Shot noise is a type of electrical or optical noise caused by the fact that electric current and light are not smooth, continuous flows. They are made up of individual particles: electrons in a wire, photons hitting a camera sensor. Because these particles arrive at random, slightly unpredictable intervals, the signal they produce fluctuates slightly from moment to moment. Those tiny random fluctuations are shot noise.
Walter Schottky first identified this phenomenon in 1918 while studying vacuum tubes running at low currents. He called it the “shot effect” (schroteffekt in German), and the name stuck. A century later, shot noise remains one of the fundamental limits on how precisely electronic and optical instruments can measure anything.
Why Discrete Particles Create Noise
The standard analogy for electrical current is water flowing through a pipe, smooth and continuous. But current is actually carried by individual electrons, each with a tiny discrete charge. If you could watch electrons passing a single point in a circuit, you’d notice that a few more drift by in one moment and a few fewer in the next. The variation is completely random. You can predict the average number of electrons passing per second, but never the exact count for any given interval.
For shot noise to appear, two conditions need to hold. First, the charge carriers must be flowing in one direction, as they do when crossing a junction in a diode or transistor. Second, each carrier’s arrival must be independent of every other carrier’s arrival, a purely random event with no coordination between particles. When those conditions are met, the statistics governing the process follow a pattern called a Poisson distribution, where the variance (the spread of the fluctuations) equals the mean number of particles counted. This is the mathematical fingerprint of shot noise.
The Core Formula
Shot noise has a remarkably simple relationship to the current flowing through a device. The noise power is proportional to the average current multiplied by the charge of a single electron. Specifically, the noise power spectral density equals twice the electron charge times the DC current. In practical terms, this means that if you double the current, you double the noise power. But because noise is usually expressed as a fluctuating current (the square root of power), the noise amplitude grows more slowly, proportional to the square root of the current.
This square-root relationship has an important consequence for signal quality. In a system where shot noise is the dominant source of interference, the signal-to-noise ratio equals the square root of the signal itself. So collecting 100 photons gives you a signal-to-noise ratio of 10, while collecting 10,000 photons gives you 100. You gain one order of magnitude in clarity for every two orders of magnitude in signal strength. On a graph, this traces a straight line with a slope of one-half.
How Shot Noise Differs From Thermal Noise
Thermal noise (sometimes called Johnson-Nyquist noise) is the other major source of fundamental noise in electronics. It arises from the random thermal motion of electrons in any conductor and depends on temperature. Cool a resistor down and its thermal noise drops. At absolute zero, thermal noise vanishes entirely.
Shot noise does not work this way. Because it stems from the discrete, particle-by-particle nature of current flow rather than from heat energy, lowering the temperature does nothing to reduce it. The only factor that matters is the current itself: more current means more shot noise; less current means less. Both types of noise are spectrally flat, meaning they spread their energy evenly across all frequencies (making them forms of “white noise”), but their physical origins and the ways you can control them are fundamentally different.
Shot Noise in Photography and Imaging
If you’ve ever taken a photo in dim light and noticed a grainy, speckled texture, you’ve seen the effects of photon shot noise. A camera sensor works by counting photons that hit each pixel. When plenty of light is available, each pixel collects thousands or millions of photons, and the random variation between pixels is a tiny fraction of the total. The image looks clean. In low light, each pixel may collect only a few dozen photons. The random Poisson fluctuations then represent a large fraction of the signal, and the image looks noisy.
This same principle affects fluorescence microscopy, astronomical imaging, and any situation where photons are being counted. Camera manufacturers can engineer away other noise sources: thermal noise drops when the sensor is cooled, readout noise can be reduced with better electronics, and digitization noise improves with finer analog-to-digital conversion. But photon shot noise is inherent to the quantum nature of light. No amount of electronic design can eliminate it. The only way to reduce its relative impact is to collect more photons, either by increasing the light intensity or by extending the exposure time.
Where Shot Noise Matters Most
Shot noise becomes the dominant concern in systems that operate at very low signal levels. In fiber-optic communications, photodetectors converting light pulses into electrical signals deal with shot noise as a fundamental sensitivity floor. In semiconductor devices like diodes and transistors, shot noise appears wherever current crosses a junction. Precision analog circuits, such as those in scientific instruments and medical devices, must account for it when amplifying tiny signals.
At higher signal levels, shot noise doesn’t disappear, but other noise sources (thermal noise, interference, amplifier noise) tend to overshadow it. The transition point depends on the specific system. In a well-designed, cooled camera sensor imaging a bright scene, shot noise may still be the largest remaining noise source simply because everything else has been minimized.
Reducing Shot Noise in Practice
Because shot noise scales directly with current, the most straightforward way to reduce it in an electronic circuit is to use smaller currents. This is a real design constraint in precision analog electronics, where engineers deliberately keep bias currents low to minimize noise at sensitive input stages.
In imaging and photon-counting applications, the strategy flips. You want more signal, not less, because the signal-to-noise ratio improves with the square root of the photon count. Doubling your exposure time, for instance, doubles the signal but only increases the noise by a factor of about 1.4, so the image gets cleaner. Using a sensor with higher quantum efficiency (one that successfully converts a larger fraction of incoming photons into electrons) has the same effect.
In some quantum-mechanical systems, correlations between charge carriers can actually suppress shot noise below the level predicted by simple Poisson statistics. This is called sub-Poissonian noise, and it occurs in certain nanoscale devices and quantum-optical setups where particles are no longer fully independent. The ratio of observed noise to the standard Poisson prediction is known as the Fano factor: a value of 1 means standard shot noise, below 1 means suppression, and above 1 means enhanced noise. These effects are mostly relevant in research settings, but they reveal that shot noise is not just a nuisance. It carries information about how particles interact and correlate inside a device.