[ProAudio] The High-Resolution Challenge

[Mike]

On 2/14/2020 5:28 PM, Dan Lavry wrote:
 > I am not saying that 48KHz is better or worse than 96KHz. Clearly at 
44.1Khz
 > the theoretical bandwidth is 22.05Khz. That is already a couple of dB 
loss at
 > 20Khz. Add a 20Khz mic for another 3dB, and perhaps a speaker for 
another
 > 3dB.. pretty soon you find it to be difficult to accomodate the 
20-20Khz hi fi
 > standard.

At the time when this pipe-at-a-time organ company was in business, 
Sennheiser was touting a mic that was only a couple of dB down at 50 
kHz. And if you're also supplying the speaker system for the organ, 
there are a number of transducers that get up into the ultrasonic range 
to take over where the cone loudspeakers poop out. I'm not arguing that 
this is all a great idea, but it's feasible. And when the ultrasonics 
mix in air, you get something in the audible range.

 > It seems to me that too much is attributed to figuring sample rate, 
which is a
 > step removed from the question of what we need for bandwidth.
 > I undestand that it us not intuitive to understand that you only need 
to reach
 > some minimum number of samples for a given bandwidth, and more samples
 > do not add any information for perfect reproduction of the original.

Well, in this case, the folks building the organ had a rational case for 
needing extended bandwidth to do what they wanted to do. I record 
fiddles and banjos, and I don't miss anything at 44.1 kHz.

One case for higher sample rate is that for any given frequency, the 
samples are closer together, so for the standard audio bandwidth, you 
have twice as many samples as Shannon and Nyquest say you need in order 
to perfectly reconstruct what went into the system. In a universe where 
we do a lot of digital number crunching to emulate filters and dynamics 
processors (ahem! non-linear elements) you get less rounding error when 
you average more points. This is an explanation (not sure if it's the 
correct explanation) of why up-sampling can make a standard sample rate 
converter sound better.

 > I wrote a whole 32 page paper Sampling Theory, based on Nyquist work, 
but
 > some folks can't shake the notion that the more is better....

I read that years ago, probably still have the PDF somewhere. I remember 
thinking that you were absolutely right then, and you still are today. 
The point is that because we can sample faster than Nyquist says we need 
to, we can come up with situations where there's content above the 
standard audio band that, when taken out of isolation, does make a 
difference when sounds interact.

[added to my reply to Dan after reading the next blast of digest]
Apologies to Dick and JJ regarding the non-linearity requirement for the 
supersonics to combine and make a difference (both meanings). That's how 
they explained it to me and I didn't know enough to argue about it. I 
didn't realize that air was that linear at respectable SPL. But if they 
weren't convinced that it mattered, they could have saved a lot of money 
on hardware and disk space, as this was before the days of the $29.99 
2-terabyte drive.

-- 
For a good time call http://mikeriversaudio.wordpress.com