sampling rate 44.1 or 48 kHz

Version 4.0 had defualt 48 kHz sampling rate.
Now it is back to 44.1 kHz. ???

Which is best 16bit/44.1kHz or 16bit/48kHz ?
(Then burn to CD.)

I am not a beliver in 24 bit recording.
(It is a well known fact that 12bits is all it takes.)
But 44.1 kHz maybe is just below optimum.
At least CD-players may have a hard time
converting to analog in a “nice way”.

But here of course We have to convert from
48 to 44.1 before burning to CD.

On the other hand soundcards works at 24/96
so maybe here its better to convert fom 96 to 48.

/Goran Sweden ???

What soundcard? Important to know, since some ones have troubles working on another sample rate than 48Khz.

About bit depth, i recommend you to record and mix at 24 bit, if you have a machine fast enough, is more important than a high sample rate. Make some search in the forum, there are several interesting discussions about that.

My soundcard is

12 bits is enough for the final sound !

/Goran Sweden

'suppose 12 is enough, but:

If the CD is in 16, why would you want to ‘upsample’ anyway ?

And (if my understanding is correct) 12 should be enough if you have an absolutely perfect gainstaging accross board and everything runs at peak all the time.
24 bit gives you the headroom to breath a little in the dynamic range without losing quality.

If you don’t think that 24 bit is better for you then run 16bit at 44 or 48. I would do 48 and then after the final mixdown use something like r8brain to dither to 44.1 onto cd.

But then I also believe that recording at 24bit works better for me, so take it as you see it…


Here are some interesting previous threads to read, if you want, about advantages using 24bit, even if you go down to cd quality:
Discussion 1
Discussion 2

I do not think I could get lower noise levels with 24 bits.

When I use my mics (ADK Hamburg and Rode NT2-A on my acoustic in M-S style)
I can’t get the guitar - noise(room+mics) below 60dB (can You ?). That is 10 bits.

Now if I keep recording levels(attacks) at say -9db (between -12dB & -6dB)
I will have at least 14 bits. That is 2 bits more than 12 and 4 bits more than the signal/noise.

So I do not need more.

I think You could even lover the recording levels to -18db (or so) without any real loss of anything.

/Goran Sweden

It’s not just about the noise floor. More important for processing after recording. More elbow room, less rounding/truncation noise.

But this has been discussed before.

Your “well known fact” is neither well known nor a fact. I’ve worked with 12-bit digital and know quite well the issues and limitations. There is no doubt a truth underlying what you’re saying, but I think you’re overgeneralizing it.

You can make great recordings using 16 bits. You can more easlily make great recordings using 24 bits, leaving lots more headroom. But even with my hearing loss, the tails (fadeouts) on 12-bit recordings have obvious and annoying quantization noise. Note that these tails have less than 12 significant bits. If there is no dynamics in a recording, 12 bits might be sufficient. But I don’t care for music without dynamics.

For your soundcard, you should record in 44.1 to avoid resampling at the end. Some soundcards don’t work well in 44.1k, and for that reason, should be run at 48k and resampled to 44.1 at the end.

However, if you use plugins that work using variable delay lines (e.g., most chorus & flanger plugins), you’ll get noticeably better results in 96kHz. Now, whether the improvement is worth the extra disk space and processor usage – well, that’s a judgement call. With all the argument and fuss about high sample rates, I’ve never seen anyone address this particular issue but it’s quite obvious both theoretically and practically. I suspect it’s at least partially responsible for the big difference between the theorists’ usual argments and the results-driven assessments of engineers.

Goran, you’re ignoring the enormous qualitative difference between uncorrelated (e.g., white) noise and quantization noise. Our ears and brains are fine tuned to mask out white noise. On the contrary, we hear QN noise at much lower levels than we hear white noise. QN noise sounds musically terrible, whereas white noise doesn’t.

Furthermore, when the S/N ratio equals 1, this does not mean there’s no signal. Quite the contrary, our ears and brains can recognize and even appreciate signals that are buried way under the noise level, when it’s white noise (or nearly white noise). This is something you can test for yourself. IIRC, we can recognize music at S/N ratios of 1/8 or lower. However, when the noise in question is quantization noise due to truncation or rounding, the signal is lost entirely when it’s below the noise level. When we use dithering instead, the signal is still detectable but lost much more quickly (MUCH more quickly) than a fully reproduced signal with added white noise. Or even “added Q noise”, though I should figure out exactly what I mean by that before saying it!

So, your approach to thinking about this in terms of significant bits is useful, but there are a number of very important facts that you have to add in order for your results to be really meaningful. This is a useful thing to think through, though, and can be rather illuminating. In particular, this approach can teach us alot about dithering, in a practical (listenable) way. If you’re intersted in seeing the things I’m saying demonstrated with actual wave files where we can actually hear what’s going on, start a new thread with your argument above and we can discuss it.

PS: I agree that you don’t get significantly lower noise levels using 24 bits, at least when recording with a mike. The differences are reservable headroom and the nature of the noise. Even then, it’s a very subtle difference, and not one that is demonstrable in an A/B comparison for a single track. 16 bits is really quite good!

Finally, I started out using gear that had a 65 dB S/N ratio, where the noise was white (tape). Believe me, each time the S/N ratio increased by 10 dB, it amounted to BIG differences in the quality of mixes, even disregarding additional record/playback stages (bouncing, mastering, etc). Even though (a) it was “nice” white noise – believe me it sounds better than quantizatoin noise!) and (b) when summing two tracks, uncorrelated noise increases by only 3dB.


Thank You Jeff

I take your advice.
I will record in 16/44.1

/Goran Sweden