Elliptical Filter Theory vs SImulation

[dknguyen]

I used some filter design software to make an elliptical filter, and then I used PSPICE to simulate it with my op-amps and actual component values I could get my hands on.

I'm wondering why there isn't that dip on the high-frequency stopband like there is in the theoretical design (it does seem to appear on the low-frequency stopband in the simulation after all)

 

[Roff]

Your op amp(s) probably don't have enough bandwidth at the frequency of the notch, and the phase shift/delay of the op amp at that frequency could also "smear" the response on the high end. Due to the rising closed-loop output impedance of the op amp with frequency, some lowpass filters will actually begin to have less attenuation as frequency rises. This is caused by coupling from input to output through feedback capacitors in the filter.

 

[dknguyen]

Is there a way to try and find the maximum attenuation for a given frequency (similar to product gain bandwidth but for attenuation? It's probably playing around with the gain bandwidth product somehow, but I'm not sure what it is).

 

[Roff]

If you mean, which op amp should you choose, I suspect it's the one with the highest GBW, but that might be overkill. Does Pspice have a parameterized behavioral op amp model? LTSpice has UniversalOpAmp2, a 2-pole model, with the following tweaks:

Avol=1Meg GBW=10Meg Slew=10Meg
ilimit=25m rail=0 Vos=0 phimargin=45
en=0 enk=0 in=0 ink=0 Rin=500Meg

These are the default values, and all can be changed. Some of these are self-explanatory. All are described in the online manual.
This part can be very handy, and it generally simulates faster that a manufacturer's model. You can switch to a mfr's model after you find the values of the parameters that are important to you.
Pspice probably has a similar part.

 

[dknguyen]

THe model I was using was the OPA743 with a GBWP of 7MHz. I tried to cook up something using the GBWP and I seem to get 33GHz bandwidth to get what I want (well at least something that resembles the ideal graph). In actuality, the way the op-amp gives out and drops on the high-frequency stop band gives me the almost as much average attenutation as there is in the ideal elliptical filter.

 

[Roff]

THe model I was using was the OPA743 with a GBWP of 7MHz. I tried to cook up something using the GBWP and I seem to get 33GHz bandwidth to get what I want (well at least something that resembles the ideal graph). In actuality, the way the op-amp gives out and drops on the high-frequency stop band gives me the almost as much average attenutation as there is in the ideal elliptical filter.

So what was your previous question about? I didn't understand it.

 

[dknguyen]

So what was your previous question about? I didn't understand it.

You know how you can use the gain bandwidth product to figure out the bandwidth an op-amp can have for a certain gain? Is there a way to do the same thing, but for a filter?

 

[Roff]

You know how you can use the gain bandwidth product to figure out the bandwidth an op-amp can have for a certain gain? Is there a way to do the same thing, but for a filter?

Are you talking about a filter with gain, or do you mean how do you determine the required GBW of an op amp when it is used in a given filter, or something else entirely?

 

[dknguyen]

or do you mean how do you determine the required GBW of an op amp when it is used in a given filter

That's what I mean.

 

[Roff]

That's what I mean.

It depends on how close to theoretical you want to be, but I don't know the answer. **broken link removed** is an app note on the subject as it applies to lowpass filters. The math is fairly hairy, but it gives you some ideas, like predistortion. This is where I get empirical. Simulate the ideal filter (you can use a voltage-controlled voltage source as an infinite bandwidth, very high gain op amp), then sim it with a "real" op amp model, and play with the component values until you get close to what you want. Understanding and doing the math (if I could) just takes more time and effort than I would be willing to invest.