5
\$\begingroup\$

Sometime ago I had to implement an on-chip (CMOS 65nm) low-pass filter (first order) with very low cutoff frequency (lower than 100mHz). I used PMOS pseudo resistors and momcap, and had to check all the corners to make sure the cutoff freq is low enough. Obviously the output of this filter feeds the gate of another MOS transistor. However, the PDK did not include the gate leakage current, so I used IO devices with large Vth (and other tricks to solve the voltage headroom issues) in the hope of lowest possible leakage current. But that didn't help much and the DC voltage drop across the pseudo resistors made serious issues.

I'm just wondering if someone has experience designing on-chip, very low frequency, filters where leakage current is present in the process.

\$\endgroup\$
  • 1
    \$\begingroup\$ Couldn't you do some of the filtering digitally? \$\endgroup\$ – Justin Sep 15 '15 at 20:29
  • \$\begingroup\$ That seems a very weird design. Why not using counter or something? \$\endgroup\$ – Gregory Kornblum Sep 15 '15 at 20:48
  • \$\begingroup\$ Well, this was an analog chip, current consumption was a big concern too. So I was looking for a very simple design. Implementing a sigma-delta modulator with a simple counter to smooth out the output, as you suggested, is my next approach. Still working on the details though. So I wanted to reach out to analog experts and find out my options. \$\endgroup\$ – Ali Sep 15 '15 at 21:21
  • 1
    \$\begingroup\$ Way outside my knowledge area, but have you considered a switched capacitor filter? I seem to remember that they were held up as an example of a filter which can be implemented in an IC. en.wikipedia.org/wiki/Switched_capacitor \$\endgroup\$ – mkeith Nov 26 '15 at 6:53
1
\$\begingroup\$

try use some followers and cascodes to buffer the S, D & B of the output device. That way there'll be very little voltage across its gate oxide.

High VTH devices won't help (unless you're worried about subthreshold leakage) -- if you have any thick oxide devices, those might work better.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Could you elaborate a little more? A hand-drawn schematic of your suggestion would be great. And yes, it's the subthreshold leakage giving me headach! \$\endgroup\$ – Ali Dec 6 '15 at 2:45
  • \$\begingroup\$ I forgot to mention that usually I/O devices have thicker oxide \$\endgroup\$ – Ali Dec 6 '15 at 2:46

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.