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

Are Anti-Aliasing Filters always analog?

Discussion in 'Recording' started by 100WChris, Jan 13, 2009.

  1. 100WChris

    100WChris Guest

    Im a newbie and im looking for an answer: i tried to understand the process of AD-Conversion and i wonder if i am right: the anti-aliasing-filter must be analog?
  2. RemyRAD

    RemyRAD Guest

    Sep 26, 2005
    One might think that should be the case? But when you realize that it has to switch for 22.5kHz, 31.5kHz, 44.1kHz, 48kHz, etc.. Brick wall filtering of that type is rather involved. So in that situation, it may be digital filtering? Brick wall filtering in analog requires a fair amount of componentry. You don't see that in little $80 interfaces at Best Buy, is too much componentry would be required. So for 44.1kHz recording your filter has to be brick walled at 20kHz. And there's good & bad ways to make sophisticated filters to do that. Why does digital sound harsh? It's the brick wall filtering that's the largest offender.

    You have to worry about that kind of filtering with the almost defunct DSD (direct stream digital) system which has a frequency response clean to 100kHz. No filtering equates to much better sounding stuff. Almost like analog. Too bad we're not at the technological point of making it more affordable, yet. That's the only digital recording sound I like so far.

    Waiting to get back to analog sounding digital
    Ms. Remy Ann David
  3. dpd

    dpd Active Member

    Sep 29, 2004
    I guess I would state it this way:

    If converting from analog to digital, the AAF MUST be analog

    In between digital processing signal processing stages, AAF will be digital
  4. Boswell

    Boswell Distinguished Moderator Resource Member

    Apr 19, 2006
    Home Page:
    The simple answer is yes, by definition. Anti-aliaising filters must be used to band-limit the analog signals prior to their being digitized. Anything done in the digital domain is not anti-aliaising.

    In the early days of digital audio when ADCs sampled at the required output word rate, the anti-aliaising filters had to have switchable cut-off frequencies to match the available sampling rates. It was a tall order to design analog filters to do this job. I know - I designed lots of them.

    The big problem for audio designs was not how to get the necessary amplitude response, but how to tame the phase response horrors that were a by-product of the steep cut-off. It's always been my belief that the antagonism to early "digital sound" came substantially from the mauling that these ugly phase characteristics did to the signal. As far as I am aware, nobody had ever carried out the experiment of returning the output of the anti-aliaising filter to a couple of tape tracks and performing an A-B comparison with the original tracks.

    In the last 20 years or so, ADC designs have moved from SAR (successive approximation) types to sigma-delta types, which work by over-sampling the input and performing digital manipulation on the single-bit bitstream output. Using 64x oversampling at 48KHz means the actual analog sampling rate is 1.33MHz, and many new types run at 128x or 256x rates. Although an anti-aliaising filter is still required, it can be a gentle roll-off in the high tens or even hundreds of kilohertz, with a mild-mannered phase response. All my recent audio designs have been of this type. Reduction of the bandwidth to the necessary 20 or 40 KHz is done by digital filtering, along with noise-shaping to push quantization noise up into frequency ranges that get filtered out. This filtering operation is NOT anti-aliaising.

    Just to complete the picture, in the opposite direction, a D-A converter must be followed by a reconstruction filter to generate a band-limited analog output, and also to correct for the inherent sinc function (sin(x)/x) in the frequency response. This re-construction is a form of anti-aliaising, and is loosely, but incorrectly, referred to as such.
  5. 100WChris

    100WChris Guest

    Thank you very much!
    You´ve answered all my question :D
    Great Forum!
  6. 100WChris

    100WChris Guest

    Thank you very much!

    You have answered all my question :D
    Great Forum!


Share This Page