I've been working on a circuit for a while now. It's a circuit consisting of 4 Silicon Photomultipliers, MicroFJ-60035-TSV, in parallel all going into an op amp circuit. I posted questions about it before regarding separate issues about it, and I learned a bit more about about it by studying how the circuit functions. You can see it in the image below along with one of my simulations. Apologies for the messy part layout.

Schematic and Simulation

In the circuit, I've connected the cathode to ground, and I'm reading the current from the anode. I wanted to get the correct pulse shape as shown in documentation, so I'm following the recommendation of reading from the anode. I am planning to read from the cathode once the circuit is finalized and not in simulation. I'm using voltage sources attached to the photomultiplier parts to control when they 'fire'. The 1 microOhm resistors are just there so that I can read the current coming from the sensors. I figured a low resistance wouldn't affect the circuit much, but ideally, there would be no resistance there.

I'm kind of facing two issues right now. The first issue is with the current coming from the photomultipliers. You can see in the simulation results on the left that when one of the photomultipliers fire, some current also flows back into the other photomultipliers. What can I do about it to remove it or control it so that this doesn't happen?

My next issue is with the feedback circuit. Through reading, I found that the op amp circuit converts the current signal into a voltage signal, and the feedback circuit acts as a low pass filter to hopefully filter out high frequency noise. You can see another test below where three photomultipliers fired at once.

Test Results with three SiPMs firing at once

I've increased the capacitance to try and reduce the noise from the 4 parallel photomultipliers, but I'm having issues with the rise time and clipping. The time where three were firing had a faster rise time compared to the one which fired only once. However, the voltage seems to have clipped, staying the same after a certain point. Is there anything that can be done about it? There always seems to be a trade off between the cutoff frequency and the rise time.

UPDATE: I changed my schematic so that I'm now reading from the cathode instead of the anode. The current from the sensors looks a little messier, but the output is now positive as I was planning. The circuit does not require biasing. It's already done in the model of the part. My team is now looking for better signal-to-noise ratio for the circuit, so what can I do in my circuit in order to improve the SNR?

Updated Schematic, Read from Cathode

  • \$\begingroup\$ Do you have a link for the data sheet for the Si pmt? \$\endgroup\$ Apr 26, 2017 at 19:53
  • \$\begingroup\$ Here's a link to the product page which has links to the user manual and datasheet: sensl.com/products/j-series \$\endgroup\$ Apr 26, 2017 at 20:23
  • \$\begingroup\$ Your outputs now appear to be clipping - you should reduce your transimpedance gain. \$\endgroup\$ Apr 26, 2017 at 21:23
  • \$\begingroup\$ I ran some numbers with MATLAB, focusing on the feedback resistance (100 and 50 ohms) and capacitance at 12pF. If I reduce the resistance, and reduce the output gain and improve clipping, the cutoff frequency would be affected, reducing it from around 7.8 MHz to 1.56 MHz and the overall bandwidth. However, with the relatively large capacitance, and with the cutoff frequency being in the range of MHz, it really shouldn't matter too much in terms of noise cancellation? \$\endgroup\$ Apr 27, 2017 at 15:02
  • \$\begingroup\$ Sorry, I screwed up. Reducing the resistance from 100 to 50 ohms would increase the cutoff frequency from 78 MHz to 156.9 MHz, widening the bandwidth and allowing more noise. \$\endgroup\$ Apr 27, 2017 at 15:38

3 Answers 3


Glad to see somebody else is using silicon photomultipliers! They are nifty devices, I've been playing with them recently on a project.

First off, I'm not seeing a bias across the device - remember, you have to bias the silicon photomultiplier past its avalanche breakdown point to get any avalanching in response to photons. While the SPICE model may give you a current pulse when you trigger the fire input, this is not the real behaviour of the device. In your case, the grounded cathode should instead be raised up to about 27.5V to bias the SensL device into operation.

Unless cost is a significant issue, I would use four transimpedance amplifiers to read them out, one for each channel. This will solve your backfeed problem. Something like the OPA4354 (4 amplifiers, 250MHz GBWP, $8 in singles) might work for you. Since the SiPMs themselves are quite a bit more expensive than this, it probably won't break the budget. If you need to add the channels together in analog, this can easily be done afterwards with a resistive summer.

Keep in mind that if you read out at the anode with a transimpedance amplifier, the signal from your transimpedance amplifier will be inverted. This isn't an issue if you have bipolar supplies or use a virtual ground, but I thought I should mention it. If I recall correctly, anode readouts from the SensL devices are preferred for speed since the cathode is tied to the package ground and has greater capacitance.

Noise should not be a problem with a SiPM setup, since the current pulses are discrete (1 pulse height = 1 photon) and the peaks are well above the noise floor in a properly designed circuit. I would skip the feedback capacitor in the opamp circuit altogether unless you're having ringing problems. As you have likely surmised, adding the feedback capacitor in parallel with the feedback resistor creates a lowpass filter, which is going to directly impact your rise time. So, only do this if you need to combat ringing.

Finally, if you need extremely sharp risetimes, I would look at using the capacitively coupled fast output of the SensL device instead. This is described in the C-series manual.

  • \$\begingroup\$ Thank you for your reply. SensL said that their model didn't require biasing in simulation. It's done in their model, so there's no need for an external bias at the cathode. My team and I have already done a circuit where four SiPMs (anode biased at -30 V) had their own individual amplifier. We were wondering how the output would be if they went into one amplifier. Apologies, I modified my circuit so that it's read out from the cathode, so the output is positive. We need to have a high SNR ratio to combat noise as you said, so there's a tradeoff here. I'll have to speak with my team about it. \$\endgroup\$ Apr 26, 2017 at 20:23
  • \$\begingroup\$ @user101402 Sounds like you're on the right path then! You could probably make something work with one amp by putting ballast resistors from the SiPM to ground and doing a summing amplifier, but that'll introduce noise and you'll have to have everything well balanced. Considering how cheap a quad opamp is compared to the cost of a SiPM, I'd say four separate circuits is worth it :) \$\endgroup\$ Apr 26, 2017 at 20:28
  • \$\begingroup\$ Oh, also - since you're concerned about sharp rise times, physical layout is important. Build the design on a PCB, and take high-speed layout principles into account. \$\endgroup\$ Apr 26, 2017 at 20:35
  • \$\begingroup\$ Thank you for your comments. I'm unfamiliar with ballast resistors, so I did a quick search on them. They seem rather big and bulky for a small circuit. If it will introduce more noise into the circuit, then it would seem that this approach might not be worthwhile. What kind of ballast resistors were you looking at? The ones I saw seem too big for a small PCB board. \$\endgroup\$ Apr 27, 2017 at 15:40
  • \$\begingroup\$ I should clarify - by "ballast resistor" I am referring to the role of the resistor (a current limiting resistor), not a particular type of resistor. A resistor used to limit the current to a small indicator LED might be a tiny 0.6 x 0.3mm 0201, but it would still be functioning as a ballast resistor. I will draw a diagram of what I am proposing after work. However, I still think that using an $8 quad op-amp with a resistive summer after it is the best approach, especially since those SiPMs are like $100 a pop. \$\endgroup\$ Apr 27, 2017 at 16:51

Your first issue would probably be best corrected by buffering the output from each multiplier into their own voltage then summing those into the final amp.

The second issue would be better handled using a classic RC bypass filter rather than the integrator type you have chosen.

  • \$\begingroup\$ Thank you for your reply. I apologize; I'm not entirely familiar with amplifier circuits aside from basic ones, so this is a bit of a learning experience for me. How would I go about buffering the output for the multipliers with passive components? \$\endgroup\$ Apr 26, 2017 at 20:04
  • \$\begingroup\$ @user101402 you can't with passive, you would need to add an op-amp for each channel. or potentially a transistor circuit. \$\endgroup\$
    – Trevor_G
    Apr 26, 2017 at 20:05
  • \$\begingroup\$ Thank you for your reply. I was hoping to avoid using more op amps. Why do you say that a classic RC bypass filter would be better? Is it because I'm reading from the anode? I changed my schematic so that I'm reading from the cathode now, and an image of my new schematic is in the post. \$\endgroup\$ Apr 26, 2017 at 20:34
  • \$\begingroup\$ @user101402 The feedback filter is mean to be used on AC signals that have no bias and integrate out to zero. \$\endgroup\$
    – Trevor_G
    Apr 26, 2017 at 20:36
  • 1
    \$\begingroup\$ @user101402 I understand your desire to only use a single op-amp, but it makes life so much easier when you can isolate things that way, especially when things are not working the way you expect. Sometimes that alone is worth the extra few $. \$\endgroup\$
    – Trevor_G
    Apr 26, 2017 at 20:38

You need a voltage buffer (assuming that you really want to isolate the diodes in your model, that may or may not be how the real world works).

Use a universal op amp or an ideal op amp found in the Devices\opamps folder. Remeber the characteristics of ideal opamps, they can source infinite current ect.

If you just want to 'copy' a voltage to a node, use a B-source or in this case four of them

  • \$\begingroup\$ Thanks for your reply. How would a voltage buffer fix the issue of current from firing sensors going back into other sensors? It's a current signal coming from the photomultipliers, not a voltage signal. I tried using an op amp voltage follower for just one of the sensors, replacing the resistor with an ideal op amp. It made the output voltage stay constant for a really long time, and the fall time was too long. You can see this in the link below: drive.google.com/file/d/0B8_U5aaAFBjwWDd5aWRaNmE5XzA/… \$\endgroup\$ Apr 27, 2017 at 14:22

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.