I'm trying to improve on a transimpedance amplifier design by changing out the op-amp that I'm currently using for another one. The issue I'm facing now is that I'm unsure as to how to pick a good op-amp based on my needs.

My transimpedance amplifier circuit is shown below. I'm using the SensL MicroFJ-60035 silicon photomultiplier (SiPM), and the amplifier I'm currently using is the TI OPA656.


simulate this circuit – Schematic created using CircuitLab

SensL requires you to give your email for their datasheet, but I'll try and put some info on the device here. If you need more information, please let me know, and I'll add more info here. The spectral range of the sensor ranges from 200-900 nm, and its dark count rate is 80 kHz/mm squared. It has a typical gain of 5.3*10^6, a dark current parameter of 4.1 uA, and a rise time of 300 ps. Also, the sensor is rated for a max current of 15 mA.

The OPA656 is a voltage feedback amplifier with a slew rate of 295 V/us, and it has a gain bandwidth product of 230 MHz and a bandwidth of 500 MHz. It has an input bias current of 1 pA, and it has an input offset voltage of 100 uV. Also, the capacitance at the anode output is 4000 pF. The datasheet never said about the capacitance at the cathode output.

Right now, the issue I'm facing with my design is the speed. Currently, it's too slow, with the rise time at 14 ns. My team and I have managed to reduce the rise time of the circuit down to around 6.5-9 ns by reducing the feedback resistor from 470 ohms to 25 ohms, but the gain is so tiny, with the max amplitude being around 160 mV. I am trying to find a new op-amp that would allow for a faster rise time without the necessary drop in gain.

As I did some research, I don't necessarily have a full understanding of the nuances of op-amp design, so I focused primarily on the slew rate. However, I found that as the slew rate increases, the bandwidth decreases. Here's my first issue. My team needs a certain bandwidth requirement, but if the relationship between the slew rate and the op-amp bandwidth is inversely proportional, how do I determine the amount of bandwidth I need for my application? Is it based on my SiPM, or is it from the test environment on which the SiPM is tested?

My next question also relates to the bandwidth of the op-amp. A smaller BW means that the amplifier won't work on high frequency signals outside of its range, but how does that affect the output voltage? I found some SPICE models of my SiPM as well as the OPA656 and some possible replacement amplifiers (specifically the AD8014, ADA4895, ADA4860, and the AD8001). These replacements have a faster slew rate, but the bandwidth is smaller. Would a smaller bandwidth result in a smaller but faster output voltage? If this is the case, then this isn't too big a problem. We don't necessarily need a large voltage signal; having one at around 600 to 850 mV is enough for our needs.

  • \$\begingroup\$ What is the amplitude of the signals you actually want to process? \$\endgroup\$
    – The Photon
    Commented Nov 7, 2017 at 16:38
  • 2
    \$\begingroup\$ An important parameter of the photodiode that you haven't given is its capacitance. \$\endgroup\$ Commented Nov 7, 2017 at 16:44
  • \$\begingroup\$ Typically slew rate and BW go up together. (but not always.) You usually care more about the BW. (what's the voltage of your 14 ns rise time? Is it close to 300V/us... ~ 4V/14ns?) My two fav books on photodiodes and such. Mark Johnson's "photo detection and measurement" And Phil Hobb's "Building electro optical systems" \$\endgroup\$ Commented Nov 7, 2017 at 16:56
  • \$\begingroup\$ @ThePhoton The max amplitude of the signal with a feedback resistance of 25 ohms is around 160 mV. My team would like it to be around 600-850 mV. \$\endgroup\$ Commented Nov 7, 2017 at 17:08
  • \$\begingroup\$ @GeorgeHerold With 470 ohms as the feedback resistance, the 14 ns rise time gave us an output voltage of approximately 1.75 volts. Given the requirements, I was hoping to sacrifice some of the output voltage in exchange for the speed to be faster. Thanks for the book suggestions. \$\endgroup\$ Commented Nov 7, 2017 at 17:14

2 Answers 2


With 1mA input across 470 ohms, that Vout of 0.47 volts will hardly challenge the slewrate; 0.47v is more in small signal region.

A high-capacitance sensor will kill your speed. And a slow-test-photon-pulse will confuse the measurements. What are your params?

And is the 50_ohm load the problem?

Consider, to debug the circuit, dropping in a common-base NPN transistor. If that is slow, then the sensor and the parasitic Vin node capacitance is the impairment.


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ Where did you see the value "1 mA" in the question? \$\endgroup\$
    – The Photon
    Commented Nov 7, 2017 at 16:39
  • \$\begingroup\$ Thanks for your reply. I added info about the capacitance from the datasheet of the sensor, though it only told me about the capacitance at the anode output, being 4000 pF. Also, I don't think that the 50 ohm resistor is the issue here. It's there to match the coaxial connector impedance. \$\endgroup\$ Commented Nov 7, 2017 at 17:23

Here is a good resource on optimizing high-frequency response of TIA circuits:


To summarize, I would look at three areas:

1) You may have increased performance just by using a better op-amp. The LTC6268IS6-10#TRMPBF has better specs on paper, but it too will suffer if there is too much parasitic capacitance on the input to the amp.

2) Have a detailed look at parasitic capacitance in your PCB layout. The guide I linked to above shows a handful of ways to decrease parasitic input capacitance. You can take some of this out of the equation by lifting the leg of the input pin of your IC and "dead bug" wire the output of the SiPM into it via a wire above the board. Similarly, solder your feedback resistor on top of the IC, and wire up to appropriate pins. This would be the best case scenario with very little feedback or input capacitance.

3) Examine your test setup. 300ps is pretty fast. Can your scope handle it? You may want to run a test signal in and see what the fastest signal you can measure is so that you can get a baseline for your measurements. There are many signal sources you can use, but here is one I can vouch for:


You can get a fast rise-time module for this and get sub-nanosecond (100ps?) 5V signals out of it. At $5k, it's not cheap, so others may have a cheaper way to test your setup.


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.