new pro tools is not as good

[mightyduck]

Hi,

Not to be a downer or anything, but I see and hear a lot of spin, from Avid, and from others, about how the "new 32 bit floating point" mixer is an improvement.

It isn't.

One glaring example of the spin is where they claim "thousands of total dB" [or something like that] in so-called "headroom". Now if you add up all the dynamic range available for all the possible channels you would come up with some huge number. But that's useless and irrelevant. They are implying, with the usual "clever" wording, that you would have some purported benefit of thousands of dB of dynamic range on individual channels, which is not the case.

Floating point is inherently not as good for audio as is fixed point. Floating point is just more convenient for programmers. The only advantage floating point math has inherently is dynamic range. But you don't need dynamic range for audio applications, because the usable dynamic range for audio is small. What you need for audio applications is precision, and floating point is not as good as fixed point when it comes to precision. FACT.

Anyhow, HD was previously 24 bit, with a 48 bit fixed point mixer [56 bit accumulator]. That's better than a 32 bit float system. I know they say that parts of the path are 64 bit now, but its not all the parts that were 48 bit fixed before.

When / if they ever get it working in 64 bit float, then it will potentially be better than the old 48 bit fixed mixer. But for now I see no reason to "upgrade". What they have now is some sort of kludge intermediary thing going on in order to transition. That should be free, because there is really a less good mixer than there was before, just so that they can continue to accommodate RTAS and what-not in the interim.

Anyhow, hopefully they will get a true 64 bit system working.

[juhanni]

"that should be free"?

[Chris Lambrechts]

Originally Posted by mightyduck

Hi,

Not to be a downer or anything, but I see and hear a lot of spin, from Avid, and from others, about how the "new 32 bit floating point" mixer is an improvement.

It isn't.

One glaring example of the spin is where they claim "thousands of total dB" [or something like that] in so-called "headroom". Now if you add up all the dynamic range available for all the possible channels you would come up with some huge number. But that's useless and irrelevant. They are implying, with the usual "clever" wording, that you would have some purported benefit of thousands of dB of dynamic range on individual channels, which is not the case.

Floating point is inherently not as good for audio as is fixed point. Floating point is just more convenient for programmers. The only advantage floating point math has inherently is dynamic range. But you don't need dynamic range for audio applications, because the usable dynamic range for audio is small. What you need for audio applications is precision, and floating point is not as good as fixed point when it comes to precision. FACT.

Anyhow, HD was previously 24 bit, with a 48 bit fixed point mixer [56 bit accumulator]. That's better than a 32 bit float system. I know they say that parts of the path are 64 bit now, but its not all the parts that were 48 bit fixed before.

When / if they ever get it working in 64 bit float, then it will potentially be better than the old 48 bit fixed mixer. But for now I see no reason to "upgrade". What they have now is some sort of kludge intermediary thing going on in order to transition. That should be free, because there is really a less good mixer than there was before, just so that they can continue to accommodate RTAS and what-not in the interim.

Anyhow, hopefully they will get a true 64 bit system working.

Well - interesting post you made there mightyduck and I'm not even thinking of getting in a technical discussion with you on some of the stuff you are insinuating.

Mainly - I admit - because my technical knowledge of that stuff is simply not good enough to get into an educated discussion about it.

There have been tons of floating versus fixed versus bits discussions all over the internet. Most of them give me a headache by the end of page 27. But overall the consensus is that 32 float is the way forward.

That said - have you actually listened / used the system is 32 bit float ?
I personally find it a lot easier to stay out of clipping trouble for example

I think you underestimate what the extra headroom can do for you in the real world - not just on paper :

Here's an example I just found on youtube :

http://www.youtube.com/watch?v=Qt-EJhDDHUI

Chris

[Greg Malcangi]

Originally Posted by mightyduck

Not to be a downer or anything, but I see and hear a lot of spin, from Avid, and from others, about how the "new 32 bit floating point" mixer is an improvement.

Host based PT systems have always had a 32bit FP mixer, so unless you think over a decade is "new" you've got your facts a little mixed up. Actually quite a few of your statements are not quite accurate. For example, the TDM mixer is not 48bit throughout, most data paths are only 24bit, which can result in a lot of truncation (or dithering if you're using the dithered mixer).

I'm not going to argue the pros and cons of the 32bit FP mixer vs the fixed point TDM mixer, that's been well covered in a number of other threads on the DUC.

G

[panamajack]

Originally Posted by Greg Malcangi

...Actually...the TDM mixer is not 48bit throughout...

There is a KB article about this, apparently the issue with portions of the path dropping out of the 48 bit fixed created the impetus to make it 32 bit throughout.

There are also several threads on dither, and the summary impression is that dithering is less critical with 32 bit.

[mightyduck]

Originally Posted by juhanni

"that should be free"?

Well, its really just avid's side-step / "side-grade", meant to keep people on board the pro tools train while they get the 64 bit up to speed. Why should folks pay for that? Folks are just accommodating avid while avid transitions from tdm into the 21st century.

[mightyduck]

Originally Posted by Chris Lambrechts

Well - interesting post you made there mightyduck and I'm not even thinking of getting in a technical discussion with you on some of the stuff you are insinuating.

Mainly - I admit - because my technical knowledge of that stuff is simply not good enough to get into an educated discussion about it.

There have been tons of floating versus fixed versus bits discussions all over the internet. Most of them give me a headache by the end of page 27. But overall the consensus is that 32 float is the way forward.

That said - have you actually listened / used the system is 32 bit float ?
I personally find it a lot easier to stay out of clipping trouble for example

I think you underestimate what the extra headroom can do for you in the real world - not just on paper :

Here's an example I just found on youtube :

http://www.youtube.com/watch?v=Qt-EJhDDHUI

Chris

Yeah, I've listened to a gazilion 32 bit float host based systems. They are all basically the same since they are based on apple's 32 bit scheme, which is not really designed for audio, but, rather, for general computing. The "new" system is no different in that respect than the same old thing they've been using in pro tools le since magellan's time.

Respectfully, that you find it "easier to stay out of clipping trouble" in a 32 bit system is not a valid reason. That's just about basic engineering skills. Its not difficult to keep from clipping if you understand and use proper gain staging. Even 24 bits is a huge dynamic range, far beyond what humans can tolerate. Also, using 32 bits' extended dynamic range to "go way into the red" without clipping will inevitably create other problems. That's one reason why the metering on host based systems still reads as if it were a fixed point system.

[mightyduck]

Originally Posted by Greg Malcangi

Host based PT systems have always had a 32bit FP mixer, so unless you think over a decade is "new" you've got your facts a little mixed up.

I don't have my facts mixed up. I know 32 bit float is old technology. That makes my point. It is known to be problematic.

64 bit float, done right [and we have yet to see if avid will do that], can potentially outperform 48 bit fixed. But we're still waiting for that in pro tools land.

Originally Posted by Greg Malangi

Actually quite a few of your statements are not quite accurate. For example, the TDM mixer is not 48bit throughout, most data paths are only 24bit, which can result in a lot of truncation (or dithering if you're using the dithered mixer).

The mixer plug-in is in fact 48 bit throughout, with a 56 bit accumulator.

Nobody stated that the entire data path was 48 bit. It doesn't have to be. Moreover, the audio will have to be handed off as 24 bit fixed to the interface in any event.

Originally Posted by Greg Malangi

I'm not going to argue the pros and cons of the 32bit FP mixer vs the fixed point TDM mixer, that's been well covered in a number of other threads on the DUC.

G

It has been shown definitively that 48 bit fixed beats 32 bit float. The TDM fixed point system, including the 48 bit fixed point mixer is superior in the hands of someone who actually understands and knows how to use it.

The only advantage of 32 bit float systems has been cheap, easy programming and the fact that its kind of "idiot proof" for novice "engineers" [who tend to have a hard time with proper gain staging].

[mightyduck]

Trust me. The current incarnation of Pro Tools is a "side-grade" of sorts, for the purpose of keeping everyone on board while they get the 64 bit stuff worked up.

Look at the way they marketed it.

Sonically, the "new" thingy is not as good as a big TDM system was.

Hopefully the real 64 bit version will kick. But I see no reason to upgrade until that happens. You can still use the new interfaces and the software will work with the old cards for the moment, right?

[mightyduck]

Originally Posted by panamajack

There is a KB article about this, apparently the issue with portions of the path dropping out of the 48 bit fixed created the impetus to make it 32 bit throughout.

Well, the fact that parts of the path were not 48 bit was impetus for making the whole thing 64 bit. But the current 32 bit scenario is just a place-keeper while the 64 bit version is being prepared.

Originally Posted by panamajack

There are also several threads on dither, and the summary impression is that dithering is less critical with 32 bit.

I don't know what you mean by the termnology "summary impression", but its not that dither is "less critical" or "less necessary" with 32 bit float, its that its "less possible". In fact its not possible at all to really dither floating point.

So they probably tried to spin the issue by leading people to believe the ridiculous idea that its "less necessary".