2
\$\begingroup\$

The Microchip (ex-Atmel) ATSAMD10 has a 12-bit SAR ADC with the option to use differential input. I'm generating a virtual ground (\$V_\mathrm{mid}\$) at the midpoint (512 mV) of the reference voltage (1.024 V), with the input signal centered around that.

enter image description here

My question is regarding the reference plane for the analog section preceding the ADC. There is a simple 2nd order low-pass Sallen-Key for anti-aliasing, and a placeholder for another RC filter after the Sallen-Key. Should the plane beneath the analog section be ground, or \$V_\mathrm{mid}\$? My feeling is that it should be \$V_\mathrm{mid}\$ as the filters are all connected to that reference voltage. The only ground connections in that region are for the opamp power.

enter image description here

This is for a two layer board, currently with a ground pour on the bottom layer (green) with some \$V_\mathrm{mid}\$ traces through for routing. All the analog inputs are on the bottom of the microcontroller at the top left.

enter image description here

The system is not particularly demanding, but (notional, given the ADC's performance) LSB performance would be nice.

  • Source impedance: 18 \$\Omega\$
  • \$V_\mathrm{LSB}\$: 250 \$\mu V\$
  • \$f_c\$: 1.25 kHz
  • \$f_s\$: 4.096 kHz
\$\endgroup\$
2
\$\begingroup\$

It (almost) doesn't matter what voltage the plane is at, as it will only be capacitatively coupled.

The important thing is to use the lowest impedance domain so it can rapidly absorb/source charge. This is usually your "ground".

12 bits, at your voltage and frequency...use ground and you should be good to go.

\$\endgroup\$
8
  • 1
    \$\begingroup\$ My analog days were dual supply, so I was uncertain about this - makes sense, it's just dumping things to any fixed low impedance. I'll wait a bit before accepting in case there are other thoughts, but otherwise all good :) \$\endgroup\$ – awjlogan Jan 26 at 15:17
  • \$\begingroup\$ @awjlogan I think gnd too. Your mid-bias is going to have a decoupling caps anyways, right? From your divider output (or opamp output buffering the divider)? \$\endgroup\$ – DKNguyen Jan 26 at 15:33
  • \$\begingroup\$ @DKNguyen Yes - thanks, that's a good point, there's no decoupling on the output from the voltage reference at the moment. I don't want to put a capacitor on the output of the buffer opamp, as it'll need a series resistor for stability and I don't want to up its impedance. So 0 V ground is the current front runner. \$\endgroup\$ – awjlogan Jan 26 at 15:41
  • \$\begingroup\$ @awjlogan I meant to say "decoupling cap from divider/buffer output to each rail" \$\endgroup\$ – DKNguyen Jan 26 at 15:43
  • 2
    \$\begingroup\$ @awjlogan I don't think necessarily wrong. I put those on dividers generating a mid supply but maybe not needed with opamps due to the low output impedance. Yeah, maybe not. \$\endgroup\$ – DKNguyen Jan 26 at 15:51

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.