2nd order butterworth MFB active filter design

Do you have an amplifier in mind? 5Mhz would be plenty for this target

Ok, assuming the OPA377 is close to what you want, I had to do some repair work on this design tool I developed some years back - but I think I got it working again. You say optimum? not really a good word here, but there are better and worse solutions here - this one 

1. Accounts for amplifier GBP

2. Tries to reduce noise using lower R values and shifting the C ratios to reduce in band noise peaking

3. finds a best fit standard E96 R's and E24 C's - that effort lets the gain go off one standard value if it gives a better Fo and Q fit -as it did here. 

And one of the key things to look at is the resulting loop gain, quite high here at the peaks (about 40dB)  kind of saying the OPA377 might be excessive GBP. The filter shape is always easy to get with a vast range of RC values - these other things are more 2nd order issues to shift slightly in a better direction. 

And as long as I have this open and working again, here is this design in TINA. Using split supplies now, but that is easy to change, perfect fit to 20kHz F-3dB, 

And then running the output spot noise, that little hump is what this design flow reduces. It never goes to zero so a post RC would help flatten the integrated noise off. 

And this startinng point file, 

20kHz MFB LP.TSC

And then, if you wanted some background on this - this 1st app note gets it started, but there I had made a mistake in the cubic coefficients. 

https://www.ti.com/lit/an/sboa114/sboa114.pdf

And here I fixed that error, (published, I had fixed it much earlier for an online filter design tool I built for Intersil - gone now I think)

https://www.planetanalog.com/include-the-op-amp-gain-bandwidth-product-in-the-rauch-low-pass-active-filter-performance-equations/

And in confirming that the OPA377 might be a good fit, I noticed a link to the online TI filter designer where it delivers this solution - not sure how to change the op amp selected, but modifying my previous OPA377 design to this set of RC values to these will give an output noise comparison, This design flow came from the old BurrBrown FilterPro developed in the late 80's. There the feedback C value is chosen as a starting point then here it delivers an exact C to ground not standard value. 

Here you can see the effect of this design flow, same filter shape but higher flatband output noise and more peaking, this incrementally increases the output integrated noise which may not be of much interest in some cases, but never hurts to reduce if its free. 

and, I had not put that TI recommended op amp into my MFB tool. The OPAx734 has disable, the OPAx735 does not but both have a staggering high input voltage noise, here I dropped it into the sim using the TI RC values and the OPA735 TINA model, apparently low noise is not in the part selection algorith for the TI online filter tool,

And to move my design down to a slower, lower Icc, but slightly higher input voltage noise design, the OPA378 looks promising, 

And this file, 

20kHz MFB LP improved RC values OPA378.TSC

Hello,

Do you have any other amplifier requirements?

Supply voltage, Output noise, package size, channel count, power budget?

We are happy to help you find the best amplifier for your application needs.

Thanks for all simulation and design help Michael! 

Best,

Jacob

We are planning to use OPA2374 since this  is a single supply circuit. 0 to 5VDC.

I completely agree with Michael's summary, TLV9062 is a great option from a performance to cost perspective. 

Please let us know if you require any additional support for your design.

Best,

Jacob 

Hey Jacob, good looking table with one error. 

I always compare flatband noise not 1kHz noise. For the TLV9062 that is 10nV not the 16nV you reported - the 1kHz thing is an holdever in the precision group from the audio days. 

Hey Michael,

Good catch, you are absolutely correct. I forgot that the tool uses noise reported at 1kHz rather than flatband. 

Let me take a look to see what it would take to use flatband rather than 1Khz noise for the table generation.

Thanks,

Jacob

Good luck with that Jacob, 

Here is that loop gain plot with the TLV9062 GBP of 9.6MHz, min LG around Fo is about 46dB with this faster part, not sure it is letting me paste that in?

Hey Michael, 

Seems like my dream of using broadband noise for the table will have to stay a dream for now. Data for the table is getting pulled from the Electrical Characteristics table spec for 1kHz noise. 

The plot looks good, seems like ample loop gain for this application.

Thanks,

Jacob

Yea, I would give up on that one - the parametric tables for the >50MHz parts in high speed are populated with flatband numbers.