RF field sensitivity of pressure transducer

[logbook]

I have a loose pressure transducer assembly, running off a bench power supply and it is remarkably sensitive to RF interference from my mobile phone. I am using my mobile phone because that is the only high power RF source I have access to. The reason for all this is that the overall equipment has such a pressure transducer and its RF field immunity is not as good as we would like it to be.

Now this pressure transducer assembly is not my design, it is bought in. It has three terminals; power, ground and output, the output being scaled to 5V for full scale. This is a strain-gauge type pressure transducer with a two stage amplifier, giving gains of x37 and x8. I have traced out the circuit to see what can be done to fix it.

The manufacturer sells a version in a solid casing with feedthrus, but this is too big and too expensive.

The front end of this pressure transducer is a two opamp instrumentation amp, which is a poor start admittedly. I have shorted across the bridge outputs and I can still see 900mV spikes coming out of this first stage when the phone approaches it. (this is a UK phone from Orange [Sagem] ). The opamp is a low power 2MHz BW type and the source impedance is 3K to the amplifier. I decoupled the bridge output to ground using 47pF caps. Basically the circuit consists of one opamp, a couple of resistors, and a 100pF feedback capacitor.

Now I know LF circuits can rectify/detect RF signals but the magnitude of this effect seems extraordinarily large. I am going to try other high power/higher bandwidth opamps to see if I can get a better one. But what I am trying to figure out is what the mechanism is exactly. It seems to me that the RF signal is getting into the opamp input, perhaps as pickup in the feedback path. This signal is rectified by the non-linear input capacitance and the resultant DC signal is amplified by the LF capability of the opamp. This whole circuit is only 1 inch in diameter so the scope for radiated pickup is not that great.

Does anyone have a better theory or some insight into the problem and its solution?

 

[buzzp]

Thats kind of a tough one. When we did similar tests we only seen micros resetting which was easily fixed with a different cap on the clr pin. Sounds like your data is getting corrupted from the RF (what power is it anyway? we used little 5W radios). The best advise I can give you is to use a differential amp to condition the signal first to take care of any common mode noise picked up on the wires to the ducer. This may not solve the problem either, especially if your putting the phone right next to the device. The only real solution, in this case, is to use a shielded enclosure with attention to gaskets and such.
I assume your getting a large pressure (at least voltage) reading when the phone draws closer? How close are we talking here?
Your dealing with two issues: one, the RF picked up before the wires enter the device and two, the RF picked up by the device itself (which the solution would be a shielded enclosure). The diff amp should take care of junk picked up by wires along with some basic filtering (bypass cap on the input pins along with your series resistance of 3k).

 

[sreid]

If stray RF gets into the opamp inputs, stange things can happen. It doesn't take too much to make the first opamp stage saturate. Generally RC Filters near the inputs will solve the problem. RF Ferrite beads are also effective if there are lead in wires.

 

[Comcokid]

Strong RF fields are rectified at the pins of a semiconductor device to result in the noise. The longer the trace connection or wire at the pin, the more likely a strong RF signal is present. The longest wire connections will be the power and output wires.

Is the case of the transducer connected to ground? If not, and you need isolation, connect the case to ground through a capacitor.

To be effective, capacitors have to be small in value so they are well below their self-resonant frequency - use leadless SMT capacitors. Your selection of 47 pf is good.

Place a 47 pf (or less) capacitors directly at the power pins to ground. Use a leadless SMT cap. Precede it with a ferrite bead(s). If possible, make a PI network of two capacitors and ferrite beads. Do the same on the output wire.

Place 47 pf directly on the bridge outputs to ground. Shunt gain stages with 47 pf across the feedback resistor. Make sure unused devices (or package pins) in the same package are properly disabled. Place a cap or ferrite bead on the + excitation terminal of the bridge (note, some current excitation circuits can't handle a capacitor).

You have to experiment - sometimes a capacitor at a input pin or on a feedback may only act to store the rectified RF resulting in offsets when RF is present.

Also realize, a very close cell phone may result in a field strength stronger than the 3 V/M that is typicial of basic immunity testing.

 

[logbook]

Buzzp, you missed the point I made that the pressure transducer is already connected to a two opamp instrumentation amp (= diff amp).

Although the whole circuit is on long wires, the pressure transducer and signal conditioning all fits in a block 1 inch cubed. I think the bridge outputs can be de-coupled effectively and the problem remains. The tracks can be more than 2cm long. At 100V/m that suggests less than 2V rms. Now mind you it wouldn’t take much imagination to see this being rectified to 20mV and that would be enough to cause the problem.

The transducer case is ceramic and hence it is not grounded.

I noticed today that this opamp has protection diodes right across the inputs. There are other types that do not have such diodes. The absence of the diodes may well mean that this 2v signal does not get rectified and therefore the non-clamped opamp may well be more immune to the effect. Hopefully the new amps will arrive tomorrow …

Thanks for the discussion.

 

[btrueblood]

You are probably getting the strain gage of the 'ducer to act as an antenna, thus the noise is coming in on the same wires as the signal. I'd expect that a metal-shrouded strain gage would be better shielded, thus more immune to noise. Can you put filter caps between the strain gage and the the op-amp? I'd agree that eliminating protection diodes on the inputs could make the noise go away.

Good luck

 

[logbook]

Thank you, but you missed the part in my first post where I said I "shorted across the bridge outputs" meaning only common-mode noise should be left, and then I grounded that bridge point via 47pF capacitors, which should have eliminated the CM as well.

Actually answering this has forced me to think again. I had to use wire-ended caps to do this CM shorting, spanning the 1cm gap between the inputs and the ground. This means I still get a pickup loop, given the high ambient field. Oh this is a tricky area! It is going to be difficult to test any ideas without re-doing the layout.

I still have to try the other opamps though.