************     I fixed your drawing    **********************

> >4. Assuming I cascade two second-order VCVS low-pass filters to make a
> >fourth-order filter:                                Sallen-key circuit
> >
> >                  +--------------|C1(--------+
> >                  |                          |
> >     +   ---~~R1--+--R2--+-------+|\         |
> >                         |        |op-amp----+----- +
> >                         |        | /        |       
> >   Vin                C2 |     |--|/         |        Vout
> >                        ---    |             |
> >                        ---    +-------------+
> >                         |         
> >      -  ----------------|------------------------- -
> >

Normalized values for a sallen key circuit with gain = 1 are 

                Stage 1            Stage 2
                 R1=1               R1=1
                 R2=1               R2=1
                 R3=OPEN            R3=OPEN
                 R4=0               R4=0
                 C1= 2Q             C1= 2Q
                 C2= 1/2Q           C2 1/2Q

Q values

4 th order:  Q(stage1) = .54 ; Q(stage2) = 1.31

Choose a crossover frequency in Hz and multiply it by 6.28 to convert it to rad/sec.
Now divide all the caps in both stages by this frequency (in rads/sec).  This is 
called frequency scaling.
All the resisters should be 1 ohm at this point.  Choose a desired resister size (say 
10K) and multiply all resisters by that number (lets call it Km).  Then divide
all cap values by Km.  This yields the same exact frequency response (cutoff freq)
as the normalized circuit except the componant values are much more reasonable.
This is call magnitude sacling.  Magnitude in this case refering the the componant 

And thats it.  play with the Km until the resistors and caps a reasonably close
to standard sizes and pick the nearest one.  

I suggest using polypro or metal film caps in the signal path.  You can magnitude 
scale for small cap values (.1uF) which are cheap.  As for opamps I suggest the 
LF347 from National Semiconductor.  This on was designed from the ground up for 
audio purposes.  You can use an 741 from radio shack to see if it all worked out 
but the sound quality won't be outstanding.  Should sound OK though.  

Gain can be whatever you want.  The higher the gain, the higher the noise floor on 
the input side though.  If your radio drives you amp to full output now,  Let k=1.

That should do it

Dan Cousin

Date: Sun, 15 Sep 1996 15:22:24 -0700

Original Subject: Re: Active filter design

Comment on this article

Comments on this article

This article is referenced in the following indexes: