How to improve my Transimpedance Amplifier

Implying a feedback resistor on the order of what, 0.1 to 10M ohms?

You didn't mention supply bypassing, is that adequate for a 90MHz amplifier? And are you being careful not to create, or to dampen any bypass-capacitor resonance?

How much total input-node capacitance do you have (the amplifier has 3pF), including all wiring and the detector? Are you adding any feedback-resistor capacitance? Keep in mind this capacitance should be greater than ~ sqrt(Cin / 2pi Rf f_T). In my experience a resistor's self-capacitance can be much less than the 0.2pF they mention in the data sheet. You don't want low-level oscillation.

Win Hill is probably right - your op amp is most likely oscillating at some extraordinary frequency. This can be sometimes be difficult to test, as the scope probe can alter your stray capacitances enough to kill the oscillation.

A possible alternative explanation is that you've got enough direct current running through your "0V" rails that there are ten of millivolts of potential difference between points that you fondly imagine to be a common 0V reference. This is a bit easier to test - you hang one of your multimeter leads on one easy-to-find and easy-to-find-again "0V" test point, then crawl over the board with a copy of your circuit diagram, measuring the voltage difference between you reference point and every other "0V" node on the circuit diagram.

This can be educational. This got burnt into my memory on a board that mixed precision analog and big, old-fashined PROMs. The PROMs drew about 100mA and dumped it into one end of a track that ran past all the precision analog bits in succession. Cutting that track at the right point and re-routing that current through a chunk of 18swg wire improved the board's performance no end.

-------------- Bill Sloman, Nijmegen

I used a 100K-500K variable feedback resistor and a 10pF-60pF variable feedback capacitor. The op amp OPA380 is driven by a single +5V power supply and the supply bypassing capacitance is 100nF. Now I encountered two puzzling problems:

1. There is a quite large input offset voltage of 80-90mV (even much larger than the output offset), which may come from the ground loop by my guess.
2. After the PCB board had been tested for some time and the feedback resistor was changed to be 5M, the output offset turned to oscillate acutely.

Thanks for any helpful instructions to find out the origin or the solution of the above problems! I'll send the circuit layout files to anyone who would love to see by E-mail.

Jiaqi Shen