1
\$\begingroup\$

I have a low level DC signal from a thermopile sensor of 0-9.1mV with 7uV increment. I feed this to an INA114 instrumentation amplifier with gain of 550 to get 0-5V output.

I tried introducing RF noise by means of GMS (phone call or SMS) with the output moving up from 0 - 1V (0 - 1.82mV input). I have a voltmeter on the input and output to monitor it at the same time. With the noise present, the input is perfectly fine while the output ruined.

I then set the gain to 1 so that input is equal to output. Still, the input is perfectly fine while the output is ruined. It seems that the interference is hitting the INA114.

Why is this happening and how would I overcome it?

\$\endgroup\$
6
  • \$\begingroup\$ Filter. filter and filter... GSM phones are vile things for interference but it's probably too late to ban them, so it's the best you can do. Start with a capacitive delta (C from one leg to the other, and from each leg to ground.) as close to the InAmp inputs as possible. If the sensor wires are long, add a ferrite common mode choke, just before the caps. \$\endgroup\$
    – user16324
    Commented Jul 2, 2014 at 8:33
  • \$\begingroup\$ @BrianDrummond, that's a good point. The noise is not likely being received by the op-amp, but rather amplified by it with a gain of 550 so it's obvious at the output but too small to notice by your multimeter at the input. I am missing how you are suggesting to wire the capacitors. Are you saying from The two inputs to ground and then a capacitor between the two inputs? The capacitors to ground I can see provide low-impedance to ground, but what does the cap across the inputs do? \$\endgroup\$
    – sherrellbc
    Commented Jul 2, 2014 at 9:42
  • \$\begingroup\$ @sherrellbc When I set the gain to 1 i.e. same reading of input and output, only the output signal is affected by the noise. The input is perfect. Both signals are equal and at mV range, but when the noise kicks in, only the output readings are messed up. \$\endgroup\$
    – user46635
    Commented Jul 2, 2014 at 9:53
  • \$\begingroup\$ The gain doesn't really matter : the RF noise is not primarily being received or amplified by the opamp, but demodulated to baseband (i.e. rectified) by it. Gain may affect how much spurious output you get, but even at gain=1, the RF will be rectified by internal circuitry. I still think banning cellphones is the best solution... (so do airlines and hospitals, to a certain extent) Am I grouchy this morning? \$\endgroup\$
    – user16324
    Commented Jul 2, 2014 at 11:21
  • \$\begingroup\$ This kind of thing is common with thermocouple amplifiers in an industrial environment.. rectification occurs in the input stage. You need series impedance and parallel capacitance that's effective at RF frequencies. If you keep the noise amplitude down to mV, then the input stage will be linear enough that it will only affect the input-referred output by a small amount. \$\endgroup\$
    – Spehro 'speff' Pefhany
    Commented Jul 2, 2014 at 12:05

1 Answer 1

Reset to default
4
\$\begingroup\$

It is well-known that Inamps can suffer from degraded performance in the presence of EMI. Here is a document from ADI that gives tips on this subject and below is a small extract: -

enter image description here

Of vital importance (of course) is the requirement for the engineer to read the data sheet and follow specific guidelines for the chip such as adequate power supply decoupling and layout methods. A lot of EMI issues are caused by bad layout and generally not following the data sheet.

I'm not saying the OP hasn't followed the guidelines in the data-sheet of course but it would be interesting to see a picture of the PCB layout. If it's a breadboard then basic EMI protection is out of the window!

\$\endgroup\$
3
  • \$\begingroup\$ Naah with a bit of effort you can do decent enough filtering on a breadboard; just use dead bug style. \$\endgroup\$
    – user16324
    Commented Jul 2, 2014 at 11:24
  • \$\begingroup\$ I'm afraid I'm still at the breadboarding stage. But I still can't get why the output is noisy while the input is not when having the in amp at gain 1. It's like the input goes in perfectly but somewhere inside the chip, it gets noisy hence the output. How did the rf noise entered the chip? \$\endgroup\$
    – user46635
    Commented Jul 2, 2014 at 11:40
  • \$\begingroup\$ Regarding that RC filter at the input, what configuration should I use if my setup is single-ended? The thermopile, which is just a series of thermocouples, is connected to the + and - pin of the INA114 with a 10k resistor from - pin to ground which serves as return path for the bias current which I followed from the datasheet. \$\endgroup\$
    – user46635
    Commented Jul 2, 2014 at 12:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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