Advantages of active vs. passive filtering (and visa versa)

Discussion in 'General Electronics Chat' started by Management, Jan 6, 2010.

  1. Management

    Thread Starter Active Member

    Sep 18, 2007
    306
    0
    Can you guys let me know what the advantages and disadvantages of active filtering vs. passive filtering and visa versa?

    I have implemented a passive 2 pole low pass filter with cutoff at 7 Hz and an active 2-pole sallen-key lpf where the latter is used for anti-aliasing and cutoff at 2400 Hz and the former is just to get the DC component (It is in a feedback loop).

    I'm trying to decide if I should just use a two device part and go with two active filters or just eliminate a part (go cheap) and use an passive filter for the 2500 Hz anti-aliasing filter.

    Thanks.
     
  2. Darren Holdstock

    Active Member

    Feb 10, 2009
    262
    11
    Subtle pros and cons to both methods. Passive filters have insertion losses in the passband and the source and load impedance form part of the characteristic. Active filters are buffered against the output impedance (but not necessarily the input impedance), but have no passband insertion loss. Some may even have some gain, though this can be tricky to implement and can compromise stability margins and sensitivity to component tolerances.

    For an anti-aliasing filter it is good practice to have an active LPF followed by a passive RC network. This is because some active filters (the S&K LPF in particular, but not the Multiple Feedback architecture) have a limit to their ultimate attenuation as the parasitic inductance of the amplifier output causes increasing output impedance with increasing frequency, causing the rolloff to ultimately shelve. Passive RC LPFs will also ultimately shelve due to stray C on the R and stray L on the C, but to a much lesser extent than the S&k active filter.

    S&K LPFs won't work with DC on their inputs - this will saturate the output at one or other of the supply rails.
     
    Last edited: Jan 7, 2010
  3. Management

    Thread Starter Active Member

    Sep 18, 2007
    306
    0
    Thank you so much for this information. I've learned quite a bit from it.

    Really!? That sucks.

    What should I do because sometimes I have DC with some AC riding on it and I need 1200 Hz and 2200 Hz (1 and 0) and it is a filter for 2500 Hz. But if there is no AC then it will only be a DC voltage ranging from around 500mV to approx. 5V.

    What can I do?
     
  4. hgmjr

    Moderator

    Jan 28, 2005
    9,030
    214
    It sounds like you need a bandpass filter. Such a filter would block the DC component and pass the AC signal component of interest.

    hgmjr
     
  5. Ron H

    AAC Fanatic!

    Apr 14, 2005
    7,050
    656
    In the case of the unity gain configuration, I don't understand why you say this.
     
  6. Management

    Thread Starter Active Member

    Sep 18, 2007
    306
    0
    Yeah that made be quite confused.

    I don't want to block the DC component .... I want it. I want a low pass filter. Sometimes there are two frequencies that I want and therefore I have to have a LPF at 2.4-2.5kHz cutoff. I use a digital bpf inside an FPGA to separate them out but the LPF is before an ADC.
     
  7. Ron H

    AAC Fanatic!

    Apr 14, 2005
    7,050
    656
    Why do you have two LPFs? It only makes sense if they are not in series.
     
  8. Darren Holdstock

    Active Member

    Feb 10, 2009
    262
    11
    Ron H is correct about the viability of the unity gain DC-coupled S&K LPF, I am mistaken in my recollection; didn't think that one through. I was probably thinking of active bandpass filters, which either have an AC coupling capacitor on the input that blocks DC (not relevant here), or are DC coupled but are intolerant of DC on their inputs. A good while back I was designing a low-Q very low frequency bandpass filter on a single rail supply , and I couldn't find an active BPF architecture that would tolerate the (specified) varying DC levels on the input. AC coupling the input was out due to R and C scaling limitations, and DC servo control was overkill and underwhelming at the same time. I settled on a S&K HPF (with some cheeky X10 gain enhancement) followed by an MFB LPF.

    All that said, if your LP rolloff is only a couple of kHz, I don't think you have to worry about the subtleties of non-ideal stopbands, as they'll be far from the frequencies of interest. It may do to have more than a 2nd order response though, as aliasing artefacts can be nasty. If you're cascading 2nd order active filters to get 4th, 6th order etc then you'll find the individual 2nd order stages need to be a little different to keep the -3dB rolloff point in the same place and the damping controlled. There are filter tables available to glean the coefficients from.

    Talking of damping, you'll need to choose a response. The possibilities are endless, but they can be simplified to 3 main types:

    Butterworth (the most popular) gives optimum flatness of amplitude in the passband and a reasonable rolloff slope.
    Chebychev gives a steep rolloff but has amplitude ripples in the passband.
    Bessel gives the best response in the time domain because it's the most phase-linear, but the rolloff isn't so steep.

    All these responses can use the same architecture, it's the crucial ratios between certain component values that gives the response. Which one you choose depends on the application - post a few more spec points if you like and it should become clearer.

    In summary, even the simplest filters can be horribly complicated. I've a great collection of popular and rare books and papers on filters, and that still doesn't even come close to covering even a fraction of what I'd like to know.
     
Loading...