Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[trock]

Hi All

I have been reading some of the posts here and fingding alot of interesting info. one thing i am curious about is this

I used to use Cubase SX 3 on a MAC G5 and before that a roland 2480. i am only a home hobbyst with this stuff, no school etc, so my technical knowledge is limited. although i read alot of forums

anyway, i now use SAW Studio on a PC and it is written in 64 bit fixed? to me, it has a better, deeper, richer sound than my cubase got me. however i have been told this is not possible and that SAW is just another DAW.

So is 32 bit floating different in the end than a 64 bit fixed program?

can someone tell me if i am just hearing something i am supposedly not?? does how it is written have anything to do with sound quality? i am using the same guitars, i am the same vocalist, in the same room. i did swap an avalon M5 for an API 512 pre but that is about the extent of the changes. and i got rid of a UAD-1 studio pak and WAVES bundle for just some stock SAW plugs now for my verbs, comp, EQ, etc. so to most it seems i "downgraded" my plug quality.

anyway i was just curious. thanks much

[Ronny]

trock wrote on Sat, 13 August 2005 14:05

Hi All I have been reading some of the posts here and fingding alot of interesting info. one thing i am curious about is this I used to use Cubase SX 3 on a MAC G5 and before that a roland 2480. i am only a home hobbyst with this stuff, no school etc, so my technical knowledge is limited. although i read alot of forums anyway, i now use SAW Studio on a PC and it is written in 64 bit fixed? to me, it has a better, deeper, richer sound than my cubase got me. however i have been told this is not possible and that SAW is just another DAW. So is 32 bit floating different in the end than a 64 bit fixed program? can someone tell me if i am just hearing something i am supposedly not?? does how it is written have anything to do with sound quality? i am using the same guitars, i am the same vocalist, in the same room. i did swap an avalon M5 for an API 512 pre but that is about the extent of the changes. and i got rid of a UAD-1 studio pak and WAVES bundle for just some stock SAW plugs now for my verbs, comp, EQ, etc. so to most it seems i "downgraded" my plug quality. anyway i was just curious. thanks much

You have too many altered factors to determine if the sonic change is only from the increase to 64 bit. You'd really have to A/B raw tracks, recorded through the same mics, pre's and played back through the same DAC's. I can't tell any difference between my 32, 48 or 64 bit processors. I have a TC reverb that processes at 32 or 64 bit and I can't hear any difference when A/B'ing those two different depths on the same processor. Be sure to keep your mind open and realize that the weakest point in any non-blind audible evaluation is the auditory cortex.  

[trock]

thanks ronny

yeah i figured since it was def A/B it was a worthless comparison. i think since SAW works better for me as an envirnment and all the other factors that come with that package that i like may be influencing me vs the steinberg experience (comapny) i have had etc??

who knows

i was just wondering about the math and if it made a difference?

thanks

[Ronny]

trock wrote on Sat, 13 August 2005 16:55

thanks ronny yeah i figured since it was def A/B it was a worthless comparison. i think since SAW works better for me as an envirnment and all the other factors that come with that package that i like may be influencing me vs the steinberg experience (comapny) i have had etc?? who knows i was just wondering about the math and if it made a difference? thanks

Well you do have more calculations available with higher bit depths, thus making some processes more accurate, however I haven't found the difference to be audible on my gear. It's like we have to reach a point somewhere along the line where increasing bit depth isn't going to show any audible improvement. Like taking a staircase with 10 steps, cutting them in half to make 20, they both still get you to the top, the end product is the same. Do we really need to take twice as many steps, there's twice as much energy expended to get to the final landing and when we are thrown off our typical pace. My digi consoles mix busses operate at 32 bit, the internal eq's have a 56 bit accumulator, they don't sound any different than my older consoles that had the 44 bit accumulators, so I'm not convinced that going above 48 bit makes much difference on the end product.  

[trock]

makes sense

that is about the best description i have read by the way

and i meant to say since they were NOT accurate A/B comparisons, sorry to confuse there

anyway, thanks for the help and answers!

now i need to stop thinking about binary math and thin about some music

[Ralf Kleemann]

There are several lenghty threads in the R/E/P forums about this... in a nutshell, the differences between 32 bit floating and higher bit depths are rather minute, but become a little more apparent whenever the DAW is doing something reasonably complex. This can be something simple like summing a 20 track mix to a stereo sum. Personally, I've done a comparison between summing in the 32 bit (floating) domain and an external 80 bit (fixed) mixer, and while I believe to hear a difference, it is almost imperceptible for my ears.

Best regards, Ralf

[trock]

Hi ralf

thanks for that, cool you did that test. i honestly thought i heard a noticeable difference in quality but it could have actually been a bunch of things, not the least of which was my frame of mine with SX 2 on OSX and all the issues i was having.

anyway, interesting stuff for sure.

[cgc]

I just took a look at the plugin API for SAWStudio (some nasty stuff), and it states that the app internally uses 32 bit integers.  They say that you can convert the ints to floats and process that way, but that is a really bad idea if you want decent performance.

[Nika Aldrich]

Just thinking through it I have to think that 64 bit fixed would provide more accuracy than 32 bit float over the course of a mix, but the difference would probably be pretty negligable depending on what was happening and how it all was implemented.

Nika

[Ronny]

Nika Aldrich wrote on Mon, 15 August 2005 07:49

Just thinking through it I have to think that 64 bit fixed would provide more accuracy than 32 bit float over the course of a mix, but the difference would probably be pretty negligable depending on what was happening and how it all was implemented. Nika

Fixed int 64 bit would allow double precision 128 bit, but you've hit the nail on the head, processing versus audible differences. No doubt that there is more precision with 64 bit, but sonically little difference if the processes are printed. Stability wise running real time, it seems that 64 bit will lessen the amount of tracks, it will certainly double archive space if you stay at 64, if nothing else. There are some pretty good reasons for not recording over 24 bit that D Lavry has written about, I wonder when we'll reach the point of no return with processing at higher rez, maybe we already have.

[Eric Bridenbaker]

cgc wrote on Sun, 14 August 2005 18:43

I just took a look at the plugin API for SAWStudio (some nasty stuff), and it states that the app internally uses 32 bit integers.  They say that you can convert the ints to floats and process that way, but that is a really bad idea if you want decent performance.

Could you explain this a bit? I took a look at the API .pdf file, but not being a programmer, couldn't really find the part you are referring to.

Are you saying that any plugin instantiation in SAW (Native API or otherwise)  subjects the audio data to 32 bit integer math at some point?

Bob Lentini, the designer for Sawstudio gives a description of the internal math here:

             http://www.sawstudiouser.net/forums/showthread.php?t=600& ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;amp ;highlight=internal+depth

According to his post, all multiplies and divides including those of the the strip compressors/gates is done with 64 Bit integer math, with the exception of the filters in the EQ's, which use some 32 bit processes.

He talks about 32 Bit DWORDS. What exactly is that?

Third party plug developers for the Sawstudio API such as Sonoris use 64 bit integer math internally (once again with the exception of filters).

The plugin ports must have some 32 bit code (floating point?) in order to achieve compatability with VST and some directX plugs. In the link above Lentini cautions against "trashing" the audio by using third party algorithms that go back and forth between fixed and float too many times.

Any info on this would be most helpful.

As for the sound, one must make their own call in the end using their own ears. IMHO Sawstudio may very well be the most sonically pure software on the market today. I immediately noticed a considerable authority to its sound above the 32 float apps.

Cheers,
Eric

[trock]

interesting stuff

to me SAW did sound much clearer and richer if that makes any sense etc. but again i was slammed for saying that so i thought i would ask about the math behind all this and just see what some though or knew.

[doogle]

Hi guys, i would very much appreciate if somebody out there could tell me weather Protools Hd 48 bit fixed will give better results than Le 32 Floating, im considering getting hd, but if the results are neglible ill spend that excessive amount of money on converters and speakers. Thanks all

[Timeline]

I wonder if anyone here can explain how an app like Nuendo or SX can play SDII audio files without converting them and how differing floating 32 bit schemes, one app to another like DP to N3,  may effect sound.

Don't think this is too OT.

[Nika Aldrich]

doogle wrote on Fri, 19 August 2005 02:31

Hi guys, i would very much appreciate if somebody out there could tell me weather Protools Hd 48 bit fixed will give better results than Le 32 Floating, im considering getting hd, but if the results are neglible ill spend that excessive amount of money on converters and speakers. Thanks all

The difference between LE and TDM is not in the sound resulting from the difference in math.  It comes from many other places.  In the end they very well may sound different but it will have almost nothing to do with the 32 bit floating vs. 48 bit fixed issue.

Nika

[cgc]

Eric Bridenbaker wrote on Mon, 15 August 2005 23:02

[Are you saying that any plugin instantiation in SAW (Native API or otherwise)  subjects the audio data to 32 bit integer math at some point? He talks about 32 Bit DWORDS. What exactly is that?

32 bits is the resolution of the internal storage.  That is the governing factor of any audio application.  Each process starts with 32 bits and ends up with 32 bits to pass to the next process.  You can do 123 billion bit math on that 32 bits but at the end the result is still 32 bits.

Basically, the resolution of the answer it what matters.  Audio DSP typically produces a lot of answers which are fed into the next set of problems, so it makes some sense to have a storage size that ends up with enough bits to have a useable answer.  If you like the answers that software gives you and don't feel the need to question them, then don't worry about it.

[cgc]

Timeline wrote on Wed, 24 August 2005 08:32

I wonder if anyone here can explain how an app like Nuendo or SX can play SDII audio files without converting them

Without converting them to what?  Another file format or 32 bit float or ???  Odds are that any sound file you play in those apps are converted to whatever the internal storage format is, and that's probably floats.  Some apps have preview features that might send the integers right to the audio HAL (ASIO, CoreAudio) without conversion.

[Timeline]

Quote:

Some apps have preview features that might send the integers right to the audio HAL (ASIO, CoreAudio) without conversion.

Yes sgc I think that sounds like what may be the process but I have problems using DP SDII audio files in Steinbergs app Nuendo.

Here's what I tried.
In DP4.51 I create a non embedded OMF of my 40+ 96K audio tracks and simply open this OMF in Nuendo.  Nuendo asks for the location of the folder on first open then plays the OMF and the SDII files. The Nuendo project audio files folder still shows no audio files at all so no conversion.

Problem is, I find the sound rounded and poor when playing the non embedded OMF this way.  I doubt it's Pan law issues. I also wonder about 32 bit float processes being different in each app causing what I hear.

Would this be as you mentioned 'preview' mode and would the DAW manufacturer expect you to stay in this mode throughout the production or convert the SDII's in the end to aif or BW? Also, is more CPU headroom used while playing them rather than converting the audio files to a more native format for the app?

This is a big curiosity for me.

[Chrunchy]

whats a bit?