0
\$\begingroup\$

I'm working on a circuit involving Silicon Photomultipliers. Two SiPMs are in parallel together, and they go to a feedback amplifier circuit where the current signal gets filtered and converted into a voltage signal as shown below:

schematic

simulate this circuit – Schematic created using CircuitLab

Our first design used only one SiPM in the circuit, so we decided to try and see the effects of two SiPMs in parallel together in one amplifier circuit. We decided to go with a design where we have a pair of pads that connect the two SiPMs together, so they can either be connected together in parallel when the pads are shorted, or they can stay on their own individual channel if the pads are not soldered together. For this design, we soldered the pads together so two SiPMs are in parallel together, and we left elements of the unused channel bare, with no parts soldered to it.

One issue that we're seeing is crosstalk. I'm really new in designing boards like this, and I never learned about issues such as this. I never took a design class, so issues such as crosstalk is all brand new to me. The PCB board for the circuit has already been designed, so there's nothing much we can do about the trace length and such. I'll have to keep these design issues in mind for the next board. For now, I was just wondering if there was anything I can do to reduce the crosstalk effects at the output. My first thought was to increase the feedback capacitance from 3 pF to a higher value, say around 6-12 pF. The initial circuit with one SiPM utilized 3 pF capacitance, and there were no issues with it. That was the recommended circuit values provided by SensL. Putting two SiPMs in parallel increases the amount of noise at the input thanks to parasitic capacitance, so I thought that by increasing the capacitance, it would compensate for it, and since it is still in the pF range, the RC time constant is still tiny in the nanosecond range, so the rise time shouldn't be affected very much up to a certain value. Does this sound correct, or are there other ways to reduce the crosstalk?

EDIT: I apologize for being vague in explaining crosstalk. Please let me elaborate a bit more. First off, the PCB board that we are working with contains four SiPMs, each on their own individual channel, and there a set of pads at the cathode of two pairs of SiPMs to put them in parallel if desired. You can see that in the schematic below. Our current design with two SiPMs in parallel per channel is shown. Our issue is that we used an LED pulser to 'activate' our SiPMs, and we examined output from the two different channels. One channel was being pulsed, while the other channel was being left alone. You can see it in the picture below, with the pulse in purple, the pulsed channel in green, and the non-pulsed channel in blue. The pulsed channel shows a positive peak which is what we expect, but we noticed a negative pulse in the unpulsed channel, which we believe to be is crosstalk. Again, this is all occurring on the same PCB board.

schematic

simulate this circuit enter image description here

\$\endgroup\$
  • 1
    \$\begingroup\$ Please expand on what "crosstalk" means. It doesn't make a lot of sense in this circuit. Exactly what symptoms are occurring, and why are they a problem? Does the output interfere with another circuit which you have not described? \$\endgroup\$ – WhatRoughBeast Jun 1 '17 at 18:32
  • \$\begingroup\$ Yes, we need more details about what you mean by "crosstalk." FYI, I answered a question a while back about paralleling SiPMs: electronics.stackexchange.com/questions/301947/…. My advice is generally don't do it, since the backfeeding is a problem and the extra high speed op amp is barely any extra cost compared to the cost of the SiPMs (expensive). \$\endgroup\$ – Peter Jun 1 '17 at 19:47
1
\$\begingroup\$

Assuming your crosstalk is Efield, you have MANY methods to reduce crosstalk.

Any bit of grounded metal, sticking out above the PCB, will capture some flux lines and reduce crosstalk.

Any nearby component (especially if Grounded, like BypassCaps) reduce crosstalk.

If the transmitter of energy has direct line of sight to the receiver of energy, then block that line of sight.

Any small grid of wires, or large grid of wires, GROUNDED, is a fine gatherer of flux lines and will improve your quietness.

\$\endgroup\$
  • \$\begingroup\$ Thank you for your reply. So the idea is to have grounded metal bits to capture flux lines to reduce crosstalk. With the way my board is now, I can short some pads to ground and add a bypass capacitor to help with it, though it is a bit of a hack job given the layout. Is there a general guideline as to how much capacitance should be used? Also, what do you mean by 'line of sight'? The SiPM and the output are on opposite sides of the PCB, so they shouldn't be interfering with each other. Also, based on what you're telling me, increasing the feedback capacitor would do nothing to the crosstalk? \$\endgroup\$ – user101402 Jun 1 '17 at 15:47
  • \$\begingroup\$ Efields are accurately modeled as pools of water into which you spill ink. Eventually the ink will explore ALL POSSIBLE PATHS. To keep you circuits clean, place them behind shields. The challenge is to approximately calculate the risks. \$\endgroup\$ – analogsystemsrf Jun 2 '17 at 4:05
0
\$\begingroup\$

Notice the Victim is behind 2 GROUNDED shield plates, but Efield simply goes above and around and behind and then zaps down to the victim.

Theory tells us this pattern IS LEGITIMATE SOLUTION to the Efields, because in a grid of resistors, with 4 regions (TX driven, RX to monitor what is induced, and 2 shields tied to GND to collect Efield flux) of metal in the SPICE SIM grid of resistors, current WILL go behind the shields. Run this in SPICE, and see.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\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.