32bit floating or 32bit integer whats purest for recording?

Stige · Post by **Stige** » Mon Feb 09, 2004 6:55 am

My few words to this topic...
SFP mixers are definitely 32-bit integer. You can find proof about this by just reading the documentation. SFP dynamic range is about 180dB (don't remember the exact number). With floating point, there would not be certain limit in dynamic range. Another thing to consider is that SFP mixer channels can get overloaded. This would not be the case with floating point processing. There is built-in 24dB default headroom in the STM2448 mixer.
Cubase uses floating point and SFP not. It's a bit matter of taste, whether to use floating point ASIO, as at the very moment the signal comes into SFP, it will be turned into 32-bit fixed. Floating point ASIO doesn't give any more headroom in SFP. If you 'overload' the signal in cubase, it won't be really overloaded, until it reaches the SFP. With floating point ASIO, it is as much overloaded as fixed point ASIO, when you feed it into SFP mixer.
About SFP ASIO Fixed point drivers: Yes, cubase describes them as '24-bit'. Still if you record in cubase with 32-float, you will get full 32-bit precision thru the drivers. You can test this simply by recording some SFP synth and then using the 'analyse' function in cubase SX. It always shows that 29-32bit has been used.
Another thought about floating point processing in general: As we know floating-point processing calculates very small numbers (for example 0.00034504346773...). In the present processors there is a certain limit, how small numbers they can process. After that limit is reached, there will be errors in calculation. Thus floating-point audio is not accurate. The situation is totally different when processing fixed point audio. Fixed-point calculation does not need extremely small numbers. So there will be no math resolution limits with CPU's/DSP's. Audio calculations stays always accurate.

Of course comments are welcome

I must admit that I've not been playing around with Floating point ASIO drivers too much. Just quick testing. There seems to be bug with them, as cubase level meters jumps to the max, looks ugly.

musurgio · Post by **musurgio** » Mon Feb 09, 2004 9:17 am

Nice comments.
I too think that 32bit integer is the most accurate presentation of a digital file compared to 32bit floating.
Do you know whats the actual difference if you use 32bit integer ASIO as opposed to 32bit floating ASIO under SFP when recording in Cubase and as well as mixing in SFP (Cubase plays back) ?
What would you use 32bit integer ASIO or floating as best choice.
You can divide it into Recording session and into Mixing session.
My answer would be to use 32bit integer ASIO when recording into Cubase and maybe 32bit floating when mixing fronm Cubase into SFP.
Dimitrios

samplaire · Post by **samplaire** » Mon Feb 09, 2004 12:30 pm

I never bothered those 32bit, 32 flt drivers as my sfp was always synced to my A3000 sampler which is 16bit/44.1kHz. Now I try to avoid using the sampler and do everything in SFP/Logic. A question came to my mid after reading your discussion. First I have to make a background to the question: I use SFP fx/synths and Logic native and VST plugins, I don't record anything to the analog pulsar/luna inputs. I have to add my Logic version can handle only 24bit/96kHz (due to not too fast system and not too many dsp chips I try to avoid 96kHz and stay with 44.1kHz). Is there any point for me to use the 32bit integer or flt drivers? What benefits I have from ithem in this case? Does SFP have internal processing in 32bits regardless the asio drivers?

[ This Message was edited by: samplaire on 2004-02-09 12:32 ]

Immanuel · Post by **Immanuel** » Mon Feb 09, 2004 12:41 pm

SFP is 32bit nomatter, what drivers you use.
For me, a mayor reason to use higher than 16 bit resolution is to have some room to fool around.

Zer · Post by **Zer** » Mon Feb 09, 2004 12:43 pm

for recording you might want to try samplitude, too.

samplaire · Post by **samplaire** » Mon Feb 09, 2004 1:20 pm

On 2004-02-09 12:43, Zer wrote:
for recording you might want to try samplitude, too.

If there was a Mac version...

samplaire · Post by **samplaire** » Mon Feb 09, 2004 1:21 pm

On 2004-02-09 12:41, Immanuel wrote:
For me, a mayor reason to use higher than 16 bit resolution is to have some room to fool around.

Do I get the mentioned 32bit headroom in my Logic only by using the 32bit asio drivers?

_________________
Sir Sam Plaire Scopernicus

[ This Message was edited by: samplaire on 2004-02-09 13:22 ]

zezappa · Post by **zezappa** » Mon Feb 09, 2004 4:54 pm

I'm really a dummy about all these tech issues but I've found that my best recording/mixing work flow is got when using Only & Always asio2-32 source 64 and asio2 dest-64, and no matter what bit rates are in use. Some better system performance/stability are achived too.
(I'm talking just about SX on XP here)

musurgio · Post by **musurgio** » Tue Feb 10, 2004 9:29 am

zezappa:
Are you saying that you tried also the 32bit floating drivers but you actually preffered the 32bit Asio2 ?
Dimitrios

Immanuel · Post by **Immanuel** » Tue Feb 10, 2004 9:31 am

On 2004-02-09 13:21, samplaire wrote:
On 2004-02-09 12:41, Immanuel wrote:
For me, a mayor reason to use higher than 16 bit resolution is to have some room to fool around.
Do I get the mentioned 32bit headroom in my Logic only by using the 32bit asio drivers?

_________________
Sir Sam Plaire Scopernicus

[ This Message was edited by: samplaire on 2004-02-09 13:22 ]

Inside Logic, you get what Logic can provide you with. Inside SFP/XTC, you get 32bit. I hope that answers you question?

samplaire · Post by **samplaire** » Tue Feb 10, 2004 9:53 am

Absolutely

zezappa · Post by **zezappa** » Tue Feb 10, 2004 12:50 pm

On 2004-02-10 09:29, musurgio wrote:
zezappa:
Are you saying that you tried also the 32bit floating drivers but you actually preffered the 32bit Asio2 ?

Yes I am, though I can't say if it's the best for someone else or explain it relating an expert knowledge.
I shall also say that my main work is in XTC mode.

rodos1979 · Post by **rodos1979** » Tue Feb 10, 2004 6:36 pm

Hello to all!

Well, I am definatley not an expert on the subject... however, I d like to tell my opinion and hope someone more knowledgeable will correct me!

My observations:
1) SFP is 32bit fixed-point, as we all know, while most sequencers nowadays are 32bit floating-point.
Each way has its advantages and disadvantages. With 32bit fixed-point, you have a true 32bit signal, which yields a WAY more accurate representation than 32bit floating-point, which is in essence 24-bit.
On the other hand, with 32bit fixed-point you are "limited" to about 192dB of headroom. With 32bit floating-point the headroom is "limitless" practically.
2) The best AD converters nowadays are 24bit, but still do not offer the 144dB of headroom that 24bit fixed has to offer! Thus, it SEEMS to be pointless to record to anything more than *real* 24bit, if what you are recording is the signal coming from an AD Converter. Recording in 32bit floating point will give you no advantage in terms of headroom, as you are limited by your AD Converter's capabilities. (however, see 5)
3) Matters change with sound coming from SFP! SFP provides you with a true 32bit signal, thus recording it in anything less than a true 32bit resolution will deteriorate a bit the signal (do not forget that 32bit flt point is in reality 24bit!). This is one of the reasons that SFP has some advantage to sound in comparison with native synths and effects. If the very same SFP synths and effects (with the same programming) were running natively (aka 32bt float=24 bit), their sound would be a bit inferior to what it is now.
4) I may have given you the impression that 32bit floating point is useless... This is NOT the case! The advantage of floating-point is the limitless headroom, which is of utter importance for care-free mixing!
5 and last) It is a fact that 32bit (fixed or floating, doesnt matter) puts less strain on your CPU than 24bit! I dont remember very well how to explain that, but it is true!

In conclusion, if hard disk space and transfer rate is not a matter, try to record always in 32-bit. If you mix in SFP and use lots of SFP synths, then go with 32bit fixed. If you mix natively, then 32bit float.

As I personally record mostly from my analog audio source, I prefer to record in 32bit flt and mix natively with limitless headroom in XTC mode.

The purfect however to me would be if we could record in 40-bit floating point, having a true 32bit signal (that SFP provides) plus the advantage of limitless headroom...

musurgio · Post by **musurgio** » Tue Feb 10, 2004 8:52 pm

All these sound very truthfull...
So what would you use when recording from SFP mixer (using effects too) into Cubase ?
32bit floating ASIO or 32bit fixed ASIO ?
Also after recording when mixing into SFP and playing back audio from Cubase would you use 32bit floating ASIO or fixed ?
Dimitrios

rodos1979 · Post by **rodos1979** » Tue Feb 10, 2004 9:41 pm

Hello!

So what would you use when recording from SFP mixer (using effects too) into Cubase ?
32bit floating ASIO or 32bit fixed ASIO ?

- To the best of my knowledge, SX is not capable of 32bit fixed audio recordings. So, it is not possible to capture what comes out of SFP accurately. So, taking into account what I ve said before, the only option is 32bit flt.

Also after recording when mixing into SFP and playing back audio from Cubase would you use 32bit floating ASIO or fixed ?

- There is no option to record audio at 32bit fixed in SX. If there was, I would record in 32bit fixed (given that I mix in SFP which is 32bit fixed). However, take into account that if the previous scenario was true, you shouldnt touch anything in SX! Even the slightest fader movement converts the signal to 32bit flt.

musurgio · Post by **musurgio** » Wed Feb 11, 2004 2:11 am

So you are saying that if there will be no fader movement in Cubase (by the way I use 5) the 32bit fixed will be preserved ?
That would be a nice scenario indeed.
What if I don't fader move but put 32bit effects in insert on Cubase channels will the 32bit fixed still exist ?
Dimitrios

Post by **garyb** » Wed Feb 11, 2004 3:00 am

everything in the sfp realm is 32bit which allows a bit more accurate math(headroom),however,sx is 32flt which is 24bit with room for overflow.you sound files will be 24bit at best.there is no problem with this. a well recorded track will sound good enough to win an award

.

when you use an sfp effect in sfp or xtc mode,the calculations ar in a 32bit world.this is helpful.it makes for a smoother sound.the fact that it then goes to 24bit is not so destructive as it might seem. you are going to end up in 16bit if the track is pressed to cd anyway.

still,the finer resolution when using effects will translate to a better sounding product at the final bit depth.

it's not really fair to judge sound quality between finished(final mastered) and unfinished(anything before that step)tracks. mastering is a miraculous process when done properly.that is how silk purses are made from sow's ears.......

cannonball · Post by **cannonball** » Wed Feb 11, 2004 7:24 am

hi
i have 2 questions
i have a better final sound if i mix through the asio2 32fl and i rec the mix in cubase
with the asio2 64(44.1/32bit)?

if yes the better sound is possible also if the track are rec on 44.1/24 and pass into the asio2 32fl(and of course mixer effect etc.)?

ale

[ This Message was edited by: cannonball on 2004-02-11 07:27 ]

samplaire · Post by **samplaire** » Wed Feb 11, 2004 7:58 am

I just asked the same questions in this thread and the conclusions were:
- processing in SFP is always 32 bit integer regardless the asio module.
- using asio32 flp should not cause any better result if you are using 24bit recording mode in cubase since you downgrade the bitrate. Mathematically of course. But you never know

braincell · Post by **braincell** » Wed Feb 11, 2004 8:16 am

I always wondered about which ASIO module I should use.