88000 sample rate is it good for mixing

[jones-y]

Sorry, but the auditioning equipment has nothing to do with the internal rendering quality of audio effects and instruments.

Errr... that's what I said ?!?!? Thanks for reaffirming though. It does, however affect your monitoring. In other words, what you're hearing...

And a pure digital production including mastering on a final digital media has also nothing to do with a soundcard (as long not used for recordings)...

Errr.... See above? Do you hear digits?

Bottom line is running your soundcard (the device through which all you audio runs, regardless of origin) at a higher sample rate many times makes it sound better.

[MrM]

I have been told that 88000 Hz is nice, nicer then 44100 Hz. But 96000 Hz is da bomb.

I have also been told that most soundcards behave best at the highest sample rate they can handle, cause it is designed for that sample rate, any lower is not getting the most out of your soundcard.
Besides that, recording, or importing and upsampling to, 96 kHz will be good if you plan to do any EQ or modulation, filtering and what not to the file.
The difference is said to be quite significant.

The people who told me this were the kind and friendly posters on the audioforums dot com.

[jones-y]

96 kHz will be good if you plan to do any EQ or modulation, filtering and what not to the file.
The difference is said to be quite significant.

It does sound reasonable, in an anecdotal sort of way, but the reality is that there is absolutely no quantifiable proof of this.

What the textbooks (and the DSP pros) say is that as far as the audio spectrum goes, 88.2k and 96k hold exactly the same data as 44.1k. Therefore lowpass filtering a 96k signal is exactly the same as LPF'ing a 44.1k signal. Now what's up for contention is those frequencies we can't hear, how do they affect our perception of the sound? And there is no definitive answer to that question, the pros debate this until they're blue in the face.

[mauseoleum]

there was a long thread about aliasing. in theory, you could avoid that to some extent when *forcing* the sequencer to one of higher rates while exporting, that could be called *manually oversampling* or something like that - basically allowing the sum to be limited by a higher nyquist fq. whether you use *recorded* audio, that's up to you to decide - more data and more bit depth should be more hi-fi ...

[useruseruser]

Sorry, but the auditioning equipment has nothing to do with the internal rendering quality of audio effects and instruments.

Errr... that's what I said ?!?!? Thanks for reaffirming though. It does, however affect your monitoring. In other words, what you're hearing...

And a pure digital production including mastering on a final digital media has also nothing to do with a soundcard (as long not used for recordings)...

Errr.... See above? Do you hear digits?

Bottom line is running your soundcard (the device through which all you audio runs, regardless of origin) at a higher sample rate many times makes it sound better.

Fine. That all meight be right.
But that also was not the discussion topic here.

[useruseruser]

96 kHz will be good if you plan to do any EQ or modulation, filtering and what not to the file.
The difference is said to be quite significant.

It does sound reasonable, in an anecdotal sort of way, but the reality is that there is absolutely no quantifiable proof of this.

What the textbooks (and the DSP pros) say is that as far as the audio spectrum goes, 88.2k and 96k hold exactly the same data as 44.1k. Therefore lowpass filtering a 96k signal is exactly the same as LPF'ing a 44.1k signal. Now what's up for contention is those frequencies we can't hear, how do they affect our perception of the sound? And there is no definitive answer to that question, the pros debate this until they're blue in the face.

Hmm. Thats typical.
Either they have *merely* "pure theoretical experiences" or anything is wrong with their ears...

You actually can hear it (and must not be a pro)! And most plugin filters sound *definitively* different at different sample rates (also EQs and Reverbs).

And me as a programmer say you, that is *is* a difference in quality, if you render an oscillator or a filter or any other effect with higher sampling rate. Definitively.

You "hear" on 88.200 Hz twice as much samples in the same amount of time. The sound is therefore much smoother, deeper and clearer (try it yourself).

This improvement has nothing to do with the spectral frequency content of the audio material, but rather with higher resolution/desnsity of the digitized audio information.

With other words 44100 Hz is just significant more "digital rastered" than 88200 or 96000 Hz (but may have nevertheles exactly the same spectral content), which is finally closer to the "analog world" and so definitively audible as higher quality.

If the 44.100 mark would be the final "state of the art" (since more than 10 years now), so why to f**k all manufacturers tend to increase the sampling frequencies of their products (hard and soft) all the time? Just for fun?

[peejunk]

What the textbooks (and the DSP pros) say is that as far as the audio spectrum goes, 88.2k and 96k hold exactly the same data as 44.1k. Therefore lowpass filtering a 96k signal is exactly the same as LPF'ing a 44.1k signal.

The first sentence OK, but my DSP textbooks tend to disagree with the second one. The spectrum data of the signal itself is (theorethically) the same, but digital filters have different behaviour on different sample rates.

They're, among other things, more stabile and precise as cutoff is futher from 1/2 samplerate (i.e. Nyquist) frequency, and their "distance" from nyquist will, given the same cutoff point, differ for two different samplerates.*

I don't have the book here and my practical knowledge in DSP is very limited. So I'm not sure how general is this to different types of digital filters.

But I'm sure some of DSP big dogs on the forum will set the record straight pretty soon.

* There is a mathematical proof of this in Miodrag Popovic's "Digital Signal Processing" (not available in English tho) textbook. But I'm sure there is a math proof of this in other textbooks aswell.

[cptgone]

But that also was not the discussion topic here.

yes it is: mjones4th was only reacting to someone's statement that music sounded better at higher sample rates.

[DevonB]

I tend to mix in 44 but a higher sample rate will give you more digital headroom.

No, a higher sample DEPTH will give more headroom, not sample rate. Rate only records how many times a second it plots where your waveform is. Depth (16/20/24) is how much dynamic range you have 96dB vs 120dB vs 144dB, or 6 dB for every bit that's added.

Devon

[MrM]

Suppose you describe a pure 1 kHz sine (-10 dB FS).
The duration of this sine is not relevant, but let's say it is exactly 1 s (or 1000 cycles).

Now how many samples dsecribe this signal?

@44.1 ; 44100 per second makes 44.100 samples
@48 ; 48.000 samples
@96 ; 96.000 samples

Now which samplerate will describe the given signal best?

Which samplerate will most likely be best to add subtle changes with the least amount of unwanted rounding (take a sampleposition in the frequency domain that lies not on the calculated ponit, so a nearest point has to be chosen.)

Remember, in theory one could describe the 1 kHz sine with a samplerate of 2 kHz, which gives 2000 samples to describe 1000 cycles.
But what if all samplepoints are taken at the exact moment the cycle has the zero amplitude value? You'll end up with samples that all result in the - inf dB. So in general people take at least 2.2 * f(high) makes f(sample) = 2.2 * 1 kHz = 2.2 kHz.
Have you ever seen a plot of how the output of this sampling looks? Horrible!
The more samples taken, the more precise a signal is described, and the less it is likely to end up with unwanted distortion during applied FX.

I hope this wasn't too technical

[MrM]

No, a higher sample DEPTH will give more headroom,
...[cut by MrM]...
Rate only records how many times a second it plots where your waveform is.

Devon

Devon the above quote as quoted by me (note the cut), tells that more bits describe the amplitude of a wave better. With which I absolutely agree!

The same goes for the sample frequency; it describes the wave better, but not the amplitude of it, but the frequency (or time domain).

[MrM]

I have to make a short sidenote;

Most recording is done at 16 bit or 24 bit.
32 bit is only for internal processing; edits done in software. Cause recording with 32 bits adds no value to the 24 bit recording.

The same goes for recording at very high sample rates; when it comes down to editing, that is where higher sample rates matter. So there is a 'end' or treshold frequency. Recording above that frequency will not give much difference.
To take it to an extreme; recording at 1 Ghz will not be that much better audible then recording at 456 kHz. But it will eat up a lot of disk space
That is overkill !!! (But salespeople will tell you recording at 1Ghz is da bomb*)

* totally false of course, but sales people want you to believe them; they sell you a lie, and you buy it...

[useruseruser]

What the textbooks (and the DSP pros) say is that as far as the audio spectrum goes, 88.2k and 96k hold exactly the same data as 44.1k. Therefore lowpass filtering a 96k signal is exactly the same as LPF'ing a 44.1k signal.

The first sentence OK, but my DSP textbooks tend to disagree with the second one. The spectrum data of the signal itself is (theorethically) the same, but digital filters have different behaviour on different sample rates.

They're, among other things, more stabile and precise as cutoff is futher from 1/2 samplerate (i.e. Nyquist) frequency, and their "distance" from nyquist will, given the same cutoff point, differ for two different samplerates.*

I don't have the book here and my practical knowledge in DSP is very limited. So I'm not sure how general is this to different types of digital filters.

But I'm sure some of DSP big dogs on the forum will set the record straight pretty soon.

* There is a mathematical proof of this in Miodrag Popovic's "Digital Signal Processing" (not available in English tho) textbook. But I'm sure there is a math proof of this in other textbooks aswell.

Yes. this topic can be discussed theoretical and mathematical to the point, where it flows out of our asses...

But that's also not necessary. Because everyone should be able to *hear* the difference. So a theoretical prove is completely superfluous (wathever, I think it's very obvious without being a DSP guru, that higher sampling frequencies actually cause higher audio quality).

Practical example: Try to test an FFT based Noise Reduction algorithm at 44.100 Hz and then at 96 kHz. Then you'll see very soon, what the advantage of higher sampling frequency is.
Or simply try most of the available synthesizers out there.

Digital == signal quantization (as a function of samples per time frame). Analog == natural continuosity (no signal quantization).

Our ears and real world audio waves are analog.

So higher sampling rates are logically closer to the analog world than lower ones.

Thats all, what is necessary to know.

[cptgone]

The same goes for the sample frequency; it describes the wave better, but not the amplitude of it, but the frequency (or time domain).

isn't it so that only the range of frequencies present in the recording is determined by sample frequency? so that 44kHz covers what we hear (but not what we feel when clubbing)?

Please correct me if I'm wrong.

[MrM]

See http://www-camil.music.uiuc.edu/classes ... cepts.html

The last pic on the page (sample rate vs sample width) says it all.

Figure 3 on this page: http://arts.ucsc.edu/EMS/Music/tech_bac ... es_16.html also shows the influence.