[Nemo]

I've got a circuit where the signal goes through one of two low pass
VCVS filters, and the output of one of them is selected by an analog
switch (DG419).

One filter uses 1nF 2% PPS capacitors - there are various constraints
like trying to minimise the variety of components on the board so this
value was chosen, though it could be changed if there was a good reason.
I note that the DG419 has a typical channel capacitance of 35pF, which
is 3.5% of 1nF. My question is, does this matter, i.e. will it upset the
filter: do relatively low capacitances hanging on its output affect the
feedback loop? The op amp is a NatSemi LMP7702.

Thanks,

Nemo

 

[Nemo]

What filter configuration? Sallen-Key is notoriously sensitive to
component variations,

Good question. At the time I wrote the above, I had a VCVS design. Today
I found some how-to's which implied a multiple feedback / Rauch design
would be more tolerant of output conditions, so I decided to go for a
few more component types - not many needed as it happens - and went for
that instead. I'd still be interested in advice though as I've never
done a MFB filter before. I'm aiming for a Bessel response.

Thanks,

Nemo

 

[Nemo]

If this...

http://www.k7mem.150m.com/Electronic_Notebook/filters/act_bpfil.html

is what you have in mind, both are versions of Sallen-Key.

Write out the equations and you'll see what I mean by component
"sensitivity".

However, independent of filter configuration, ALL OpAmps will exhibit
some phase shift due to capacitive loads.

What is the load after the DG419? If high impedance simply connect
from OpAmp to DG419 with a resistor to isolate the capacitance.

Hard to nail down exactly without some specs on filter frequency and
Q.

You could write out the equations ;-)

Thanks. I hadn't appreciated that MFB's are Sallen-Key types too. The
resistors will be 0.5% or better low tempco, and the caps 2%, so
hopefully the only thing to worry about is the loading. The frequency is
about 100kHz. The load beyond the DG419 is 1kohm. Sounds like I need to
knock together a test circuit, then.

 

[Nemo]

By VCVS, do you mean opamp Sallen-Key? I don't understand why the
analog switch changes the value of one of the caps. Got a schematic?

See diagram 16-19 here:
focus.ti.com/lit/ml/sloa088/sloa088.pdf

this then feeds into the input of a DG419 analog switch.

What concerned me was that this would therefore place ~35pF in parallel
with one end of the 1nF 2% capacitor from the op amp output back towards
the input. On reflection, this was probably a silly concern as the
signal at the point that the 1nF feeds back to is not going to be
affected by something happening at the output of the op amp.

One thing to be careful about: charge injection from an analog switch
can mess up an opamp for a while, microseconds to milliseconds. Even
if the switch is on the opamp output. If you switch rarely, that may
not be a problem.

Not an issue here - users will switch manually, see what "looks best"
for their application, and leave it there.

The opamp alone won't mind a 35 pF load.

Expect a fair amount of distortion using this kind of opamp in a S-K
filter, or in fact any non-inverting configuration.

Hmm. I had been using a passive LC filter but decided it was silly to
throw away half my signal that way, and the wound components were long
lead time, so I went to an active filter. Signal levels at this point
are about a volt but I've chosen film (PPS) capacitors. I thought that
would eliminate significant distortion. Is there a better way of
filtering I've not thought of? I'm after a low pass Bessel at about 100kHz.