What Sample Rate Do You Work In?

[gabriel g.]

This is realy ruff stuff. James made it clear with an easy explanation.
So If you want the best audio (in the END!!!!your CD (44.1khz and 16bit) )
go with 44.1 khz. If you want it "better" because you can: go with 88.2khz.

Why 24bit and NOT 16bit: because of your SNR: Signal to noise ratio.
With 24 bit you have a bigger headroom. This means your music has more place for dynamics before the music clips or is "noisy".
the last effect you use, to do a audio-CD is dithering. This makes the downsampling from 24bit to 16bit sounding more natural.

Be carefull before throwing names like Nyquist in the round. This is very complex stuff. When you name Niquist you have to name Shannon.

So If you want to get REALY deep in this stuff check out this three points.

How do we get music/sounds...record. How do we get this shit digital...
1.So check out Fourier.
Fourier was a math-genius!!!
2.What to we hear? 20-20.000hz. So why we fu**king record at 44.1khz or even have the possibility to record at 192khz?????
So check out Shannon.
3. If you check out shannon then go for Nyquist.

So here are some Wiki links:

1.http://en.wikipedia.org/wiki/Fourier_transform

2.http://en.wikipedia.org/wiki/Information_theory

3.http://en.wikipedia.org/wiki/Nyquist-Shannon_sampling_theorem

so after understanding the main idea, check out

http://en.wikipedia.org/wiki/Sigma-delta

then check out

http://en.wikipedia.org/wiki/Huffman_tree

http://en.wikipedia.org/wiki/Dithering

http://en.wikipedia.org/wiki/PCM

cheers

and by the way

I record with 44.1khz and 24 bit.

I think it may be internal problems within cubase or xyz that any summing sound better or worth in 48khz or 44.1khz.
A lot of pro plugins always upsample the audio automaticly (waves for example) for more percession.
I cant hear higher frequencies thatn 20.000hz so I wont hear the difference between 44.1khz/2 or 48khz/2.

After reading about fourier, shannon and Nyquist you will understand the /2

cheers

 

[Metaltastic]

Just in case there was any doubt, btw, dithering only masks the quantization issues in going from 24-bit to 16-bit, it has no effect on the artifacts left by downsampling to 44.1

 

[John_C]

Ok, i'm gonna stop theorising (i've doing a 4 year physics degree starting in september, so there'll be plenty of time for that then)

here is a 5500hz test tone sampled at 44.1khz http://dl.getdropbox.com/u/910836/tone44.1.wav

and here is the same test tone converted from 44.1khz to 11.025khz http://dl.getdropbox.com/u/910836/tone11.025.wav



MY method was this (just so people can check i haven't made some stupid mistake)

1)shove a JS tone generating plugin into reaper (running at 44.1khz), tonegenerator as a matter of fact

2)record the output of the plugin

3)render the project at 44.1khz with no dithering or soundshaping or anything, but with the EXTREME quality sampling setting

4)render the project again at 11.025khz, again with the EXTREME setting


i chose to work with 11.025 because i thought it would be likely to easier to downsample from 44.1

 

[gabriel g.]

I dont want to judge anybody!!!dont get me wrong.
But I realy dont get your point with the 11.025???

You destroyed the harmonic content. We hear this because the human can hear up to 20khz. (myspace player )
But when you do this with something recorded in 88.2khz and then downsample it to 44.1khz I dont think I will hear the difference, because the harmonic content is out of my hearing.

With 44.1khz and after shannon and nyquist you have 22.05khz left, so there is all audible harmonic content in the media.

Do you get my point?

cheers

 

[John_C]

I dont want to judge anybody!!!dont get me wrong.
But I realy dont get your point with the 11.025???

You destroyed the harmonic content. We hear this because the human can hear up to 20khz. (myspace player )
But when you do this with something recorded in 88.2khz and then downsample it to 44.1khz I dont think I will hear the difference, because the harmonic content is out of my hearing.

With 44.1khz and after shannon and nyquist you have 22.05khz left, so there is all audible harmonic content in the media.

Do you get my point?

cheers

What i was trying to find out is whether downsampling affected the audible quality of sounds. I used 11.025khz as its 1/4 of 44.1khz, just so as to simplify things

BTW the test tone is a sine wave, so there is no extra harmonic content (or at least there shouldn't be)

 

[Force666]

I just figure if you can't get stunning results at 44.1, then jumping up the sample rate isn't going to help you at all.

 

[Fox Mulder]

I plan to get i7 + 15000rpm hdd pretty soon
And would be working with 96khz from then.

 

[gabriel g.]

In theroy there shouldnt be Yeah thats right. The sine-wave is the simplest, periodic, process. But only in theorie.
You know the fourier transformation reconstruct your input with sine-waves, so MAYBE that is the point?!?!
The sequenzer maybe thought after pressing record "dude, thats a full tone" not just a sine-wave???

do you know what I mean.

Also what would be the other explanation???

It could be the nyquist frequency snapping up the 5.5khz because 10.25khz/2
to execute the Shannon theorie!!!

So make a mixdown with 22.05khz.
If it is a true sine-wave the pure tone at 5.5khz shouldnt sound different than in the 44.1khz solution?!?!

cheers,
getting very insteresting!!!!

 

[John_C]

In theroy there shouldnt be Yeah thats right. The sine-wave is the simplest, periodic, process. But only in theorie.
You know the fourier transformation reconstruct your input with sine-waves, so MAYBE that is the point?!?!
The sequenzer maybe thought after pressing record "dude, thats a full tone" not just a sine-wave???

do you know what I mean.

Also what would be the other explanation???

It could be the nyquist frequency snapping up the 5.5khz because 10.25khz/2
to execute the Shannon theorie!!!

So make a mixdown with 22.05khz.
If it is a true sine-wave the pure tone at 5.5khz shouldnt sound different than in the 44.1khz solution?!?!

cheers,
getting very insteresting!!!!

Mmhmm yeah i'll try that!

the thing is, DACs don't use fourier transformations as far as i know:

"Practical digital-to-analog converters produce neither scaled and delayed sinc functions nor ideal impulses (that if ideally low-pass filtered would yield the original signal), but a sequence of scaled and delayed rectangular pulses."

I've got too many exams coming up to be getting too heavily into this, but maybe in a few weeks i'll try and understand better

for now i'm gonna produce a set of test tones at different frequencies and sample rates with corresponding zoomed in waveforms so we can see if we can get a practical answer to this question

 

[Aaron Smith]

I just figure if you can't get stunning results at 44.1, then jumping up the sample rate isn't going to help you at all.

Sounds fair to me!

 

[KeithTidd]

I just figure if you can't get stunning results at 44.1, then jumping up the sample rate isn't going to help you at all.

true, but what if your results are noticably better at a higher sample rate?

I use 88.2 because it pushes the nyquist filter to a higher frequency than with 44.1, resulting in less passband artifacts. Bob katz explains is better. In other words, the higher sample rate makes up for the flaws in the converters.

 

[John_C]

true, but what if your results are noticably better at a higher sample rate?

I use 88.2 because it pushes the nyquist filter to a higher frequency than with 44.1, resulting in less passband artifacts. Bob katz explains is better. In other words, the higher sample rate makes up for the flaws in the converters.

Yup, in a sort of roundabout way i covered that earlier. Your converters reconstitute the digital signal into analog waves, and the more detailed the digital stuff is before you hit the converters, the better the results (the less the approximations are apparent)

 

[ahjteam]

question: why manufacturers produce sound cards to 192kHz ? who is using these samplerates rattle-mouse ?

The classical guys, but I think that is more of a "preparing for the future" kinda thing

 

[Force666]

true, but what if your results are noticably better at a higher sample rate?

I use 88.2 because it pushes the nyquist filter to a higher frequency than with 44.1, resulting in less passband artifacts. Bob katz explains is better. In other words, the higher sample rate makes up for the flaws in the converters.

True, but do you really think that your (insert prosumer interface here) converters are actually high quality enough to produce that 88.2 vs. 44.1. without errors? Not to mention the increase in disk space and read/write errors.

I agree with you that in theory it produces better results, but in practicality and in the real world it doesn't, especially when in the end your going to downsample and even more errors can occur.

Aside from that, IF (I emphasize "IF") someone like Andy's mixes at 44.1 are blowing your mixes at 88.2 away, then what's the point? It's still not making a discernable difference in the real world.

 

[Metaltastic]

I guess my only response to this thread is - if you can record at 88.2 (or 176.4), than why not?

 

[John_C]

Having taken this down to it's simplest form and worked with sine waves at particular frequencies............i conclude that there is no audible difference between sample rates when working below 1/2 the sample rate. I think. Unless i've missed something

However, you do lose A LOT of volume at frequencies near the limit when you downsample. I think Joey uses a high shelf to compensate???

i wonder if there's some maths somewhere to give us a graph of exactly how much you lose

 

[John_C]

having listened harder, i could start to hear a sort of graininess at lower sample rates, and it sure as hell looks foul. You can also see the drop in volume, although it was quite small in this example

 

[KeithTidd]

Yup, in a sort of roundabout way i covered that earlier. Your converters reconstitute the digital signal into analog waves, and the more detailed the digital stuff is before you hit the converters, the better the results (the less the approximations are apparent)

all i can say is TLDR.

True, but do you really think that your (insert prosumer interface here) converters are actually high quality enough to produce that 88.2 vs. 44.1. without errors? Not to mention the increase in disk space and read/write errors.

I agree with you that in theory it produces better results, but in practicality and in the real world it doesn't, especially when in the end your going to downsample and even more errors can occur.

Aside from that, IF (I emphasize "IF") someone like Andy's mixes at 44.1 are blowing your mixes at 88.2 away, then what's the point? It's still not making a discernable difference in the real world.

I use digi 192's regularly and do this, it definately makes a difference - is that difference the difference between a good album and a great one? no, but it does sound slightly better. My home interface is 44.1/48 max, so i mix in 44.1 at home.

 

[ahjteam]

Using pictures and theoretical text is nice, but I made an audiovisual experience for you:

 

[Force666]

Keith, I understand that. And still agree in theory.

My question is do you still hear that difference when you downsample it back down to 44.1? Maybe not in the future, but CD's are still the preferred medium for now.

Also it begs the question, are you hearing it "better" because you want to hear it better. Only a double blind test could tell for certain.


Anyways, we all know this could go on forever, back and forth. I'll leave it at that and say we should let these arguments be hashed out by the fags at gearslutz.