Feedback Gain Loss of Transimpedance Amplifier

Question

Hi all, I have designed a single-supply (3.3V) potentiostat that initiates electrochemical reactions, converts the generated electrochemical current into a voltage via a transimpedance amplifier (op-amp used was ADA4528), which then passes through a sallen-key low-pass filter and to an ADC pin on an mbed NXP LPC1768 development board. The ADC value is then sent to a PC and the current is back calculated and displayed in real-time on a graphical user interface. The circuit is similar in function to the one in the colorimeter CN-0363, which also uses ADA4528 as the transimpedance amplifier. The feedback resistors tested were 16k, 160k, 1.6M and 16M. I've been using the Keithley 6221 current generator to test my board's function and applied currents that fall within the voltage output range of the op-amp for the different resistors. My system is very accurate till towards the end of the respective current range, after which I start losing gain. There is no discernible error for 16k, but for 160k and 1.6M, the gains started falling off at around 0.2V and 0.4V from the positive rail. The 16M was the worst at almost 1V from rail. I figure the gain loss is due to the limited output swing of the op-amp, when the internal transistors get out of the saturation region. However I do not know why different feedback resistors give different output swings, especially for a rail-to-rail op-amp like ADA4528. I used the same 100pF capacitor to compensate all resistors, which based on calculations, is excessively overcompensating (though I should be operating at DC so this shouldn't be a problem?). However, I found out that when I put a 100nF capacitor with my 16M resistor, the feedback gain became very accurate, and the output swing was no longer restricted to 1V from rail. The side-effect was that there was noticeable "charging" due to a significant time constant. To sum up: 1. Why does the output swing change for the ADA4528 when different feedback resistors were used? 2. Why did the 100nF capacitor "fix" the output swing limitation and gain loss issue of my 16M resistor? Any help is appreciated! Thanks! The simple circuit diagram of the transimpedance amplifier with programmable gain and the sallen-key low-pass filter is posted. For testing, I have removed the multiplexer and only testing one feedback resistor. The filter cut-off is around 13 Hz, to remove the main hums noise. 
 Attachments:

nickgray · Answer

Your sensor is highly capacitive, with this capacitance at the input of the op-amp. The feedback resistor and this capacitance form an RC filter that delays the output as it drives the inverting input to keep that input equal to the non-inverting input. That is, it takes time for the amplifier output to drive the inverting input of the amplifier back to the voltage of the non-inverting input. By the time the non-inverting input is at the proper level, the output has gone too far, so the amplifier output goes back in the opposite direction to compensate, but the feedback to the inverting input always lags the amplifier, causing the amplifier to oscillate. The solution is to increase the feedback capacitor when you increase the feedback resistor. The larger capacitor presents a lower impedance at the oscillation frequency and allows the inverting input of the opamp to "keep up" with the amplifier output. If 100nF (0.1uF) works with a 16M feedback resistor, then it would seem that 0.01uF should be used with 1.6M feedback resistor, 0.001uF with 160k, and 100pF with a 16k feedback resistor. Of course, you might find that this does not work in all cases since we are not sure of the actual value of the 100nF capacitor and the 16M resistor you used. Remember that tolerances can cause things to be different from what we might expect.

Good luck!