Inverting input causing current draw on non-inverting input?

Question

Problem: unexplained current draw on non-inverting input when voltage on non-inverting input is above the inverting input.

Background: I have been playing around with an RC timing circuit from pg. 24 of "The Art of Electronics" (Horowitz and Hill, 2015).

The idea is to charge a capacitor, and then slowly drain it using a resistor. A comparator op-amp is comparing the voltage on the capacitor against a reference voltage, and turning on an LED while the capacitor voltage is above the reference.

^{simulate this circuit – Schematic created using CircuitLab}

I breadboarded up the circuit using a 100 uF capacitor and 1 Mohm resitor. I used a TLV2462CP op amp (because that's what I had in my box), and a LM4040-based precision voltage reference (Adafruit P/N: 2200) in place of the voltage divider.

Confusion:

My plan was to experiment with RC times, but I found the LED was shutting off far too early. When I probed the circuit at the testpoints above, I got this strange behavior. Notice the kink in V_CAP near V_REF. This kink moves upwards if I use V_REF = 4.096 instead.

What gives?

I'm clearly getting some sort of increased bias current while V_CAP > V_REF, but why? The TLV2462CD is not a comparator, but this behavior is still strange.

Solution: I took suggestions 1 and 2 of @dannyf. When the capacitor is fully charged (i.e. 5 V), the discharge current across R2 is 5 uA. According to the datasheet, the worst-case bias current for the op-amp should only be 25 nA.

However, I decided to add a buffer op amp to the capacitor (luckily, the TLV2462CD is a dual op amp!), and the circuit behavior is now as expected.

^{simulate this circuit}

Answer 1

What gives?

a few possibilities - none definitive.

1) bias current isn't linear - datasheets generally don't specify its behaviors other than a typical, max. whether that "kink" is there due to the design or by accident is not clear.

2) bias current can be significant sometimes, like in your case when the discharge current is very small to begin with.

3) assuming a typical long-tail differential input, the bias current is higher than the differential voltage is larger than 0.7v (assuming one pe junction).

4) your measurement instruments could have introduced additional behavior not present in the actual circuit. the key here is whether the desired time constant is met.

...

Answer 2

Probably back to back diodes between inputs for protection. Look at the internal schematic of the op amp on the datasheet.