3
\$\begingroup\$

I have a circuit that incorporates a 40 channel InGaAs (IR) photodetector array (https://www.hamamatsu.com/eu/en/product/type/G8909-01/index.html). I have a TIA design using a LTC6268-10 first stage. However, because I require 40 of these, it is impossible to position all of the op-amps close to the photodetector. The application potentially delivers quite low light input to the photodetector; I'm estimating ~1uA of photocurrent minimum. Fortunately, InGaAs photodiodes are less noisy than Si, but I'm concerned about noise pick-up on the traces from the wire-bond pads to the amplifier inputs.

As the Photodiode array is common-cathode, I can't connect the anode and cathode across the input terminals of the op-amps, which I believe would aid in reducing common-mode noise.

I'm wondering if I can lay the PCB traces to provide a coax-cable like performance, but I'm not sure how to go about achieving this.

Can anybody suggest some 'expert tips & tricks' that may help to retain my SNR?

\$\endgroup\$
5
  • 2
    \$\begingroup\$ How long is "long distance"? \$\endgroup\$
    – Dave Tweed
    Oct 14, 2020 at 14:35
  • 2
    \$\begingroup\$ What speed of response do you need and what noise requirement? The biggest problem is that the capacitance of the screened cable will increase the noise of the amplifier. \$\endgroup\$ Oct 14, 2020 at 14:44
  • \$\begingroup\$ I'm currently laying out the PCB to get an idea of distances. I have positioned the first 20 channels (10 on the top, 10 on the bottom) and the distance to the farthest ones are around 50mm. I estimate, continuing with this layout method, the farthest distance will be in the region of 70 - 100 mm. Speed-wise, I want to modulate the optical signal at around 10kHz. \$\endgroup\$
    – Miakatt
    Oct 14, 2020 at 15:09
  • 1
    \$\begingroup\$ Great question, I've been working on something similar in my spare time. Since the bandwidth is pretty low, I would try to lay them out with pairs of two TIAs on each side of the sensor, with the pattern mirrored on the backside (so tiles of 8 TIAs). Then I would put two tiles of 8 on each end. That means you have to fit 3 tiles of 8 along the ~1" length of the sensor. I think you could probably pull this off on a 4 layer board without the traces being unreasonably long. \$\endgroup\$ Oct 14, 2020 at 16:17
  • \$\begingroup\$ @user1850479. That's basically the direction I'm going in at the moment. I could change the layout of the op-amp block to allow the V_in pin to be closer to the PD, at the cost of longer distances to the feedback resistor (make the circuit footprint narrower). Have you come across any multiple channel TIA ICs? When I worked at CERN, we had some 12 channel packages, but I think they must have been made in-house as I can't find anything similar OTS. They would be ideal for this task. Thanks. \$\endgroup\$
    – Miakatt
    Oct 15, 2020 at 9:01

1 Answer 1

3
\$\begingroup\$

1 - Noise pickup on the leads is not what you need to worry about. If nothing else, you can put your detector in a shielded enclosure, with filtered power in and a hole just large enough for the IR to illuminate the array.

2 - What you do need to worry about is (as Kevin White commented) the effects of those long traces on the input capacitance of the op amp. Unless you properly compensate with a capacitor across the feedback resistor of your TIA, the circuit will oscillate. Fortunately, your frequency requirements (<100 kHz) are fairly modest, and I'd guess you should be able to get a decent signal amplitude.

3 - You may need to go to a multi-stage amplifier chain, with the maximum gain of the TIA (set by the feedback resistor) being limited by the feedback capacitor. Subsequent amplifiers will degrade the SNR of the signal slightly, but that's unavoidable.

4 - 40 LTC6268s in close proximity is, frankly, a disaster waiting to happen unless you are very careful about ground paths and power decoupling. Whenever you have multiple high-frequency op amps you need to take care that a signal on one does not affect a neighbor due to coupling through the power supply lines. And, at 100 MHz open-loop bandwidth, the LTC 6268 is a good example of a high-frequency op amp. You might be well-advised to go with an op amp with a lower bandwidth, such as 10 MHz.

5 - Given the fact that your photodiode array has a reverse bias limit of 6 volts, the example TIA shown on page 1 of the 6282 data sheet is an excellent place to start, keeping in mind the need for an overt feedback cap. 1 uA / 20k will give a signal of 20 mV, and you may (or may not) be able to get away with a larger feedback resistor. You can then boost the signal with a subsequent op amp. Since your bandwidth isn't too great, you should be able to provide enough gain without too much extra noise being introduced.

\$\endgroup\$
4
  • \$\begingroup\$ Comments/Questions in-line. 1. Good point. I'll add some shielding. 2. I have a 5pF cap in parallel with the feedback resistor. Not sure if that's enough, but I can replace it if a different value is req'd 3. Yes. I'm planning on taking the 6268 O/P to a LTC6244 on the motherboard. 4. Is this an issue if it is operating at only 10kHz? If my estimate of the photocurrent is off (optics under dev), I could need much more gain and hit the GBW limit. Hence the LTC6268-10 for the extra GBW overhead. 5. I'd like to keep the option of PV mode, for noise reasons. Is this useful for InGaAs? \$\endgroup\$
    – Miakatt
    Oct 15, 2020 at 8:49
  • 1
    \$\begingroup\$ @Miakatt - "4. Is this an issue if it is operating at only 10kHz?" It has nothing to do with the closed-loop gain. It has to do with the open-loop gain at high frequencies. Careful decoupling, grounding and signal/power routing will be key. \$\endgroup\$ Oct 15, 2020 at 16:01
  • 1
    \$\begingroup\$ @Miakatt - "5. I'd like to keep the option of PV mode, for noise reasons. Is this useful for InGaAs?" Sure, in that case you'll ground the common cathode. However, the effective capacitance for PV mode is much greater than for PC, so you're more likely to run into stability/gain issues and will need a larger feedback cap on your TIA. \$\endgroup\$ Oct 15, 2020 at 16:04
  • 1
    \$\begingroup\$ @Miakatt - "2. I have a 5pF cap in parallel with the feedback resistor. Not sure if that's enough, but I can replace it if a different value is req'd " You need to pay close attention on this, and look carefully at the output of the TIA, tuning the chopper step response to something like critically damped to slightly overdamped. \$\endgroup\$ Oct 15, 2020 at 16:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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