I am wondering if 32bit integer is actaully better than 32bit floating...
As I know 32bit floating is actually 24bit audio data continously variable in gain but 32 bit integer is pure 32bits ...
So I assume that 32bit integer is the most pure 32bit recording...
I use Cubase 5 (still) anyone knows if the 32bit for recording is floating ?
as for 23bit true tape I think is floating...
That worries arrise from the fact that I am not satisfied with audio results using undewr SFP 32bit floating ASiO2 drivers recording into cubase 32bit.
Even with a single track I fell that something changed (for the worse)(more soft)from when I was monitoring via 2448 mixer just before recording...
Anyone other that has similar worries ?
Dimitriios
32bit floating or 32bit integer whats purest for recording?
-
hubird
-
marcuspocus
- Posts: 2310
- Joined: Sun Mar 25, 2001 4:00 pm
- Location: Canada/France
-
Nestor
- Posts: 6683
- Joined: Tue Mar 27, 2001 4:00 pm
- Location: Fourth Dimension Paradise, Cloud Nine!
I didn't notice the diffence in quality becuase I don't have appropiate speakers.On 2004-02-08 10:44, dbmac wrote:
When I use ASIO2 32 source in SFP, CubaseSX Device setup shows the VST Inputs/Outputs as "Scope 24 bit".
When I load ASIO2 FLT, SX shows the In/Out as "Scope FLT".
The difference in mixing headroom is significant. I always use the Asio2 FLT drivers.
/dave
Now, why the ASIO1-FTL destionation 64, and the ASIO2-FTL source 64 do not match in versions? I mean, shoudn't they be version "2" and "2", and not "1" and "2"? What I'm misunderstanding here?
*MUSIC* The most Powerful Language in the world! *INDEED*
Asio 1 and Asio2 are different - I think Asio2 allows for better monitoring in Cubase, but it doesn't concern me cause I use SFP for my monitoring. I just use Asio2 because I think it might be a newer version. As long as the Source and Dest devices match, there's probably not much difference.
The difference bewteen Asio 32 and Asio FLT is something else. If I change from Asio FLT to Asio 32, my SX projects are clipping all over the place, I have to back off 3db on all the channels. Like I say, more headroom, hotter levels with Asio FLT.
I don't know why the Asio32 driver shows up as 24 bit in SX, but it definitely doesn't have the dynamic range of the FLT driver.
/dave
<font size=-1>[ This Message was edited by: dbmac on 2004-02-08 18:12 ]</font>
The difference bewteen Asio 32 and Asio FLT is something else. If I change from Asio FLT to Asio 32, my SX projects are clipping all over the place, I have to back off 3db on all the channels. Like I say, more headroom, hotter levels with Asio FLT.
I don't know why the Asio32 driver shows up as 24 bit in SX, but it definitely doesn't have the dynamic range of the FLT driver.
/dave
<font size=-1>[ This Message was edited by: dbmac on 2004-02-08 18:12 ]</font>
-
Nestor
- Posts: 6683
- Joined: Tue Mar 27, 2001 4:00 pm
- Location: Fourth Dimension Paradise, Cloud Nine!
This is a great point. As I said, I don't haver the appropiate speaker boxes to hear such subtle differencies, but I will definitely use them from now on, following your advice. I'll try it.
If you are totally sure about your point, I think this is relevant enough to build a thread with some information on "why" this happens, in the Tip s and Tricks forum, we are all searching for better retuls all the time.
Thanks for your answer
If you are totally sure about your point, I think this is relevant enough to build a thread with some information on "why" this happens, in the Tip s and Tricks forum, we are all searching for better retuls all the time.
Thanks for your answer
*MUSIC* The most Powerful Language in the world! *INDEED*
When you are recording at 24bits then I cannot see the point why you want more headroom than 24bits.
Just for recording ...
Ok for mixing this is another thing and I am sure 32bit floating is the way to go, but for recording ...???
And this is the other thing that might be of great concern,whats the difference in sound beetween 24bits 32integer and 32 floating, not just the dynamic range.
Dimitrios
Just for recording ...
Ok for mixing this is another thing and I am sure 32bit floating is the way to go, but for recording ...???
And this is the other thing that might be of great concern,whats the difference in sound beetween 24bits 32integer and 32 floating, not just the dynamic range.
Dimitrios
Nestor, that's not so much a tip as an observation, namely, that CW's Asio32 driver is handled in Cubase SX I/O as 24 bit, with significantly less headroom (for mixing) than the Asio FLT driver.
Dimitrios, I agree, recording at 32 bit has no advantage - I set the bit rate in the recording preferences of the sequencer, and leave the Asio drivers always at Asio flt.
/dave
Dimitrios, I agree, recording at 32 bit has no advantage - I set the bit rate in the recording preferences of the sequencer, and leave the Asio drivers always at Asio flt.
/dave
yes Dave but isn't it better to record at 32 bit integer at the sequencer than at 32bit floating ?
SFP mixer uses 32bit so if you are recording with say some equalisation at least then you output true 32bit files ,so if you record at 32bit floating (24bits actually) then you are throwing away 8 bits of information...
Even if you just change the vlume fader you will have a 32bit file output at your SFP mixer.
Is there an sequncer recording 32bit integer ?
Dimitrios
SFP mixer uses 32bit so if you are recording with say some equalisation at least then you output true 32bit files ,so if you record at 32bit floating (24bits actually) then you are throwing away 8 bits of information...
Even if you just change the vlume fader you will have a 32bit file output at your SFP mixer.
Is there an sequncer recording 32bit integer ?
Dimitrios
Cubase doesn't record at 32 int. only floating, that's substantially 24 bits with a trick to get more depth moving the ...point hehe. So 32 int. drivers must be used only with capable apps, otherwise they are seen as 24, I think...
I stopped upgrading at sx 1.06 (still using vst32), dunno SX 2.x though.
<font size=-1>[ This Message was edited by: alfonso on 2004-02-08 20:41 ]</font>
I stopped upgrading at sx 1.06 (still using vst32), dunno SX 2.x though.
<font size=-1>[ This Message was edited by: alfonso on 2004-02-08 20:41 ]</font>
-
paulrmartin
- Posts: 2445
- Joined: Sun May 20, 2001 4:00 pm
- Location: Montreal, Canada
I always record at 24 bit. I never thought about the results of truncating a 32 bit wav produced by SFP - I figured this would be inaudible.On 2004-02-08 19:51, musurgio wrote:
yes Dave but isn't it better to record at 32 bit integer at the sequencer than at 32bit floating ?
SFP mixer uses 32bit so if you are recording with say some equalisation at least then you output true 32bit files ,so if you record at 32bit floating (24bits actually) then you are throwing away 8 bits of information...
Even if you just change the vlume fader you will have a 32bit file output at your SFP mixer.
Is there an sequncer recording 32bit integer ?
Dimitrios
/dave
Paul
It is nice to have something to "give". When recording at 16 Bit, if you do anything, that changes the level of the recording (eq for example), then you either loose bits (when turning down) or boost something, that is not there (If you record -12dB and then boost 12dB, you actually boost the signal with 2 "empty" bits).
When recording at higher solutions, it doesn't realy matter that much anymore, because you still have bits to cut boost.
Another thing is, that a lot of processing can be done more precisely, if you have more exact resolution in the low level of the signal. With 16 bit, you have very bad resolution in the low level parts.
As for how much you can hear it - I don't know too much about it My speakers are not realy fit for the task. I record in 32bit floating. Hard disk space comes cheap these days, and don't have to worry *hit about levels I actually find it rather nice to normalice to around -8 or -12dB. This gives some headroom for when I process the signal with effects, that are not always 100% predictable. And since I record in 32 bit flt, I don't loose anything from cutting down on the signal level (well maybe a slight bit in the very low level signals - but nothing that I can hear, and nothing, that can not be heard when everything is cut down to 16 bit.
About the integer<->flt:
There is so much discussion about this, that I truly believe it is a bit academic (othervice, I think it should be easier for people to agree (again counting, that people often tend to defend their own stuff)). Integer does have more resolution, untill you get to -48dB. But I believe (this is my very own wanna-be-hypothesis) flt makes less artifacts, when doing processing - becuase it's headroom is dynamic. Again, the stuff is so low down in level, that I don't believe people will hear it (maybe unless they have 1m€ equipment - but then it is truly thier own fault
).
It is nice to have something to "give". When recording at 16 Bit, if you do anything, that changes the level of the recording (eq for example), then you either loose bits (when turning down) or boost something, that is not there (If you record -12dB and then boost 12dB, you actually boost the signal with 2 "empty" bits).
When recording at higher solutions, it doesn't realy matter that much anymore, because you still have bits to cut boost.
Another thing is, that a lot of processing can be done more precisely, if you have more exact resolution in the low level of the signal. With 16 bit, you have very bad resolution in the low level parts.
As for how much you can hear it - I don't know too much about it
My speakers are not realy fit for the task. I record in 32bit floating. Hard disk space comes cheap these days, and don't have to worry *hit about levels I actually find it rather nice to normalice to around -8 or -12dB. This gives some headroom for when I process the signal with effects, that are not always 100% predictable. And since I record in 32 bit flt, I don't loose anything from cutting down on the signal level (well maybe a slight bit in the very low level signals - but nothing that I can hear, and nothing, that can not be heard when everything is cut down to 16 bit.About the integer<->flt:
There is so much discussion about this, that I truly believe it is a bit academic (othervice, I think it should be easier for people to agree (again counting, that people often tend to defend their own stuff)). Integer does have more resolution, untill you get to -48dB. But I believe (this is my very own wanna-be-hypothesis) flt makes less artifacts, when doing processing - becuase it's headroom is dynamic. Again, the stuff is so low down in level, that I don't believe people will hear it (maybe unless they have 1m€ equipment - but then it is truly thier own fault
).I am not shure about this one. that would be something like 40 bit floating or so. I always thought, that SFP was 32 bit integer - meaning, that -6dB is effectively 31 bit. Can some developer verify?On 2004-02-08 19:51, musurgio wrote:
Even if you just change the vlume fader you will have a 32bit file output at your SFP mixer.
I also know that SFP uses 32bit integer, fixed point.
That is way better than 32 bit floating as long as you stay over -48db.
I would like though to have a sequencer that could record and work at that resolution.
VDAT records 32bit integer so I assume (never heard it) that this is why it gets so nice comments about its sound.
Could not be anything else (I guess)
Dimitrios
That is way better than 32 bit floating as long as you stay over -48db.
I would like though to have a sequencer that could record and work at that resolution.
VDAT records 32bit integer so I assume (never heard it) that this is why it gets so nice comments about its sound.
Could not be anything else (I guess)
Dimitrios