2
\$\begingroup\$

I am building a low pass filter IC in 0.25 micron CMOS. My design requires 14 Mega-Ohm resistors. I want to replace them with an active load in order to save space. If possible I would like to use a current source as a resistor.

What would be the most simple and compact design?

I've already tried switched capacitors but had various difficulties, including NOC generation.

A simple floating gate MOSFET would require a transistor length of 2000 microns. If worst comes to worst I'll work on balancing transistor size with capacitor size.

Thanks a lot.

Edit: Topology is Shallen key. Output should be a decent sine wave of frequency 20kHz.

R1 =R2=14MOhm and C1 = C2= 0.56pF.enter image description here enter image description here

\$\endgroup\$
  • \$\begingroup\$ I'm not sure this is achievable. What kind of stability do you need for 14MOhm so your filter still meets its requirements? \$\endgroup\$ – user6972 Nov 26 '14 at 18:25
  • \$\begingroup\$ interesting that your low pass filter requires multiple of one valued resistors. Are you sure your filter is correct? What are the parameters for this filter? \$\endgroup\$ – Funkyguy Nov 26 '14 at 18:39
  • \$\begingroup\$ I suppose your filter topology is based on one of the known active RC-filter structures. Did you also consider OTA-C techniques? This is one of the techniques especially suited for integrated circuits. \$\endgroup\$ – LvW Nov 26 '14 at 18:51
  • 1
    \$\begingroup\$ @Funkyguy Having equal components is fine, and ends up having a Q of 0.5. It is known as a Linkwitz-Riley filter, and is popular in audio because there is no build up when two pass bands meet. \$\endgroup\$ – Matt Young Nov 26 '14 at 19:33
  • \$\begingroup\$ @MattYoung Interesting, I've never heard of that kind of filter! \$\endgroup\$ – Funkyguy Nov 26 '14 at 19:35
1
\$\begingroup\$

If you really are doing CMOS analog design and you are using large resistors you're already heading in the wrong direction. There are all sorts of issues with resistors in CMOS processes like: parasitics (capacitance, non-linear capacitance), Low resistance values and extremely poor matching. NOC design (Non-overlapping clock) is trivial in comparison. At least you can be sure it will work in Si. I'd recommend a switched capacitor approach using differential signalling, there are even good topologies in layout that help cancel/balance stray capacitance effects.

Since you already have you floating capacitor (Feedback) you obviously are already using a MIM or PIP capacitor, you've all the modules necessary in your process.

\$\endgroup\$
0
\$\begingroup\$

R1 =R2=14MOhm and C1 = C2= 0.56uF.

If these values were used in a sallen key filter the cut-off frequency would be: -

enter image description here

0.02Hz and not 20kHz.

Try recalculating based on this website and then decide what you need to do.

\$\endgroup\$
  • \$\begingroup\$ Turns out my capacitance was 0.56pF. Really sorry about that. Updated in post. \$\endgroup\$ – Keri Nov 26 '14 at 21:02
  • \$\begingroup\$ I don't think this topology will work with 0.56pF because gate capacitance of inputs is going to be about that value (or greater) and have a significant tolerance that makes the response unpredictable. I'd definitely want to consider switched cap filters more. \$\endgroup\$ – Andy aka Nov 26 '14 at 21:25
  • \$\begingroup\$ What's the lowest capacitance I could get away with for my topology? \$\endgroup\$ – Keri Nov 26 '14 at 21:34
  • \$\begingroup\$ It depends how accurate you want the sallen key filter to be and what the variation in parasitic circuit capacitance you have. \$\endgroup\$ – Andy aka Nov 26 '14 at 22:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.