Skip to main content

With another thread brought up regarding the sh*tty sound in pro-tools...I was left thinking today as I sat at a contol-24 - Playing with the 7k mouse, I started to dwell upon the premise of that the only sound of pro-tools (other that converters and plugs)is the summing math of the software. Basically it is understandable that the internal summing of tracks is represented in a way that can translate harshly back to analog again.

Some seem to think other formats sound better - How could this be, granted the conversion can't be night and day, plus people use the "A" name converters as if they want to make their rack look like a big Barney to keep their kids busy. So how could this be? With the plugs and converters set aside, all that is left is the math of the summing - the same thing with conversion to an extent - yet this time it is in the software.

Lets think about how a fixed-bit summing bus works. Even without knowledge of binary code, or word length or bit depth, just follow my basic analogy. Each track is a bunch of groups of numbers used to decribe the sound. Think of them as different issues of a comic book series in a stack. You have seperate stacks each devoted to the different characters in the storyline ( representing the instruments in the mix). Yet all of these stacks must now fit in the place of one stack ( the completed storyline, so two stacks for stereo, part 1 and 2 ). You now have to fit all of your seperate stacks into the two final stacks - now imagine the closet you are using for storage has a fixed shelf halfway up - and eventually you run out of room, requiring you to press down on the comics - sqweezing, stomping, pushing - until they can all fit. As you wipe the sweat from your forehead, you realize you succesfully achieved your goal - yet you also degraded the quality of the comics - many of them are now wrinkled, ripped, even crushed! Geez, they were all in great condition when I was collecting them, and now that I've put them all together they have turned to crap - oh shucks, my investment, down the drain - waaaaaaaaahhh! If only you could have raised the shelf in the closet a tad or put some comics on top of the shelf, everything would have fit within a reasonable about of space, and everyone would be happy. Unfortunatly you had a fixed-but summing bus for a closet, you did get everything into it, but things got kinda f**ked up at the same time.

This is where the whole 32-bit floating-point thing comes in...

With 32-bit floating point math, you are given an extra 8-bits or word lenght. These extra 8-bits function as a headroom buffer. If you were to exceed the headroom of the original 24-bits, the extra 8 are used to describe the "over" rather than just reperesnting it with harsh digital distortion. This translates to a huge amount of virtual headroom. As In our comic book / closet storage analogy, the extra 8-bits would be a box in the closet sitting above the top shelf - storing the comics that were a bit too tight to fit in. Now set aside to be called upon when needed - rather than having them crammed in (distortion).

This increase in summing-bus headroom means no more harsh or distorted summing bus. It is a method that maintains accuracy at lower levels, and has excess headroom to handle higher ones.

Now when it comes to the headroom of an analog desk - I can't get all teckie with you, but I do know that It can overload at a certain point - yet of course sound is reliant upon circuty. The point of this post was to shed some light on why some DAWs don't sound great for mixing - and also that there are others using different math that sounds better, as in don't give up on digital mixing so fast.

anyone else care to touch upon floating point - or the headroom of the bus they sum with?

Comments

anonymous Thu, 08/23/2001 - 01:32

I've been using both PT TDM and LE softwares and as far as I know they're based on these two different technologies that you mentioned (fixed/floating).

What I have noticed is that using TDM could sometimes be painful because you have to be really careful to not overload any buses or plugins. Whereas in LE you can have little peaks here and there and sound is still good. Of course you can't push to hard on LE neither. But you don't have to mind if there's couple of peaks that goes over.

When I have translated LE sessions to TDM, the first thing I sometimes hear is overloads in some plugins. Therefore it's quite interesting that cheap LE version works much better than pricey TDM. Is future versions of TDM still based on fixed-bit?

(sorry my english :o )

Guest Thu, 08/23/2001 - 01:43

hmm, yarz I have heard Digi witchdoctors say the RTAS CAN sound better (its the maths) than TDM.

I have 2 x TDM rigs so I am 'stuck with it' so rather than carp about it's quality, my interest lies in getting optimum performance with PT.

On that subject, there is a school of thought that uses Waves or other 'high math' plug ins to control a channels volume rather than the Digi Fader. Greg M the moderator / resident Expert of Recording Org's Pro Tools forum, says (go look) this is valid ONLY for folks who patch out to mix on an analog desk (so the moderator of this forum might like to try it!) , but is of no use for folks like me who mix internaly,

Anyone got a view on that? I was using that method, but would quit if it's bogus...

And that RTAS plugs perform slightly better at this task too?

Thanks,

Jules

anonymous Thu, 08/23/2001 - 03:58

Have you guys tried the Dithered mixer.
I thought it was just hype, it takes a bit more DSP. Although I can hear the diff. I do not know why it just sounds better to my ears.I do not think it works with LE or PC PT.

I have a control24 too, just cannot get out of doing everything with the bloody mouse/keyboard.
Shit i am still running 20bit converters, I will upgarde later spending the money on better mic pres/mics etc. Also holding off as it will be 192k or 96k in 6 months time.

Hope you are all fine

anonymous Thu, 08/23/2001 - 04:30

Jeez Come on guys, This is the mixer forum. Lets talk about mixing. If I see another conversation on bits and 32 point floating algor whats you might call it I might just throw up. It really doesn't matter what media/tape you work with. I thought we are just trying to do the best with what we are presented with. If you want to get dweeb then I suggest the moderators open a panel especially for you. Julian to be moderator of the "I've got a bigger byte than you" forum
Regards,

Attached files Image removed.

Guest Thu, 08/23/2001 - 10:22

Er... back to the point of the starting thread then eh?

Internal digital architecture is ALL about mixing. in (or just) digital. (internaly or via an analog desk)

buried in the above posts are even tips on how to somewhat cure what ails Mixerman when mixing from PT masters. (the dreaded fader pull down)

P.S. I do use the dithered mixer simply out of fashion, voodoo belief and support for the folks that were begging digi for it.. I haven't yet had time to stop and compare it to the old one.

Jules

Aaron-Carey Thu, 08/23/2001 - 11:42

Were having a discussion in the Sonic Foundy Vegas forum right now along similar lines. This is a totally different product than PT (most would say inferior), but these internal summing problems seem to plague all aps.
I have duplicated one of the posts here at DAW world in the sf forum.

At the SF site I have documented many tests and results regarding phase shift, digital clone ability, and the nature of garbage in garbage out operation. I have tried the same things in Vegas, Samplitude, Cube-Endo, PTFree, Sonar and Logic, with near similar results. But no real answers.

I am starting to thing that the thinking behind the summing math itself is flawed. The designers talk about " additive summation" as a way of summing. In some cases this means that the values of each individual sample are simply added together, and the summ is the result ( or sum ). You culd QUICKLY run out of headroom here. Others use a process of adding all the individual samples together, then dividing by the number of active channels/plugs/etc...

I feel that this may NOT be how sound is summed in our ears. And most likely also not how sound is summed in an analog mixing console. I dont think they just add up like this. i think there is a very complex interraction between frequencies, volumes and phase, possibly even some masking effects.

I know I use cheap ass apps so most wont listen to me, but I think this is relevant to all of us. Anyone have a better understanding on how sounds are summed in the real world ?

anonymous Thu, 08/23/2001 - 14:29

I have an expanded mix plus system and do all my mixes "all in" as does Jules. I am using the 24 bit dithered mixer. The internal processing in Pro Tools is 48 bit. This math is then truncated or dithered to 24 bit at the out put of the mixer, depending on whether or not you use the new dithered mixer. I think it sounds pretty good, depending on what YOU the engineer decides to do.

But you guys only want to compare it to one or two MILLION dollar consoles and recorders. I think with music the main things that affect the sound are: who played it, who sang it, who recorded it, and who mixed it. To say that it's always just a pro tools problem is very short sighted.

Also, in the discussion about how much more headroom you get from 32 bit processing versus 24, mention was made that a few overs won't make much difference with 32 bit processing. Bit depht doesn't affect how loud your music can be before clipping. It changes the low level detail resolution. If you have distortion from overs, bit depht won't change things much.

Mixerman Thu, 08/23/2001 - 20:53

Originally posted by Steve Shepherd:
I have an expanded mix plus system and do all my mixes "all in" as does Jules. I am using the 24 bit dithered mixer. The internal processing in Pro Tools is 48 bit. This math is then truncated or dithered to 24 bit at the out put of the mixer, depending on whether or not you use the new dithered mixer. I think it sounds pretty good, depending on what YOU the engineer decides to do.

But you guys only want to compare it to one or two MILLION dollar consoles and recorders. I think with music the main things that affect the sound are: who played it, who sang it, who recorded it, and who mixed it. To say that it's always just a pro tools problem is very short sighted.

I've compared it to $5,000 2" machine through a $20,000 console, which is the equivelant in price to the system.

As a mixer, I am beholden to the quality of the recording. Fortunately, as I've said before, a great performance and a great song will transcend a mediocre recording medium. Given the choice I'd pick the great song and performance, over the great recording on a great sounding machine, that's a no-brainer. But in my travels, having a great recording to go with the great song and performance usually commands the best results.

Your milage may vary.

Mixerman

anonymous Thu, 08/26/2010 - 21:38

32 bit floating point summing

MicrophoneMan, post: 39352 wrote: With another thread brought up regarding the shitty sound in pro-tools...I was left thinking today as I sat at a contol-24 - Playing with the 7k mouse, I started to dwell upon the premise of that the only sound of pro-tools (other that converters and plugs)is the summing math of the software. Basically it is understandable that the internal summing of tracks is represented in a way that can translate harshly back to analog again.

Some seem to think other formats sound better - How could this be, granted the conversion can't be night and day, plus people use the "A" name converters as if they want to make their rack look like a big Barney to keep their kids busy. So how could this be? With the plugs and converters set aside, all that is left is the math of the summing - the same thing with conversion to an extent - yet this time it is in the software.

Lets think about how a fixed-bit summing bus works. Even without knowledge of binary code, or word length or bit depth, just follow my basic analogy. Each track is a bunch of groups of numbers used to decribe the sound. Think of them as different issues of a comic book series in a stack. You have seperate stacks each devoted to the different characters in the storyline ( representing the instruments in the mix). Yet all of these stacks must now fit in the place of one stack ( the completed storyline, so two stacks for stereo, part 1 and 2 ). You now have to fit all of your seperate stacks into the two final stacks - now imagine the closet you are using for storage has a fixed shelf halfway up - and eventually you run out of room, requiring you to press down on the comics - sqweezing, stomping, pushing - until they can all fit. As you wipe the sweat from your forehead, you realize you succesfully acheived your goal - yet you also degraded the quality of the comics - many of them are now wrinkled, ripped, even crushed! Geez, they were all in great condition when I was collecting them, and now that I've put them all together they have turned to crap - oh shucks, my investment, down the drain - waaaaaaaaahhh! If only you could have raised the shelf in the closet a tad or put some comics on top of the shelf, everything would have fit within a reasonable about of space, and everyone would be happy. Unfortunatly you had a fixed-but summing bus for a closet, you did get everything into it, but things got kinda ****ed up at the same time.

This is where the whole 32-bit floating-point thing comes in...

With 32-bit floating point math, you are given an extra 8-bits or word lenght. These extra 8-bits function as a headroom buffer. If you were to exceed the headroom of the original 24-bits, the extra 8 are used to describe the "over" rather than just reperesnting it with harsh digital distortion. This translates to a huge amount of virtual headroom. As In our comic book / closet storage analogy, the extra 8-bits would be a box in the closet sitting above the top shelf - storing the comics that were a bit too tight to fit in. Now set aside to be called upon when needed - rather than having them crammed in (distortion).

This increase in summing-bus headroom means no more harsh or distorted summing bus. It is a method that maintains accuracy at lower levels, and has excess headroom to handle higher ones.

Now when it comes to the headroom of an analog desk - I can't get all teckie with you, but I do know that It can overload at a certain point - yet of course sound is reliant upon circuty. The point of this post was to shed some light on why some DAWs don't sound great for mixing - and also that there are others using different math that sounds better, as in don't give up on digital mixing so fast.

anyone else care to touch upon floating point - or the headroom of the bus they sum with?

I've been mixing inside Digital Performer since version 2.7, back in 99. Then it was 16bit, due to many reasons (mainly track count). 24 bit improved the situation quite a lot, rivaling the quality of my studio's console (modified Soundtracs Solitaire).
I'm doing 32b fp nowadays, and I feel that I work more easily, ie, I find myself realizing after the fact that the mix went on effortlessly. Never a stop to lower every track by 3db to make room, more clear sound, more space, better separation, clarity, etc. I'm hooked. The difference may be subtle for most people, but I work better and like the result a lot.
The original audio is 24b, but all the processing has extra room to deploy the added harmonics and textures.

audiokid Thu, 08/26/2010 - 21:54

Cool to see this thread reopened. Welcome to RO nicog!
Mixerman is using the Dangerous 2-Bus and claims it to give even more space. I am following the hybrid DAW crowd. Good to hear your POV and experience. I still have an old Pro Tools Mix plus system sitting collecting dust and a new hybrid DAW almost ready to turn on. Are you into in any analog summing systems or are you all ITB at this point?

anonymous Thu, 08/26/2010 - 22:30

ty, audiokid, and cheers to all. Actually I toy with the idea of analog summing every now and then, but I'm staying inside for practical reasons. As a rationalization, I tell myself that I avoid da-ad as much as possible. I also work in 44.1 instead of 48 (nasty math, lots of rounding up). If I had a faster system and bigger drives I would use 88.2 instead of 96 (no decimals in the conversion).
So for now, I'm, as u said, all ITB, but I suspect Mixerman is right. We all know what good analog does. Boy, there's a TL Audio tube mixing desk that makes me drool. That would be stem mixing, not summing, tho, but... yummy