Sample Rate: 44.1 kHz, 48 kHz, or higher?

[noisetonepause]

If you've got the hardware and you really care, there's no reason to not go up...

Sample rate conversion is pretty decent these days...

[Tarekith]

I use 24/44.1, final mixdown is rendered at 24/96 though. If you have the storage space, not much harm in going with 88.2, though I personally don't hear the difference for the kind of music I do. If we were recording string quartets in beautiful churches, I think it would matter more.

[v00d00ppl]

I use 24/44.1, final mixdown is rendered at 24/96 though. If you have the storage space, not much harm in going with 88.2, though I personally don't hear the difference for the kind of music I do. If we were recording string quartets in beautiful churches, I think it would matter more.

ye, if you can't hear the difference....what difference does it make.........the only real difference is when working with samples in a 24 bit environment versus 16 bit................the 24 bit sound is noticable even for me...........when i sample vinyl on 16bit samplers its ok, but if its mroe than just drums i am chopping up and its something like a bass or string that i want to tweak.....the 24 bit is with me.

[Michael-SW]

Nyquist theory is a mathematical fact....

Yeah, Nyquist theory is a mathematical fact. But for it to work (stated in the Nyquist theorem) you have to surgically remove all frequencies higher than half the sample rate before digitizing. Ie, for your A/D converter to perfectly capture frequencies up to 20.5 kHz you have to have an infitiely steep mathematically perfect filter at 20.5 Hz. In the real world, you will have an imperfect filter with a slope. That will mess up your A/D process and can introduce artefacts and/or reduce the real frequency range down a couple of kHz.

A little knowledge is a dangerous thing...

(Now, I don't think this matters for most real world applications. 44.1/24 bits works fine for me. But people that state bullshit like it were an absolute fact pisses me off)

[dancerchris]

Yes it is. That is why on a lot of audio forums there are a bunch of ignorant individuals spouting falicies about how the signal gets distorted because they don't understand sinx function summing and the mathematical decomposition of bandwidth limited specta analysis. Ergo the statement about Nyquist theorem. I could be wrong but I believe the original theorem did not include aliasing and was part of later work by another person.

The filtering your refering to is the same brickwall filtering that I've mentioned in my previous post. Steep rolloff filters introduce signal fluctuations in the high frequency range prior to the cutoff frequency. Hardware designers use different techniques ranging from analog brickwall filters (that introduce fluctuations in the high frequency data near the cutoff value) to internal oversampling with simple filtering. If someone chooses to use 96k to eliminate high frequency artifacts due to antialiasing technique (as mentioned in a previous post), they may have inadvertently used a redundant technique.

Short of writing a dissertation on the subject I think anything past this is beating a dead horse, but then again as you say "A little knowledge is .......

[Tone Deft]

Nyquist simply states that to sample a signal at x Hertz you have to sample at more than 2*x Hertz, it's nothing more than that.

The rest of this dick dancing debate is pointless. Filter design depends on the software/hardware you're using and doesn't really matter in this thread. Put your calculators away, go smoke a blunt.

I like discussing this stuff but you guys are already jumping around like a bunch of apes beating your chests.

[Dear Ayça]

dancerchris is 100% spot on.

Worth noting though that the cinema industry tends to use 48 or 96 Khz.

I have read somewhere that it is a "better" conversion from 88.2 to 48 than it is from 96 to 44.1, so if you are making something that could be used for both purposes 88.2 khz is the way forwards.

Makes sense:
88.2 / 48 = 1.8375
96 / 44.1 = 2.1768707483