Simulating Bandpass Filter

[Coceth]

I'm simulating a MFB bandpass filter and given my part selection, I should be getting a center frequency of 421.1kHz according to Okawa's filter calculator. An AC sweep in LTSpice shows a center frequency of 187kHz though. I'm certain I wired the circuit up right, but the simulator doesn't lie... Can someone take a peek?

[LvW]

I don`t know where the errror might be - however, try another (better) design and start with R2 infinite.
As you can see, R2=20 ohm is extremely low in comparison to the other resistors.

More than that, because of the high Q-value, another variation of the filter structure is recommended:
Some positive feedback (resistive voltage divider to the pos. oamp input).  For high Q values, this modification was recommended by Deliyannis long time ago.
Also in this case: R2 infinite !
As a consequence, this modification gives a higher midband gain - is this acceptable?

Hint: This MFB alternative is considered in some filter design packages - perhaps also in Okinawa`s program?

[ogden]

You definitely shall try this: http://www.analog.com/designtools/en/filterwizard/
and/or this: http://www.ti.com/design-tools/signal-chain-design/webench-filters.html

[Coceth]

I appreciate your advice LvW. I agree that R2 is very low (nearly a short). I'll play with the values a try make it >100 ohms at least.

Also thanks for directing me to the modified MFB topology - I'll look into it!

[T3sl4co1l]

What opamp is that?

Tim

[LvW]

If you select the Deliyannis Modifikation (with pos. feedback, factor k) the design is very simple:

R2>>>infinite
C1=C2=C; R3=R1=R
k=Ro/Rf (Rf: Feedback resistor to the pos. input, Ro between pos. input and ground)

Midfrequency wo=1/RC;
Quality factor: Q=1/(2-k)
Midband gain: Ao=Q(1+k)

As you can see (formula for Q) the circuit is rather sensitive to the factor k (tolerances) - in particular for High Q values (as in your case).
Therefore, it would be wise to use one of the other bandpass topologies which are better suited for large Q values - for example a GIC structure (if you can afford to use two opamps).

[Coceth]

It's "UniversalOpamp2". I've subbed existing generic op amps in Analog Devices' catalog, and the bandwidth gets slightly worse. So my simulations were best-case I guess.

[LvW]

It's "UniversalOpamp2". I've subbed existing generic op amps in Analog Devices' catalog, and the bandwidth gets slightly worse. So my simulations were best-case I guess.

I hope you are aware that - for hardware realizations - the opamp must be selected very carefully with respect to GBW as well as slew rate (which very often is the limiting quantity).

[Wimberleytech]

It's "UniversalOpamp2". I've subbed existing generic op amps in Analog Devices' catalog, and the bandwidth gets slightly worse. So my simulations were best-case I guess.

I hope you are aware that - for hardware realizations - the opamp must be selected very carefully with respect to GBW as well as slew rate (which very often is the limiting quantity).

Correct.  I ran a simulation using a VCVS and got close to the correct response. Did not investigate the additional peaking however (did you design it for 6dB gain at midband?).

update...took another look and zoomed in to the passband peak.  Looks like a little over 9dB at the peak.

[Coceth]

Thank you for all the responses! It helped me realize that "UniversalOpamp2" has a default GBP of 10MHz, which may be too low. I turned up the GBP, and the higher I made it, the the more the center frequency converged to the calculated value. I'm lowering the Q and gain of the filter so the GBP of the opamp doesn't have to be so high.

Thanks again