filters

[Tony Clark]

> Ummm... the MF10, MF40, MF60, etc are NOT switched cap filters. They
> are rather like the MAX7-something in that they are programmed via
> resistors and may (depending on model) have "built in" caps. 

   Could have fooled me, Maxim has the MF10 labeled as, "Dual Universal 
Switched Capacitor Filter".  That's what I remembered them being when I 
got samples from them.

> Analog Devices also has appnotes on using these with D/A convertors to
> create "digitally programmed" filters. Seems one could just as easily
> substitute OTAs to create "voltage programmable" filters.

   Glad you brought this up.  It doesn't take a lot of circuitry to 
generate a good filter, but to create a similar "kind" of filter, i.e. a 
6 to 8 pole active filter from opamps and OTA's is a more difficult 
task.

> Problem is, the damn things are expensive. If you're ALREADY using
> OTAs (for the VC part) you might as well just go ahead and use a
> single opamp and a coupla discrete caps to make a VCF; it'd sure be
> cheaper.

   There is a nice relationship between circuit board real estate and 
chip cost.  It tends to be "You pay to save space".

> This is completely false. Switching ("chopping") in the time domain
> does NOT guarantee "stairstepped" outputs (some of the most expensive,
> highest performance opamps are "choppers" - look in your AD databook).
> Hell, you shouldn't even be getting "stairsteps" from a properly
> designed DAC interface - but that's another discussion.

   I never mentioned Analog Devices, which I find to be a company that 
is highly specialized and therefore very pricey.  I tend to avoid 
their products if I can help it.
   As for my reference to the A/D D/A thing, of course it was just a 
comparison.  Of course a properly designed DAC will not get 
stairstepping, but usually there is some sort of filter doing some work 
in there.  
   As for the MF10, let me quote from the Maxim datasheet:

   "Figure 21 is a simple inverting switched capacitor integrator, where 
the R of the standard integrator has been replaced by a capacitor and two 
analog switches, S1 and S2. As S1 and S2 alternately open and close at 
the rate set by the clock input, first C1 is charged to the input 
voltage, then C1 transfers its charge into the capacitor C2.  This 
creates a series of voltage steps on the integrator output, with each 
voltage step having a value of -Vin X C1/C2.  If the value of C1 is small 
compared to C2, the series of steps at the integrator output approximates 
a ramp with the slope <Huge equation deleted> where Fclk is the frequency 
of the clock input."

   Now granted I didn't do a great deal of testing with my MF10's, but 
the above sure as hell indicates that the chip can have a stairstepped 
output.  Now whether or not this will be "noticeable" and whether 
or not it affects the overall quality of the analog signal being 
passed through it, I can't tell you.

> Switched filter caps are as "analog" as a BBD delay line. They
> discretize TIME, but they do NOT quantize level. A switched cap filter
> merely switches one or more caps in and out of the circuit many
> times/sec as a means of setting "effective impedance" of the filter. 

   Yes, they are analog.  I'll certainly back that up.  No, they probably 
don't quantize like a D/A converter would.  The stairstepping 
"levels" would be totally dependent on the input analog level when the 
capacitors are switched in and out and would change as the signal 
frequency changes.

> Many companies DO make these things, and many are very easy to use 8
> pin DIPs. They use a switching freq typically many times the bandpass
> (>100KHz for 10KHz response) and you DO need to "post filter" (as well
> as "pre filter") these to properly control aliasing.

   Yes, aliasing is an inherent problem with switch capacitor filters.

> Switched cap filters can be VERY high performance. The "off the shelf"
> devices are limited in a synth application because of the difficulty
> varying Q, and (quite possibly) because they're just "too much filter"
> (most run fourth to TENTH order response - WAY too much to be of use
> in a synth). They might also just be too durn *clean* to be any fun.

   I'm sure there are some good filters of these types.  And yes, they 
are limited in synth applications.  You might find them useful for EQ's 
or something similar perhaps?  Vocoding?  I dunno. Anyway, I have a few 
MF10's that aren't going to be used anytime soon, that's for sure.
   This has been a rather fun discussion.  Look's like I'll go read a few 
more datasheets to get the "larger" view on switched capacitor filters.  
Maybe we all should...

   Tony

-----------------------------------
I can't drive (my Moog) 55!
-----------------------------------
Tony Clark -- clark at andrews.edu 
http://aupe.phys.andrews.edu/~clark
-----------------------------------