Active twin t notch filter oscillation

[DThodoris]

Hi, i have designed a tunable notch filter circuit for an application where i need to filter out a specific frequency.

It is the well known active twin t, i just added a few multiturn trimmers to make it tunable in frequency and Q.
The circuit is not very easy to tune, but once you tune it it stays there and achieves a good deep and sharp notch.I used 1% resistors and 1% C0G capacitors.

The circuit exibits a 'weird' behaviour when i apply power to it.It oscillates for some time and then operates normaly.

The duration of the oscillation is directly proportional to Q and inversely proportional to the supply voltage.For example:
(@Vcc=9V)
When i set the Q too high (trimpot R11 at 100%) it oscillates for quite a long time (tens of seconds).
When i set it at 80% the duration becomes 1-2 seconds.

(@Vcc=16V) The oscillation times reduce by a factor of 10

The frequency of the oscillation is in the audio band,about 5-10% above the notch frequency.
These measurements were done with the input shorted, and the output directly to the oscilloscope.

Can anybody please explain this behaviour and maybe suggest a solution?Thank you in advance!

[nctnico]

Did you do a stability analysis? I think you want the filter to be way too sharp with only a few components.

[f5r5e5d]

sounds like too much positive feedback - other places I have seen only 10-20x Q multiplication with the twin T

http://www.tronola.com/moorepage/Twin-T.html

you really need to think over your real requirement for how narrow/high Q the notch needs to bee and use the minimum value to avoid long settling transients

also the TL072 are poor at these performance levels - at least step up to OPA1642 if not OPA827 op amps - "best" would be supply bootstrapped composite op amps rather than singles

[T3sl4co1l]

Why did you reference input and output to GND when you have a perfectly serviceable "VB" there?

Tim

[MagicSmoker]

Personally, I'd be looking at a switched capacitor filter IC if I needed to adjust the Q and frequency of a higher order filter independently (or you could go the really modern route and implement a digital filter on a MCU or DSP or FPGA). Something like the venerable (albeit not so great performing) Texas Instruments MF10 or the newer Maxim MAX7490.

The reason why your twin-t filter is oscillating is the same reason any active op-amp circuit oscillates: there is 360 degrees of total phase shift when loop gain is 1 or higher.

[DThodoris]

Hi guys thank you all for your time! 

As most of you suggest it seems that i am pushing the circuit beyond its inherent limitations...The reason i chose this topology in the first place is it's simplicity (easy design equations and low component count) and the high Q values it can achieve (or so i thought ).

From a practical point of view though, the circuit does the business; at the lowest Q setting (R11 @ ~80%)  that satisfies the application AND Vcc=16Volts it oscillates at 4kHz but only during power up (duration of oscillation is less than ~100ms) and then works fine. I was just wondering if there is a way to push this topology further and get rid of the oscillation completely.Anyway, thanks a lot guys!

PS The reason i needed such a filter in the first place is a topic in itself!I ll open a new thread in the appropriate category for that and maybe we can discuss it there

[DThodoris]

Why did you reference input and output to GND when you have a perfectly serviceable "VB" there?

Tim

Its because the source is already referenced to GND (the same power supply feeds the previous stage and the signal is GND referenced)
Does this have to do with the oscillation in some way?

[T3sl4co1l]

Can you rewire that stage, then?  ...And so on?

The oscillation described doesn't sound like it's necessarily the direct cause, but startup transients will definitely excite the system's resonance every time.  You can at least reduce the amplitude by arranging the circuit to not have an initial voltage offset; namely, by grounding everything to VB, so the supply acts like a split supply.

Tim