Tom
brett wrote on Wed, 24 August 2005 14:42 |
Tom
Quote: | Just read the pdf to double check, sorry man Pro tools LE is not 32 bit float it's 24 bit.
|
I don't know about the pdf...but plugin based bit meters say otherwise.
Chris Cavell wrote on Wed, 24 August 2005 17:24 |
Quote: | Just read the pdf to double check, sorry man Pro tools LE is not 32 bit float it's 24 bit.
|
I don't know about the pdf...but plugin based bit meters say otherwise.
|
Do you mean the plugins are 32bit? or are you reffering to P.tools LE DAW bit depth?
The DAW's bit depth. Of course, you have to modify the signal in some way to register the float values on a plugin based bit-mater...a simple 0.1dB change in level on the mixer is enough to do that though...or throw in any plugin before the bit meter.
(The plugin's bit depth is dependent upon the individual design...but all RTAS plugs that I know of are *at least* 32 bit.)
Chris Cavell wrote on Wed, 24 August 2005 17:41 | The DAW's bit depth. Of course, you have to modify the signal in some way to register the float values on a plugin based bit-mater...a simple 0.1dB change in level on the mixer is enough to do that though...or throw in any plugin before the bit meter.
(The plugin's bit depth is dependent upon the individual design...but all RTAS plugs that I know of are *at least* 32 bit.)
|
heres a link for you to check out every other one i've checked says the same... http://www.dolphinmusic.co.uk/d2/page/shop/news_story/a/news _id/e/14/a/r/e/google
also checked out the specs on the actual digidesign site, still got 24 bit processing not 32 bit float.
With the LE sytems, be it 001, 002 or M-box they are all set to record at 16 to 24 bits, sure all types of plugs operate at different depths but the LE depth is 24. Most waves stuff dithers internally down to 24 bits on LE as do others.If you send a link to show me an page stating LE does operate at 32 bit float then cool it's just I've yet to find one.
I think it's important when discussing a DAW's internal processing math to realize that often the bit depth of the temporary internal processing math can be different than the bit depth of the figure that is returned by such calculations. From my understanding - in general current apps allow a plugin to do its internal processing math at its full level but then either dither or truncate the returned number to a lower level before sending it to the next process. So a plugin with 64bit processing would indeed use that full math level no matter what app it was loaded in. I believe in many cases whether the result is truncated or dithered depends on the plugin also. So the question would be whether PT LE is sending either 24bit integer or 32bit floating point returned figures to the next process.
Best regards, Steve Berson
TotalSonic wrote on Wed, 24 August 2005 18:00 | I think it's important when discussing a DAW's internal processing math to realize that often the bit depth of the temporary internal processing math can be different than the bit depth of the figure that is returned by such calculations. From my understanding - in general current apps allow a plugin to do its internal processing math at its full level but then either dither or truncate the returned number to a lower level before sending it to the next process. So a plugin with 64bit processing would indeed use that full math level no matter what app it was loaded in. I believe in many cases whether the result is truncated or dithered depends on the plugin also. So the question would be whether PT LE is sending either 24bit integer or 32bit floating point returned figures to the next process.
Best regards, Steve Berson
|
I can't seem to find any documentation on this, every page on the web i've tried so far say 24 bits and does'nt refer to whether the LE systems are sending 32 float in any process.
But hey this orig thread was from dongle wanting to know why samp or tools shows up differences with the same wav with the same settings on the processing and I'd rather not hijack his thread to start a debate on bit depths... sorry dongle
There's a simple test for floating point operation.
Take a 16 bit source Turn the volume up 60 dB past clipping Turn the resulting signal back down 60 dB Measure the number of bits in the output.
If it's floating point, there will only be 16 bits and the signals will null.
Pro Tools le passes this test. It only reads and writes 24 bit audio files and not 32 float so it's a good idea to dither the output to your files or output.
Digidesign just doesn't go out of their way to point out the advantages of le over TDM.
What have I started? Funny though. Thanks Bob, as always, for your clarity.
Bob Olhsson wrote on Wed, 24 August 2005 18:11 |
Digidesign just doesn't go out of their way to point out the advantages of le over TDM.
|
So you are saying the sound quality of processing in LE exceeds that of TDM. There are so many pro's that I have seen writing posts here on these threads that say the TDM versions of Plugs sound better than the Native version including the LE versions. I use the L2 for example. Numerous Me's wrote that the TDM version was cleaner than the Native Version. I can't back this up, just letting you know what I read.
But, I don't really see how truncating a 32bit float will sound better than a 48bit fixed split evenly by two. There is no quantizing needed to return to the buss. I guess the quantizing and truncation goes on as it reaches the limit of the fixed 48 bits and that could be it. Please clarify your remarks.
Gerald Leitch wrote on Wed, 24 August 2005 11:51 | With the LE sytems, be it 001, 002 or M-box they are all set to record at 16 to 24 bits, sure all types of plugs operate at different depths but the LE depth is 24. Most waves stuff dithers internally down to 24 bits on LE as do others.If you send a link to show me an page stating LE does operate at 32 bit float then cool it's just I've yet to find one.
|
Not true. You can look in just about any waves plugin manual and read with little variation from plugin to plugin:
from the Waves Renaissance compressor manual | New to the processor is dithering for the final output. TDM output is dithered to 24-bit (fixed point); native output is dithered to 32-bit (floating point). In both cases the signal is "handed back" to the application being used. In TDM, the internal resolution is 56-bit fixed point; in native, it is 64-bit floating point.
|
The entire LE signal path is 32 bit floating point. LE is a "native" system. Every insert point throughout the system has 32 bit floating point I/O, and the stream doesn't get converted back to 24 bit fixed until after the last insert point on the master fader.
Chris Cavell wrote on Wed, 24 August 2005 19:54 |
Gerald Leitch wrote on Wed, 24 August 2005 11:51 | With the LE sytems, be it 001, 002 or M-box they are all set to record at 16 to 24 bits, sure all types of plugs operate at different depths but the LE depth is 24. Most waves stuff dithers internally down to 24 bits on LE as do others.If you send a link to show me an page stating LE does operate at 32 bit float then cool it's just I've yet to find one.
|
Not true. You can look in just about any waves plugin manual and read with little variation from plugin to plugin:
from the Waves Renaissance compressor manual | New to the processor is dithering for the final output. TDM output is dithered to 24-bit (fixed point); native output is dithered to 32-bit (floating point). In both cases the signal is "handed back" to the application being used. In TDM, the internal resolution is 56-bit fixed point; in native, it is 64-bit floating point.
|
The entire LE signal path is 32 bit floating point. LE is a "native" system. Every insert point throughout the system has 32 bit floating point I/O, and the stream doesn't get converted back to 24 bit fixed until after the last insert point on the master fader.
|
I really don't think we should continue hijacking dongles thread like this but if you insist... Waves L2 IS dithered to 24 bits in LE as is linear phase EQ and linear phase multiband if the dither control is on and 32 bit float Only if the DAW is 32bit float, this I have just read again from the pdf's. I'm still waiting to see a link describing Pro tools LE's bitdepth as 32 bit float? By no means am I questioning your authority on this matter, however I always like to see some form of official documentation on these matters I will admit if I am in the wrong and need corrected. cheers.
Gerald Leitch wrote on Wed, 24 August 2005 14:24 | I'm still waiting to see a link describing Pro tools LE's bitdepth as 32 bit float? By no means am I questioning your authority on this matter, however I always like to see some form of official documentation on these matters I will admit if I am in the wrong and need corrected. cheers.
|
Totally understood...really not trying to be argumentative or a hi-jacker on this end. Just trying to clear up a point of contention.
If you have the 6.4 Ref guide, check page 430, right hand column, 2nd paragraph, first sentence.
If you have th 6.9 ref guide: p. 473, left column, 2nd par., 1st sentence p. 534, the two bulleted items at bottom left of page p. 546, the section titled "24-bit Input and Output"
I think maybe these will sort the matter out for us.
Chris Cavell wrote on Wed, 24 August 2005 20:39 |
Gerald Leitch wrote on Wed, 24 August 2005 14:24 | I'm still waiting to see a link describing Pro tools LE's bitdepth as 32 bit float? By no means am I questioning your authority on this matter, however I always like to see some form of official documentation on these matters I will admit if I am in the wrong and need corrected. cheers.
|
Totally understood...really not trying to be argumentative or a hi-jacker on this end. Just trying to clear up a point of contention.
If you have the 6.4 Ref guide, check page 430, right hand column, 2nd paragraph, first sentence.
If you have th 6.9 ref guide: p. 473, left column, 2nd par., 1st sentence p. 534, the two bulleted items at bottom left of page p. 546, the section titled "24-bit Input and Output"
I think maybe these will sort the matter out for us.
|
Don't have the 6.4 ref guide just the individual pdf's for each plugin.I know what your saying though that the plugins are dithered to 24bit in a 24bit daw and 32 bit float in a 32 bit float daw with the dither off. But It's mainly about LE being 32 bit float that puzzles as I have not seen any docu regarding this, all I've seen is 24bit.
Chris Cavell wrote on Wed, 24 August 2005 20:39 |
If you have the 6.4 Ref guide, check page 430, right hand column, 2nd paragraph, first sentence.
If you have th 6.9 ref guide: p. 473, left column, 2nd par., 1st sentence p. 534, the two bulleted items at bottom left of page p. 546, the section titled "24-bit Input and Output"
I think maybe these will sort the matter out for us.
|
And on 6.7, there is a reference of it on page 455 2nd paragraph.
For those who don't have access to these, it goes as follows.
"Pro Tools TDM systems process all audio internally at 24-bit, and Pro Tools LE process internally at 32-bit, floating."
Gerald, here are links to the Pro Tools Reference guides for your viewing pleasure...check the areas I have pointed out to you in my previous post...this is the documentation you've been seeking:
http://akwww.digidesign.com/support/docs/69/Pro%20Tools%20Re ference%20Guide.pdf http://akwww.digidesign.com/support/docs/PT_6.4_Reference_Gu ide.pdf
Cheers for the links Chris, though already had the pdf. Checked it out and...oh well, I stand corrected I've had tools for years and use HD2 at studio and totally over looked that paragraph! Always thought internaly it was working at 24bit as Digi don't make an issue of this other than a small paragraph in the pdf. Thanks again.
Does'nt say whether it dithers output to 24 bit or truncates though?
I'm not sure what it does...
I don't think it dithers though...
If you were to bounce a sine wave, then import it back into the session, flip the phase of one, and play them back...if it dithered, shouldn't there be a little noise down around -140dB? In LE, if you do this, there isn't any (remaining ultra low level noise that is).
high guys, dont feel bad for highjacking, this is a topic im much interested in, i mean the avenue you goes went down. I must say im boggled now, i thought tdm was 48bit fixed. Im gonna give up on this. Its one big mind f@#k, bloody digidesign, they could market terds effectly with there cunningness, and skull duggery. So what was the final highjack result guys.
dongle wrote on Wed, 24 August 2005 22:38 | high guys, dont feel bad for highjacking, this is a topic im much interested in, i mean the avenue you goes went down. I musted say im boggled now, i thought tdm was 48bit fixed. Im gonna give up on this. Its one big mind f@#k, bloody digidesign, they could market terds effectly with there cunningness, and skull duggery. So what was the final highjack result guys.
|
Hi Dongle, turns out that P.tools IS in fact 32 bit float internal BUT 24 bit on the output. BUT wait...none of that still answers your orig question as to why the same wav in samp and P.tools sounds different so sorry man. I think it has to be the way both programs are written to handle the audio (i.e the algorithms) it's all I can guesstimate at this time. By the way have you tried the sonalksis 315 comp yet dongle? I think you'd like it
No i havent, il see if i can get a look at a demo, thanks for the tip. Hows the comp windup going man. With pt being 32 float and 24 at the output does that mean it would be a good idea to dither to the 24th bit, say when finished with a mix, put an insert in the master fader, dither set to 24 bits. Also is HD 48 fixed, cause didnt you say you read that it is something else. Thanks guys
dongle wrote on Wed, 24 August 2005 22:49 | No i havent, il see if i can get a look at a demo, thanks for the tip. Hows the comp windup going man. With pt being 32 float and 24 at the output does that mean it would be a good idea to dither to the 24th bit, say when finished with a mix, put an insert in the master fader, dither set to 24 bits. Also is HD 48 fixed, cause didnt you say you read that it is something else. Thanks guys
|
Nearly finished the comp test... if you mean when finishing a "mix" e.g a song, then in theory as the internal process is 32bit float then technically you should dither on output to 24bit - but this is only speculation as I am not certain whether pro tools dithers itself to 24bit or truncates.Chris said he tested this in LE without dither on the output but noticed no additional noise.So I would say it's not needed.
To Chris... "pro tools reference guide" page 473 para 2.
Quote" pro tools TDM systems processes all audio internally at 24 bit" pro tools reference guide - page 546.
quote" pro tools TDM systems support full 24bit audio input and output signal paths,with 48 bit internal mixing and processing!"
confusing?...
It's been a really long time since I kept up with the innards of TDM, but it is my understanding that it works like this:
The summing buss is entirely 48 bit (or dual precision 24 bit).
All the insert points have fixed 24 bit single precision i/o streams (which made the master fader inserts function as expected, they're after the fader, and in TDM this means that the signal path after the master fader is 24 bit fixed).
Now here's where it gets confusing...and digi hasn't quite wrapped their heads around it themselves yet.
RTAS plugins need a 32 bit floating point I/O...so when they're placed on a track's insert, they change that insert point and all those that follow it on a given track to 32 bit float (AARGH!!!), and convert back to 24 bit fixed at the output of the last insert point. This is why there are such huge restrictions regarding the placement of RTAS plugins in a TDM system...digi hasn't quite worked it out yet...but I'm sure they're doing their darndest to give the TDM guys just as much RTAS functionality as the LE guys.
When the audio comes out of the last insert point, it is passed as a fixed single precision 24 bit stream and placed in the middle of a dual precision 48 bit stream: the summing buss. this leaves some empty bits above, and some below, leaving a pretty large amount of headroom as well as usually way more than enough lower bits to prevent quantization distortion within the mix buss itself. The master fader then acts as a range selector across those 48 bits, choosing which range of 24 from within the 48 to send on to the master fader plugin section and/or on through to the converter.
Chris , thanks for communicating that so clearly. I was trying to explain it but wasn't doing a very good job. Didn't know the terminology, and wasn't confident enough to tell people they were worng, but this exact topic has gone around before. When I first starting using PT I was dumbfounded why you couldn't use RSTAS on the bus or Aux sends. It was that conversion reason from native to TDM that they didn't work out in earlier versions. I use 6.1 with mix plus. From what I remember on the DUC RTAS was implimented in 6.9 for HD users. Correct me if I am wrong.
Any computer tech will tell you, your native processes are at the native processing buss's speed. So when you were saying LE was 24, I wasn't buying it. With the advent of G5's and 64 bit pc's, we will see new software supporting 64bit float. Native looks better and better. I am interested to see how next generation TDM systems will work. I am sure they have 64bit cores on the way when people start looking to native, they bring that stuff to the forefront. Right now people are still buying HD systems. When the matket demands higher processing, the market will get it. 48 bit obviously works. There are plenty of high end digital hardware devices that use 48bit fixed.
As for the sound variance that started this thread, I say it again, the bus algorythms and sound are unique to software just like they are unique to any other mixer, beit analog or digital. The sound may be diferent from one app or hardware mixer to the other.
I can tell you from using Cubase 32 and then using Cubase SX years ago, I notice a marked improvement in my audio quality. They both processed at 32bit float and both recorded 24bit audio. The change was in the switch to the Nuendo mix architechture and audio engine. So when you say LE doesn't sound as good as a Samp bounce, I say leave it be and use Samp!!!!!!
Quote: | From what I remember on the DUC it was implimented in 6.9 for HD users.
|
Digi has managed to slowly increase the RTAS functionality within their professional lines. The "what can follow what" restrictions remain iirc, but you can now place rtas plugs on aux tracks (a big bold + for producers who use alot of softsynths and samplers). I don't think you can put them on a master fader yet.
FWIW the only native Waves plug-ins that dither to 24 bits or less are the L-1 and L-2. L-1 has been double precision since I think version 4 which was when 24 bit dither first became an option.
I've had a hunch for 5 years that TDM will be replaced by some kind of a native co-processor scheme that gets around the latency issues. TDM is amazingly long in the tooth and 56k assembler code is prohibitively expensive to write.
brett wrote on Thu, 25 August 2005 14:21 |
Any computer tech will tell you, your native processes are at the native processing buss's speed. So when you were saying LE was 24, I wasn't buying it. With the advent of G5's and 64 bit pc's, we will see new software supporting 64bit float.
|
Brett - It's important to note that buss speed is not the same as teh bit depth of an internal processing calculation routine. You can easily have 64bit floating point math calculations done with a 32bit OS. Theoretically the calculations will get done faster with the 64bit OS.
Best regards, Steve Berson
Bob Olhsson wrote on Thu, 25 August 2005 14:38 | FWIW the only native Waves plug-ins that dither to 24 bits or less are the L-1 and L-2. L-1 has been double precision since I think version 4 which was when 24 bit dither first became an option.
|
L3 also has these same dithering options.
Best regards, Steve Berson
true!
A lot of floating point processors operate at more than 64 bits too.
Gerald Leitch wrote on Wed, 24 August 2005 16:19 | Does'nt say whether it dithers output to 24 bit or truncates though?
|
I believe that that is the main difference between the PT dithered mixer and non-dithered mixer.
masterhse wrote on Sat, 27 August 2005 11:14 |
Gerald Leitch wrote on Wed, 24 August 2005 16:19 | Does'nt say whether it dithers output to 24 bit or truncates though?
|
I believe that that is the main difference between the PT dithered mixer and non-dithered mixer.
|
A word length reduction is a truncation, regardless of whether dither is applied or not. It's not either, I truncated or I dithered, being two opposing events, it's I truncated without dithering or I truncated with dither applied. A dithered word reduction is a truncation, because regardless of applying dither or not, you are still quantizing bit data.
Ronny wrote on Sat, 27 August 2005 11:44 |
A word length reduction is a truncation, regardless of whether dither is applied or not. It's not either, I truncated or I dithered, being two opposing events, it's I truncated without dithering or I truncated with dither applied. A dithered word reduction is a truncation, because regardless of applying dither or not, you are still quantizing bit data.
|
True Ronnie, I assumed that the original post was a shorthand way of asking does PT dither before truncation.
Shamless plug here but when I come across these type of threads, I find the people that might be interested in what I do. I am writing a book all about plug-ins and the DAW technology called "Plug-in Power: the Complete DSP Reference Guide" coming out this fall/winter.
< http://www.amazon.com/exec/obidos/tg/detail/-/1592009530/qid =1125588731/sr=8-2/ref=sr_8_xs_ap_i1_xgl14/103-6411458-44262 18?v=glance&s=books&n=507846>
I happened upon this thread doing research on pan law among other things and will have much information on this and other topics that come to the surface of the boards all the time. Hopefully my efforts will consolidate much of this information and will be helpful to one and all wishing to improve their DAW audio results.
To add to the topic at hand, an earlier post described the PT bit depth chain and was accurate for the TDM systems. The plugins are connected by 24-bit busses and the mix buss is a custom scaled 48-bit systems that is reduced to 24 bit for output to the master plugins and analog converters. This is as per Bobby Lombardi, who handles plug-in design for Digi. I am double-checking all my info for the book as I go along...
PT LE is a native app and runs internally at 32-floating point until final output which MUST be reduced to 24 bit fixed point in order to be converted to analog! This is something that gets confused often. 32-bit floating point files cannot be directly played back via analog converters. Something must quantize the file to 24-bit fixed point in order to work.
Both fixed point and floating point math involves some errors during typical mixing. These errors crop up at different times and with different effects. I'm sure those differences will be debated as long as possible out here on the boards. Both systems are capable of tremendous work and if understood, can be maximized for each systems potential while avoiding any weaknesses.
The question I would like to add is unless you are using dynamic or moving pans in your mix, do you think that the pan law will truly effect the outcome? I believe it won't. If a sound is panned in one spot, you will adjust your level accordingly regardless of the pan law setting.
Now, importing and exporting between DAW's that have differing pan laws will surely change the sound, but if the pan law does not change, how does that affect the mix for non-moving sound sources?
-ashley shepherd
|
|
|