Precision with opamp in transimpedance amplifier configuration with LED as photodiode

Question

I found this circuit in searching for a circuit that uses LED to measure absorbance

What I want to ask is: in order to "accurate" the readings from \$ V_{out} \$, for example: assume that the current through LED is \$ -0.1 µA \$, \$ V_{R} = 10 kOhm \$, then \$ V_{out} \$ should be \$ 1 V \$ but the actual reading is \$ 1.303 \$ for example. I know this is due to many parameters such as: devices, offset current, bias current, offset voltage, blah blah ...

I just wonder how can I reduce the error and obtain the result more accurately?

Answer 1

You start with "how accurate do I need?" then move on to "How do I get it?" Never start the second until you know the first.

The example you cite shows a 0.3% error. You can work for a decade without needing that sort of precision.

For this particular case, to get better than that, you would need to start with a very precise \$1.000M\Omega\$ resistor. Then, you need an opamp with small bias voltages and currents.

Answer 2

then Vout should be 1V but the actual reading is 1.003 for example .

that would be an extremely accurate reading. if that's indeed what you get.

I just wonder how a real "accurate" readings could be obtained ?

through calibration mostly.

Answer 3

There is real problem here:

assume that the current through LED is −0.1µA, VR=10kOhm, then Vout should be 1V but the actual reading is 1.303 for example.

Apparently you "assume" the current is 0.1 uA, but you measure the output voltage at 1.303.

Unless you measure the current as being 0.1 uA, there is no way to tell if it really is. Then you have no way to tell that the output is actually wrong.

I'd suggest building the following circuit

^{simulate this circuit – Schematic created using CircuitLab}

Here the critical parameter to consider is the input bias current current of the op amp. Make sure that it is on the order of 1 nA or less. Also fairly important is the input offset voltage. Since the nominal voltage across the resistor will be about 100 mV for an assumed LED current of 0.1 uA, you'll want an offset of 1 mV or less if you want 1% accuracy.

With this circuit, you can establish the actual current being provided. Only then can you start determining what your system error actually is.

Note that this circuit essentially does the same thing as the TIA you built for the first stage, and doesn't have the possible stability issues which make the feedback capacitor necessary. So why use a TIA? It's faster for large gains. The feedback capacitor compensates for the LED capacitance which will limit the follower circuit.