4 replies to this topic

[spazmatron]

Interesting discussion on sampling rates et al...

in a blind test between SACD/DSD and regular CD, respondants picked the correct unit approximately 50% of the time: ie., randomly.

http://mixonline.com...s_new_sampling/

Further interesting point to note the part about SACD's "sounding better" because the they are engineered for a niche market and not hard limited for top 40 consumption etc.

[rhythmboy]

Very interesting - glad I read it to the end before getting outraged!

Very well balanced article too, and he did his research, he cites the best of the best, those who should know...

I have heard the difference between 44.1 and 192 and 192 did sound tighter, clearer and more 'solid' - no noticeable change in frequency response though. And I was in a proper control room, listening to a solo'ed cowbell (yay!) in a direct A/B test. Hardly the real world and hardly a complex recording.

From the sound of it, I'd put my neck on the block and say that in theory, the higher sample rates do get closer to the analog continuum because the time interval between samples are shorter than a human brain's own 'sample rate', but the anomalies of our acoustic environment - and the comb filtering, time and phase shift that occurs when we simply move our head a bit - make discerning it in any useful way so difficult as to be largely meaningless to most people.

[em]

To my understanding, a sample rate at 96k will benefit if the ADC has poor resolution; the noise produced by modulation will be spread over twice the bandwidth of 44.1k, therefore lowering the noise floor.

[rhythmboy]

Quote

To my understanding, a sample rate at 96k will benefit if the ADC has poor resolution; the noise produced by modulation will be spread over twice the bandwidth of 44.1k, therefore lowering the noise floor.

For lower quantisation resolution perhaps , eg if it's an 18-bit converter or something. But most 96k ADC's are at least 24 bit these days. However the critical issue for the ADC is the stability of the crystal clock to capture the samples. In most sound cards and converters, the clock is not stable enough to do good sampling in the first place.

A few years ago a student of mine did some experiments with sine tones at 96k and 192k and found that with his semi-pro sound card, the higher he sampled the more triangular the sines became. So potentially any gain in noise floor is lost in harmonic distortion, which is worse because it can't be removed.

The bottom line is poor converters and high sampling rates should never mix. It's just too fraught with errors...

[em]

Yes, nobody wants a crappy ADC. The improvement could easily be achieved with a good quality converter at a lower sr. Yet in theory, this would happen; not arguing for a higher sampling rate.