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

Dithering?

Discussion in 'Recording' started by Threeskinmusic, May 5, 2005.

  1. Howdy guys, although I know what dithering is, I am not sure if most people use it or not. Here is my situation. I record into Cubase at a high bit rate, mix and export at that same bit rate. I will then open the file in Wavelab, do some final touches, and render that to a 16 bit file to burn to a cd. Should I actually be doing any dithering instead of just saving it down to a 16 bit file? Am I losing anything by not using the dithering section of Wavelab? As always, thanks for your help on this.
     
  2. David French

    David French Well-Known Member

    You should be applying dither when going from 24 to 16. This will give you an apparent resolution of 18-20 bits on your 16 bit CD.
     
  3. Thanks David, would I actually hear a difference in the sound?
     
  4. David French

    David French Well-Known Member

    Try and see. With decent monitoring and dynamic music, you probably will.
     
  5. karbomusic

    karbomusic Active Member

    To expand on what David says....

    What he says is correct. In Wavelab you do your dithering by using the "Render" option in the Master Section. Either use the built in dithering that wavelab supplies or others that you may have installed via plugins.

    Also, I would suggest rendering/dithering to a new file (an option in Wavelab) leaving your original mixdown from CBase intact for later mastering experiments etc. Its always nice to keep an untouched master file right out of your multitrack application. Also, only use 16bit versions where you need 16bit versions like a CD IMHO.

    There are also some decent explanations/suggestions in Wavelab's help files on dithering best practices etc.

    And yes you should hear a difference, especially dynamic range wise (unless you used lots of compression on the mix or it is severely lacking in dynamics) as you have just theoretically cut it in half.... from 32 to 16 bit.

    Best regards-

    Karbo
     
  6. dpd

    dpd Active Member

    David - I'm inclined to disagree, here. Care to explain further, please? Dithering will maiximize the linearity of the 16 bit output and is very important.
     
  7. David French

    David French Well-Known Member

    Sure dpd,

    Dither allows you to encode values that are below the least significant bit because that signal modulates the dither signal such that its average value over time is not random, but instead is that of your low level signal content.
     
  8. dpd

    dpd Active Member

    David

    The dither, when added to the signal, occasionally raises the next value to above the value for the LSB , thereby causing the LSB to toggle. Certainly, the improvement in resolution is highly dependent on the dithering signal but I still have a hard time believing one can increase the resolution down that many bits.

    And, since the signals are transient in nature, the resolution increase would be minimzed for these and maximized for longer-term signals, as I see things.
     
  9. David French

    David French Well-Known Member

    I'm honestly not the one to prove it to you rigorously; i'm simply sharing what I have learned.

    That's an interesting point about transient signals...hmmm...
     
  10. Nika

    Nika Guest

    Paul,

    Dither maintains the infinite dynamic range of which a digital system is capable, though much of this range is below the noisefloor.

    Think of the following for a moment:

    Take a digital 24 bit data stream of pure DC - all zeros. Now let's add some data only to the 24th bit in the stream in some sort of pattern, say, 1 1 0 0 1 1 0 0, etc. Now generate random numbers that are 9 bits in depth and of triangular probability distribution. Add these numbers to bits 16 through 24 of your data stream. Now lop off bits 17-24.

    If your data stream is long enough and if you use random enough numbers, you can do a mathematical analysis on the resultant 16 bit file and indeed find the 1 1 0 0 1 1 0 0 pattern repeated in the 16th bit, though with some random behavior added. A Discreet Fourier Transform on the results will show this behavior manifested in the 16th bit. The behavior, however, will be significantly masked by the noise, of course, because the amplitude of the random behavior that dwarfs it is 48dB higher, but the signal is still in there.

    And the same would hold true if we toggled the 48th bit and then added dither - or ANY bit and then added dither. Dither effectively brings that data forward to the least significant remaining bit. Without dither all of that is lost. This is why dither is imperitive - it maintains ALL information in the digital system - even below the LSB - though that information gets increasingly burried beneath the noisefloor the further back the information is.

    Nika
     
  11. Nika

    Nika Guest

    Regarding this, remember that all transients get filtered and end up with "ripples" bleeding out in one or both directions from the initial impulse. This exposes any transient for what it really is - a "longer term signal" that will be as affected by dither as any other.

    Or we can try this another way - if the transient is seemingly not carried forward by the dither to the remaining LSB, that's merely because the instantaneous amplitude of the noise at that moment in time dwarfed it, which will happen on occasion and randomly. The transient is still there and we have to consider that it is still represented by the LSB, just with noise that seemingly eradicates it.

    Nika
     
  12. David French

    David French Well-Known Member

    I was hoping you'd show up Nika! Thanks for saying what I couldn't.
     
  13. FIMseth

    FIMseth Guest

    OK quick question.....My dither plug-in has 2 options: 24-bit dither and 16-bit dither. If I wanted to use dither when converting from 24-bit to 16-bit which one would I use, the 24 or the 16-bit? I using DP 4.12
     
  14. Nika

    Nika Guest

    16 bit.

    24 bit dither is for use when producing a 24 bit result - in order to accommodate for the fact that the internal bit depth of the software operations such as mixing and plugins is 32 bit floating point. The 24 bit dither therefore goes 32bit float -> 24 bit fixed. The 16 bit dither goes 32bit float -> 16 bit fixed.

    Nika
     
Similar Threads
  1. jhagertybhs
    Replies:
    5
    Views:
    2,619
  2. Alécio Costa - Brazil
    Replies:
    5
    Views:
    965
  3. Kasim Sulton
    Replies:
    2
    Views:
    1,003
  4. blueinkpen
    Replies:
    10
    Views:
    1,518
  5. ThirdBird
    Replies:
    7
    Views:
    575
Loading...

Share This Page