1
\$\begingroup\$

My formal education was in mechanical engineering and I’m well trained in some CS as well. Yet here I am on an electrical project (that will eventually turn CS). I have been lost and confused for many months now and could really use some specific help.

Project scope/goals: Measurements from the sample to photodiode should be in the 0.3 to 3 picoAmp range. An array of photodiodes will detect this small amount of light. Right now I’m just trying to get a circuit working for a single photodiode at that extremely low light range. Time response can be long or short because we can hit the sample with the laser for a long time without any problem. The laser rep rate is ~76MHz versus the approx. detector bandwidth ~23KHz so there shouldn’t be a problem with signal decay.

Materials:

  • S10355 photodiode (we are looking to replace, suggestions would be awesome)
  • Powered breadboard with +5V, +15V, and -15V supplies
  • NI USB-6211 for measurements
  • Op Amps we have around(happy to buy more as needed)
    • TLC271 (0 to +15)
    • TL031 (-15 to 15)
    • MCP603 (0 to +5)
    • MCP6002
  • Large resistors: 10Meg, 22Meg
  • Resistor kit (100 ohm to 900k ohm, lots of values)

‘Ideal’ Circuit simulation (that doesn’t really work but is good for visualization):

schematic

Adapted from book:

pages

I chose my resistance total based on amplifying 0.3picoAmps to 1 Volt. The book also describes a way to do this with a single op amp but I haven’t gotten that one to work even a wee bit. The only other way seems to be a tee-network but this causes proportional noise gain as well which wouldn’t be appropriate for my application from what I can tell.

Major problem: There appears to be so much noise even at ~no light~ conditions (there’s never no light but I can get close), that the voltage readings are maxed. Since we are looking at subtle changes this makes it pretty unclear. I read about making a pseudo-ground for the cathode of the photodiode and set one up at 1V from the +5V of the breadboard. This doesn’t seem to do as the forum post indicated or maybe my gain is so high I need a mV or pV pseudo-ground? Not sure. I have also read that choice of Op Amp is critical, and I’m not sure I have the right ones for the job right now, could this be hindering linear response and ground levels?

Power turned off/no light Vout:

scope

Power turned on/no light Vout:

scope

Future problems: I know to eventually implement any of this on a PCB board (for the full array of photodiodes) I will have to be exceptionally wary of noise for detecting such low current levels. Do you have some tips for this? There’s a nice noise analysis in the book but I don’t fully understand the ee concepts I am working around to be honest.

\$\endgroup\$
12
  • \$\begingroup\$ Your photodiode has a dark current of 100 pA typical with 10 mV reverse bias. I think trying to measure 3 pA signals with this guy is a non-starter, even in photovoltaic mode. \$\endgroup\$ – The Photon Aug 10 '18 at 23:06
  • 1
    \$\begingroup\$ Start over. Define objective, optical method, specs for power or current levels expected, range or path loss, and response time ( Bandwidth) Engineers are more productive with Specs before starting any design . \$\endgroup\$ – Tony Stewart EE75 Aug 11 '18 at 0:31
  • \$\begingroup\$ Melchomps can you do this ? I’m sure there is a better way with matched optics, bandwidth and low cost to do what you want. \$\endgroup\$ – Tony Stewart EE75 Aug 11 '18 at 1:12
  • \$\begingroup\$ Wait, what? You're using a 76 MHz rep rate? On a 4 MHz op amp? Good lord, of course you're having problems. "Signal decay" be buggered, you're causing the input stage to have fits. Not to mention @laptop2d's concerns. And trust me, if anything he's underplaying the unsuitability of a bread board for this application. At the least, you should put a passive low-pass on the photodiode, which will require switching to photovoltaic mode. Since you claim (without providing numbers) that speed is not an issue, it's possible. But not likely at these current levels. One magic phrase - input optics. \$\endgroup\$ – WhatRoughBeast Aug 11 '18 at 1:49
  • \$\begingroup\$ @TonyEErocketscientist Thank you for taking a look at my problem. The project scope/goals section defined the current I am expecting within a pre-defined optic pathway after the laser hits our sample. We are inserting the photodiode into the optics path. Does that make sense? The laser isn't directly hitting the photodiode, it's hitting a sample, dispersing photons, going through optics then hitting photodiode. Sounds like I have some major issues with both breadboard and current photodiode bias. \$\endgroup\$ – melchomps Aug 11 '18 at 16:59
1
\$\begingroup\$

I read about making a pseudo-ground for the cathode of the photo diode and set one up at 1V from the +5V of the breadboard. This doesn’t seem to do as the forum post indicated or maybe my gain is so high I need a mV or pV pseudo-ground? Not sure. I have also read that choice of Op Amp is critical, and I’m not sure I have the right ones for the job right now, could this be hindering linear response and ground levels?

The things to know about building these circuits are:
1) Breadbaords aren't going to cut it, this automatically adds more than 10pF to your circuit and way to much inductance, use perf board and solder the components. Or use wire wrapping. 10pF*1e6Ω = 10Hz so if you use a breadboard, 10pF will automatically cut your bandwidth to 10Hz with no gain capacitor.

2) You'll need an amplifier with a lower input bias current than that of what you want to measure. An op amp with an input bias current in the fA range would be appropriate. Input bias current means the current flowing into the opamps + and - terminals. If you want to measure pA keep in mind that even large resistive materials can source pA of current, FR4 (PCB material) drops to 10e8Ω when damp and if you put 1V on the other side of the PCB, you get 1pA of leakage (and offset). Guard traces can be essential.

3) Select an amplifier with a larger open loop gain, and larger bandwidth than you require. Make sure you understand that when you put capacitance and gain, they affect the gain bandwidth product

4) Use two amplifier stages to gain up your signal, if you need 1e9 of gain then gain up the first one with 1e5 and the next one with 1e4 (for example)

\$\endgroup\$
4
  • \$\begingroup\$ "if you use a breadboard, 10pF will automatically cut your bandwidth to 10Hz with no gain capacitor." And without a feedback cap the circuit will oscillate. \$\endgroup\$ – WhatRoughBeast Aug 11 '18 at 1:29
  • \$\begingroup\$ @WhatRoughBeast so don't use a breadboard is the jist of it. When I tried mocking up some transimpedance amplifiers I learned this lesson. I suppose you could go for a multi gain stage on a breadboard at the cost of noise and other problems \$\endgroup\$ – Voltage Spike Aug 11 '18 at 5:28
  • \$\begingroup\$ @laptop2d Thank you so much, this was a very helpful post. I didn't realize breadboards had these problems although it makes sense. I really appreciate the feedback on lots of my problems! \$\endgroup\$ – melchomps Aug 11 '18 at 17:05
  • \$\begingroup\$ If you like the answer upvote and mark the answers as answered and participate in the community. meta.stackexchange.com/questions/126180/… \$\endgroup\$ – Voltage Spike Aug 11 '18 at 17:29
0
\$\begingroup\$

As a useful rule of thumb, your amplifier bias (and offset) current should be an order of magnitude less than your operating current. Since you want to measure 0.3 pA (300 fA), you should be in the market for an op amp with input currents of about 30 fA. As you can guess, your present amp is not even close to adequate.

As it happens, there are such op amps around - they are generally called electrometer op amps.

As an example, you might take a look at the ADA4530. Bias currents are 20 fA, and the gain-bandwidth is 2 MHz, so it's not that much slower than your present choice. But there are other models and other manufacturers, so shop around.

And while you're at it, you need to start thinking about just how you're going to deal with these low currents, too. You're getting down near the range where leakage currents live, and at the least you need to start thinking about scrupulous cleanliness. Things like fingerprints in the wrong place can really ruin your day. You might consider reading some Bob Pease articles ("What's All This Femtoamp Stuff, Anyways?" and "What's All This Teflon Stuff?" are good reading.)

And you'll need to think about thermal control of your circuit, too. With 22 Mohm in the TIA, 3 pA will give you a nominal 66 uV. A good op amp will have input voltage offset drifts on the order of 1 uV/dec C or more, so that can be a problem. Note that ADA4530 has a worst-case drift of 2.8 uV/degC, which is about 4% of your reading per degree. And if, as you stated, you're considering 0.3 pA, that's a 40%/degC drift. It's true that you can get much better drift performance (see Zero-Drift op amps), but they don't have the current performance you're looking for.

While I'm at it, you might consider not using PDs at all. How about a good CCD? In effect, you'd finesse the whole current issue by integrating detector current and feeding the total out in a much shorter, but bigger, readout pulse. You'd need a focussing lens and alignment would be important, but you can choose integration times to get the gain you need, and they typically have dark cells at the ends of the CCD so you can compensate for temperature drifts. 3 pA is about 2 million electrons/sec, so an integration time of 10 msec will give you about 20,000 electrons/sample, which is quite reasonable. As I say, though, you'd need careful attention to optics. Well, nothing's free, right?

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.