Analog filter wizard design oscillates!

bandpass.zip

Ok, think I figured out how to attach files... the json of the design.

Thanks Stephen for attaching the design.  I was able to open it.  I'm going to do some more investigation on this tomorrow, but I wanted to tell you what I found this morning.

In filter design, we deliberately add in some instability to create peaking.   This is measured with something call the Q of the stage.  If you go to the stage view, you can see it for the different stages.

More details on Q here.

The bigger the Q, the closer the stage is to instability.  Bandpass filters have considerably higher Q than highpass or lowpass filters, and are more prone to instability.

You can see this in the tolerances view, where even with 5% capacitors, we have a huge envelope on where the filter can perform:

Higher order bandpass filters are inherently fragile and so something small can send it over the edge.  

Things you can do:

1) Use a faster op amp if you can afford the power.  I think our lower power optimization is too agressive here.   The LTC6259 is having a hard time with these requirements.  Below is comparison of the LTC6259 vs LT6233 (default if low noise optimization is chosen.)  We may need to go tweak this some in our algorithm.  

2) Reduce the Q requirements on your filter:

  - Increase passband width relative to the center frequency

  - Change to Bessel filter rather than Butterworth (i.e. move filter response slider in specifications page all the way to the right)

3) Specify your components really tightly:

- Use 1% capacitors.  An example.

- Use 0.1% resistors.

----

I'm still going to investigate some more tomorrow.  For example, I don't yet understand why your second stage would ring at 28kHz rather than 3 kHz

Matt

Thanks Stephen for attaching the design.  I was able to open it.  I'm going to do some more investigation on this tomorrow, but I wanted to tell you what I found this morning.

In filter design, we deliberately add in some instability to create peaking.   This is measured with something call the Q of the stage.  If you go to the stage view, you can see it for the different stages.

More details on Q here.

The bigger the Q, the closer the stage is to instability.  Bandpass filters have considerably higher Q than highpass or lowpass filters, and are more prone to instability.

You can see this in the tolerances view, where even with 5% capacitors, we have a huge envelope on where the filter can perform:

Higher order bandpass filters are inherently fragile and so something small can send it over the edge.  

Things you can do:

1) Use a faster op amp if you can afford the power.  I think our lower power optimization is too agressive here.   The LTC6259 is having a hard time with these requirements.  Below is comparison of the LTC6259 vs LT6233 (default if low noise optimization is chosen.)  We may need to go tweak this some in our algorithm.  

2) Reduce the Q requirements on your filter:

  - Increase passband width relative to the center frequency

  - Change to Bessel filter rather than Butterworth (i.e. move filter response slider in specifications page all the way to the right)

3) Specify your components really tightly:

- Use 1% capacitors.  An example.

- Use 0.1% resistors.

----

I'm still going to investigate some more tomorrow.  For example, I don't yet understand why your second stage would ring at 28kHz rather than 3 kHz

Matt

that's a great answer - thank you for how complete it was!

Problem #1 - I think I figured it out for the first stage. The REF voltage that should be supplied to the positive feedback pin needs to have R3 populated in order for this to happen. I just happened to notice the ref designators in the popup:

and R3 is not in the schematic - I populated that with a jumper and the first stage is behaving more sane. The second stage is however still sick.

And I have worked out the issue with the second stage - there was a poor solder connection on one of the feedback caps and a bad value on the REF resistor (681K instead of 681).

The third stage just worked (thank god) and the whole chain is now up and running. I appreciate the help!

Great to hear!