What Sample Rate Do You Work In?

from the wikipedia article on the nyquist-shannon sampling theorem

"The conclusion that perfect reconstruction is possible is mathematically correct for the model but only an approximation for actual signals and actual sampling techniques."

that article goes on to explain that it is only possible to reconstruct the original waveform in theory

Real DACs use an approximation to the reconstruction formula, and hence it is handwavingly shown that sample rates above 44.1khz DO have a theoretical advantage within the bounds of current audio equipment and human hearing
 
96k recording

44.1 mixing


Why? Because I can. lolz

No srsly. I record at 96 because the amount of samples taken a second gives a clearer representation of the waveform as it was in the analog domain. Then I just edit in 96 and convert to 44.1 (so the SRC to 44.1 has more samples to chose from)

+1

this is what i try to explain to everyone, no one listens to me

i also dont give a shit if __insert a big shot's name here__ is doing 44 / 24.
 
Still seems like a better idea to work at 88.2 or 176.4 though, since they're direct multiples of 44.1 and thus the downsampling is cleaner...
 
96k recording

44.1 mixing


Why? Because I can. lolz

No srsly. I record at 96 because the amount of samples taken a second gives a clearer representation of the waveform as it was in the analog domain. Then I just edit in 96 and convert to 44.1 (so the SRC to 44.1 has more samples to chose from)

why don't you mix in 96 and dither to 44/16 in master bus ?
 
From a combination of my own research and torniojaws' explanations:

Your Digital to Analog converters take the digital signal which is only an approximation of what it would be in the analog domain, and try to piece it back together accurately.

They are quite good at it, but if you work at 44.1khz, they are getting to the limits of what they can do when they get towards the highest frequencies of human hearing

However if you worked at say 192khz all the weay through a project (the maximum on the profire 2626) then the converters will be getting lloooaaaadddsss of samples to work with even up to 20khz

This means they will be able to produce a truly accurate representation of the high frequency content.




However, currently most releases don't get the privilege of being printed to SACD or ADVD and as such end up at 16 BIT 44.1khz anyway. It is still worth working at higher sample rates though, as your plugins and outboard gear will be reaping the benefit (not to mention summing etc..)

I say: IF YOU CAN WORK AT HIGHER SAMPLE RATES..........................THEN DO!


EDIT: In purely mathematical terms, 44.1khz is enough for everything human, unfortunately until we get a computer that can sum to infinity (ie never) the maths is actually not possible to impliment
 
by audio resolution im talking about the amount of analog signal being sampled per second

its increased the higher the sample rate

this means more accurate representation of the analog signal, because analog is "infinite" as far as digital representation is concerned.

thanks for the explanation!:kickass:
 
I say: IF YOU CAN WORK AT HIGHER SAMPLE RATES..........................THEN DO!

But also remember that when you work in 44,1k, you don't have to downsample - what you hear is what you get. But when you work at a different sample rate, you have to downsample and there will be aliasing/quantizing errors due to approximations. Those errors may or may not be heard.
 
But also remember that when you work in 44,1k, you don't have to dither - what you hear is what you get. But when you work at a different sample rate, you have to down-sample and there will be aliasing/quantizing errors due to approximations. Those errors may or may not be heard.
dithering is a process for maintaining apparent resolution while bringing the bit resolution down from 24 to 16 bits... and you still have to do that, irrespective of sample rate... as opposed to simply truncating.

but yeah, down-sampling from 48 to 44.1 requires some funky math that potentially causes more issues with your audio than the theoretical gains from working in 48. the only way to avoid that is to print your mixes to an analog medium prior to mastering.
 
In my simplistic way of looking at things, the 3 samples i mentioned before if you join the dots you get a triangle wave. Which is a pretty shocking approximation of a sine wave

About this, yes, in a way it is a triangle (or square) wave, but included in the D\A conversion is a lowpass filter (after the conversion) at half the sample rate. A 22 kHz triangle wave consists of a 22 kHz sine plus some higher components, so when you low-pass at 22 you get back only the sine. (This is what the Nyquist theorem is about, you can reproduce all the components up to half the sample rate, but lose the rest.)

The Nyquist theorem unfortunately assumes you have unlimited precision for the samples, but in practice 24-bit should be enough.
 
About this, yes, in a way it is a triangle (or square) wave, but included in the DA conversion is a lowpass filter (after the conversion) at half the sample rate. A 22 kHz triangle wave consists of a 22 kHz sine plus some higher components, so when you low-pass at 22 you get back only the sine. (This is what the Nyquist theorem is about, you can reproduce all the components up to half the sample rate, but lose the rest.)

The Nyquist theorem unfortunately assumes you have unlimited precision for the samples, but in practice 24-bit should be enough.

Interesting......

This all gets rather complicated, doesn't it.....
 
i think too, the more information, the better it is, the plugins have more numbers to work from, and also in the mixing stage for summing, and this can postivle effect the result.

but on the other hand, the file sizes increase drastically too,

so when you have alot the tracks the workability with much bigger files need more power from your computer / harddisc

i have some project files, that contains like 5GB of audio files (44,1/24) its still ok to work with, but sometimes jumping from one point to another forces a little delay, until this part is loaded into the memory..
i dont want to imagine how it would be if i would use a much higher settings ;)

cheers
chris