Author Topic: Designing Sallen-Key stages in S domain  (Read 832 times)

0 Members and 1 Guest are viewing this topic.

Offline DmeadsTopic starter

  • Regular Contributor
  • *
  • Posts: 171
  • Country: us
  • who needs deep learning when you have 555 timers
Designing Sallen-Key stages in S domain
« on: March 25, 2020, 06:36:10 am »
Hello! Hope yall are healthy!

Im a digital guy working on some DSP stuff, but Im building a high performance (relative to my knowlege) anti-aliasing filter using the S plane. Still learning the analog stuff, but its really cool!

My filter is a 6 pole LP butterworth (at least I think it is butterworth, there is no stop/passband ripple).

fc = 100KHz.

I found my ampilification factors from trig using the s plane circle. ( all poles are evenly spaced on left side) They can be view on top next to the Bi-quad instance number.

During simulation, i found some interesting results. After the first sallen key stage (first 2 poles), the -3db point was at 50 KHz, after the 2nd SK stage, 68 KHz, then after the third stage, 81 Khz.

I know filters arent perfect but it seems I should get a better -3db point (closer to actual cuttof of 100 KHz).

Does anyone know what I did wrong? Is it calculation error?

Thanks.

-Dom
 

Offline Wimberleytech

  • Super Contributor
  • ***
  • Posts: 1134
  • Country: us
Re: Designing Sallen-Key stages in S domain
« Reply #1 on: March 25, 2020, 11:38:27 am »
Try R6 = 10K

Also, you will need an opamp with more GBW than the 741.  Experiment with a generic opamp and dial in various GBW values.
« Last Edit: March 25, 2020, 11:59:08 am by Wimberleytech »
 
The following users thanked this post: Dmeads

Offline Siwastaja

  • Super Contributor
  • ***
  • Posts: 9336
  • Country: fi
Re: Designing Sallen-Key stages in S domain
« Reply #2 on: March 25, 2020, 12:33:00 pm »
Remember to simulate with component tolerances. I recommend you play around with Analog Filter Wizard ( https://tools.analog.com/en/filterwizard/ ); it's easy to use and will draw the component tolerance envelope for you.

The more complex the analog filter is, the more critical the component choices become, including opamp parameters and the tolerances of the passive components; sometimes even the parasitics (such as capacitive coupling).

It may so happen that designing a simpler 2- or maybe 4-pole filter and just using an ADC with higher sample rate, to use a steeper filter in digital domain, ends up being not only more accurate, but possibly cheaper as well. The only function for the analog filter being satisfying the Nyquist criterion, you can move that further away by just sampling at higher freq, thus make the less steep transition region of a lower-pole filter acceptable. Then, in the digital domain, it's trivial to make an almost brick-wall filter, avoiding all component nonidealities.

How important the stopband attenuation is for you? I.e., do you have actual strong, disruptive signals to stop, or is it just to prevent some low-level noise from aliasing?
« Last Edit: March 25, 2020, 12:37:20 pm by Siwastaja »
 

Offline DmeadsTopic starter

  • Regular Contributor
  • *
  • Posts: 171
  • Country: us
  • who needs deep learning when you have 555 timers
Re: Designing Sallen-Key stages in S domain
« Reply #3 on: March 25, 2020, 06:57:09 pm »
stopband attenuation isnt that important, also i could probbly tolerate as low as -20db/dc rolloff too, i was just playing around
 

Offline DmeadsTopic starter

  • Regular Contributor
  • *
  • Posts: 171
  • Country: us
  • who needs deep learning when you have 555 timers
Re: Designing Sallen-Key stages in S domain
« Reply #4 on: March 25, 2020, 08:46:27 pm »
Also, you will need an opamp with more GBW than the 741.  Experiment with a generic opamp and dial in various GBW values.

you're right, I got it all dialed in. thanks.
 
The following users thanked this post: Wimberleytech


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf