1. Register NOW and become part of this fantastic knowledge base forum! This message will go away once you have registered.

TDM II

Discussion in 'Recording' started by Greg Malcangi, Jan 18, 2002.

  1. Greg Malcangi

    Greg Malcangi Member

    There seems to be a bit of confusion over whether anything has changed with mix bus and if it will sound better. I've gone over the docs in quite a bit of detail and here is what I've garnered:

    The most obvious improvement with TDM II is the drammatically increased number of time slots. However, it's not just the number of time slots but how they are managed that has changed. Looking at it, I think it's going to be almost impossible to max out with time slots, even if you are deliberately trying to!

    Basically Digi have slightly improved the mixer in a couple of places but it remains much as before. I can't say that I'm at all surprised or bothered by this. Despite all the rumours to the contrary there really is nothing wrong with the mix bus. The mix bus maths are provably correct and I don't see, and obviously neither do Digi, how you can improve upon that. The two changes that have taken place that could improve the quality of your mix are:

    1. If previously you had a lot of channels in your mix, which would require more than one DSP chip, then the output from one chip would be summed, truncated to 24bit and passed on to another chip. This truncation no longer happens and no matter how many channels you're running the whole thing stays at 48bit.

    2. The dithered mixer. My complaint about the old dithered mixer was that under certain circumstances summing together all that additional noise completely negated it's potential benefits. The new dithered mixer uses uncorrorelated (I think I OD'ed on the "r" key there) white noise and should in theory half the amount of perceptable noise that the dithering adds. I think I'll certainly be using the new dithered mixer more than I did the old. AFAIK, I am the one who bought this problem to Digi's attention. I mention this only because I would like you all to fully appreciate how clever I am!! :)

    Greg
     
  2. I''m a bit confused on the details of the first change that are described here.
    The posting states that in the old system, when using more than one DSP chip, the "output from one chip would be summed, truncated to 24-bit..."
    It goes on to note that in the new system, "This truncation no longer happens and no matter how many channels you're running the whole thing stays at 48-bit."
    Are you saying that, in both the old and new systems, the audio data that is recording at 24-bit is processed internally at 48-bit? I'm not quite sure how a signal recorded at 24-bit gets to be 48-bit and then needs dithered or, even worse, truncated.
     
  3. Greg Malcangi

    Greg Malcangi Member

    Hi DD,

    First off let me say that it looks like my interpretation of the first change may not be correct. The jury is still out but most likely there is no change in this regard from TDM I to TDM II. Change 2 is definately correct though. In the meantime I'll have a large fries and a McCrow burger! :)

    Are you saying that, in both the old and new systems, the audio data that is recording at 24-bit is processed internally at 48-bit?

    I explained this a while ago here on RO but I can't seem to find it so here goes again:

    Bear in mind that bit depth equates to volume resolution. Let's say you have a summing mixer (or mix bus) that is only 24bits wide and you have a 24bit track that's being output to that mixer. If the fader level of the track is at unity no problem but lower the fader level and your 24bit track is now lower and therefore outside the 24bit window of your mix bus. Anything outside of this 24bit window will get discarded (truncated). In other words if you were to lower the fader to half way between unity and infinity it would in effect only be 12bit resolution. The way that Digi solves this is to create a 48bit "window" so even if you pull the fader way down on a particular track you don't loose resolution. Great so far apart from one minor problem, there's no gear out there that can handle 48bit resolution files, everything is either 16bit or 24bit (occasionally 20bit). So to get your mix out of PT and into anything else it has get back to 24bit resolution. Using the standard 24bit otimised mixer in PT will truncate the 48bit to 24bit and using the dithered mixer will obviously dither it down to 24bit. Bear in mind though that it is mathematically impossible to directly hear this truncation as it happens in the -288dB to - 144dB range which is below the noise floor of any 24bit converter on the market. There is a possible theoretical argument that under certain conditions the truncation could cause quantisation errors that are just within the noise floor range. Very dubious though.

    In short the summing mixer/bus is 48bits but come off that bus, going to inserts for instance or external sends, and you're back to 24bit.

    Greg
     
  4. ckevperry

    ckevperry Active Member

    I have to admit I am surprised at the 24 bit truncations they left in TDMII. Now the math does not bother me, I think plugin quality(and overuse) has always been the PT "sound" everyone hates and not the mix bus.

    But still, Digi has to know they are going to catch hell for this from the less informed. I wonder how much they can tweak the mixer plugin with the new hardware?

    I have read on the DUC of a chap from England who has tested HD and says it sounds really much better. But he thought it was due to newer better converters and a much better clock....in his words "lowering THD and greatly improving imaging."
     
  5. Greg Malcangi

    Greg Malcangi Member

    Now the math does not bother me, I think plugin quality(and overuse) has always been the PT "sound" everyone hates and not the mix bus.

    I'm with you on this one, although I would add that obviously the sound going into PT from the converters has quite an effect. According to the math, individually plugins should sound better at 88.2kS/s or 96kS/s and they should sound better when you stack them up.

    Greg
     
  6. Dedric

    Dedric Guest

    Originally posted by Greg Malcangi:


    I'm with you on this one, although I would add that obviously the sound going into PT from the converters has quite an effect. According to the math, individually plugins should sound better at 88.2kS/s or 96kS/s and they should sound better when you stack them up.

    Greg


    The "sound" of PT mixing (or at least $$$ required to get it sounding good) was one reason I didn't get a PT rig a couple of years ago (Paris instead). Paris is now officially done, and so I'm looking elsewhere for 96k/surround - hence my interest in this thread.

    Looking at the manual, I too am a little confused if HD is really running 48bit busing or just a different arrangement of the TDM time slots and busing between chips. I am hestitant to believe the 48dB of headroom vs. the prior 30dB makes any audible difference as it really only affects (according to the manual) when clipping occurs.

    I would have thought a redesign would have retained full 48 bit fixed or even 56 bit floating point processing from start to finish with dithering at the final stage (but I guess that would be completely non-compatible with older projects). This would eliminate the need for dithering at each DSP/plugin, and leave it for the final output only (which could be a much cleaner sounding design, IMHO).

    I guess time will tell whether HD can stand up to the scrutiny and expectations for mix sound quality that seem to be floating around.

    Regards,
    Dedric
     
  7. Greg Malcangi

    Greg Malcangi Member

    Hi Dedric,

    Looking at the manual, I too am a little confused if HD is really running 48bit busing or just a different arrangement of the TDM time slots and busing between chips. I am hestitant to believe the 48dB of headroom vs. the prior 30dB makes any audible difference as it really only affects (according to the manual) when clipping occurs.

    Here are a few clarifications:

    All busing in PT is 24bit, the only exception is the summing bus which is at 48bit. The dithered mixer provides 30dB of headroom. Dithering only occurs at certain points, not neccesarily when going to a plug or DSP. For example let's say that you've got an audio track and you want to stick on an EQ plug. Your 24bit audio file is fed straight into the plug without any dithering because the audio file is only 24bit to start with. If the EQ plug processes the audio and creates information below the 24bit "window" that information is dithered back to the 24bit window. If nothing below the 24bit window is created, nothing is lost. However, if you instantiate a plug on the masterfader then the information is coming from the summing bus (48bit) and so has to be dithered to 24bit before being passed to the plug.

    None of this has changed, it's the same in TDM II as it was in TDM. Bear in mind that this dithering is occuring in the bit range from -25bit to -48bit, ie. From 1bit to 24bits below the least significant bit in your 24bit converter (-288dBFS to -144dBFS). Bear in mind also that the least significant bit on your converter (and probably one or two more significant bits) are just system noise. It is therefore impossible to hear these dithering operations.

    The only change in TDM II is that the dithering mixer now uses uncorrelated white noise. The effect of this is that the dithering process should now sound as if it produces about half the amount of noise of the previous (TDM) dithered mixer.

    Greg
     
Similar Threads
  1. jeronimo
    Replies:
    9
    Views:
    802
  2. Alécio Costa - Brazil
    Replies:
    14
    Views:
    2,316
  3. trock
    Replies:
    3
    Views:
    1,471
  4. vagelis
    Replies:
    0
    Views:
    801
  5. audiokid
    Replies:
    0
    Views:
    439
Loading...

Share This Page