0
\$\begingroup\$

I just looked over a circuit designed by my predecessor, which is used to interface with a pressure sensor. It works, but I have some doubts about the way the input filtering is done. The simplified schematic looks like this:

schematic

simulate this circuit – Schematic created using CircuitLab

The pressure sensor has a differential, low level and relatively high-impedance output (up to 25kOhm) which is scaled up into a larger differential signal by those amplifiers. A differential amplifier at the output (not shown) then turns this signal into a single-ended one. In other words, this is part of a normal three-amp instrumentation amplifier.

The gain setting resistor is part of the pressure sensor, and it is factory-calibrated (between 2.5kOhm and 12.5kOhm) so that it compensates the variation of the full-scale output in each individual sensor. This way, replacing the sensor will not change the output much even though each sensing element has a different full-scale output. In order for this to work, R1 and R2 need to be 100kOhm.

This is all according to an applicaiton note (TN-003 by ICSensors) which I unfortunately couldn't dig up on the internet just now. However, my predecessor added the capacitors C1 and C2, probably to suppress interference issues we saw in tests - the sensor is connected via an unshielded cable inside our device, and since the feedback line to the gain resistor has such a high resistance going to it, it looks like a good candidate for picking up noise.

The output does not need to be fast (~1Hz is fine), so rejecting high frequencies seems like a good idea. However, seeing a large capacitor between a pair of op amp inputs makes me uncomfortable. I plopped this circuit into LTSpice and saw that it is stable (even with much larger capacitors), but it rings a lot due to a resonance at ~4.2kHz.

Now I'm wondering, wouldn't it be better to put C1 and C2 parallel to R1 and R2 instead? That way, the resistance of the gain feedback line would be lowered at high frequencies, and the amplifiers would somewhat attenuate high frequencies, too. The simulation looks good, but since I'd still consider myself a novice I'd appreciate some confirmation whether this is a good solution or if I'm overlooking something.

\$\endgroup\$
  • \$\begingroup\$ Is there any way you could provide us with the .asc file, or is it company secret? With that we could see what you see, but currently we can only imagine what you possibly might have seen. \$\endgroup\$ – PlasmaHH Nov 27 '14 at 15:26
  • \$\begingroup\$ If you insist, you can get the .asc file here: pastebin.com/ptwggx3J - V2 only changes the common mode input voltage, V3 the differential mode. V4 is supposed to simulate EMI. C1 and C2 are placed according to my suggestion, you can connect C4 and C5 as shown above to get the current state of things. \$\endgroup\$ – Medo42 Nov 27 '14 at 15:39
  • \$\begingroup\$ Since I don't know if your emi/noise model is a good way to simulate it, I am not confident enough for an answer, but the "old" version really looks fishy (in ac as well as transient), and your version looks much better. I don't know if you know this little trick, but I like to do "component here or there" tests with a .step param Ro list 1n 1e300 and then put resistors in series with the caps and give them the values {Ro} and R=1e300-{R0} respectively, which basically means it "steps" through both versions; this makes it easier to compare them imho. \$\endgroup\$ – PlasmaHH Nov 28 '14 at 10:34
0
\$\begingroup\$

Yes, putting C1 and C2 in parallel to R1 and R2 provides good RF rejection on the gain resistor lines.

I built the circuit and did some "poor man's EMC testing" by attaching a long cable to the sensor and wrapping it around my mobile phone, then calling the mailbox. The result of the modified circuit were very good, showing only a small increase in noise, while an alternative configuration showed readings all over the place.

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

Just as a second opinion putting caps across the inputs of each op amp is one good method to stop noise. The more noise the more exotic the filtering needs to be. As long as the circuit works at the current noise level no more filtering is needed. Putting Caps in parallel with R1 and R2 would increase negative feedback at high frequencies, but why change what works.

\$\endgroup\$

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.