Bit rates

[muzicman]

How do the different bit rates affect sound from recording through mix down and sharing? Any adverse affects to be aware of? It sounds logical to record everything in the highest bit rate, but I would hate to go through all that and hit a wall.

[Telefunky]

bit rate means the bandwidth of a digital transmission: the more bits can be transferred per (time) unit, the better the reproduction - 128 kbit/sec mp3 is less detailed than 384 kbit/sec

but probably refer to the bit depth, eg 16, 24, 32 bit audio data
this defines the maximum dynamic resolution of the signal (in the first place)
it does not define how detailed the spectral content will be reproduced
the 'sound' of 16 or 24 bit (using identical converter chips) is exactly the same

16 or 24 bit is only relevant at the analog->digital and digital->analog conversion stages
'within' the system it's usually anything from 32-80bit and may even change depending on processing - but that's entirely the programmer's job ...

as it's about dynamic, the output part is fairly simple: 16bit can handle some 90dB range
if you turn your stereo up just as much that you can hear the most faint part of the track, then a 16bit record can add up to 90dB 'more sound', which will end up pretty loud...
you'll find that probably in some ambient recordings of thunderstorms, but no regular tracks
those may have a dymanic range of (say...) 10-20 dB, usually less than 3 dB (loudness maximizing)
there's simply no need for 24bit because the source doesn't provide enough details

at the input side it's quite different, as digitzing low signal levels (as with microphones) may suffer from something called 'quantization noise', when (simplified) the converter fails to detect the proper level.
this is quite evident with many 16bit converters and 24 bit gives some extra range which results in a noticably 'smoother' recording.
with enough input (like a guitar/bass DI signal) it doesn't really matter
but an acoustic guitar tracked with a ribbon mic over a distance of 10-20" is a completely different story...

important sidenote:
when I refer to 16bit I mean quality converters (studio grade)
there's a lot of consumer stuff labeled 16bit, which actually performs more like 12bit devices

basically this also applies to 24bit converters - real performance may vary a lot
(don't trust any specs unless you measured the gear or had the chance to listen yourself)
the most important parts for a good result are clock stability and analog noise/powersupply
it's not even the chip itself - the same one may perform completely different on 2 different circuit boards

cheers, Tom

[muzicman]

Thank you Telefunky. So with this daw I'm good at 16? If I understand you correctly. That's a lot of info for an analog guy!

[pwnified]

If unsure, definitely go for 24 bit for recording. It will give you the extra headroom in case things get clippy. The thing to remember is to keep the recording level well below the red (on the IN meter), as you are probably aware.

I agree with everything Telefunky said. The amount of dynamic range of most recordings these days is probably 10db due to all the dynamic compression. It really seems silly to have so much quality in the recording stage just to have it all smashed out in the mastering stage.

If you do go with 16 bit, multitrack will still sound good, because when converting to floating point to do the processing and then back, a triangular dither is applied to any signal which drops bits (like going from float to 16 bit or even 24 bit integer)

But as far as samplerates go, 44.1khz is fine, this is the target samplerate for CD format. This is because we are making music for humans. 44k is also the most supported for other apps and things like compatibility with IAA and Audiobus.

[Telefunky]

in Multitrack DAW you'll usually set the project to 24 bits, as there's no choice for individual file settings and with microphones or very dynamic signals it's the best choice anyway

but you can track any synth, keyboard, drum-machine, DI electric guitar/bass without loosing quality in 16bit if the converter is a quality device, say an old Adat or 'vintage' gear from Digidesign, Korg etc
(ok, not class compliant anyway - may apply to IOS only if the interface features Adat optical ports)

pawnified mentioned improved headroom
that used to irritate me for some time... as every converter has a fixed analog upper limit
what you get effectively is more lower bits, but those become more and more unreliable the lower their 'position' (due to laws of physics)
so there's still good reason for proper level setting
(certain strategies of oversampling and noise shaping in current gear seem to help, but I'm not too deep into tech details)

when tracking with 16bits I assumed that the converter would toggle just the last bit at low levels
(which would be expectable and perfectly ok)
but real world results were more like that bit hopping around in a 3 bit range
at signal levels around -70 to -80 dB it's not terribly noticable on it's own, but it makes the sound grainy, in particular if dynamic processing follows

best advice is to quickly check it yourself by recording an empty channel and add tons of gain afterwards to lift the noisefloor to convenient listening level - also helps to check proper grounding and such stuff...
as there are differences in analog gear, so there are differences in digital - it's not all created equal

cheers, Tom

[muzicman]

Thanks guys. I'm recording on an Apple IPod I bought last year. I'm gonna try the noise floor thing at 24bit. Ya know, this very different from tape! Hahaha! But really, I'm already impressed with how clean it all sounds. I'm missing some of the tape warmth and hiss, but the clarity makes up for it I think.

[pwnified]

pawnified mentioned improved headroom

I didn't mean to imply that, I simply mean that if you don't want to worry as much about matching the ranges, go with 24 bit and turn it down well below the red. At 16 bit, you'll need to be more careful about how each component 'packs up' along the chain. If you don't maximize the headroom at 16 bit, there are a lot of usable bits being thrown away, but at 24, you'll capture the whole range. Anyway, good discussion.