rendering in 24bit?

kabelton · Post by **kabelton** » Tue Jan 24, 2006 4:29 pm

hi,
two questions,
my audioloops are always in 16bit/44,1kHz, does it make sense to render them in 24bit/44,1kHz after using a lot vsts/ableton fx? does it make sense to do the mixdown in 24bit?

thanks

clipperer · Post by **clipperer** » Tue Jan 24, 2006 4:49 pm

i quess not, for example take a very low rate mp3 and convert it to wav, youll hear that quality hasnt changed, i think same happens with everything else

stinky · Post by **stinky** » Tue Jan 24, 2006 4:51 pm

As long as you plan on using dither if you want to listen to at 16bit

kabelton · Post by **kabelton** » Tue Jan 24, 2006 4:56 pm

clipperer wrote:i quess not, for example take a very low rate mp3 and convert it to wav, youll hear that quality hasnt changed, i think same happens with everything else

i thought because of the fx after the loops, perhaps their output is 24bit or 32 bit floating point.

thanks

icedsushi · Post by **icedsushi** » Tue Jan 24, 2006 5:05 pm

kabelton wrote:perhaps their output is 24bit or 32 bit floating point.

Does anyone know which one it is? I've been hearing some buzz (from very reputable audio app programmer sources, IMO) that 32 bit floating point doesn't need to be dithered when going directly down to 24 or 16 bit in one pass, because floating point re-quantizes the bits on the fly.

Whether this is OK to omit dither in Live or not --- going from 32 bit float down to 16 in one pass without dither, I would like to know myself. I know everyone has beat into their head to "always dither" but I've been hearing this 32 bit float thing is starting to change that.

tomperson · Post by **tomperson** » Tue Jan 24, 2006 5:26 pm

The thing is: in your final mixdown (the one that you will be mastering later) you should always go for the higher bit depth, that way you get the highest resolution possible. It doesn't matter if your wav files are 16 bit, all the internal computations are made at 32 bit float, so when rendering to 16 bits you lose resolution.

Do the reduction to 16 bits on the final master, with dithering as someone posted.

icedsushi · Post by **icedsushi** » Tue Jan 24, 2006 5:29 pm

tomperson wrote:The thing is: in your final mixdown (the one that you will be mastering later) you should always go for the higher bit depth, that way you get the highest resolution possible. It doesn't matter if your wav files are 16 bit, all the internal computations are made at 32 bit float, so when rendering to 16 bits you lose resolution.

Yep, I agree with this part above to answer the original question.

tomperson wrote: Do the reduction to 16 bits on the final master, with dithering as someone posted.

So you have no comment on my question about whether dithering is necessary or not coming directly down from 32 bit float to 16 bit for a CD?

boomklik · Post by **boomklik** » Tue Jan 24, 2006 5:43 pm

I do all mix down at 24 bits no matter what the resulting media will be.
I have tried using 16 but there is a noticeable difference in sound quality
after many, many operations have been performed on each track
(i multi track out everything in live to pro tools)

As far as I know, you should apply a dither when moving from higher to
lower bit rates...per Mr Katz, this should be the absolute last step in the
process.

tomperson · Post by **tomperson** » Tue Jan 24, 2006 5:44 pm

icedsushi wrote:
tomperson wrote: Do the reduction to 16 bits on the final master, with dithering as someone posted.
So you have no comment on my question about whether dithering is necessary or not coming directly down from 32 bit float to 16 bit for a CD?

I haven't heard of that before to be honest...Could be true, but I can't imagine how the 32bit to 16bit resolution reduction could not be avoided without dithering.

When you lower the bit depth, basically what is done is to keep the most significant bits. For instance, when going from 24bits to 16bits, the lowest 8 bits are discarded. Basically, dithering adds a noise floor that is added to the original signal so that the tiny details get preserved when the bit discarding takes place. It moves some of the detail that lies in the lowest 8 bit (in the example) to the upper 16 bits.

I guess the same could be applied to a 32 bit signal, but as live works 32bit floating point (not integer) the question could be tricky. Don't know.

tomperson · Post by **tomperson** » Tue Jan 24, 2006 5:46 pm

boomklik wrote:I do all mix down at 24 bits no matter what the resulting media will be.
I have tried using 16 but there is a noticeable difference in sound quality
after many, many operations have been performed on each track
(i multi track out everything in live to pro tools)

As far as I know, you should apply a dither when moving from higher to
lower bit rates...per Mr Katz, this should be the absolute last step in the
process.

Absolutely. But we are dealing here with 32bit floating point, not integer, so the bit reduction algorithm maybe works differently. Discarding the "16 less significant bits" makes no sense here.

Angstrom · Post by **Angstrom** » Tue Jan 24, 2006 5:50 pm

tomperson is right about it being good to go to 24 even if your loops are 16bit- because Lives effects will benefit from the increased bit-depth.

re: dither.
yes, dither.

I started writing a bit about dither here then remembered bob katz has a massive article about it. (an extract from his book)
http://www.digido.com/portal/pmodule_id ... age_id=27/

essentially by dithering you are 'encoding' an average of what was present in the 24bit version. If you imagine 16 bit as only having 16 values (yes I know it doesn't!)...

if an input volume curve wobles between values 2 and 3 in our pretend 16 bit encoding scale,
EG: 2.11, 2.321, 2.143. 2.123 ....

all that will come out of 16 bit output is '2',
EG: 2 , 2 , 2 , 2 ...

not very accurate !
By adding dithering noise to the input signal the signal is made to wobble around the 16 bit range a lot more ..
EG:our above figures becom : 3.11, 1.002, 2.483, 1.3023 !!

but the averaged value from this stream comes out as more accurate than the raw 16bit!
EG: 3, 1, 2, 1, ..
Imagine those points plotted on a graph and 'smoothed' , you could now read floating point numbers off the graph that were longer than 16 bit allows. Magic!

So the answer is - yes dither , but keep a 24 bit copy!

tomperson · Post by **tomperson** » Tue Jan 24, 2006 6:03 pm

I have the book, it's really interesting, if a bit dry at times (and I'm quite techhead). Interestingly enough he never talks about 32bit floating...He talks about 32 bit "words", so one must assume he refers to 32bit integer arithmetic, not 32bit floating point, where some of the 32bits are used for the integer part and some others for the numbers after ".".

The idea you mention Angstrom is correct, but in 16bit integer you wouldn't have 2.11, 2.321, 2.143. 2.123, you would have integers 1,2,3,4, etc up to 65536.

Angstrom · Post by **Angstrom** » Tue Jan 24, 2006 6:45 pm

yeah,
thats why I said we were working with an imaginary form of 16 bit. I dont really think that 16 bit has only 16 values of loudness!

the point I was making was that the real world signal will fall somewhere in between the quantized values however they are enumerated, also that a 24bit value will fall in between the values of a 16 bit system. By dithering you can encode and decode some of that higher bit-depth.

tomperson · Post by **tomperson** » Tue Jan 24, 2006 7:29 pm

That's it

I wonder if someone can come up with a good link talking about 32bit floating and wether we should dither or not. My first bet is that we should. In fact, I do it.

kabelton · Post by **kabelton** » Tue Jan 24, 2006 7:45 pm

Angstrom wrote:tomperson is right about it being good to go to 24 even if your loops are 16bit- because Lives effects will benefit from the increased bit-depth.

but when i render my 16bit loop added with fx, and after that, i never use fx on that loop again, is it still better? i only render it again with the whole mixdown, if it is a song and not my liveset.

thanks