2
\$\begingroup\$

In working on a spectrophotometer design, I have to make a design decision for implementing the photodiode front-end amplifier and ADC. One of the requirements I have set is a dynamic range for the input of 10^5 in terms of radiant intensity, translating directly to photodiode current of, say, 10pA to 1uA, so, at least 17 bits of resolution. I have settled for the MCP3550 22-bit S/D ADC with an appropriate reference and likely an LMP7721 ultra-low Ib opamp for the TIA amplifier.

The design requires that the photodiode is positioned away from the electronics by about 0.5-1m. It will be enclosed in a full metal case with a window for the light input. The analog electronics will also reside in a shielded enclosure, uC interface will be galvanically isolated, as will be the analog power rails. Sources of interference include a large (150VA) mains transformer and a NEMA 17 stepper motor and driver, also in ~0.5m vicinity.

My question is whether it is better to convert the photocurrent to a voltage immediately at the detector and send the output voltage (e.g. 10uV to 1V) via a shielded cable, which also carries the supply voltages, to the ADC in the enclosure, or if sending the photocurrent via a triaxial cable back to the enclosure presents any advantage. I have illustrated the two designs below.

Option 1: Design 1, direct TIA preamp in detector head

Option 2: Design 2, preamp in main enclosure

The relevant questions are as follows:

  1. Is it preferable to send a single ended (or pseudo-differential, as shown) uV-level voltage signal along a shielded twisted pair or to use a double-shielded coax and transmit pA-level photocurrent?
  2. In design 2, what is the effect of cable capacitance on input capacitance of the TIA? The inner "guard" shield appears bootstrapped, but is it in reality.
  3. In design 2, to gain some extra dynamic range, I have shown the ADC's "-" input tied to Vref, such that I can measure a 0-5V signal as -2.5 to +2.5V from the perspective of the ADC, but now the input is truly single-ended. However, given the close proximity of the opamp and ADC in this design, this shouldn't come at a penalty?
  4. In both designs, I have shown the shield as connected to a separate ground, which I plan to be protective earth (PE). Is this meaningful in a fully isolated system with high DC and AC impedance to PE? Thank you in advance,
\$\endgroup\$
2
  • \$\begingroup\$ I think the cable capacitance in option 2 is going to make the transimpedance amplifier oscillate. Or at least I would simulate that carefully to see what it is going to do your signal. \$\endgroup\$ Commented Mar 11 at 17:27
  • \$\begingroup\$ Cable capacitance can be compensated with a feedback capacitance, which of course decreases bandwidth, but I am not aiming for speed anyways. It is more a question of interference. \$\endgroup\$ Commented Mar 13 at 6:28

2 Answers 2

Reset to default
2
\$\begingroup\$

I don't have much experience with triax so maybe it's doable, but I shudder at the thought of running a signal with miniscule currents, quite low voltages, and tremendous gain on a long cable if it can be avoided.

The cable will add to stability concerns due to input capacitance -- maybe you don't care if you are running low bandwidths.

I think amplify first so you have a low-z output with reasonable voltage levels, and propagate the signal thereon.

I would choose design #1 as a default unless otherwise forced to do differently.

\$\endgroup\$
1
  • \$\begingroup\$ Rohmeooo, thanks. Bandwidth is limited by the ADC sample rate of 12.5Hz, which for the electromechanical system I have in mind is about as fast as I can go, even if I average two readings. I have previously run photodiodes over regular coax over 20-30 cm without issue, but the system had less sources of EM interference. I have already done some tests, which clearly indicated the need for isolating the frontend, since even judicious star grounding resulted in tens of uV of noise at the ADC when the stepper motor was running. I wanted some confirmation either way before committing things to PCB. \$\endgroup\$ Commented Mar 13 at 6:41
2
\$\begingroup\$

You need to do an error budget calculation to see what you are likely to get in the scenarios you mention. For instance, if the temperature at the photodiode changes more than the temperature at the end of the cable (where it could be controlled) then, it's likely that the TIA op-amp will be best suited to the calmer environment.

This is based on the op-amp having an input offset voltage of +/- 150 μV and a drift of up to 4 μV per °C. See extract from op-amp data sheet below: -

enter image description here

Those numbers strongly jeopardize your aims hence, a proper error budget needs to be made before any decision can be made. You need to make that error budget calculation to predict what the likely accuracy achievable is.

In short, all four of your numbered questions cannot be properly considered until you do the error budget calculation. You also need to factor in op-amp noise and this can only be done when you decide on the bandwidth requirements for your signal.

\$\endgroup\$
6
  • \$\begingroup\$ Hi Andy, thank you for the comment. \$\endgroup\$ Commented Mar 13 at 6:29
  • 1
    \$\begingroup\$ (edit) Of course, an error budget is mandatory for any meaningful design. That I why I have calculated one, any why it was not the subject of my question. I may be mistaken and would love to be corrected, but offset voltage can be accounted for in software (or even in-circuit by trimming, although I don't think it makes sense). Offset voltage drift is a problem, yes, but my question is more about making a design choice based on susceptibility to interference. I have considered other sources of noise, such as cable triboelectric effect, thermocouple effects, etc \$\endgroup\$ Commented Mar 13 at 6:35
  • \$\begingroup\$ @NikolaDyulgyarov if we are done here, please take note of this: What should I do when someone answers my question. If you are still confused about something then leave a comment to request further clarification. \$\endgroup\$ Commented Mar 13 at 8:11
  • \$\begingroup\$ Hi Andy, perhaps my previous comment was unclear. I did comment on both responses to my question. Regarding your particular response, I am unsure how an error budget would help answering questions #2, 3 and 4. To rephrase, for instance, in design 2, is the cable inner shield bootstrapped w.r.t. the core, given the configuration, and if not, why? \$\endgroup\$ Commented Mar 14 at 16:37
  • \$\begingroup\$ I think your questions are open-ended without you stating what errors you expect to get. Q2 is really open-ended. \$\endgroup\$ Commented Mar 14 at 16:50

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.