Detecting water (via resistance) between two metal probes using 3.3v MCU's ADC

Question

For the past few days, I've been going in circles trying to plan a circuit to detect water between a pair of metal probes (or more precisely, between several pairs of stainless steel probes approximately 1 inch apart connected in parallel, and two "water sensor" circuit boards that consist of two zig-zagging 50-mil traces separated by 50-mil space).

The fundamental problem is, I don't really understand the finer points of Ohm's Law (or at least, the parts that might deviate from what I suspect is actually a simplified special case... when you aren't just talking about a simple resistor limiting current between +3.3v and ground).

Consider the following circuit (in series), using a 3.3v ESP32:

• GPIO pin that's normally left floating as an input (to minimize galvanic corrosion of the probes), but configured for output & taken high (to 3.3v) shortly before triggering ADC conversion.

• 82-100 ohm resistor (to limit the maximum possible current flowing between the GPIO and ADC pins to 40mA or less, even if there ends up being a dead short and zero ohms between the probes

• a pair of probes (normally, with air insulating them... potentially, with tap water insulating them a bit less)... and two or more such pairs connected in parallel (so there could be current flowing between none of them, between one pair, or between multiple pairs)

• ADC input pin on the ESP32

The things that have me scratching my head:

• If you assume 3.3v on the GPIO pin and only 82 ohms between it and the ADC pin, how many volts will the ADC sense/measure/see?

• Does a 12-bit ADC actually have enough "headroom" to meaningfully sense the difference between all-but-infinite resistance (82 ohms + an inch of open air) and whatever resistance you'd likely encounter between two stainless-steel probes immersed in tap water? Or at best, would you end up with readings that just jittered between something like 0 to 5 and 2 to 7 (out of 4095), and were so close and mostly-overlapping, you almost might as well randomly flip a coin to decide whether or not there's water present?

I suspect that the impedance of the ADC itself is an important detail that I don't know the value of.

I'm kind of embarrassed to be asking this, because this is apparently considered to be an absurdly-simple circuit... but all the examples I've seen basically consist of a voltage comparator whose output directly triggers a piezo, as opposed to using a MCU to measure it through a proper ADC. I'm not sure whether that's because the amount of voltage at the far end would be too low to reliably measure with a 12-bit ADC, or just because prior to just a few years ago, the very idea of using a MCU and ADC for something like this would have been prohibitively expensive overkill (instead of, like, $9 for an ESP32-C3 that you need anyway for the wi-fi link).

Do I actually need to do something like have the ESP32's 3.3v GPIO trigger a MOSFET switching 5.0v, then amplify the other end with an op-amp and/or run it through an optocoupler or something? Or am I way overthinking something that isn't nearly as complicated as I'm making it?

Answer 1

If you assume 3.3v on the GPIO pin and only 82 ohms between it and the ADC pin, how many volts will the ADC sense/measure/see?

The 82 Ω to the ADC will form a potential divider with the input impedance of the ADC. I would expect this to be in the order of 100 kΩ to several MΩ. The datasheet should give this value. In any case the if the ratio was 82 Ω : 82 kΩ (1 : 100) the ADC would see 99% of 3.3 V (which is fine). A higher input impedance would improve this further.

Does a 12-bit ADC actually have enough "headroom" to meaningfully sense the difference between all-but-infinite resistance (82 ohms + an inch of open air) and whatever resistance you'd likely encounter between two stainless-steel probes immersed in tap water?

"Headroom" isn't quite the right word as the dry reading will be 100%. "Resolution" or "span" of readings might be better. This would be very easy to test with a 3.3 V supply, your multimeter and test probes. Check the open-circuit voltage (3.3 V) and then apply the various test conditions to your probe while measuring the voltage.

I suspect that you'll find that 82 Ω is quite a stiff pull-up and that a wet probe may struggle to pull the voltage down significantly so be prepared to check with 1 kΩ and 10 kΩ.

Or at best, would you end up with readings that just jittered between something like 0 to 5 and 2 to 7 (out of 4095), ...

More likely > 4000, if anything (he said without doing any tests).

I suspect that the impedance of the ADC itself is an important detail that I don't know the value of.

It will be high relative to what you're measuring so I wouldn't worry about it (unless you end up with very high pull-up resistors). That information must be available from the datasheet?

^{simulate this circuit – Schematic created using CircuitLab}

Figure 1. Equivalent circuit. R_PU is the pull-up resistor.

So you're saying that the circuit would NOT be, GPIO (outputting 3.3v) -> resistor (limiting current) -> probe -> water -> probe -> ADC pin?

No. That wouldn't work. You want to create a potential divider. The voltage at the ADC input will be

$$ V_{ADC} = 3.3 \frac {R_{SOIL}} {R_{PU} + R_{SOIL}} $$

Now it should be clear that the ADC input impedance makes very little difference to the reading.