View Single Post
Old 14th June 2019
Gear Maniac

Originally Posted by IanBSC View Post
Here is an example: I couldn't find the specific passage so I'm assuming he is talking about oversampling in a converter and not a plugin. Virtually all modern converters oversample at 128 or 256 times, WAY beyond x2. Lavry doesn't like how 192khz effects HIS converters. And it's a whole different question whether or not his converters sound the best...I've found them to be very good, but I don't favor them above all others.

Ironically, Lavry makes a case for 96khz, but he tends to get cited most often by advocates of 44.1.

This is true AND its also why Greg Wells doesn't post here. He started a thread about how claimed his Antelope clock improved the sound of his mix and people came out of the woodwork calling him a fraud and a shill, or simply deluded by his gear purchase.

Dave Derr made a large contribution to this thread early on, and he isn't around anymore either.

I think you'll find most converters have lower latency with higher sample rates.

Let's differentiate between DACs and ADCs.

The MHz input stage of a AD converter works at low bit resolution. It is just an intermediate step to make it so that the analog anti-aliasing filter that precedes it doesn't need to have a steep roll-off.
That being said, while it is true that (in most cases at least) all sample rates that are multiples of each other, 44.1-88.2-etc vs. 96-192-etc, have the same 'father', it is also true that they start to differ depending on the filter on the decimation stage that brings the digital signal to the final sample frequency required. That filter is at the father's sample rate, but it has (or should have) different cut off frequency points (half of the final sample rate required).
To filter at lower cut off frequency one needs more taps (because of less wide transition band), so more latency.
I guess if you need to do stuff on the fly like in the case of sound reinforcement every bit of latency saved counts. That's hardly a situation where the most accuracy of sound is the main goal, but fine, let's save that fraction of ms. Yay!

Where people get hung up is the mixing phase (as in studio mix, not live mixing console). Higher sample rate means more space for the harmonics created by the plug in to 'live' without getting in the aliasing zone. Ok... Oversample, then! You can do that just fine with the addition of 0s in between samples. The low pass filter in that stage (that I previously forgot to specifically mention) is the one that can be made easy on the CPU, because of the signal having 0s in between the original sample frequency values. After that, you have even more space for the harmonics to live in, and you will have generated less of them.

BUT... it does depend on the design of the plug in. If it's not done right, it might very well be that lower sample rates will sound like crap, after you put a few plug ins in series.

Same oversampling principles.
You start with low sample frequency, you oversample, you filter, you feed it to the DAC stage that can have a relaxed analog filter at the output stage.
Again, the filter can be made easy on the DACs CPU because it has samples known to be 0 in between 'real' samples, so it can be done faster than with a random signal at the oversampled rate.
Not that it matters, because when you are listening, what's the hurry that you can't wait 0.5 sec (which is a loooong time, practically overkill) to make sure the signal is filtered right? Unless you are in the sound reinforcement situation above, which is not one that warrants accuracy above all else.

So what's the point of going above 44.1 (or 48)? Only in the case one NEEDS low latency. In that case the digital filter can be made more 'relaxed', which means less taps, which means less latency.
I can't think of any situation where that need coexists with maximum transparency and purity of sound. And for sure none of the people that are saying that 92 kHz sounds better than 44.1 kHz have done their tests in such a situation.

To sum this up:
I'm not saying most consumer or even some pro products actually take the time to let the digital filters do things optimally.
I'm willing to bet there are DACs (and ADCs, and plug ins) that, for the sake of low latency, will do it less than egregiously.

But is it an INTRINSIC flaw of the lower sample rate? NO!
Could they take their f***ing time and do it right at the lower sample rate in most cases, and absolutely in EVERY SINGLE ONE of the cases that people posting on this thread have employed them? YES!

Originally Posted by IanBSC View Post
And yet nobody has definitively encountered this distortion in 20 years. I concede that it is possible that your analog chain could have problems if you have tons of ultrasonic junk, but where is the actual evidence that this has been a problem? I've done a number of projects at 192khz and I haven't run across it. The only real hazard I found was older converters with less precise clocks are a bit less defined at these rates, which seems to be a jitter issue not intermodulation.
Of course they have! Non linearities in DACs and amps (and don't even get me started on speakers) are measured all the time. Mostly they are bundled up in figures that are somewhat indirect, like THD and the likes, but non linearity is absolutely measured for audio components. I'm surprised you didn't know about that, to be honest.
You have definitely run into it in your 192 kHz projects. It could have been measured, although it's program specific and quite hard to do. Whether you might have actually heard it, or even preferred it, is a whole different subject.