Technical Article

Silicon Photomultiplier (SiPM) Structure, Characteristics, and Applications

November 21, 2019 by Amos Kingatua

Learn about the structure, characteristics, and applications of a silicon photomultiplier (SiPM).

A silicon photomultiplier (SiPM) is a solid-state, high-gain radiation detector that produces an output current pulse upon absorption of a photon. These P-N junction-based sensors with single-photon sensitivity can detect light wavelengths from near-ultraviolet (UV) to near-infrared (IR).

Generally, the compact, solid-state SiPM provides a better alternative to bulky photomultiplier tubes and is suitable for sensing, quantifying, and timing all levels of light down to a single photon.


SiPM Applications and Benefits

Major benefits of the SiPM include high gain, low voltage operation, excellent timing performance, high sensitivity (down to a single photon), and immunity to magnetic fields. These characteristics make it a good choice for light detection applications from single and up to several thousand photons.

SiPMs are compact devices with the ability to withstand mechanical shocks. Their excellent performance makes them suitable for a wide range of photometry (light detection) applications, especially in situations where precise timing is necessary.

Typical SiPM applications include biophotonics, LiDAR and 3D ranging, high-energy physics, aero particle physics, sorting and recycling, hazard and threat detection, fluorescence spectroscopy, scintillators, medical imaging, and more.

Silicon photomultiplier market sectors include industrial, aerospace, automotive, oil and gas, electronics, and information and communications technology.


Flow cytometer application. Image used courtesy of Hamamatsu

Manufacturers often customize an SiPM’s physical size, design, and other parameters according to the application and target light. For example, UAV applications use miniaturized sensors while field gamma spectroscopy operations rely on physically larger assemblies. Also, there are RGB SiPMs optimized for visible light and NUV SiPMs for the near-ultraviolet region.


SiPM structure 

An SiPM consists of an array of hundreds or thousands of self-quenched, single-photon avalanche photodiodes (SAPDs), also referred to as pixels or microcells.

Each SAPD, designed to work when biased above the breakdown voltage, has an integrated series quenching resistor, an anode, and a cathode for the standard SiPMs.


Standard SiPM structure; SPADs connected in parallel 

Some manufacturers, such as SensL, have a fast output SiPM with a third output terminal in addition to the anode and cathode. This has an integrated fast output capacitor at the SPAD anode.


SensL fast output SiPM. Image used courtesy of ON Semiconductor 


In practical applications, the SiPM consists of hundreds or thousands of microcells in parallel. This gives it the ability to detect multiple photons simultaneously and is useful in various light and radiation detection applications. The electrical output directly correlates to the number of photons that the pixels absorb.


Basic Operation of a Silicon Photomultiplier

The micron-sized SAPD microcells are designed to operate in the Geiger mode reverse-biased condition, just above the breakdown voltage.


Biasing the SiPM. Image used courtesy of ON Semiconductor 


The figure below shows an equivalent circuit of the APD. Generally, the P-N junction acts as a photon-operated switch. Without light falling on the microcell, switch S is open and the voltage on the junction capacitance CJ is VBIAS


Equivalent circuit of a SiPM. Image used courtesy of Hamamatsu


When a photon lands on the microcell, it generates an electron-hole pair. One of the charge carriers then drifts to the avalanche region where it initiates a self-sustaining avalanche process and current flows. If unquenched, the current will flow indefinitely.


SiPM output current pulse from a micro-cell upon absorbing a photon. Image used courtesy of First Sensor


The switch S instantly closes upon the initiation of the avalanche and CJ discharges from VBIAS to VBD (breakdown voltage) through Rs (APD internal resistance) with a time constant of RSCJ.

As the quenching occurs, switch S opens and VBIAS recharges CJ with the time constant RQCJ. The APD is in its recovery phase and resets back to Geiger mode awaiting the detection of a new photon.


Characteristics of SiPMs

Photon Detection Efficiency (PDE)

The photon detection efficiency or PDE quantifies the ability of the SiPM to detect photons. This refers to the ratio of the number of detected photons to those that reach the SiPM. The PDE is a function of the overvoltage ΔV across the terminals of the APD and wavelength λ of the incident photon.


Breakdown Voltage

The breakdown voltage (VBD) in a SiPM is the minimum (reverse) bias voltage that results in self-sustaining avalanche multiplication. When VBIAS is above VBD the SAPD outputs a current pulse. The difference between VBIAS and VBD is the overvoltage ΔV which controls the operation of the SiPM. Increasing the overvoltage ΔV improves PDE and SiPM performance. However, there is an upper limit beyond which the noise and other disturbances, which increase with overvoltage, start interfering with SiPM operation. 

The breakdown voltage depends on the temperature and other SPAD characteristics. As such, the datasheets usually specify the breakdown voltages for different temperatures. 



Recovery Time

This is the time it takes between the quenching of the avalanche and when the microcell fully resets and gains the ability to detect an incoming photon. During the recovery time, the microcell slightly loses its ability to detect new incoming photons. The time constant of the recovery phase is RQCJ.


Temperature Characteristics

Temperature directly influences the breakdown voltage, gain, junction capacitance, dark counts and the photon detection efficiency. In particular, the breakdown voltage is higher at elevated temperatures and will affect the gain and photon detection efficiency which are directly proportional to the overvoltage. Higher temperatures will also increase the probability of dark events due to thermally generated charge carriers.


Noise in Silicon Photomultiplier

Semiconductor impurities and other factors often cause random output pulses both in the presence and absence of light.


Primary Noise – Dark event

Thermal agitation and other factors often lead to the generation of random electron-hole pairs and carriers. If the random carrier enters into the avalanche region of the APD’s depletion region, it travels through the high-field region where it triggers an avalanche Geiger discharge and an output current pulse. The generation of the pulse in the absence of light is known as a dark event. The dark count rate refers to the number of dark events in a specified period and is expressed as counts per second (cps).


Correlated Noise

Correlated noise refers to the output from the secondary avalanche discharges triggered by a previous photon or dark event. The two main types of correlated noise are the Afterpulsing (AP) and Optical Crosstalk (OC) events.



Afterpulsing occurs when carriers trapped during the avalanche multiplication in the silicon are discharged during the recovery phase of the SAPD. The carriers end up generating a new secondary current pulse of a lesser magnitude than the original.


Normal SiPM output pulse and afterpulsing noise output graph


Optical Crosstalk in a SiPM

Optimal crosstalk (OC) occurs when a primary avalanche in one microcell triggers a secondary avalanche in adjacent microcells. The net effect of the secondary discharge (avalanche) on the output current pulse is that it increases the amplitude of the output signal, such that it is higher than that produced by the incident photon. 

The probability of optical crosstalk (OC) increases with overvoltage.




Silicon photomultipliers are compact, solid-state optical sensing devices with high gain and ability to detect light down to the photon level. The technology is finding applications in a wide range of fields and industries but has a few drawbacks, such as noise, that can limit its performance. However, SiPM technology is still improving and has great potential as it matures.

  • P
    pyroartist January 21, 2020

    So when using this type of APD is the light value computed by counting pulses over a known time period or is there just a capacitor charging circuit that accumulates all the many very fast pulses to a voltage?  Please provide details and a circuit. There is very little written about how APDs are used in their avalanche breakthrough, single photon counting mode. For example what is a typical count of average room lighting over say, 1 microsecond. Is it millions of photons or just hundreds? Thanks for writing this.  Where can I purchase one of these devices? Are they expensive?

    Like. Reply
    • P
      pmd34 January 22, 2020
      I have played with some of the SPMs myself, and various amplifying circuits. You can buy them from Farnell such as: Quite reasonably priced, but they are not quite as ideal as presented in the article: There are A LOT of "dark counts" tens of thousands per second. Each of the dark counts is, mostly, from a single channel so the pulse height corresponds to just 1 channel firing. So for detecting you have to implement a minimum height threshold, and rely on triggering more than one channel at once for the events you actually want to measure. You could potentially cool the SPM down to reduce the dark counts, but whether you can cool it enough to reduce the dark counts completely, and still have a functioning SPM and amplifier (as this is usually right next to the SPM to reduce noise and capacitance) I am not sure. For a typical amplifier the pulses out of the counter are about 100-200ns long with a very fast rise and exponential decay. Normally some form of pulse shaping (differentiator, integrator etc.) is used to allow you to measure the height, simple pulse counting is easy enough however. I have seen some work published that has shown it is also possible to use the pulse duration as a simpler way of equating the number of channels "fired" rather than trying to do a very fast pulse height measurement. For the number of photons see something like this: Basically its a lot, and something like a PTM or SPM goes into saturation long before this kind of intensity, effectively you have to have enough time for one pulse to have partially died away before the next one comes, other wise you would use a photo-diode type detector. You also need to bias the SPMs with a very stable noise free voltage of about 27V.
      Like. Reply