B. A. BryceWhen I wrote the first analog design log entry I proposed 4 topologies that might be used to solve the problem.
I decided quickly that not having a constant reverse bias on the diode could lead to sticky issues with the response function also being a function of the input power.
So I started to look at buffered designs based on transimpedance amplifiers. All well and good except I overlooked something at 3 AM and that is that the amplifiers have finite output resistance (of course). They can only source or sink a few mA or perhaps 10s of mA. Indeed the chosen zero offset amplifier from LT can only sink 2 mA for a supply around 3 V. For the power range we would need to sink up to 100 mA. The good news is that LT6230 which was chosen for the last topology can source/sink at least 25 mA. With a typical value of 40 mA. OPA378 is similar at 30 mA so all together this means that the feedback resistance will need to be set to something more on the order of 100 ohms rather than 50.
OPA376 also looks like a good alternative. It has a typical offset voltage of 5 uV and noise of 0.8 uVpp in 0.1-10 Hz. It can source 30 mA and sink 50 mA. It has higher offset drift than OPA378. AD8615 looks like a possible choice as well though at much greater offset, it can drive a massive 150 mA! AD8615 is probably a better choice for the 4th topology and OPA376 is probably the best part I have found for the 3rd. Flattening the sensitivity function of the detector by choosing a detector coated to do better in the near-UV would improve the high power headroom as well.
If we wish to self-limit the amplifier to a drive of 30 mA with a supply of 3.3 V then the feedback resistance should be nominally 110 ohms rather than 50. For a sensitivity function peaking at 0.7 A/W, this would correspond to a laser power limit of just over 40 mW (280 mW for the best case). If the noise floor in band is of order 1 uV, then this corresponds to current of 9 nA or about 3 times the dark current. For a sensitivity of 0.1 A/W this is a should mean a minimum detectable power of about 90 nW (12 nW for the best case).
Discussions
Become a Hackaday.io Member
Create an account to leave a comment. Already have an account? Log In.
Such a buffer adds noise. If it is a high current opamp it certainly adds noise, if it is something like PNP transistor then it will decrease the headroom of the circuit and make it not rail-to-rail. If the circuit that does the buffering has a threshold then the offset voltage has to increase to produce the bias-voltage for the transistor which should be the needed bias voltage divided by the open loop gain. If the open loop gain is a million and the bias is of order 1 then the offset has to go up by order uV. Everything comes with a trade-off. I am considering a discrete current gain stage but really the other amps I found with drive from 25-150 mA should be enough. At the powers those are equal to I am worried about heating the diode anyway.
Are you sure? yes | no
You just need to buffer the output of the zero offset opamp and run your feedback from there. The buffer can either be a discrete solution or a high current opamp. The feedback loop will take care of the offsets.
Are you sure? yes | no