Goals This laboratory will help you become familiar with the operation and characteristics of two detectors commonly used for optical radiation: silicon photodiodes, and photomultipliers.

Background Quantitative optical detectors generally fall into three classes: thermal detectors, photoconductive or photovoltaic detectors, and photomultipliers, which actually multiply electrons, not photons. In addition, there are several techniques used for special experimental situations, such as photoacoustic and optogalvanic detection. The basic theory of optical detectors can be found in many texts; a particularly good reference is Detection of Light: From the ultraviolet to the submillimeter , by G.H. Rieke. Other sources are Verdeyen, Wilson and Hawkes, Yariv, or Swindell. Also, useful practical information can often be found in manufacturer's catalogs and application notes. [6, 7, ] Thermal detectors have the advantage of relatively simple absolute calibration that is constant over a very broad spectral range. They have a very slow response time and are used primarily either for low power cw radiation, e.g., solar radiation, or for single-shot energy measurement of low repetition rate lasers. Sensitive disk thermopile detectors can be quite fragile, and are often reserved just for calibrating other detectors. You will not be using a thermal detector in these experiments. Photodiodes When a photon is absorbed within a pn-junction, an electron-hole pair is created. If the generation takes place within the junction depletion region, or if the carriers diffuse to the depletion region before recombining, the electron and hole move to opposite sides of the junction under the influence of the intrinsic field in the depletion region, resulting in a current. If the device has no external connections, or only a very high impedance connection, this current must be balanced by a diffusion of majority charge carriers against the intrinsic field. This results in a lowering of the voltage across the junction that can be measured. The voltage change is a logarithmic function of the photon-induced current, and the response is slow. This method of photodiode operation is called the photovoltaic mode; the Newport 818 detector head operates in photovoltaic mode. If an external reverse bias is applied, the additional field suppresses the majority carrier flow, and a net current flows through the bias circuit. The photon-generated carriers produce a current in the external bias circuit that can be measured, usually by measuring the voltage across a load resistor. This is called the photoconductive mode of operation. Personally, I think this is a poor name, because there are true photoconductive detectors: resistors whose value changes with illumination. A better term might be photocurrent mode, since it is the actual photon-produced current that is measured. The time constant of a biased photodiode depends on both the internal structure and the external circuit. A simple model of the system is shown in Fig. 1. The current source represents the photo-induced current. The reverse-biased diode can be modeled as an ideal diode plus a parallel plate capacitor with the area of the junction, and a separation equal to the depletion region. The time response of the measured voltage is just the RC time constant of the external load resistor and the junction capacitance. Lowering the load resistance decreases the time response of the circuit, but at the cost of a lower voltage for a given light input, since the diode is a current device. Thus, low junction capacitance is an advantage. Small active areas reduce capacitance, but at the cost of reduced response (smaller sensitive volume) and experimental difficulty (harder to align the small area). Higher reverse bias (within limits) results in a wider depletion region and a lower capacitance. Also, the width of the depletion region is often enhanced in photodiodes by including an intrinsic region between the p- and n-doped regions; these are referred to as PIN diodes. The larger active volume of the PIN structure also increases sensitivity. Thermally generated carriers in a large junction volume can produce measurable noise current in the circuit, called dark current. Photodiodes are available in a wide variety of configurations that make different trade-offs in these parameters for different applications. Finally, the spectral response of photodiodes depends on the optical properties of the semiconductor used, such as the energy gap and absorption. We will be using silicon photodiodes, which have a broad response covering the visible and near ir, but other materials are available for other regions. Photomultipliers Photomultipliers rely on the photoelectric effect to convert photons into electrons at a photocathode. The work function of the photocathode material determines the spectral response of the detector, and the formulation of broadband, sensitive photocathode materials is an art. Electrons from the photocathode are accelerated and impact on the next element, a dynode. Impact ionization at the dynode produces several electrons, which in turn are accelerated into the next dynode. The dynodes are biased and shaped in order to accelerate and focus the electrons through the tube, from element to element. The supply voltage, usually several kilovolts, is divided by a resistor chain to provide the dynode bias voltages. Typical photomultipliers have from 8 to 14 stages of electron multiplication, leading to total current gains of 10^6 or more at the anode. Electron multiplication is probably the lowest noise method of obtaining gain, which makes photomultipliers the best choice for low level light detection. The equivalent circuit for a photomultiplier is the same as for a photodiode (see Fig. 1, with a gain much greater than 1; the capacitance results from the internal dynode structure. The dynodes are spaced relatively far apart, but there are a lot of them. The end result is that photomultipliers usually have a somewhat slower time response than photodiodes. The fastest tubes have rise times of a few nanoseconds.

Photodiode Experiments In these experiments you will investigate the properties of two similar silicon PIN photodiodes. The size of the active area of the two diodes are different, as you can see just by looking at them. The large one has an area of 13.7 mm^2, and the small one has an area of 1 mm^2. Since the detector capacitance is a function of area, very high speed photodiodes often have much smaller active areas; the New Focus 25 GHz photodiode is only 25 micron in diameter. The diodes are mounted in a Thor mount that consists simply of a mounting socket, a 22 V bias battery, and a switch; you have to supply the load resistor to complete the bias circuit and to convert the photocurrent into a voltage. We have constructed convenient load boxes for this purpose with a selection of resistors. (Obviously, this is not the optimum configuration of a load for high speed measurements.) Calibration Calibration of a photodiode consists simply of applying a known optical power to the active area and measuring the resulting current/voltage. There are two general ways to do this: we could flood the diode with an optical field having a known power density, or we could focus a beam of known power onto the active area. We will use the first technique since it corresponds more closely to the usual experimental situation. Plus, focusing the beam to a small area on the detector can cause local saturation, and there is always some uncertainty in whether all the focused beam is actually collected by the active area. First we need to create a beam of known, constant power density large enough to fill the detector area: Establish an experimental axis on the table and align the HeNe laser using two steering mirrors as you learned earlier. Expand the beam using a negative lens, focal length -73 mm, placed about 30 cm from the last steering mirror. The lens must be centered on the beam axis so that the beam propagates along the same axis as it did previously. Use an aperture or a mark on a screen to locate the beam before you insert the lens, and then adjust the lateral position of the lens to recenter the expanded beam on the mark. The beam should expand to a ~ 3 cm diameter about 80 cm from the lens. If you place a screen there, it is obvious that the beam intensity is not uniform; it fact the power density has a gaussian distribution. A small region near the center, however, is reasonably uniform. Place an iris ~ 80 cm from the lens; center it in the beam. Close the iris to a diameter of about 5 mm; you can either measure this with the ruler, or close the iris around a precision metal dowel pin of known diameter. Try the 9/16 inch (4.76 mm) dowel. (One could also drill or punch a hole in a thin metal sheet or card.) The resulting beam should appear very uniform. Estimate the uniformity based on a gaussian distribution for the whole beam. Measure the power transmitted by the aperture using the Newport meter, and calculate the power density. Now you have a beam of known power density. Of course, the beam will continue to expand in size beyond the iris, so you need to work close to the iris, or remeasure the beam size at the location of your detector. These photodiodes will measure the ambient light as well as the input beam, just like the Newport detector, and you must subtract the ambient signal manually. It is convenient to use a chopper wheel when making measurements, since it provides alternate on and off periods for easy determination of the minimum and maximum voltage levels. The difference corresponds to the light you are trying to measure. Place the chopper either right after the laser or in between the steering mirrors, so it blocks the laser but not ambient light that may enter the detector during a measurement. Connect the large area photodiode through a load box to the oscilloscope; set the load to infinity so you are using the 1 MegaOhm input resistance of the oscilloscope, and make sure the input channel is set to dc. Align the diode so it is centered in the beam. It may be easier to move the diode about 60 cm away from the iris where the beam expands to about 8 mm and fully covers the face of the diode can (but remember to recalculate your power density). Measure the voltage. Check for saturation and use attenuators if you need to. Calculate the diode response in terms of amps per watt. Use lower values of load resistance and compare your calibration. Comment on any differences. Repeat the process to calibrate the small area diode. Compare the two responses to the differences in areas. Linearity and Dark Current Check the linearity of the detectors by inserting attenuators in the beam and measuring the response. You should be able to add attenuations of 3 OD (small area detector) to 4 OD (large area detector) and still have a measurable signal into a 1 MegaOhm load. Since these light levels are very low, you will need to use a custom light shield, a.k.a. a cardboard box. Place the box over the detector and put the attenuators right against the input hole of the box. At the highest oscilloscope sensitivities you should see random voltage spikes generated by thermal electron-hole pairs. You will have to turn off or dim the room lights, even with the shield. Confirm the source of these spikes by covering the input hole of the box; notice the effect of shining a flashlight near or at the box edges. Are the spikes simply electrical noise picked up by your cables? What is the minimum detectable signal, both in terms of watts and photons per second? Time Response In many applications photodiodes are not used to make quantitative power measurements, but rather to determine the time behavior of an optical source. In such cases, it is important that the time response of the detection system be faster than any variation of the measured signal, otherwise an integration takes place. For example, the detector in a 1 Gbps optical communications link must have a sub-nanosecond response time. The purpose of the next set of measurements is to determine the time response of the detector and compare it to our simple model of Fig. 1. We will need a much faster light source than the chopper to test our detectors. We will use a light emitting diode (LED) driven by a fast pulse. You can turn off the HeNe laser and the chopper. Set up the SRS Digital Delay/Pulse Generator to produce pulses at a 1 KHz rate, 100 microsec long, and 4 V in amplitude at the AB output. (You should terminate unused generator outputs.) Connect the AB signal to Channel 2 of the oscilloscope in order to monitor the pulse and trigger the scope. Connect the pulse generator AB output to the LED in the modified Thor mount; this mount contains no battery or switch; it just provides a convenient mechanical mount and 50 Ohm load resistor and electrical connection for the LED. You should be able to see the LED light up when the pulses are applied; if not, reverse the polarity of the LED in the mount. Place the large area photodiode right up against the LED in order to collect as much light as possible. Connect the output of the photodiode through the load box to Channel 1 of the oscilloscope. Determine the rise and fall time of the detection system as a function of the load resistance. The LED's light output essentially follows the electrical pulse and has rise and fall times of about 5 ns (trust me for now), so the waveform you observe with the oscilloscope corresponds to the response of the detector system. Start with a 1 MegaOhm load; you should observe a pulse of amplitude around 3 volts with rise/fall times of 300 microsec. You can set up the oscilloscope to automatically measure and read out these pulse parameters; it saves a lot of time. See oscilloscope module. Switch to lower load resistances and notice the change in response time. For loads of 10 kiloOhm and less you will want to shorten the pulse length to 10 microsec or less so you can increase the sweep rate and still see the entire pulse. What is the lowest load resistance, and fastest rise and fall times you can measure? You will find that you run out of signal before you can come close to determining the true time response of the LED pulse. BUT, how would you know this if I had not told you that the LED pulse was very fast? Describe a simple experimental test for determining whether you are observing the true time behavior of the light or the time response of your detection system. From your measurements, determine the equivalent shunt capacitance of the detector system. What load resistance would be necessary to observe a 5 nsec rise time? What voltage would be available for measurement? Optional. If you have the time and inclination, repeat the measurements using the small area photodiode. It should be faster according to our model, but you will not be able to observe much of a difference because you will run out of signal level even sooner (higher load resistances) because of the smaller light collection area. In summary, low load resistances are necessary to get fast response times, but they can result in signals that are too small to measure reliably. One needs either more light or more detector gain. You can try using lenses to collect more light from the LED and to focus it onto the detector. This is not easy; I was only able to gain about a factor of two, but sometimes two is a lot. You can also try driving the LED harder to get more light: the connectors on the back panel of the SRS pulse generator produce 1 microsec--long pulses with ten times the amplitude. You might destroy the LED (I have not tried this), but they are cheap. Next, we will investigate a more sensitive detector.

Photomultiplier Measurements The 1P28 photomultiplier (PM) is described in the handout. It is a convenient general purpose tube with a response throughout the visible. Like most things made of glass, photomultipliers are fragile. The mechanical shock from, for example, dropping the tube on the table or knocking the mount over, can damage the internal components, even if the glass envelope is not broken. So please use care handling the PM. The tube is mounted in a housing that provides some mechanical protection, shields the tube from stray light, and makes electrical connections to the base of the tube. You can remove the top of the housing to examine the tube itself. High voltage for the tube is generated in the housing base by a dc-dc converter, so there is no external high voltage. There are two connections on the housing base; the multiconductor cable connects the tube to the control box and provides power and monitors the tube voltage. The supply voltage applied to the PM is adjusted using a knob on the front of the control box, and the voltage on the tube is indicated on the digital display. The gain of the PM is approximately an exponential function of supply voltage. The BNC connector on the housing base is the signal output. The outer shell is grounded and the center conductor is connected to the anode of the tube. (The BNC connectors on the rear and front of the control box are connected directly together, and are just for convenience of mounting.) Like silicon photodiodes, photomultipliers produce a photocurrent proportional to the input light and require a load resistor. Since electrons flow out of the anode to ground, the signal voltages produced across the load resistor will be negative. Photomultipliers are much more sensitive than photodiodes, and can be damaged by excessive light levels; actually, it is the resulting current that damages the tube elements by overheating them. The maximum average output current for this tube is 0.1 ma . NEVER expose a PM directly to room light when the voltage is on, and always begin measurements with the minimum supply voltage (minimum sensitivity) and then turn up the voltage while monitoring the current. Note that some tubes also have a maximum photocathode current, which is independent of the gain or voltage. Exceeding this value damages the photocathode material and reduces the response, or destroys the tube. Photomultipliers designed for very low noise, low light level detection should be shielded from strong light even when off to protect the photocathode. Response Calibration In the next set of experiments you will determine the response of the PM as a function of operating voltage. The photocathode has sufficient area to accept the entire HeNe laser beam, so you will not need the negative lens and iris. Just measure the HeNe power with the Newport power meter. You will want to use the chopper again to provide a convenient reference level. Make a rough table indicating the maximum allowable signal voltage for various load resistors. Use the specification sheet to estimate the maximum allowable input optical power for the 1P28 PM at full sensitivity/voltage. Have sufficient optical attenuators available to reduce the HeNe laser output to this level. Make sure the PM control box is off and check the connections between the PM housing and the control box. Connect the signal cable through a load box to the oscilloscope. Set the load box to 10kS, and set the supply voltage knob to the minimum value, fully counterclockwise. The oscilloscope should be set for DC detection; convenient starting scales are 1 msec/div and 10 mV/div. Align the PM so that the HeNe beam enters the photocathode, lock it down, and then shield it from stray light with a box. Mount the 633\nm spike filter right at the input to the shield to further reduce ambient light. You can seal the filter directly to the box with black tape. Turn on the power supply and adjust the supply voltage to 100~V and note the signal level. Increase the voltage to 150 V, and then to 200 V. Why is the signal waveform distorted? Return to 150 V and add a 1 OD attenuator in the beam just before the spike filter. This sequence will provide data to calibrate the filter, if you need to. Then increase the voltage to 200 V again and determine the PM response. Determine the PM response as a function of voltage by continuing in this manner, adding attenuators as necessary to avoid saturation and the output current limit. Increase the voltage in 50 V steps up to perhaps 400 V, where you will probably need to dim or turn out the room lights. Then use 100 V steps up to the tube limit of 1200 V. Above about 800~V you will begin to see high frequency noise on the signal; speculate on the source. Next, experiment a little to get a feeling for the maximum sensitivity of the PM. First turn off the supply voltage (just turn off the power to the control box) and then block the HeNe beam and cover the input to the box with a card or black tape. Now turn the supply voltage back on and observe the oscilloscope while you change room lighting conditions, shine a flashlight at the box, etc. Try larger and smaller values of load resistors: larger for higher sensitivity, smaller for faster time response. Can you detect thermal dark noise; single photons? After lab, make a plot of the PM response versus supply voltage and compare it to the specification sheet. Also compare the maximum sensitivity to the photodiode response. Time Response You will again use a pulsed LED as a fast light source to test the time response of the PM. Set up the LED as described in Sec. 3.3. First, measure the LED power output using the Newport power meter and setting the pulse generator for a 50% duty cycle. Place the Newport detector right up against the LED so as to collect all the light, and cover the combination with a box to exclude room lights, or turn off all the lights. Next, set the pulse generator to produce a 50/±sec long pulse. Place the LED as close as possible to the PM photocathode, lock both down to the table, and cover the combination up with a light shield. As with the photodiode, you should determine the time response (rise and fall times) of the detection system as a function of load resistance, but now you will also vary the PM supply voltage and the pulse length. It will probably be convenient to collect your data in the form of a rough table in your notebook with columns for the variables, data, and any notes. We are primarily interested in low values of load resistance, so you can start with a load resistor of 10kS to save time. Feel free, of course, to investigate larger values. Adjust the PM voltage to minimum and then turn on the control box. Slowly raise the PM voltage to get a good signal, perhaps 0.5 V. Note the rise and fall times, the general pulse shape, and the signal level. Decrease the load to 1kS and increase the PM voltage as necessary. You will probably find it convenient to shorten the pulse length to a few microseconds as the time response improves. A note on current limits: the 0.1 ma limit is a time average value. Now that the duty cycle is very low, higher peak currents are acceptable, but care is still warranted. Increase the PM gain just sufficiently to obtain a reasonable signal level, say a tenth of a volt. In addition, the bias network for this tube is not optimized for pulse operation. Large peak currents, even if not damaging, extract large currents from the final stages of the bias network and can disturb the voltage distribution, changing the tube gain. This results in time dependent, pulse shape dependent saturation that can be very difficult to detect. Many experimenters have been fooled by this effect. Don't ask me how I know. Observe the pulse shape with the lower load resistances, 200 and 50S. You will need supply voltages of nearly 500 volts, and the rise and fall times will still be greater than 100 ns, which is slower than most pulsed laser outputs, for example. You may be able to gain some additional time response by lowering the load resistance further, but clearly signal level is again becoming a problem. Try changing the oscilloscope input impedance to 50S so that the total is now 25S. Given your observed time response, what data speed communications system would be possible? The answer is not very impressive. In some applications, such as digital communications, it is not necessary to fully resolve the pulse shape, but only to detect the presence of a pulse and perhaps the timing of the leading edge. Thus, considerable integration of the pulse energy by the detection system is acceptable. In fact, this is one of the advantages of digital systems compared to analog systems. Even so, the detection system must be sufficiently fast to recover before the next pulse arrives, which could be in only a few nanoseconds. Shorten the pulse driving the LED to see how short a pulse you can detect. A better criteria of ``detection'' than just observing the pulse on the screen is to see if the oscilloscope can trigger reliably on the detected signal. Is this possible for a 10\ns long pulse? Minimizing the capacitance of your entire measurement circuit can improve performance. Turn off the PM voltage, and connect the PM signal output on the housing directly to the 50S oscilloscope input with the shortest possible cable. Replace the light shielding, and observe the time response of the system. Now how short a light pulse can you detect? Another approach would be to place a high gain transimpedance amplifier (current-to-voltage) right at the PM anode, in the housing, to drive the capacitance of the cables and detection system. You should be able to detect a 1 or 2\ns long pulse using a PM supply voltage of 900 V; you will have to turn out the room lights and the signal voltage will be less than 50 millivolts. Shining a flashlight near the light shield will produce about the same size pulses. Approximately how many photons are in your 1\ns pulse? Of course, we have no way of knowing if the light from the LED is really following the input pulse for nanosecond pulses. Obviously, we can't measure it with this detection system. At the least, the LED efficiency is probably reduced for short pulse lengths. If you wish, you can check the ability of this detection system to resolve two closely spaced pulses. The pulse generator will produce two pulses of arbitrary length and separation, one on the AB output, the other on the CD output. Experiment with combining them to drive the LED, and observe the resulting detected signal. These experiments should give you some appreciation of the detection problems involved in building a 1GHz optical communications system. Current optical telephone links operate at 2.5, 5, and even 10Gbps rates. What kind of detector do you think they use? Check some references to find out.