Noise in Semiconductor Detectors

As was true for that of photoeffects, the objective of this discussion of noise mechanisms is to acquaint the reader with the broad concepts of noise in detectors without deriving in great detail the appropriate equations. See Van Vliet [2.141] for a detailed treatment. Nevertheless, it will be necessary to present certain equations which describe the dependence of noise upon internal material parameters and external system parameters. The discussion will consider initially noise in semiconductor detectors, followed by noise in photoemissive devices. 

In the absence of electrical bias, the absolute minimum internal noise exists, termed Johnson noise, Nyquist noise or thermal noise. This form of noise arises from the random motion of the current carriers within any resistive material and is always associated with a dissipative mechanism. The Johnson noise power is dependent only upon the temperature of the material and the measurement bandwidth, although the noise voltage and current depend upon the value of the resistance. 

Any other form of internally generated noise must depend upon bias. Since they add (quadratically) to Johnson noise, all other types of noise are referred to as excess noise. Three principal forms of excess noise exist. One amenable to analysis which is found in photoconductors is generation-recombination or g — r noise. A second, also amenable to analysis, which is found in photodiodes, i.e., p - n junctions and Schottky barrier diodes, is referred to as shot noise of diffusing carriers, or simply as shot noise. The third form of excess noise, not amenable to exact analysis, is called l//(one over/) noise because it exhibits a 1// power law spectrum to a close approximation. It has also been called flicker noise, a term carried over from a similar power law form of noise in vacuum tubes. 

The topics included here are limited to the usual types of noise in the common types of infrared photon detectors. Noise in thermal detectors, such as temperature noise in bolometers, is not included. Noise associated with the avalanche process is omitted. The detailed noise theory of phototransistors, an extension of shot noise in photodiodes, is not included. Modulation noise, an example of which arises from conductivity modulation by means of carrier trapping in slow surface states, is not included. Pattern noise, due to the 

The analysis of g—r noise has been the subject of study for nearly 40 years. The most rapid development of the theory came during the period 1950-1960, principally by Van der Ziel [2.142,143,148], Van Vliet [2.149,150], and Burgess [2.151-153]. Long [2.154] has clarified the application of the theory to infrared detectors. Many forms of the g—r noise expression exist, depending upon the internal properties of the semiconductor. Two of the most useful are those 

---

In comparison with the other sources of uncertainty discussed above, electronic noise in scintillation detector systems is a minor problem. More important, as we shall see, is gain drift caused by instability in the high-voltage supply. The priorities when selecting electronic modules for scintillation counting are somewhat different from those which determine a system for high-resolution (semiconductor) spectrometry. 

Noise arises in semiconductor detectors from several mechanisms. Johnson noise is found in all resistive elements. It has already been discussed in coimection with thermal detectors [see Subsection 5.1 l.b and Eq. (5.11.20)]. If the load resistance in the circuit is larger than the detector resistance, the Johnson noise of the detector element dominates because load and detector act electrically in parallel as far as the noise properties are concerned. 

Recent advances in semiconductor materials have made feasible room-temperature solid-state detectors made from crystals such as mercuric iodide. At present they are not competitive in noise and dynamic range with advanced scintillation or proportional detectors, but may become so in future. 

The angular precision required is substantial. The most important factor is for the detector axis to track the specimen axis accurately and continnonsly, to better than an arc second for most semiconductor work, corresponding to a reciprocal space resolution of 5><10 A. Random errors in the tracking result in noise in reciprocal space, while systematic errors give rise to systematic distortions of the reciprocal space map. Artefacts due to backlash and eccentricity of gear trains are noticeable, and direct axis encoders are mnch preferred. Absolute... 

The width F is indicated as electronic noise in Fig. 12.41. Of the three types of X-ray detectors mentioned—scintillation, proportional, and semiconductor counters—the Si(Li) detector has the best energy resolution for X-rays. This fact is demonstrated in Fig. 12.42, which shows the same energy peak obtained with the three different detectors. Notice that only the Si(Li) detector can resolve and lines, an ability absolutely necessary for the study of fluorescent X-rays for most elements above oxygen. The manganese fluorescence spectrum obtained with a Si(Li) detector is shown in Fig. 12.43... 

Only for particular molecules, e.g. ammonia because of its strong lines in the 20-40 GHz region, or water at 22 GHz because there is no other line until 183 GHz, would spectral considerations force the worker to lower frequencies. The 20-40 GHz band is also attractive, however, because of the cheap sources and low-noise semiconductor detectors, manufactured for movement sensors and short-path wireless links. The projected automobile collision-avoidance radar systems will make cheaper sources and detectors available for the 60-70 GHz region within the next few years. The 60 GHz across-office circuits for wireless data links could provide useful narrow-band sources for oxygen determination. The 35 GHz and 94 GHz close-range radar bands provide a useful reservoir of components and sources for the potential manufacturer of MMW spectrometers. 

The semiconductor detector is also based on ionization. The ionization takes place in the p-n jvmc-tion of two semiconductor materials. At room temperature the signal-to-noise (S/N) ratio of semiconductor detectors is poor but increases when the detector is cooled, e.g., by liquid nitrogen. The semiconductor detector has a very good energy resolution. 

Because of the electronic noise characteristics of semiconductor detectors, the maximum count rate of an ED system is usually no more than 30 000-50 000 counts per second, compared with in excess of 10 counts for a WD spectrometer, acting as a single channel analyzer. The reason is the relatively long count time constant that must be selected for ED preamplifier circuits (typically 6 ps), compared with < 1 ps for counter preamplifiers used on WD spectrometers. As a consequence, equivalent analytical precision can only be achieved by extending ED spectrum count times to a significant extent compared to WD. 

Semiconductor diodes are used in current mode to measure charged particles and are known as surface barrier detectors. They have very linear responses and are available with thin entrance windows. Surface barrier detectors are good beam monitors when used with low-noise current amplifiers. To understand the action of the particle detector, we will have to understand the basics of the semiconductor detectors. 

---