1

So I have a bit of an annoying issue. I have a circuit that measures a 4-wire PT100, using a 24 bit AtoD (ADS1244). After following some reference designs to get me started, I have a circuit which works and gives me accurate results. The (simplified for ease) input circuit looks like this:

schematic

simulate this circuit – Schematic created using CircuitLab

Please note I have excluded filtering and other circuitry irrelevant to the question.

So, as stated before, this circuit works well and gives me exactly what I need. However, this product allows for a 3 and 4 wire probe to be used (that part of the circuit not shows) so this means at some points there is the chance of an open circuit probe.

I have found that during startup, if no probe is attached, then occasionally, when a probe is inserted, it gives me all sorts of rubbish as an answer and will sometimes give me a random reading during startup (it is supposed to read ERR when out of range).

I managed to narrow down this behaviour to the auto calibration of the AtoD IC during startup. It requires a minimum of 0.5V on the reference to work correctly (according to the DATASHEET) and on the occasions that the product is switched on when no probe is attached, it does an auto cal with a floating reference which ends up giving me rubbish answers until a new auto cal is done with a probe attached.

I managed to get around this problem by adding an NMOSFET to the VREF- and during startup, this is pulled low to ensure there is always a valid VREF voltage and the auto-cal is successful.

This worked fine until I would very rarely get strange readings when no probe was attached. Maybe 1 in 100 times. This turned out to again be the reference as once the probe is removed, there is again a floating reference so there is the chance for undesirable behaviour.

I can't seem to think of a way around this, or detect that the probe has gone open circuit, especially in the circumstances where a "valid" reading is produced due to the floating reference. Does anyone have any idea for things I could try to get around this?

SamGibson
  • 17,231
  • 5
  • 37
  • 58
MCG
  • 11,809
  • 4
  • 34
  • 70

1 Answers1

2

Normally a high-value resistor conected to the pins which could be completely open-circuit is sufficient to get a reading which will reliably detect the open-circuit condition. The resistor might go to ground or your reference voltage rail.

This slightly compromises the "ideal" behavior but a few-MΩ resistor with a maximum wiring resistance of (say) 100Ω does not produce much error. Since you have an MCU you could also turn the resistor off post-detection via an inexpensive analog switch (or, say, connect it to GND rather than Vref) so the effect is orders of magnitude less again. Of course that does not allow for detection of some modes of probe connection failure during operation unless you deliberately include such a test periodically.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • Ideally, I need to be able to detect the open circuit condition once it happens so I can ensure to give an ERR reading. I can force the IC to auto-cal itself which would be the way I do it at startup (use the MOSFET to pull VREF- to GND) so that would satisfy the periodic checking, but ideally I need a way to detect it once it happens – MCG Oct 24 '22 at 10:34
  • In controller designs I typically used a 10MΩ resistor so (in my case) around 500-800nA of break protection test current. We also wanted to be able to detect a break in cases where there was a bit of leakage in the wiring, so there was a lower limit to the test current. – Spehro Pefhany Oct 24 '22 at 10:50
  • 1
    I could look at maybe a high value resistor holding the VREF- go GND, and see if calibration can take out any errors it induces – MCG Oct 24 '22 at 10:54
  • Tried it, to no avail. Did the math (should have done it before tbh) and with the resistors in the chain, a 10M or even 1M resistor wouldn't allow 500mV at VREF so I still get random results in this situation. I can't change the upper resistor values as I have them to give me the correct VREF across the entire measurement range and enough ADC resolution to have a 0.01°C/F resolution across that range too. – MCG Oct 24 '22 at 14:07
  • 1
    It might help to carefully redraw the circuit, including a copy and paste version for 3-wire, showing explicitly the wires and their resistances. – Spehro Pefhany Oct 24 '22 at 15:26
  • I left them out because it didn't make a difference. When open circuit, the "sense" pin on the microcontroller that checks for 3 or 4 wire assumes a 4 wire measurement. The wire resistance will be essentially removed from the 4 wire measurement. The extra circuitry is irrelevant to the problem, which I am certain about having repeated tests on PCBs that had only certain parts of the circuit fitted, so I boiled it down to essentially the circuit you see above only and saw the issue – MCG Oct 24 '22 at 15:39