Gain bandwidth product for filter design.

[s3c]

I'm not real fond of all the math involved for designing filters so I'm using filterpro to design a 125KHz notch filter. I haven't built the circuit yet but simulating with an ideal op amp model looks real good, now filterpro lists the minimum gain bandwidth product as 13MEG and 27MEG for the 2 stages, why do you need a GBP that high for a filter with an operating range that is much lower?

[DJPhil]

It sounds like you're stipulating some high performance criteria for the filter type you're using. Filterpro's default settings are sort of cranked up for the few things I've worked with (audio). Knowing what to trade off on is the hard part of filters. It all depends on what you're trying to accomplish.

What are the settings you're using, and what's the purpose of your filter?

Some docs you might find useful:
Filter Design in Thirty Seconds - TI
FilterPro Design Program - TI Page 12 - Opamp Selection

[s3c]

Thanks for the links DJPhil, the second document was particularly insightful. I'm trying to suppress the 125KHz carrier in an RFID application, I've done some Octave simulations on a received raw input stream from an RFID antenna and get the best results with something like the following:

Center Frequency: 125KHz
Passband Ripple: 1dB
Passband Bandwidth: 120KHz
Stopband Bandwidth: 20KHz
Stopband Attenuation: -30dB

I'm using a 4'th order Butterworth filter with MFB topology. This wasn't my original design but I had to change it a little bit to get a acceptable response with 5% components.

I'm using a notch filter instead of a standard low pass to try and retain some of the modulated harmonics, this gives me a pretty clean square wave as output and I'm trying to keep it as precise as possible since this is a RFID debug application.

I was hoping to do this with a cheap general purpose op-amp like the MCP6001 but suspect that's not going to happen, I want to bring down cost since this project is going to be using a couple of filters.

In short, it's not possible to design a filter like this with low end parts?

[DJPhil]

In short, it's not possible to design a filter like this with low end parts?

I don't believe so. I think you'll keep bumping into the ([100 to 1000] * GBP * f) rule of thumb. I plugged in your settings and used 5% resistors and 20% caps to see what I could accomplish. I managed to get a slightly lower (~5MHz) drop in GBP by lowering the passband slightly, but it's not enough to even start saving you money. Further reductions make things worse by changing the Q. I don't see a way around a high GBP if you're wanting to preserve a big chunk of post notch passband, but I'm a long way from a filter expert. If anyone sees where I'm off here, please let me know. 

I'm not well read on RFID, but it seems like they use lowpass to filter the carrier and then bandpass for the mark and space frequencies. The example I found used 12.5 kHz and a 15.625 kHz and left them as (mostly) sine waves until feeding it into a uC comparator. I think the design is intended to keep the filters low frequency and therefore cheap, and to restore the square wave shape post filtering.

I'm guessing you need to know more about the return signal than a standard reader can tell you (debug application) so the above method isn't going to help. Assuming there's no easy solution to the high GBP opamp requirement you might consider an opamp twin-t notch filter. If you're handcrafting these you can cut the cost slightly by using a single opamp, but you'll have to use precision resistors (E96) and caps (E12) and hand match them for highest performance. It's possible that you don't need that level of precision and a little bit of experimenting will help you decide. Naturally, TI has a paper on this too and covers the simulation vs. reality of implementing these. You're still stuck using a high GBP opamp for the twin-t though, the one they use is (I think) rated at 100MHz.

I did stumble on an article that might be relevant, but it's behind a paywall. I've included it in case you have access that I don't.

I hope that helps some, I'm a bit out of my league here I think.

[s3c]

Your research skills are top notch though, thanks for all the help, I'm not going to put any more design time into keeping the budget low. There probably won't be a lot of these boards made so it's not that bad. Test equipment always seem to have a bunch of requirements that creep in, I do actually need these features though.