anyone working at 96k?

[digitaldrummer]

My not-quite-20-year-old converters are Lavrys, you sure you wanna take that bet?

ok, yes, there are exceptions. I assumed that anyone that is asking this question ("anyone working at 96K?) does not have mastering grade converters or they'd not need to ask, am I right?

[I'm Painting Again]

https://lavryengineering.com/pdfs/lavry ... theory.pdf

sampling theory

here's the insight of one of the greatest converter designers

enjoy

[AndersonSoundRecording]

If it's solely for video, then I do 48k; but anything where the audio-only version is going to be important, it's 96k, so long as I can keep the track count at or below 32. For higher track counts than 32, my rig is still limited to 44k/48k

So have you had projects where the 96k and the 44k/48k are side by side, and do you perceive a difference?

I haven't been able to do any comparison side-by-side, so anything I say regarding subjective quality would be purely anecdotal. I do feel like the records I've done at 96k for the most part do sound "better", but there may be a lot of reasons for that having nothing to do with sample rate.

I will say that for operations where I need to do TC/E to squeeze an edit, I am never able to get away with even a little bit at 44k or 48k without artifacts; but at 96k I can sometimes get away with it with no loss of fidelity. We're talking classical music here - not time correcting drums or bass on a grid that are part of a busy mix.

[AndersonSoundRecording]

There is an OPTIMAL sample rate; fast enough to accommodate everything we hear (the audible range). But exceeding this optimal sample rate will only reduce audio accuracy.

I think what Dan is saying is to pick the sample rate that is just enough to capture the frequency range that you want to hear for the program material

anything faster or slower can be shown scientifically to reduce fidelity -"higher speeds/ more bandwidth reduces the accuracy of the conversion" details he says are in his white paper titled "sampling theory"

the more bandwidth you capture apparently the accuracy goes down

This is NOT the case. At extremely high sample rate PCM (think 300k on up), the electronics may have difficulty recovering in time and/or heating of the electronics & such come into play.

But I'm pretty sure even Lavry designs were "oversampling" i.e. they do Sigma Delta on the front end (2.8 Mhz/1 bit) and decimate that down to whatever sample rate you set it for. That's a pretty common design for decades now.

The real limitation has always been the anti-aliasing filter, not just at the conversion stage, but also in processing, rendering, &c. The steep slope requirements of an anti-aliasing filter at 44k or 48k will almost ALWAYS have some effects in the pass band. Working at higher sample rates relaxes the slope on the anti-aliasing filter and allows for processing and generation of harmonics above the Nyquist limit of the final sample rate (assuming 44.1 or 48k), thereby reducing aliasing & filtering artifacts all around.

Aliasing doesn't always manifest itself as the nasty artifacts we can "really hear" - it can manifest as a "cloudiness" (often apparent in cheaper converter designs), or even as colouration in the upper mids. It's not always apparent that it's aliasing that is causing it, but the effect of relaxing the filters or using oversampling processing can be seen on frequency analysers - especially if you like using saturation plug-ins or other processing that generates upper harmonics.

Unless you are really short of hard drive space or processing power, there's really no reason not to record at 88kHz or 96k. Personally, I think anything above 96k is overkill, unless you need to do some serious time stretching to the material being recorded.

[I'm Painting Again]

There is an OPTIMAL sample rate; fast enough to accommodate everything we hear (the audible range). But exceeding this optimal sample rate will only reduce audio accuracy.

I think what Dan is saying is to pick the sample rate that is just enough to capture the frequency range that you want to hear for the program material

anything faster or slower can be shown scientifically to reduce fidelity -"higher speeds/ more bandwidth reduces the accuracy of the conversion" details he says are in his white paper titled "sampling theory"

the more bandwidth you capture apparently the accuracy goes down

This is NOT the case. At extremely high sample rate PCM (think 300k on up), the electronics may have difficulty recovering in time and/or heating of the electronics & such come into play.

But I'm pretty sure even Lavry designs were "oversampling" i.e. they do Sigma Delta on the front end (2.8 Mhz/1 bit) and decimate that down to whatever sample rate you set it for. That's a pretty common design for decades now.

The real limitation has always been the anti-aliasing filter, not just at the conversion stage, but also in processing, rendering, &c. The steep slope requirements of an anti-aliasing filter at 44k or 48k will almost ALWAYS have some effects in the pass band. Working at higher sample rates relaxes the slope on the anti-aliasing filter and allows for processing and generation of harmonics above the Nyquist limit of the final sample rate (assuming 44.1 or 48k), thereby reducing aliasing & filtering artifacts all around.

Aliasing doesn't always manifest itself as the nasty artifacts we can "really hear" - it can manifest as a "cloudiness" (often apparent in cheaper converter designs), or even as colouration in the upper mids. It's not always apparent that it's aliasing that is causing it, but the effect of relaxing the filters or using oversampling processing can be seen on frequency analysers - especially if you like using saturation plug-ins or other processing that generates upper harmonics.

Unless you are really short of hard drive space or processing power, there's really no reason not to record at 88kHz or 96k. Personally, I think anything above 96k is overkill, unless you need to do some serious time stretching to the material being recorded.

I think I have heard Dan say that exact same thing that over 96 is not needed though the writings he did there *seem* to highlight the nyquist limit and bandwidth being the factor for best accuracy - so yea I dunno

i can't even tell the difference between any of them

[jimjazzdad]

There is definitely information up 'there' in the 20 KHz to 40 Khz zone (says he who can't measurably hear above 12 KHz...). I can hear a difference between the 24/96 file played from my DAW and the 16/24 render played back on the same system. Is it significantly better? Not really, but the difference is audible. High sample rates are useful if you're creating samples by dropping very high frequency sounds recorded at 96K+ down to the audible range through downsampling. But it is also important to note that D/A conversion higher than 20 KHz played through some older analogue equipment can cause oscillations in amplifiers that were not designed for such high frequencies. Anyway, storage is cheap these days so I always record at 24/96.

[Colorblind]

Your AD16 can do 16 channels at 96k. It uses SMUX over the 4 lightpipe outputs, 4 channels each. However, you need something that can receive 4 lightpipe inputs, for instance: a Ferrofish A32 Dante or such.

I actually just did this on a session a few weeks ago. Hard to believe these nearly 20-year-old converters still sound better than a lot of the ones currently available.

Ah, good to know! Thanks.