Details on the signal chain?

Discussion in 'Archived Threads 2001-2004' started by brian a, Mar 2, 2002.

  1. brian a

    brian a Second Unit

    Joined:
    Jan 29, 2000
    Messages:
    448
    Likes Received:
    0
    Trophy Points:
    0
    With all of the talk about the quality of DAC and ADC in the new units on the market, could someone please go into a little greater detail about the signal chain in the units in various configs? I get the high level stuff of:

    analog source - ADC - processing - DAC - amp - speakers

    I assume:

    digital source - processing - ADC - amp - speakers

    and

    analog source - analog pass-through - amp - speakers

    What I'd like to get is a greater understanding of the functionality of the DAC and ADC process and how the 24/96 and 24/192 specs translate into functionality and performance.

    brianca...
     
  2. Jagan Seshadri

    Jagan Seshadri Supporting Actor

    Joined:
    Nov 5, 2001
    Messages:
    528
    Likes Received:
    0
    Trophy Points:
    0
    Heh heh.
    That's when it's time to go to the library and get a book on the subject, because there's a lot of theory behind different DAC and ADC techniques.
    Do you have any specific questions to start off with? I assume you know how you go from analog-to-digital, and that you know what 24 and 96 in 24/96 refer to and why...
    Help me out here [​IMG]
    -JNS
     
  3. brian a

    brian a Second Unit

    Joined:
    Jan 29, 2000
    Messages:
    448
    Likes Received:
    0
    Trophy Points:
    0
    Well. My understanding of the process is that the 24/96 refers to bits/sample rate. So the significant difference between 24/96 and 24/192 is the sample rate. Double the sample rate means less information needs to be created due to the smaller gaps between measurements (which I am assuming are filled in by either the DAC or some kind of DSP unit?)

    I suppose that the 24 bits refers to the sensitivity of the measurement. A 32 bit # should allow for finer granularity if that's how it's used. I'm not sure about this. I don't completely understand why the 24 bit side of the DAC didn't grow, or isn't as important to improve on as is the 96Khz side.

    That should give you an idea how vague my understanding of this process is. If you just want to recommend a good book or url, that would be great as well.
     
  4. Jagan Seshadri

    Jagan Seshadri Supporting Actor

    Joined:
    Nov 5, 2001
    Messages:
    528
    Likes Received:
    0
    Trophy Points:
    0
     
  5. brian a

    brian a Second Unit

    Joined:
    Jan 29, 2000
    Messages:
    448
    Likes Received:
    0
    Trophy Points:
    0
    That's fantastic. Thanks for the help. I'll check out the book.

    Couple of quick questions. It might be too much for a quick answer, in which case I'll just hope to get it from the book later..

    1) How does measuring taking twice as many samples per second increase the peak of the sine wave measured? Or are you saying any individual point on the 96Khz wave can be measured by the 96Khz DAC, but the entire wave can't be reconstructed from those measurements?

    2)What is it about a 192Khz DAC that enables it to make better conversions of a 20Khz signal than a 96Khz DAC? I see that you note that this is the case, but I don't quite follow how.

    brianca..
     
  6. Jagan Seshadri

    Jagan Seshadri Supporting Actor

    Joined:
    Nov 5, 2001
    Messages:
    528
    Likes Received:
    0
    Trophy Points:
    0
    The DAC takes the 24-bit measurements of the samples and spits out voltages in a machine-gun fashion. These voltages are short pulses (like the samples themselves), not continous waves yet. To de-pulsify (?) the voltages, a low-pass (reconstruction) filter is used. Imagine the voltage pulses smacking around a big beach ball, and we follow the smooth curve of the beach ball to get our analog wave back.
    The filter must do two things:
    1) Reject all sharp pulse transitions (reject all frequencies above x Hz), and
    2) Ensure that any processing delay through the filter be equal for all frequencies so that the original analog signal's shape is preserved (this is called a phase-linear filter).
    Basically, if you cheap-out and sample at the bare minimum 2x rate, then it is hard if not impossible to practically build such a filter (you can build a theoretical filter that will work in this case, but it requires past and future(!!) samples in order to create a continous analog signal in the present). However, if you oversample, then you can build a very good low-pass (reconstruction) filter.
    This link says a bit more about oversampling, but more from a Compact Disc perspective (which uses a slightly different way of oversampling than today's never digital audio systems).
    There you go. I've avoided the math because it's not as intuitive as pictures, but there are plenty of books that go through the math. The explanation I just gave should hold you for now [​IMG]
    -JNS
     
  7. Saurav

    Saurav Cinematographer

    Joined:
    Feb 15, 2001
    Messages:
    2,174
    Likes Received:
    0
    Trophy Points:
    0
    Jagan, great explanations. It's nice to see someone with such a clear grasp of the issues involved.

    Brian, here's another angle on the same thing that Jagan touched upon, but didn't elaborate. Like he explained, to accurately sample x Hz, you need a sampling rate of 2x Hz, otherwise you get alaising and the reconstructed signal is of a different frequency than the original. If you look at this statement from the other direction, here's what it means: If you have a sampling frequency of Y, you better make sure that the highest frequency in your signal is Y/2, otherwise you're going to run into trouble.

    Once you understand that, take a look at regular CD - that has a sampling frequency of 44.09 KHz. That means the highest frequency that can exist in the signal is 22.045 KHz. The industry assumes that it only needs to reproduce signals till 20 KHz. So, before the signal goes through the sampler/ADC, you need to pass it through a filter which is flat till 20 KHz, and then drops to zero by 22 KHz - which is an extremely sharp filter. And this filter has to be analog, since it acts before the signal has even been digitized. As you can imagine, having such a high order filter doesn't do much good to the sound. That's another advantage of a 96 KHz sampling rate - your filter can now be up in the 48 KHz range, and any effect it might have on the sound there would be largely inaudible.

    Also, the 20 - 20K "rule" is more of a generalization - there are people can hear beyond 20 KHz, just like there are people cannot hear above 16 KHz. Again, having a filter at 48 KHz helps, because the signal remains untouched around the 20 KHz region, and some people can hear the difference.
     
  8. brian a

    brian a Second Unit

    Joined:
    Jan 29, 2000
    Messages:
    448
    Likes Received:
    0
    Trophy Points:
    0
    Ok. That makes sense. So would you say that anyone listening to a 24/96 recording on a machine with 24/96 DACs isn't reaping the rewards of the recording? Or a 24/192 recording on 24/192 DACs?
     
  9. Jagan Seshadri

    Jagan Seshadri Supporting Actor

    Joined:
    Nov 5, 2001
    Messages:
    528
    Likes Received:
    0
    Trophy Points:
    0
    I don't follow your question.

    If the recording is 24/96 and it is played back on a 24/96-capable DAC, then all is well. Same with a 24/192 recording on a 24/192-capable DAC.

    -JNS
     
  10. Saurav

    Saurav Cinematographer

    Joined:
    Feb 15, 2001
    Messages:
    2,174
    Likes Received:
    0
    Trophy Points:
    0
     
  11. RicP

    RicP Screenwriter

    Joined:
    Feb 29, 2000
    Messages:
    1,126
    Likes Received:
    0
    Trophy Points:
    0
     
  12. brian a

    brian a Second Unit

    Joined:
    Jan 29, 2000
    Messages:
    448
    Likes Received:
    0
    Trophy Points:
    0
    Go it, thanks. Next question would be why some units have different quality DACs and ADCs. An analog source comes in and get sampled at 24/192, processed, and converted back to analog at 24/96. Am I wrong in remembering that there are mis-matched components like this in some processors?
     
  13. RicP

    RicP Screenwriter

    Joined:
    Feb 29, 2000
    Messages:
    1,126
    Likes Received:
    0
    Trophy Points:
    0
    Brian,
    The reason most systems have 24/192 DACs nowadays is for oversampling. Basically oversampling just means taking the original signal and processing it at a higher rate. To not get too technical, this helps with noise-shaping and allows the DAC to produce a "cleaner" analog signal than could be accomplished with the steep analog filters that would be needed if the signal were processed at its original rate.
    One good analogy I like to use to explain sampling rate is the following:
    The higher the sampling rate, the lower the occurance of aliasing. Aliasing means that the signal was not captured properly and in the subsequent reconstruction of the waveform can leave information out that was in the original analog signal but lost due to a lower than needed sampling rate.
    You can see the effects of aliasing visually as well. Take a motion picture for example; film is technically a sampled signal as well. A frame of film is shot 24 times a second, so the "sampling rate" of film is 24fps.
    Now, one effect that's very noticeable is the "Spoked wheel effect". If a spinning wheel is filmed at 24 fps, it will appear to be moving forward as long as the rate of rotation is slower than the rate of the film. As soon as the wheel's rotation becomes faster than the film can shoot (sample) the wheel appears to slow down, stop when it makes one full rotation every 1/24th of a second, and then eventually appear as though it's moving backwards.
    The virtual impression of the wheel moving backwards gets faster as the rate of rotation gets faster, and this backwards motion is caused by the aliasing of sampling at too low a rate. Now the wheel is not really moving backwards, however it appears to be because the film is an incorrect reconstruction of the original signal caused by sampling too low. The same thing happens with an audio signal. [​IMG]
     
  14. Saurav

    Saurav Cinematographer

    Joined:
    Feb 15, 2001
    Messages:
    2,174
    Likes Received:
    0
    Trophy Points:
    0
     
  15. RicP

    RicP Screenwriter

    Joined:
    Feb 29, 2000
    Messages:
    1,126
    Likes Received:
    0
    Trophy Points:
    0
     
  16. Saurav

    Saurav Cinematographer

    Joined:
    Feb 15, 2001
    Messages:
    2,174
    Likes Received:
    0
    Trophy Points:
    0
     
  17. RicP

    RicP Screenwriter

    Joined:
    Feb 29, 2000
    Messages:
    1,126
    Likes Received:
    0
    Trophy Points:
    0
     
  18. John Morris

    John Morris Guest

    I'd say that the better DAC is the one from the manufacturer that has never, ever, mis-estimated a delivery date or a date of availability. It may require some searching to find this out, but most folks here on HTF will affirm that that is the most important factor when considering which choice to actually buy. Good Luck!!!
     
  19. Saurav

    Saurav Cinematographer

    Joined:
    Feb 15, 2001
    Messages:
    2,174
    Likes Received:
    0
    Trophy Points:
    0
     
  20. John Morris

    John Morris Guest

    Ah, shit! Just when I get really pissed, and go smartassed in my answers, I see an answer by Ricp or Suarav and then I have to really give a good answer, which coresponds with them. So, I do. Those guys know their stuff... and they are also right with their answers. Email them for more info!
     

Share This Page