Mastering at 24bit or 16bit

[bridgealantee]

The great Bob Katz can put it better than I can.

Keep the bits. Cumulative digital processes if improperly performed can be very
degrading to sound. The reason (and many engineers are not aware) is that almost
every DSP computation adds additional bits to the wordlength. The wordlength
can increase to 24-, 56-, or even 72-bits. The right thing to do is keep your newly
"lengthened" words as long as possible, until the final stage, where we will dither
them down to 16-bits for the CD. 16-bit dither should be reserved as a one-time
only process at the end of the chain.

[davepermen]

CDs are 44.1k 16 bit, do they sound bad? do you think you can hear the difference between 44.1k and 192k? between 16 bit and 24 bit? do you know how low the audible level of those differences are?

seems that often the choice of sample rate should be made to choose one that matches your other processing steps to minimize SRC processes.

you can't hear the final audio any better than 44.1k and 16bit. but while processing audio, you want artefacts to be smaller than HALF of the smallest represented step the audio has, both in time, and in volume. => you want theoretically to have at least 88.2k, and at least 32bit. that way, any rounding error will get removed completely afterwards when downsampling to 44.1k 16bit audio cd.

this is about the ninquist theoreme or how ever that's called. that you need more than twice the detail to have errors (at that detail) to not be visible in the final sampling anymore. something like this.

if you play videogames, which you might be, then you might have heard about HDR, high dynamic range renderin in games. they render the games at 16 / 32bits per pixel per color on the gpu right now to have the (with luck and tweaking sometimes) no banding artefacts in the final image. the final image, then, has only 8bit per pixel. but the whole lighting and shading needs higher presision to not get results in the end for the 8bit image. (8 bit per color, together that's rgba (a being alpha), a.k.a. 32bit color (which is really only 24bit).

so while working on stuff you want the highest precicion possible to not have any accidental artefacts.

oh, and they discuss 64bit rendering right now for certain effects actually.

[bridgealantee]

Also

One way to maintain quality is to follow this important rule: Source
recordings and masters should have higher resolution than the eventual release medium.
There's always a loss down the line, due to cumulative processing and lossy transmission
techniques. For example, consider a lossy medium like the analog cassette. Dub to
cassette from a high quality source, like a CD, and it sounds much better than a copy
from an inferior source, like the FM radio. In other words, the higher the audio quality
you begin with, the better the final product, whether it's an audiophile CD, a multimedia
CD-ROM, or a talking Barbie doll.

[bridgealantee]

I want to keep posting more, but peeps should just read this.

http://docs.google.com/gview?a=v&q=cach ... xUrEXOIG3w

[Tarekith]

Wish I could master for talking Barbie dolls.

[shatzer]

I would HIGHLY suggest using a higher sample rate than 44.1. Use at least 48, but higher is always better. Just remember to dither if you are going to bounce to 16 44.

True, but if you're sending it to a mastering engineer then you should NOT dither. And leave atleast 3 - 6 db of headroom with no eq, compression, or limiting on the master.

But if you are recording at 24 48 and you are giving your mastering engineer a CD, you have in fact lost your only chance to dither. I would ideally like to get a drive with 24 48 files and dither them myself, but often the artist just wants to send you a cd. I agree with the rest though.

The mastering engineer should apply the dither not the client. If in deed they are sending you a CD instead of a DATA CD then they probablly don't even know what dithering is.

[Tarekith]

And just to clarify, it's not the dithering that's so bad as much as the noise-shaping involved with dithering that we're trying to avoid until the last step.

[Tone Deft]

CDs are 44.1k 16 bit, do they sound bad? do you think you can hear the difference between 44.1k and 192k? between 16 bit and 24 bit? do you know how low the audible level of those differences are?

seems that often the choice of sample rate should be made to choose one that matches your other processing steps to minimize SRC processes.

you can't hear the final audio any better than 44.1k and 16bit. but while processing audio, you want artefacts to be smaller than HALF of the smallest represented step the audio has, both in time, and in volume. => you want theoretically to have at least 88.2k, and at least 32bit. that way, any rounding error will get removed completely afterwards when downsampling to 44.1k 16bit audio cd.

this is about the ninquist theoreme or how ever that's called. that you need more than twice the detail to have errors (at that detail) to not be visible in the final sampling anymore. something like this.

if you play videogames, which you might be, then you might have heard about HDR, high dynamic range renderin in games. they render the games at 16 / 32bits per pixel per color on the gpu right now to have the (with luck and tweaking sometimes) no banding artefacts in the final image. the final image, then, has only 8bit per pixel. but the whole lighting and shading needs higher presision to not get results in the end for the 8bit image. (8 bit per color, together that's rgba (a being alpha), a.k.a. 32bit color (which is really only 24bit).

so while working on stuff you want the highest precicion possible to not have any accidental artefacts.

oh, and they discuss 64bit rendering right now for certain effects actually.

interesting. we've had this discussion many many times on this forum, I appreciate another voice.

yes, Nyquist's theorom. and yes I've had exposure to some of the details of HDR, the next generation of displays is SICK, makes even the nicest of Hi Def TVs look lame.

sometimes people post in here with the craziest interpretations of this topic, glad you're on point.


OP - honestly you should ask the engineer you'll be passing them off to, right?

[davepermen]

another thing that i've not noted in here (and it's the same for 3d graphics related, and much much other things)

only work in powers of two. if your source material is at 44.1khz, don't move to 48khz in the project, and then resample it in the end again. while computation works at a higher precision, the conversion from 48 to 44.1 is rather lossy. the only conversions that don't introduce problematic artefacts are power-of-two conversions. not that people would really hear it, most likely they don't, but it should go like this:

44.1khz stuff should get edited at 44.1khz, 88.2khz, 176.4khz or higher (or even 22.05khz would be okay in that it would not create 'smearing artefacts', it would lose all higher frequencies of course.

with bitrate, it's a bit different, actually. the move from 16 bits to 24bits is lossless, as it adds data "at the end of the value", means just zeroes after the dot.

I made a visual example of the technical artefact in question. so, hosted..

[Rave]

The reason why I brought the topic up was because I was using 16bit 44khz samples.

My masters were sent off yesterday at 24bit 44kz no dither

Good images those Dave. I have a Physics degree (Graduated 1992) and specialised in FFT techniques applied to sound) but sadly I have forgotten it all LOL

[davepermen]

is that guy masturbating in your sig? haha

yeah, i normally work at 44.1khz as my soundcard doesn't allow 88.2khz. and in the end, the audio most likely end in 44.1khz (except if i would play out live). as long as i don't create music for dvds, at least.

but i use the highest bitrate available to work with, till saved to wave.

[Rave]

is that guy masturbating in your sig? haha

.

Yeah

[simpli.cissimus]

32bit at 44.1 Hz, that's my way to go...!

It's also recommended in the Ableton-Manual...!

Since I do it this way, all my music sounds the same after exporting.

Why, because Live internally works always in 32 bit floating.

I use 44.1 because it gives me enough quality...!

Need more Quality? Then double it up, to 88.2 , but never go for 48Hz.
This is for DVD only, and you can double this up to for more Quality.

Never use this for CD, converting from 48 down to 44.1 is a pain and you loose at the end.

I don't understand why all the producers still go for 24 bit...!

Makes no sense at all. Converting them down to 16 bit you have to put three 24bit parts together and cut them down to 16 bit(That's the way a converter do it for you). So you get it ''glued'' at two places and cut on three others.
Too many artifacts will always be made by that.

32 bit just needs one cut and you have two 16bit parts, and less artifacts then you can get at this high quality.

Everyone still using 24 bit is wrong...!!!

...Cheers...!!!

[davepermen]

actually, 32bit is normally floatingpoint, a.k.a. 24bit + 8bit exponent.

but it doesn't matter, live knows best how to work artefact free internally. and using those same (and thus suggested) settings for the rest means one conversion less

[evon]

The great Bob Katz can put it better than I can.

Keep the bits. Cumulative digital processes if improperly performed can be very
degrading to sound. The reason (and many engineers are not aware) is that almost
every DSP computation adds additional bits to the wordlength. The wordlength
can increase to 24-, 56-, or even 72-bits. The right thing to do is keep your newly
"lengthened" words as long as possible, until the final stage, where we will dither
them down to 16-bits for the CD. 16-bit dither should be reserved as a one-time
only process at the end of the chain.

+1