STM32L151VE ADC reads 20% lower value on one of the ADC inputs (channel 18, pin B12)

[Pavel Krupets]

Not sure if I messed something up. Signal is okay (on the pin). No differences with layout compared to other inputs. ADC is at full speed (16 MHz).

On image it is sensor 22 (reads ~2500). Sensor 23, 24, 25 all read normal values (~3000).

Voltages on all those pins is ~2.8V.

VREF, VDDA, VDD = 3.3V.

[Uwe Bonnes]

Your layout does not help to identify what Port Pins you use. How fast is your sampling? Maybe this is some issue with fast versus slow channels when sampling fast.

[AvaTar]

Exactly, there is a charge-up of the S&H capacitor involved.

Too short sample times create a "drag over" or pseudo-crosstalk effect.

[Pavel Krupets]

16 MHz (1 mega samples), 4 cycles sampling time.

Only 1 channel out of 24 which are all read in sequence reads lower value (consistently). I moved it around in a sequence but it didn't change anything.

Correction: sampling it multiple times does change the reading.

[Pavel Krupets]

Similar problem with 16 cycles sampling time.

[MikeDB]

Can we see the top side of the PCB ?

[Pavel Krupets]

[MikeDB]

Actually that's the best connected of the lot. Something the other end of the track maybe ?

[MikeDB]

Have you examined the PCB with a magnifying glass and checked you don't have a manufacturing fault on that pin ?

[Pavel Krupets]

I checked signal on oscilloscope and it looks just fine. Both ends. Signal is low frequency, analog.

Not a lot of noise in signal lines.

[MikeDB]

Yes that looks fine. Faulty IC maybe ? Do you have a second PCB loaded ?

[Pavel Krupets]

Checked under stereo microscope (well ones you use for soldering surface mount), didn't see anything suspicious.

Was curious why this stuff can happen, might read it multiple times, it seems 2nd read is better than 1st. Similar to how 1st channel in sequence reads high consistently (10% higher than it should be). This is solved by reading the same channel again immediately after 1st read.