24 bit.....where to begin??

[headquest]

So, in practice does this mean that I can pump the levels up in Tracktion and have no fear of clipping at the master output?

I would avoid doing that Steve. (partly because I don't totally trust the level meters in T1, but they are going to be improved in T2 )

Otherwise I do the same as you propose - i.e. record into Tracktion at 24/44.1, render from Tracktion at 32-bit (which ties in with Tracktions internal floating bit-point), open up in Audition (still at 32-bit) and dither down once the mastering is completed in Audition.

I find this works well here.

Good luck with the new sound card - hope all goes smoothly!

[valley]

Just found this interesting quote:

"The PCM format provides its optimal resolution when signal levels are at their very highest. As signal levels decrease to lower levels, resolution deteriorates, leaving quiet cymbals and string instruments sounding typically sterile, dry, harsh, and lifeless. The more bits you have available to you in the process of quantizing the amplitude of a waveform at any given sampling, the more accurately a lower level signal can be represented." (24 bit recording FAQ)

That's basically it.

The following is a very dry, and not very well thought out explanation of sampling theory:

sample rate, or frequency:

If you've ever used one of those webcams that has a frame rate lower than 10 frames per second, you are probably aware that the period in between frames can miss a lot of action. Kind of like moving in a strobe light - it almost appears that people travel from a-c without ever passing c. This is basically what sampling rate describes. Sampling basically means taking measurement of the amplitude of an audio source at discrete time intervals. How frequently the amplitude is measured is called the sample rate.

Because things can happen in between each sample, if you want to record a certain frequency, you need to be sure that you are taking measurements at a slightly higher frequency. COnsider a strobe light - if you move your hands in a circle fast enough, they appear to be staying still (yeah, I went to too many raves...). Now because an audio waveform is made of both a positive and a negative cycle, you need *two* measurements to capture it. This is why the sampling rate needs to be at least double the maximum desired frequency.


sample resolution, or bit depth

that's the frequency taken care of, so what effect does bit depth have? Basically, the bit depth is a the resolution with which teh amplitude is measured. Kid of like a tape-measure. If it only shows inches, you aren't going to get as good a measurement of a length as you would if the tape-measure also showed 16ths of an inch.

What is interesting about sampling though, is that in many ways it is like having an elastic tape measure. You stretch the tape measure to the desired length, then count the ticks. Rather than having a fixed length measuring device.

[headquest]

Nice one Valley

[IIRs]

The following is a very dry, and not very well thought out explanation

You got that bit wrong, but otherwise a great post!

There are two things that make high bit-depth settings more critical than high sample-rates:

Headroom and Compression.

In a real world recording situation it is not possible to predict accurately the loudest peak any performer will produce in advance, so it is essential to leave some headroom as a safety measure.

6db of headroom means only 15 bits are used to represent the audio, so if this is normalised to full scale the bottom bit will contain no useful information.

Compression is used in almost all modern productions, and will make matters even worse as reducing the dynamic range will make the grainy low resolution quiet bits even louder and more obvious.

If you record at 24 bit, you can leave 6db of headroom, compress the signal by another 6db, and still have 22 bits of useful information left.

This ensures that the final version is accurate in all 16 bits, and also provides extra low-end detail which can be statistically squeezed into the bottom bit through proper use of dither.

The equivalent of both these issues in the frequency domain (sample-rate) would caused by the routine slowing down of samples to lower than their previous pitch.. obviously if you halve the speed a 48KHz recording becomes a 24KHz one, and if you need the slowed down version to contain frequencies above 12KHz, you will need a higher sample-rate to start with.

However, I don't think most people slow all their music down to half-speed as often as they compress it. (to death!)

[Jeremy_NSL]

Good advice here. Just to reiterate, I can't think of a single reason to use 48khz sampling - unless of course you own a creative card which requires it. Actually, i can't think of a really great reason to use 88, 96 or 192khz either, except for oversampling synths, if you plan on doing granular mangling, or extreme timestretching...

24-bit OTOH is highly useful in virtually all recording situations.

[C00kie]

I can't think of a single reason to use 48khz sampling - unless of course you own a creative card which requires it.

What about the target medium being Sony minidisk? I recall that (and DVD audio?) uses 48kHz...

[Jeremy_NSL]

Good points. I don't think much of anyone is targetting minidisc as a serious medium, but some DVD-Audio and DVD-Video audio is indeed 48khz (why they chose 48khz I don't understand...) Shows I don't do soundtrack work

[P.T.]

How does all of this (especially bit depth) relate to working with Vst instruments?

I usually make my tunes with synths and samples played from a midi controlled sampler.

I sometimes will render a track to work on it in wave form, but I usually have already done a rough mix while still completely in midi.

The question I have had for a while is;

Should I render at a high bit depth (24, or 32) and should I render at the volume that track is currently mixed at, or should I raise the level up close to 0 db to get the hottest signal possible, short of clipping by a few db?

I have been thinking that the level of the signal is irrelevent as I am not recording through a mic so no noise should be added. The recording noise is already a part of the sample I am using, and any noise a vst puts out is an unavoidable and unalterable part of the sound.

What do you think?

[C00kie]

Yes (render to 32bits float)
Yes (keep peaks near 0dB, should do that in the mix already)

[IIRs]

Should I render at a high bit depth (24, or 32) and should I render at the volume that track is currently mixed at ?

Yes, and yes.

Using a high resolution means there is no need to raise the gain, as low level signals are much better represented.

[P.T.]

Thanks, platinumears. That's what I thought.

Yes (render to 32bits float)
Yes (keep peaks near 0dB, should do that in the mix already)

COOkie, how can I mix everything at 0db and call it a mix? I want each sound to be at a certain level compared to other sounds. That's what a mix is. Unless we're not understanding each other.

[C00kie]

COOkie, how can I mix everything at 0db and call it a mix? I want each sound to be at a certain level compared to other sounds. That's what a mix is. Unless we're not understanding each other.

Ofcourse mixing is about getting appropriate levels of the tracks relative to each other. What I do is set back the levels of all tracks to -12dB or so. If you get clips then adjust the level of all tracks and set them back another 3db.

Or don't bother with individual track volumes and use the master fader for that. I guess that will work fine also nowadays. Maybe I'm a bit "old-school"

edit: but do keep an eye all times at the final level meter, that is what you hear. And you don't want it to be clipped. There is a setting somewhere at what level 0dB on the final meter is. I recall it is default -6dB. That means the soundcard will clip if the meters show +6dB.

[P.T.]

One of the nice things about Tracktion is that you can adjust all of the track levels together by shift/clicking on all of them and then you can move them all together. No need to move them individually and risk screwing up the relative volumes.

[.jon]

Thanks a million, PT! hot top tip

All in all a very good topic, I've got nothing to add but thanks to other contributors, got my bits and depths all sorted out now.

[IIRs]

shift/clicking on all of them

or R-click -> select all other filters of this type.