Protecting ADC input from overvoltage and reverse polarity

[bonkers]

I am working on a circuit which is required to measure signal in the range of 0V to 10V. To shift the signal to an acceptable range (0 - 3.3V) I'm using a 3-resistor divider, which also adds a little offset. The circuit is also supposed to be used to measure resistance by injecting current via D4.
My question is as follows - if the input protection adequate, and can you point me to a better way to do the job (the job being measuring voltage in the 0 - 10V and resistance 0 - 500 ohms).

Many thanks

[Kleinstein]

The diode D2 should not go to the reference voltage, but more to a supply (if there is enough current drawn) or a zener or similar clamp to ground.
Ideally there should be one more resistor (e.g. 1 K range) from between D1 and D2 to the ADC input to limit the current flow. The diodes still allow for a slightly negative voltage and in case of high current spikes this many still be a bit over 0.5 V even with a schottky-diode.

The constant current source would need to withstand a negative voltage, as much as D3 allows.
R4 is not really helping much, but is included in the resistance measurement. If needed to protect the current source additional resistance would be more directly in series with the diode. So not shared in the voltage reading path.

Depending on the expected faults one could also consider to use a PTC / polymer fuse, e.g. to limit the fault current to D3. As shown D3 may fail if exposed to mains.

[MasterT]

What kind of the ADC? If no build-in buffer, than 10k would definitely create linearity problems. In this case low voltage opamp (rail-to-rail outputs) needed, and consequently D1&D2 could be omitted, opa has its own protection circuitry.

[bonkers]

Thank you for the detailed reply. Vref is actually a LDO on the board, not a mere reference.

R4 is included so that the circuit will measure some voltage when a 0 ohms resistor is connected to the input (in resistance-measuring mode), as you observed.

The circuit is required to withstand +/-50V inputs when the board is not powered on. I'll put in a fuse and build and test it.

@MasterT - I'm using ESP32S3, and was thinking about a buffer as well, but the signal I am measuring will be very slow, so I thought I might get away without it.

Thanls

[David Hess]

With C1 you can probably get away with the roughly 3 kilohm parallel resistance.  The 10 kilohm series resistance plus the diodes should be fine for overload protection depending exactly on your requirements.

[dobsonr741]

What accuracy are you after? What source impedance are expected as maximum? ADC resolution?

The protection will vary a lot if you start going beyond 0.1%

[Kleinstein]

A LDO for the reference may not be good in sinking current. Having the diode towards the supply has the advantage of powering the circuit when the circuit is otherwise not powered. A fixed clamp (e.g. to a zener diode) would not help against getting above some Vcc+0.5 V (the usual limit for the input to chips), when VCC is still 0.
One may want both a diode to a clamp and a diode to Vcc, if VCC can't sink enough.

[Kanbus]

Do a buffer before the ADC and limit its supply voltage so its max output swing is within ADC’s limits?

[bonkers]

Thanks all for the replies.

Adding a buffer seems a good idea - I've removed the clamps and added in an opamp.

[dobsonr741]

I think the D1-D2-D3 diodes are still required. A usual solution is not to use Vref or Vcc to clamp, rather a thermally coupled pair of PNPs to cut close to Vcc or Vref. Otherwise, the +50V input can pull up Vcc or even more so the Vref to create damage.