Question about filtering noise (1/f)

Discussion in 'General Electronics Chat' started by daviddeakin, Oct 25, 2013.

  1. daviddeakin

    Thread Starter Active Member

    Aug 6, 2009
    In this Texas app note on current sources, on page 12 it talks about reducing noise by filtering it. But it states:

    "Without the capacitor, noise from the current source would
    feed directly through to the output. The capacitor filters the
    noise at a –3dB frequency of 1/(2 • π • R• C), or about 30Hz
    in this example. Filtering below this frequency will not
    reduce noise further, since the 30Hz pole is already below
    the 1/f corner of the current source, and noise can not be
    reduced by filtering in the 1/f region."

    But why would filtering below the 1/f corner not reduce noise further? Surely the more filtering we have, the better things get, whether it is 1/f noise or otherwise?
  2. Veracohr

    Distinguished Member

    Jan 3, 2011
    That was a new one for me, I'd never heard of a 1/f 'corner'. But I found some useful reading sources and it seems that filtering in the 1/f region is practically difficult. I don't know if this is right, but my guess is that, because 1/f noise itself has a falling (spectral density) slope of 10dB/decade, the slope of a (single-pole) filter has less of an effect than it does on white noise, which is flat across the frequency spectrum. Kind of like the noise is keeping half a step ahead of the filter in a merry chase.

    Some more reading for you:
  3. crutschow


    Mar 14, 2008
    I believe in general you can filter 1/f noise but the circuit response can become very slow due to the very low frequencies you are trying to filter. In the particular case of the article you referenced they are referring to the specific FET design in Figure 31 where the intrinsic 1/f noise of the FET cannot be filtered with that circuit configuration.
  4. daviddeakin

    Thread Starter Active Member

    Aug 6, 2009
    Yes, that's what I thought. Filtering still improves the noise, but the improvement not as dramatic in the 1/f region, since the filter drops at -20dB/decade and the noise rises at 10dB/decase, giving only a -10dB/decade improvement.

    The noise of the FET is, of course, not affected by the filter.