[amsat-bb] Active Filter Design

[John <aa1ye@xxxxxxxxx>]

										18 January 2004
	KD5QGR de AA1YE   BT,

At 02:37 1/18/04 -0600, you wrote:
>Hi all,
>
>I'm currently building a system that samples downconverted rf signals at 
>500khz.  Consequently I need a 250khz lowpass filter to prevent 
>aliasing.  I thought that the easiest way to implement this would be to 
>use opamps. (other ideas?)  I don't know what type and order of filter I 
>need.  My goals are sharp cut off and reasonably flat passband.  Phase 
>response isn't all that significant so long as the phase is fairly 
>constant over a 3khz wide segment.  Where might I look to find out what 
>topolgy might best satisfy my requirements?  Thanks for any help,
>
>David Carr
>KD5QGR
>		++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
	David,

  1.		Active Filter Design Handbook.

		G.S.Moschytz and P.Horn

		John Wiley & Sons.

	ISBN:	0-471-27850-5.


  2.		Basic filter design data can be found in "ZVEREV".

  		Handbook of Filter Synthesis.

		Anatol I. Zverev

		John Wiley & Sons.

	Library of Congress Catalog Card Number:	67-17352.	


	A few words of wisdom, if I may.

  1.	No analogue filter is "perfect". Hence your theoretical 2:1 frequency
ratio, to prevent aliasing, is not sufficient in our real world. A factor
of 2.5:1 or up to 3:1 is a better design aim. 

  2.	Also remember the stopband corner frequency, of the above ratios, is
the frequency at the stopband attenuation level, not the end of the
passband. Thus, sampling at 500 KHz, you should design the filter's
passband corner frequency to be in the region of 150 KHz to 200 KHz
depending on your "non-aliased" dynamic range requirement. For example, if
you are using an 8-bit digital system, you should design for 48-50dB peak
sidelobe stopband levels. A 12-bit digital system requires 72-75dB stopband
filter levels.

  3.	I recommend a 0.1dB Tchebyscheff filter characteristic. This gives you
a reasonably sharp filter skirt without demanding un-reasonably high Q
factors of the elements. You should "break" the polynomial into "pole
pairs" and cascade the pole pair sections to achieve the overall filter.
Moschytz and Horn give you the pole pairs on page 136.

  4.	A note about "op-amps". Your frequencies of interest are much too high
for run-of-the-mill (741 style) op-amps. You must splurge up to $5-$10 each
for good high frequency op-amps. Your theoretical design assumes zero phase
shift (or 180 degrees) in the active stage. You won't get this with 25-50
cent op-amps at these frequencies.

  5.	Finally have you considered a LC filter? At these frequencies you can
achieve coil Q-factors of 400-500 with Ferroxcube pot cores and Litz wire.
These values are high enough. In this case, you build the filter as a
single assembly using the data in Zverev, et al, without having to break
the polynomial into pole pairs.	


	The performance of this filter will determine the overall performance of
your radio system(s). You need only one elegant filter at the input of your
sampler; a simple filter at the output of each down-converter will be
sufficient for each of those sub-assemblies. Don't cut corners.  Good luck;
you'll need a little, probably a lot!

	73s,   de John, AA1YE,  amsat 439.      SK

		
----
Sent via amsat-bb@amsat.org. Opinions expressed are those of the author.
Not an AMSAT member? Join now to support the amateur satellite program!
To unsubscribe, send "unsubscribe amsat-bb" to Majordomo@amsat.org