal generator software), I'll plan to do english edition.

- Klaus
 
ravon said:


Oh, it can be very useful to use sine waves (among other signals) in listening tests.


No person has the brain power to hear minute dB deviations over the audio spectrum.
You can’t listen to a freq sweep and then say … “Okay guys! We have a peak of 1.5dB at 2.5 kHz.”

If you can then you are abnormal and should put your brain power to better use than entertainment - rather free energy and world peace! :dead:

On the other hand when you listen to a full spectrum recording you can get a better idea of how the system is performing. Best is to A/B with a live recording vs live performance.

D
 
nigelwright7557 said:
There is modern equipment that woudl have no problem doing this.
DSP's run at many times 1uS.
Agreed. But what of the first part, that of correlation input to output?

And in reality, it's a two input port, two output port problem with both correlation and cross correlation..

One input cannot alter the other system. Ground loops do not care which system is the victim, nor which is the aggressor..

Cheers, John
 
jneutron said:

Agreed. But what of the first part, that of correlation input to output?

And in reality, it's a two input port, two output port problem with both correlation and cross correlation..

One input cannot alter the other system. Ground loops do not care which system is the victim, nor which is the aggressor..

Cheers, John


If you use the same mic to measure both signals then you would get exact comparison.
 
KSTR said:
For example I have created a series of complex test signals, trains of sine bursts with very specific characteristics, which allows the user to judge/verify/optimize the bass response/"character" of his system (speaker+room, predominantly) in greatest detail, without the need to measure something (the ear is the detector), since most people don't have measurement gear and don't wont to bother with that anyway (for good reasons).
- Klaus

Klaus - okay, if you can do a combination of signals with tests designed then I will take them as valid but just listening to a pure sinewave sweep is nonsense.

Any site to read more on this - my German is not that bad... :clown:

D
 
SY said:
Once you've reduced a soundfield to a discrete number of channels, it's all illusion. That's far and away the biggest error. If there's a better way to do it, that strikes me as a much more fruitful line of research than (say) wire directionality.
SY, you should give "Trinaural / Optimum Linear Matrix" a try. Quite probably you won't go back to listen to "flat earth" style anymore, conventional 2-ch playback of 2-ch material. 3 properly rematrixed playback sources do very much for increased illusion with like 99% of the 2-channel source material.

- Klaus
 
KSTR said:
For those who can read german I'd recommend http://sengpielaudio.com, he's a top notch authority in the field. Bottom line on the ITD/ILD issues is that there is not exact (as such) mathematical relationship between ITD/ILD and localization ("position" and "size" of the phantom source), it' all very empirical, different researcher got different results. It depends on many secondary factors and of course is bound to the 2-ch playback in the equilateral triangle setup. When I tested "Trinaural/OLM" and other rematrixing playback setups with 3 speakers (L-C-R) I found (with music) and verified (with specific test signals, for that matter @GlidingDutchman) some interesting differences in the way ILD/ITD relationships a percieved, especially in the "size" parameter. "Direct" (mainly ILD) sources get more "direct", "diffuse" (more ITD) sources get more "diffuse", which partly accounts for my faible for this reproduction variant (which still is very compatible, execpt for HRTF-based localization trick -- Q-Sound -- which cannot and does not work).

- Klaus

The problem must be seperated into four components first.

1. What localization parametrics arrive at the human ear? IOW, what are ITD/IID (or ILD) for an object in space which emits sounds? I'd limit to point source, as it provides the fastest IID dropoff (1/r^2 as opposed to 1/r for linear and unity for planar).

2. What is required to duplicate this soundfield in intensity and vector for a listener?

3. What algorithm is required to modify the human response to a two source simulation of a single source? (owing to the lack of an absorbtion septum down the midplane..

4. Given all the above, what system rigidity is required to provide the end user a specific localization error profile? (assuming a 2-D gaussian, depth and width)

German, eh? Oh vell..

Cheers, John

edit..sheesh, can't even count to four...:eek:
 
GlidingDutchman said:
[...]but just listening to a pure sinewave sweep is nonsense.
I have different experience. Simple sines can be quite revealing for specific tests. And, I dare say a low to mid frequency mono sine sweep (30Hz to like 600Hz or so) is the hardest test I know of for pin-point LF center image and LF imaging in general. 90% precent of the setups I've heard failed on this, and those which didn't were excelling on this with real music -- or better said, once you heard how much the center image will break out from smack center by big amounts you know why ie male singer's voices sound too big and blurry on not so well centered systems.

- Klaus
 
GlidingDutchman said:
No person has the brain power to hear minute dB deviations over the audio spectrum.
You can’t listen to a freq sweep and then say … “Okay guys! We have a peak of 1.5dB at 2.5 kHz.”

I have helped openminded people who have no knowledge of signal analysis solving bass problems in their listening room using sine sweeps and continuous sines. For that purpose it is not necessary to know dB levels, the only thing really important is sine frequency at resonance and that is not difficult to estimate, not even for unqualified people who appeared to be able to solve their own bass problem.

If you can then you are abnormal and should put your brain power to better use than entertainment - rather free energy and world peace! :dead:
I feel allright, thank you :D

On the other hand when you listen to a full spectrum recording you can get a better idea of how the system is performing. Best is to A/B with a live recording vs live performance.
D
Perhaps.

As I said earlier, it can be very useful to use sine waves among other signals in listening tests.

Another example: The listener in an ABX test is a measuring instrument which needs some sort of calibration. In my experience test signals seem very useful for that purpose.

Need more examples?
 
KSTR said:
I have different experience. Simple sines can be quite revealing for specific tests. And, I dare say a low to mid frequency mono sine sweep (30Hz to like 600Hz or so) is the hardest test I know of for pin-point LF center image and LF imaging in general. 90% precent of the setups I've heard failed on this, and those which didn't were excelling on this with real music -- or better said, once you heard how much the center image will break out from smack center by big amounts you know why ie male singer's voices sound too big and blurry on not so well centered systems.

- Klaus

Yes Klaus, you are right it could be helpfull for setup (which is very important) but certainly it won't help for testing cable differences.

I like to use a mono pink noise signal to test the setup also, work quite nice to show problems.
 
KSTR said:
SY, you should give "Trinaural / Optimum Linear Matrix" a try. Quite probably you won't go back to listen to "flat earth" style anymore, conventional 2-ch playback of 2-ch material. 3 properly rematrixed playback sources do very much for increased illusion with like 99% of the 2-channel source material.

- Klaus

Back-compatible with stereo recordings?
 
jneutron said:
The problem must be seperated into four components first.

1. What localization parametrics arrive at the human ear? IOW, what are ITD/IID (or ILD) for an object in space which emits sounds? I'd limit to point source, as it provides the fastest IID dropoff (1/r^2 as opposed to 1/r for linear and unity for planar).

2. What is required to duplicate this soundfield in intensity and vector for a listener?

3. What algorithm is required to modify the human response to a two source simulation of a single source? (owing to the lack of an absorbtion septum down the midplane..

4. Given all the above, what system rigidity is required to provide the end user a specific localization error profile? (assuming a 2-D gaussian, depth and width)
Hi John,

ad 1: This is covered by research on natural hearing. ITD/ILD is only some crude approximation to two variables, in reality we have to deal with HRTF (head related transfer functions). BTW its ILD (Level), not IID (Intensity is not the proper physical quantity). And the point source's free field SPL falls with 1/r, not 1/r² (this error is all over in the literature, even the most recpectable university professors have that wrong). All that are things I learned from Mr. Sengpiel ;)

ad 2: This can only be done properly with WFS (wave field synthesis).

ad 2: Here we are tricking our brains. The ITD/ILD relations used/created with stereo miking techniques etc and reproduced with the known speaker setup do in no way resemble any sound field a real source would have created and have nothing to do with the ones at work with natural hearing. It really is a wonder that phantom sources work at all, and it's quite a bit of a learning process involved (I experienced this preconditioning -- having learnt "how stereo sounds" -- with my mentioned rematrixing experiments). As I said, the proper "ITD/ILD to phantom localization transfer curves" have been found completely empirically, and are very signal dependent.

ad 4: I have problems to understand what you mean... all I can say that IME trinuaral is less prone to "imaging errors" from things like listeners head shift/turn that conventional stereo.

- Klaus

PS: aintz, tswy, dry, fear, ....
 
Andre Visser said:


Yes, certainly, else show me how you do it. :devilr: (Talking about audio here)

I suppose you are familiar with the ABX test? That's an excellent method for investigating the audibility of cable differences. But as with all instruments it must be used in the right way otherwise the results may be invalid.
The listener in an ABX test is a measuring instrument which needs some sort of calibration (and training). In my experience test signals seem very useful for that purpose.
 
SY said:
Back-compatible with stereo recordings?
Yes, to an astonishing degree... but there usually is a little "relearning/accomodation" period, at first that center speaker is a bit irritating... and like everything else it's a matter of personal taste (some people just cannot get used to it). The net illusion improvement one might experience is also greater the less "perfect" your normal stereo is, if you already have highest class speakers and paid minute attention to room acoustic details and such, the net gain could be less spectacular. But I still would say it always is rated a leap forward one order of magnitude, not only a tiny (but still valuable, at any rate) improvement you get from switching to better system components (let alone cables), for the people who have tested it extensively. I personally don't know of anybody who ever switched back, exept temporarly for some very specific occasions, listening to material with a big amount of Q-sound effects etc and a few classical recordings which seem to be a bit more incompatible than what is tolerable to them. With most pop/rock/jazz recordings it works exceptionally well.

Key point is to understand that it is not anything like an artifical processing that really changes any data/information (the algorithm is dead simple and transparent/reversable, "lossless"), rather it is just a different approach to image a soundfield into a room, a different projection mechanism.

- Klaus
 
jneutron said:


The mere act of conversion from time varying waveform to a magnitude spectra loses ALL timing information.

There are an infinite number of waveforms that can have the exact same FFT spectra.

There must be correlation between the input and the output.

And don't forget, it's gotta sample sufficiently fast to get into the 1 uSec domain, as well as below 1dB differences.

Cheers, John

What's wrong with measuring a Transfer Function?
 
KSTR said:
Hi John,
And the point source's free field SPL falls with 1/r, not 1/r² (this error is all over in the literature, even the most recpectable university professors have that wrong).

I'm speaking of an omni point source, propogation in full space..I calculated it thusly:

The surface area of a sphere is 4 Pi r^2.. For a specific energy front, the area is the square of the radius.. the energy per unit area follows 1/r^2...

Why would it be 1/r?


ad 2: we agree..

ad 3: we agree in general.
KSTR said:

ad 4: I have problems to understand what you mean..


I figured...I've not explained very well, sorry..

Assume the signal content has been perfectly modified to present the exact soundfield required for fooling the human into perceiving the exact images intended.

What measured parameters must the electronics meet in order to guarantee that the perfect soundfield is not altered such that a human can detect an apparent shift in any of the source content.

If for example, the power amp is reproducting the 5 Khz signals accurately that the image is at center for that signal, if suddenly one channel has a large 50 hz note added to the signal, can that channel maintain the same time and amplitude relationship for the 5khz contend such that the image remains solidly in the center? If the center is composed of a large harmonic content signal, will ALL of the harmonics remains solidly in the center, or will some drift to one or the other side.


KSTR said:

PS: aintz, tswy, dry, fear, ....
OOOOOOOOH...I deserved that...;)

Cheers, John
 
ravon said:


What's wrong with measuring a Transfer Function?

Not a thing.

My discussion centers around the criteria for the TF.

But testing the TF requires both channels operational at good power levels, and both must maintain unweildy levels of accuracy in time and amplitude while both channels are driven with content where some is correlated and some is not.

To a microsecond..

But nothing difficult of course...:eek:

Cheers, John
 
Status
Not open for further replies.