OPA2244: 2nd order Sallen-Key Butterworth lowpass filter

Hi Benoit,

What tolerance of resistors and capacitors are you using? In simulation the component values are ideal but on the bench caps can have a 10% or worse tolerance which will effect the cut off frequency.

Did you measure the input signal to the circuit to make sure that it is actually 6Vpp like you are expecting?

Can you decrease your input signal frequency to 10Hz or at a lower frequency that is in the flat region of the filter and then increase the frequency until the amplitude begins to decrease? This will help determine where the circuit is beginning to roll off at. 

Lastly, do you have a gain/phase analyzer that you can use to see the gain of the circuit over frequency?

Thank you,

Tim Claycomb

Hello Tim,

Thank you for your message. Please find my comments below:

« What tolerance of resistors and capacitors are you using? » I’m using 1% resistors and 10%. capacitors. If I used correct equations, by taking tolerances into consideration, attenuation should be in the range of [0.628 ; 0.794]. Attenuation I had in practice is about 0.587. Not sure it will help but I have ordered earlier today capacitors with a lower tolerance.

« Did you measure the input signal to the circuit to make sure that it is actually 6Vpp like you are expecting? » Yes, before making output measurement I ensure input is 6Vpkpk by adjusting my signal generator pkpk output. Injected signal is checked by measuring Vin with an oscilloscope in 16 samples averaging acquisition mode.

“Can you decrease your input signal frequency to 10Hz or at a lower frequency that is in the flat region of the filter and then increase the frequency until the amplitude begins to decrease? This will help determine where the circuit is beginning to roll off at.” Thanks for the suggestion. I haven’t done this test yet. I get back to you with figures as soon as I go to the lab.

“Lastly, do you have a gain/phase analyzer that you can use to see the gain of the circuit over frequency?” Unfortunately no but previous test should give us valuable information.

Best regards, 

Benoît,

Hi Benoit,

I look forward to seeing the results. Good luck!

-Tim

Hi Benoit,

you can easily simulate the effect of manufacturing tolerances of filtering components in TINA-TI by using the "Select Control Object" method:

benoit_opa2244.TSC

Also take care, some ceramic dielectric materials are not well suited for active filtering circuits because their capacitance heavily depends on applied voltage:

So only use class 1 ceramic caps (NP0). Another option is to use good plastic film caps (PS, PP, e.g.).

I would not take +/-10% toleranced caps here, by the way, but only +/-5% or even better. Even +/-0.1% toleranced resistors are easily available today :-)

Kai

Hello Tim, Kai

Kai thank you your answer. I’m not familiar with cap dielectric and your answer helps. I’ll also look in more details today the “Select Control Object" method. Thanks.

Tim, I have attached collected data in an excel file.

• First table is the default configuration using X7R 10% capacitors for C1 and C2 (1% for resistors). This table shows Fc around 111,7Hz. Taking into account component tolerances, computed Fc range should be [112.52; 140.30] Hz. Lab result is slightly below this range but within Kai’s simulation results. I’ll review my excel sheet.
• I have not received all the components yet but only 82nF COG 5% capacitor (C2). Second table shows results that are more consistent with simulations and computed data.

So further to lab measurements and Kai’s answer it seems that I’m not having an issue with the OPA2244.... but I have to document myself on capacitor dielectric and secure capacitors specifications in production. Thanks again for your time and advice.

Best regards,

Benoît

data.xlsx

Hey Benoit, I was having a similar discussion on this thread. Each year, the cost comes down on higher precision C0G (NPO) dielectric MLCC caps to higher values. Once you start running monte-carlo's with even 5% C's, the results can be dissappointing. I basically default to 2% caps and 0.5% R's these days for any active filter design where higher Q stages need even better probably. An important point is the low cost ones are in the E24 and E96 steps, but at higher precision - that gets into a standard value fit optimization issue, but just a purchasing point here. 

If you care about nominal and over temp filter accuracy, C0G (also called NPO) dielectric has the best tempco and 2% in volume are sub .06 each up through 47nF it seems (today, that will go up in value range as time goes by). 

Hi Benoit,

you could modify the circuit to decrease the number of necessary precision cap types :-)

benoit_opa2244_1.TSC

Kai

Thanks for the link and information very helpful.   

Thank you kai. This values open more flexibility in capacitor selection. I have ordered some parts and I'll test this configuration next week. I'll keep you informed. 

Good luck on your testing!

Hi Benoit,

I'm glad you were able to find a working solution.

Happy New Year!

-Tim Claycomb