Designing a BP filter

Frank,

You're correct, the BW of the filter should be 1.7MHz. That's the passband bandwidth, you can center it at 1MHz.

Based on the design FilterPro gave you, you'll note that you need a GBW of ~500MHz for the op amps (it drops to ~300MHz if you center it at 1MHz). Suitable choices might be the OPA820 or OPA690 - but note that since these are high-bandwidth devices, you should probably scale the resistor values down. Attached is a design centered at 1MHz with the impedances scaled so it would be suitable for these high bandwidth amplifiers. You could also use the OPA355 if you center it at 1MHz. 

Any of those amps should drive an ADC fine, and the input impedance shouldn't be an issue if you're driving from another op amp.

Simulate your choice in TINA - I found with the OPA355 the rise time was ~264ns, maybe a bit too long. Didn't try the faster amps, but with an ideal op amp the 1MHz centered filter was coming in around 234ns.

0638.cheb1MBP.pdf

Rick,

I didn't understand your comment about scaling. Why do we need to scale and how do we scale? 

Also, how did you reach to the 234ns for OPA355? 

Thanks,

F. 

Frank,

If you read the data sheets for the OPA690 or OPA820, they recommend not using resistances around them higher than ~1Kohm. If I recall correctly, there's some discussion about why in the data sheets. Pretty typical for wide bandwidth amplifiers - high value resistors at high bandwidth means more noise and the parasitic capacitances can sometimes be problematic. Depends upon a lot of factors, so it's best to read the data sheet and follow the advice given.

In FilterPro, you can change any component value in a stage by clicking on the component value, typing in a new one, and hitting the enter key. The component values for that stage will be recalculated - so I just took the highest value resistor in each stage, and changed it to 1K; FilterPro did the rest.

The numbers I arrived at for the OPA355 were from a TINA-TI simulation I did of the circuit. I've attached a TINA schematic file using an OPA820, which is giving pretty good results. Hope that helps.

8267.Cheb1MBP.TSC

Rick,

I have another newbie question. I have built the filter we have discussed using OPA355. Both in simulation and on my board, I see the following behaviour and I think I understand it but I like to double check. 

My input (signal) to the filter is a square wave with some duty cycle. (Something like 600nsec on 400nsec off). I see that the rise time meets what I have intended, so I think the bandwidth of the filter is sufficient. (Centered around 1.1MHz with 1.4MHz bw). However, my square wave is now somewhat similar to a sine wave, I it seems the filter cuts the square nature of the signal and rounds it to sine. Is this expected? (Basically signal looks like, rise ok, when it reaches to the stability it starts to fall down). I am suspecting the filter treats the stable signal after the rise as DC and cuts out but my background on the subject is limited so I don't fully understand. 

Also, would it be possible to do something so that the filter doesn't cuts the stable region?

Thx,

Frank

Frank,

Yes, the behavior is as expected - you are filtering out higher frequencies which are what contribute to the square shape of the signal - remove higher frequencies and the shape becomes less square; and since you're also filtering out lower frequencies, this will affect the shape as well.

You could try using a Bessel or Gaussian filter which will help preserve the shape a bit more, but it will continue to be "rounded", just less so since they don't remove as much frequency content for a given order of filter.

A filter for this application is not straightforward. To determine the frequency content of your non-continuous square wave, you might run a FFT on the waveform so that you know what its frequency spectrum is. You will probably need to take a large sample to get sufficient resolution.

Prepare to be surprised at the result.