Hi, we have been using the ADL5304 in a previous product to measure signals using a small photo-diode. We are now looking to use the same part with a much larger area detector. Hence this new part will have a much higher shunt capacitance (amongst other things). Typically when using a linear TIA this would introduce 2 issues:
The first issue is the stability of the amplifier would degrade and may become unstable without reducing the bandwidth (typically by choosing a larger Cf in the feedback path). I do not see any mention of stability in the datasheet of the ADL5304 in regards to the input capacitance. My understanding is the TIA works in a way such that the BJT essentially forms a "resistance" (V/I) that is logarithmic with current via the Ebers–Moll model. But the only mention I can see of a feedback capacitance to stabilise this "resistance" is on p19 of the datasheet of about 0.3pF when looking at how the bandwidth reduces with input current. I can imagine a very very wide bandwidth op-amp on the front end could resolve this issue and make the system infinitely stable if the feedback (possibly parasitic) capacitance causes the signal to attenuate before reaching the op-amp roll off, but I want to confirm this is the case and we do not need to worry about stability issues when sourcing current from large capacitance sources?
The second issue is related. In a traditional linear TIA the additional input capacitance forms a zero in the feedback loop. This will start to amplify the noise on the input pin. At low frequencies the noise spectral density is amplified by `1+Rf/Rshunt` where Rshunt is the photo-diode shunt resistance, and Rf is the feedback resistance and I assume in the ADL5304 it would be the small signal equivalent resistance from the BJT in the feedback? However this low frequency noise is usually dwarfed by the high frequency noise response as the bandwidth increases. The high frequency noise gain would typically be given as `1+Cshunt/Cf` in a linear TIA. Is this still valid in the case of the ADL5304 using the small signal model of the BJT's, and if so should I consider the `Cf=0.3pF`? Also if this noise is large owing to the ratio of the `Cshunt/Cf` would it be advisable to limit the logamp bandwidth further by adding additional feedback capacitance in parallel across pin 32 and 4?