Sample Rate Poll

Processaurus

@thenonanonymous said:

@Processaurus said:
Latency is a product of sample rate, AND processing power

Isn’t this a little misleading? Processing power is a factor for sure (for tolerable audio), but if you have a faster device, the latency rate at 44.1kHz/16bit/256frames is the same as a slow device at 44.1kHz/16bit/256frames is it not? Correct me if I’m wrong.

You’re right. I was thinking achieving a minimum latency, without crackles, is dependent on sample rate, and processing power. Just was trying to dissuade people new to digital audio from messing with their sample rate settings, trying to get lower latency, when they want to be adjusting the latency/buffer size setting.

Unless, they are rocking a high sample rate, like 96khz, or 192khz, and getting crackles, the sane thing to do would be to take the sample rate down to 44.1, or 48khz.

CPU is generally the bottleneck, that causes crackles.

Telefunky

@Max23 said:

the jump from 44,1 to 48khz is audible sometimes,
from 48khz to 96khz I can't tell the difference except I create some really extrem example scenarios ...

you may hear a difference between 44.1 and 48khz depending on the native clockrate of the digital system (as there can only be one), so samples with a non-matching rate are converted on the fly.
This is quite a demanding process, which may introduce artifacts.
The aliasing between 44.1 and 48 khz is slightly different, which may be more pleasant - but 48k aliasing still falls into the audible range - which never can happen with 96k.
Aliasing isn't all that bad, a lot of dgital synths (the original Waldorfs and drum machines like the EMU SP1200) have it as their core feature

As mentioned the data width of 16/24 bit is a noise-floor issue in the first place.
A microphone delivering 85dB SNR is a very good one and falls into the theoretical range that 16bit would cover. But as the least significant bits are uncertain by nature, 18-20 bits are more appropriate, which was pro standard for quite some time.
Today it's 24bit for simplicity (and the better selling higher figure). Few gear can digitize reliably beyond bit 20 and you need a hell of an analog chain to provide basic conditions for even 20bits.

This is a completely different thing from internal processing, it affects only the input.
Digitally synthesized sounds are practically noise-free (unless you add noise as a source).
Your Model D's output will sound exactly the same if saved as a16bit or 24bit file.

Using 24bit files today is a matter of convenience, as some audio systems seem to have trouble to handle mixed file formats consistently.

thenonanonymous

this Apple discussion provides a very strong argument for almost always using 48kHz (states there’s a few rare exceptions when using higher sample rates is beneficial):

https://discussions.apple.com/thread/8061449

Tarekith

Agreed, here's more good info on why higher sample rates are usually overkill with no benefit, and in fact my actually sound worse:

https://people.xiph.org/~xiphmont/demo/neil-young.html

[Deleted User]

The user and all related content has been deleted.

CracklePot

Dang, you all are some smart MF-ers!