John Hadstate wrote:

> Tim Wescott wrote:
>
>>I'm trying to think of examples where using an anti-alias filter is a
>>bad idea, or must at least be approached with extreme caution. I
>>already know about control systems and video applications, and I believe
>>that this is a big issue with EKG machines.
>>
>>Does anyone else have any examples that they can share?
>>
>
>
> I believe that this would be a consideration any time transient or
> impulsive phenomena are being digitized. However, Nyquist still holds,
> and if the anti-aliasing filter is interfering with your recording,
> it's an indication that you aren't sampling fast enough and the filter
> bandwidth needs to be wider.

Nyquist holds, but it can be difficult to go from a real-world (i.e.
inherently infinite bandwidth) signal to a bandlimited one without
ringing -- I'm discussing this, in the context of how you need to be
aware of the issue when establishing your sampling rate.
>
> Some people try to use overly-sharp (eg: Butterworth or Chebyshev)
> analog AA filters to push the cutoff closer to Nyquist. These filters
> have a tendency to ring, even when implemented at low orders--not good
> for sampling transients.

No kidding!
>
> Why would EKG recorders be especially problematic? If they are, then
> maybe EEG and polygraph recorders too?

AFAIK they aren't especially problematic, it's just one that I'm
familiar with being an issue -- goodness knows why, since I've never
worked with medical electronics. I just recall discussions about the
impact of sampling on EKG waveforms. I do know that if you're going to
have problems with ringing or rounding that big old spike in the middle
of an EKG waveform is going to make it happen -- particularly since
there are interesting features of the waveform that happen just before
and after the spike, and which would be obscured by ringing.
>
> One trick that I used in process control instrumentation was to
> oversample by a factor of several hundred, do the main anti-aliasing
> digitally and then decimate to a more appropriate sample rate. This
> means that the analog anti-aliasing filter can be very cheap and when
> the sampling can be synchronized to a multiple of the power line
> frequency, you can get a double whammy.
>
I do that too. Doing the anti-aliasing as an average of all the samples
leading up to the current decimated sample passes DC unmolested, but
puts a notch at all the harmonics of your decimated sampling frequency
-- so any noise that _would_ alias into your system's passband gets
notched, at the least cost in phase shift in the anti-alias filter.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/

"Applied Control Theory for Embedded Systems" came out in April.
See details at http://www.wescottdesign.com/actfes/actfes.html