On 2012, May 16, at 12:21 PM, Hébert Sylvie wrote:Dear All,
Some thoughts . . .
Perhaps the question is a little more complex if the source of the testing signal is digital. It appears from the tables that I have that the DACs and sampling rate would need to be 96kHz or even 192kHz, or higher.
An experience reported to me by Eldad Tsabary while teaching an electroacoustics ear-training class was about JNDs and sampling rate. He produced sample 'clicks' of different frequencies [in effect, a few cycles long], and the differences created by different sampling rates seems to be quite clearly audible for those whose hearing goes much above 4kHz.
The problem is shown in the table below.
Working at 44.1kHz the frequencies between [eg] 4.7kHz and about 5 kHz will all be represented by 9 samples; from 5.9kHz to 6.6kHz, by 7 samples. There is slight improvement with a 48kHz sampling rate. At 96kHz sampling rate, the error is 'simply' shifted up an octave. This is important since the question is about JND up to 14kHz.
A [not so simple] test of this would be to use two sine tone:
one of 7.2kHz; one of 7.3kHz
• four cycles of each
and do this at 44.1kHz, 48kHz, 96kHz and 192kHz. Two 'tones' -- in effect high frequency clicks, at four sampling rates.
[At 44.1 and 48kHz, this works out to be about 24 samples±. It will probably be necessary to adjust the last few samples to avoid an abrupt drop [or rise] to the zero crossing.
The abruptness of the return to the "0" state could cause ringing in the loudspeaker.]
It may be necessary to work at 384kHz for there not to be 'differences'.
The literature I have read on this -- web press commentary, calls such ideas ridiculous and [in effect] heretical.
A [golden eared - my description] audio mastering person I have worked with reports hearing "differences" in material mastered at 96kHz and 192kHz. Having done some critical listening on his system, I believe that this is likely perceptible.
That aside, another step in this chain is to get the loudspeakers to be used - I expect they would need to be in the $1000+ range, and place them adjacent to each other, and then repeat the test listening for differences in loudspeaker responses and coloration.
On 2012, May 16, at 12:21 PM, Hébert Sylvie wrote: