Re: bandpass filter oscillations

tschoepflin@xxxxxxxxx wrote:

Hi all

I'm using the dual-amplifier bandpass filter to implement a tunable
3-stage Chebyshev filter. In essence, the DABF is serving as variable
inductor via the generalized impedance converter configuration. See p.
5.93 of Analog Device's "Op-Amp Bible" by Jung
(http://www.analog.com/library/analogDialogue/archives/39-05/Web_Ch5_final_PtB_F.pdf)
for circuit details.

Anyone have experience with this filter? I am experiencing a small
oscillation at its resonant frequency (200 kHz) even with zero input.
The "envelope frequency" is about 10-20 kHz. This becomes quite
noticeable (100mVptp) after my gain stage of 128. Obviously, we could
redo the whole board to put the amplification first (which we should
have done in the first place). However, I wondered if there was an
obvious solution or some things we could try to make sure we understand
this issue before revising the entire board.

The previous design worked great and used a Toko variable inductor
instead of the generalized impedance converter.

My other suspicion is that I am using two 1000pF ceramic caps (NP0/C0G)
and I've read about microphonics being a problem. Anyone with
experience here? Could they generate an oscillation like what I'm
experiencing? I'm wondering if the 200 kHz is inherent to the filter
and whether the 10-20 kHz "envelope" is due to microphonics.

My workhorse op-amp for the filters is the Analog Devices OP467 and my
gain stage uses the AD829. I simulated the whole thing with LTSpice
and the ADI circuit models and wasn't able to reproduce the
oscillation, so I'm starting to doubt that the op-amp is at fault.

Everything everyone has said so far, plus:

In a linear circuit any instabilities result in oscillations of infinite magnitude, so any finite-magnitude limit cycles must involve circuit nonlinearities. _Small_ magnitude limit cycles usually involve small nonlinearities. Small nonlinearities are hard to model, and may not be in your SPICE component model.

Small signals appearing in band pass filters may be from oscillation, or they may be from noise pick up. It can be hard to tell the difference between oscillation and noise, particularly when the output signal is small, so you'll need your thinking cap.

Do what you can to verify that the signal isn't noise pick up and not really a circuit instability at all.

I would go at this by carefully eliminating the possibility that it's noise pickup first. If I got to the point where the input was grounded with just that sub-circuit alive and powered from batteries and there was _still_ a signal there then I'd believe oscillation (or evil magnetic forces -- is your bench next to an MRI machine?).

If it _is_ really oscillation I'd start asking myself how the actual op amps differ from the circuit model. Obvious places to consider are the environment (board layout) they're in, and subtle nonlinear effects that are too much of a pain for ADI's marketing department to include in their models.

It looks like this circuit sets the bandwidth with R1, which implies the need for a really low impedance source -- I bet that an open circuit at the input terminal would result in oscillation, and an op-amp with some crossover distortion driving that input _might_ result in oscillation. Long cable runs may do the same thing.

I don't know if distortion in an op-amp _in_ the circuit would cause oscillation, but I'd check.

Doing a poor job of bypassing your amplifiers might cause this problem as well. Ditto for poor power supply regulation or parasitics. Although parasitics shouldn't be a terrifically big deal at 200kHz you are right in considering things like microphonics (although this would show up in spades if you just tap on the board).

--

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
.

Quantcast