Read AP304_Lab_1.pdf text version

AP304 Laser Laboratory Course, Lab 1

Optical Radiation Detectors & Noise

Carsten Langrock David S. Hum Mathieu CharBonneau-Lefort Eleni Diamanti January 22, 2002

Abstract In this lab, we investigated the responsivity, time response, and noise performance of various photodetectors, including thermal detectors, silicon photodiodes, and photomultiplier tubes (PMT). A standard helium-neon gas laser (HeNe), operated in the red (632.8 nm), was used as the light source for the measurements of the silicon photodiodes. A flashlight served as a radiation source to measure the time response of the thermal detector. The measurements clearly showed the different advantages of the various detectors and their limits of operation.

Author of this lab report

1. Thermal Detectors

AP304 Laser Laboratory Course, Lab 1

1

Thermal Detectors

To measure absolute optical power accurately, thermal detectors such as the Scientech detector used in this laboratory experiment have proven to be very useful. This detector is based on a thermopile, which creates a voltage proportional to the incident optical power heating a black body absorber. Since this detector is based on actually heating a piece of metal, the time response will is very slow. Furthermore, we cannot expect a high sensitivity, since a large number of photons will be required to cause a measureable voltage drop. The advantage of this type of detector clearly lies in its flat frequency response regarding the incoming radiation. This makes it useful as an allround power detector for various lasers and radiation sources. Another plus is the fact that it can handle high average power without the need for attenuators. Measuring several watts of optical power is therefore not a problem for this kind of detector. Other types of thermal detectors include bolometers, pneumatic detectors (e.g. Golay cells), and pyroelectric detectors. The advantages of thermal detectors mentioned above became the reason for our inability to measure the output of the HeNe laser with the available thermal detectors. Since the HeNe puts out only a few mW of optical power, the available Scientech power meter was not sensitive enough, since it's most sensitive setting is 300 mW full scale. Lacking the ability to use a high average power laser, such as the Nd:YAG laser mentioned in the laboratory notes, we could only measure it's time response using a small flashlight, which seemed to emit enough power to cause a small ­ but measurable ­ voltage drop. Several repetitions of the measurement using an analog oscilloscope and a digital multimeter (DMM) resulted in an estimation of the 10%-90%-rise time as being r 15 s.

2

Silicon Photodiodes

Photodiodes are quantum type detectors, meaning that they are theoretically able to detect single photons. They are based on the fact that a photon can create an electron-hole pair in a semiconductor pn-junction, which will cause a photocurrent to flow. Photodiodes are therefore current sources. There are a plethora of different types of photodiodes which can be categorized by detectable spectral range, response time, and damage threshold. In general, photodiodes cannot be used directly to detect high power radiation without using neutral density attenuators or beam splitters to

2

Author: Carsten Langrock

AP304 Laser Laboratory Course, Lab 1

3000 60

2500

50

Voltage across resistor (mV)

Transmitted Power (µW)

2000

40

1500

30

1000

20

500

10

0

0

Experimental Data (50 load) Linear Regression (slope=0.01945) 0 500 1000 1500 2000 2500 3000

20

40

60

80

100

120

140

160

180

200

220

240

Analyzer Angle (degrees)

12000 14 12

Incident Power (µW)

10000

Voltage across resistor (mV)

Voltage across resistor (V)

Experimental Data (10k load) Linear Regression (slope=3.9015) 0 500 1000 1500 2000 2500 3000

10 8 6 4 2 0 Experimental Data (100k load) 0 500 1000 1500 2000 2500 3000

8000

6000

4000

2000

0

Incident Power (µW)

Incident Power (µW)

Figure 1. Responsivity measurements taken with the Thorlabs DET110 photoreceiver terminated into different load impedances. The upper left graph shows the measured polarizer transmittance.

sample the beam. Another major difference to thermal detectors lies in the fact that semiconductor photodiodes have a fairly narrow spectral response due to the nature of their operation. Different types of semiconductors have to be used to cover different spectral ranges due to their intrinsic band-gap energy Eb which causes a cutoff for wavelengths above c = hc/Eb . Some photodiodes such as the avalanche photodiode (APD) come with internal gain, which makes them more sensitive to weak signals. We will not go into a detailed noise analysis of photodiodes and photoreceivers, but it should be noted that the noise properties of APDs strongly depend on the semiconductor material being used. For the use in long haul telecommunication links operating at 1.55 µm, Ge or InGaAs have to be used which turn out to be excessively noisy. Therefore, it is sometimes advantageous to use a PIN photodiode in combination with an external low-noise amplifier (LNA) instead.

Author: Carsten Langrock

3

2. Silicon Photodiodes

AP304 Laser Laboratory Course, Lab 1

In this laboratory experiment, we measured the responsivity and time response of two commercial photodetectors (Thorlabs DET110 and DET210). To get accurate results, we first made sure that the incident optical power was of a known value. This was achieved by using an adjustable attenuator in front of the HeNe laser and measuring its transmittance with respect to the attenuator setting using a commercial photodiode power meter (Newport 835). Since the output of the HeNe is well polarized, we could use a simple synthetic linear polarizing filter as an adjustable attenuator. The result of this measurement can be seen in the upper left graph of Fig. 1. The expected cos2 dependence of the relative polarizer-to-beam angle can be seen quite clearly. This measurement served as a calibration for our light source and made the following reRF sponsivity measurements straight forward. We started by measuring the Thorlabs DET100 Vout terminated into its canonical impedance of 50. PD Linearity was observed over the entire range of OP-AMP input powers (see Fig. 1). A simple linear reVbias gression calculation resulted in a measured responsivity of 50 = 19.45 V/W= 0.398 A/W, Figure 2. Typical photodiode tran- which turns out to be in good agreement with simpedance amplifier. the manufacturer's data. The same measurement was repeated using a 10k load resistance. Again, we couldn't see any deviation from linearity over the range of input powers and calculated a responsivity of 10k = 3901.5 V/W= 0.390 A/W, close enough to the previous value to be consistent and within measurement error. To observe the effects of limited dynamic range and saturation, we terminated the photoreceiver into a large load impedance of 100k.

CF

The lower right graph in Fig. 1 shows nicely that the detector circuit's dynamic range is limited by the large load impedance. A few hundred µW cause the detector to drop the entire reverse bias voltage of 12 V across the load impedance. It is obvious that one has to design a photoreceiver with respect to the expected input powers and the required sensitivity and bandwidth. Tradeoffs have to be made to satisfy all required criteria as well as possible. High sensitivity, high speed and a large dynamic range can be achieved using a slightly more complex photoreceiver design, incorporating for example a transimpedance amplifier as shown in Fig. 2. The same set of measurements was performed using a similar photodetector with

4

Author: Carsten Langrock

AP304 Laser Laboratory Course, Lab 1

3000 70 60

2500

Voltage across resistor (mV)

Transmitted Power (µW)

50 40 30 20 10 0 Experimental Data (50 load) Linear Regression (slope=0.02163) 0 500 1000 1500 2000 2500 3000

2000

1500

1000

500

0

20

40

60

80

100

120

140

160

180

200

220

240

Analyzer Angle (degrees)

12000 14 12

Incident Power (µW)

10000

Voltage across resistor (mV)

Voltage across resistor (V)

Experimental Data (10k load) Linear Regression (slope=4.1132) 0 500 1000 1500 2000 2500 3000

10 8 6 4 2 0 Experimental Data (100k load) 0 500 1000 1500 2000 2500 3000

8000

6000

4000

2000

0

Incident Power (µW)

Incident Power (µW)

Figure 3. Responsivity measurements taken with the Thorlabs DET210 photoreceiver terminated into different load impedances. The upper left graph shows the measured polarizer transmittance.

a smaller active area (see Fig. 3). The Thorlabs DET210 photodetector's responsivity was measured to be as follows;

50 = 21.63 V/W = 0.433 A/W 10k = 4113.2 V/W = 0.411 A/W Once again, these values are in excellent agreement with the manufacturer's data sheet and the fact that the responsivities of the DET110 and the DET210 should be very similar. Terminating the photodetector into 100k showed the limited dynamic range of this setup and agrees well with the equivalent measurement performed using the previous photodiode. Measuring the time response of photodiodes is certainly more challenging compared to thermal detectors, since these devices can exhibit bandwidths up to a few

Author: Carsten Langrock

5

2. Silicon Photodiodes

AP304 Laser Laboratory Course, Lab 1

GHz. There are two different instruments that can be used to measure the speed of photodiodes. One is a high frequency spectrum analyzer, the other one a fast sampling oscilloscope. In either case, we have to inject a delta-function-like optical impulse onto the active area of the photodiode. With the spectrum analyzer we can directly see the absolute value of the frequency response and it doesn't pose any problems to determine the electrical 3-dB bandwidth of the device under test (DUT). A fast sampling oscilloscope will provide even more information, since it will display the transfer function (i.e. impulse response) of the DUT which contains phase and frequency information. Now, we have to say a little bit more about the delta-function-like optical impulse that we mentioned earlier. One very elegant way to create these impulses is to use pulses created by a modelocked Ti:sapphire laser. These lasers are capable of producing pulses with a duration of a few femtoseconds (10-15 s). Even high-speed photodetectors can be analyzed with these pulses. Needless to say that we didn't have a Ti:sapphire laser to measure the time response of the Thorlabs photodetectors. We had to approximate impulses by chopping the output of the HeNe laser with a mechanical chopping wheel. Since this method is not suited to create very short pulses, we had to spoil the response time of the photodetectors by loading them with an ultra-high impedance. This will naturally slow down the detection speed by increasing the RC time constant. To be sure that we were measuring the photodiode's time response and not the beam profile of the HeNe laser, we increased the chopping frequency until we saw a reduction in signal height. At this point, the photodetector couldn't follow the signal impinging on its active area. The theoretical electrical 3-dB bandwidth of a simple photodetector circuit as described in the manufacturer's data sheet is given by f3dB = 1/(2Rload Cdiode ). The diode's capacitance Cdiode is quoted in the spec sheet of the photodiodes. We have to keep in mind that any stray capacitance and impedance will reduce the bandwidth below this theoretical value. As an example, let's compare the theoretical value for the rise-time, using Rload = 50, to the value mentioned in the spec sheet. Rload = 50 Cdiode = 20 pF for Thorlabs DET110 Tr 6.3 ns (calculated) Tr 20 ns (quoted)

6

Author: Carsten Langrock

AP304 Laser Laboratory Course, Lab 1

The calculated value is the absolute maximum of what can be expected from the photodetector and great care has to be taken to come close to that value. Rugged, commercial, and fairly inexpensive packaging of the photodetector will not allow the user to achieve the theoretical limit. Therefore, it is not surprising that our measured values came out high compared with the theoretically predicted values. The measured and calculated rise-times for the DET110 are Trmeas 0.9 ms into 1M load impedance Trcalc 0.126 ms into 1M load impedance Trmeas 90 µs into 100k load impedance Trcalc 12.57 µs into 100k load impedance One obvious reason for this discrepancy can be seen by taking into account the additional capacitance of the BNC cable that was used to connect the photodetector to the oscilloscope. The same measurement was performed using the DET210 photodetector. The results are Trmeas 0.8 ms into 1M load impedance Trcalc 11.31 µs into 1M load impedance This result is one order of magnitude worse than the one for the slower detector. Reasons for this can be an inadequately low chopping frequency, and carrier creation outside the photodiode's active area. Since we didn't focus the beam onto the active area, this might have had a detrimental effect on its impulse response. Loading the detector with a 100k resistor, we couldn't observe a reduction in signal height at the highest possible chopping frequency. It is therefore possible that the mechanical chopper couldn't provide sufficient fast chopping even in the 1M setup.

3

Photomultiplier (PMT)

Photomultipliers resemble APDs and vice versa in the sense that these types of quantum detectors have a variable built-in gain. The gain of the PMT can easily exceed 106 , which makes them very useful for the detection of weak signals. However, compared to photodiodes, PMTs are fairly slow, bulky, and very expensive. Since they

Author: Carsten Langrock

7

3. Photomultiplier (PMT)

16 14 12

AP304 Laser Laboratory Course, Lab 1

PMT Dark Current (nA)

10 8 6 4 2 0

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

Bias Voltage (kV)

Figure 4. Measurement of the PMT's dark current as a function of the applied bias voltage.

are based on the external photoelectric effect, appropriate cathode materials have to be found that allow measurements of various spectral ranges1 . Materials that allow detection at telecommunication wavelengths are not readily available and high bias voltages prohibit an uncomplicated integration into commercial systems. Nevertheless, PMTs have to be used in experiments, where only a few photons make up the signal. Their excessive noise due to the large bias voltage and dark current2 can be partially overcome by using lock-in amplifiers to reject the noise by decreasing the detection bandwidth. In this lab experiment, we were supposed to measure the dependence of the dark current on the bias voltage. Since the dark current is being generated due to events such as thermionic emissions and ohmic currents, we would expect the dark current to rise with increasing bias voltage, since the gain, and therefore any signal current, increases with increasing bias voltage. The data3 shown in Fig. 4 verifies this intuition. The increase in dark current seems to happen in at least two steps. The initial slope at fairly low bias voltages might be explained by the amplification of

Similar to the choice of semiconductor material for photodiodes. Here, we rely on the internal photoelectric effect. 2 More precisely, the shot noise associated with the dark current. 3 Provided by another group of students.

1

8

Author: Carsten Langrock

AP304 Laser Laboratory Course, Lab 1

the "background" dark current, which consists of electrons that have been created by thermionic emissions and high energy background radiation (e.g. cosmic rays). The sudden increase in the slope might indicate that another source of dark current has been excited. This could be due to additional electrons escaping the dynodes' surfaces. For further experimental data on the PMT, such as its responsivity characteristics, please refer to Sarah Katherine Braden's lab report.

4

Noise Measurements

The understanding of noise and it's origins is of utmost important if one wants to distinguish between real signals and random fluctuations. Since the understanding of noise is a rather advanced topic, I would like to mention only a few common noise sources and their properties. For further information about the measurement of laser amplitude and phase noise (i.e. timing jitter), please refer to this recently published article [1]. Reading this article will make clear why it is a challenging task to measure the noise performance of light sources and photoreceivers using a simple network analyzer. First, let us reflect on the notion of noise. What is noise anyway? Noise is everything that is not the signal we would like to measure. This might seem a trivial fact, but serves perfectly as a definition of noise in general. There are many different types of noise, some that are artificial and can possibly be eliminated (or at least attenuated), some are inherent in the detection process or nature itself. Artificial sources might be of acousto-mechanical or electrical nature. In how far these sources influence the measurement depends entirely on the detection scheme. Inherent noise sources might be created by the quantum nature of light, which will cause a statistically distributed generation of charge carriers in the pn-junction of a photodiode (shot noise). Thermally excited movements of electrons inside dissipative elements, such as resistors, will cause fluctuations of voltages and currents (Johnson noise). Unlike the previous mentioned noise sources, there's one type of noise that is inherent in each and every system, be it mechanical or electrical in nature. This noise is called 1/f -noise, since it is predominant in the lower frequency range and falls off like 1/f , where f denotes the frequency of the observable. To say it once more, this noise is not due to our inability to construct the perfect data acquisition system. Even if we had a noiseless receiver, the observable would show 1/f -noise. This can be best

Author: Carsten Langrock

9

4. Noise Measurements

AP304 Laser Laboratory Course, Lab 1

thought of as a law of nature, just like Newton's law. Nobody really understands the origin of this noise, but it is very important to keep the existence of it in mind, if one wants to measure observables at low frequencies (see the LIGO project, for example). After this short introduction to noise, let's look at some experimental results. We wanted to characterize the noise properties of several radiation detectors using a low-frequency spectrum analyzer. We were especially interested in the dependence of the noise floor on the incident power level and the observation of 1/f -noise. As explained in [1], a network analyzer is certainly far from optimum to measure noise in optical systems. A high-frequency spectrum analyzer with built-in noise measurement capability (e.g. HP 8591E) would be more appropriate for this kind of measurement if one doesn't want to spend too much money on a dedicated noise measurement test set (e.g. HP 3047A). Since the simplest model for various noise sources assumes the noise to be "white" (i.e. flat over a very large frequency range), measuring as far away from dc as possible should result in an estimate of the shot noise floor. To characterize a commercial photoreceiver (New Focus Model 2000), we connected the detector to the network analyzer and varied the incident optical power. As a first step, we wanted to qualitatively observe the photoreceiver's system noise by eliminating any kind of carrier generation due to photons. Unfortunately, the network analyzer's noise floor turned out to be higher than the photoreceiver's noise floor. At least that's what we thought in the beginning, since we couldn't see any change in the displayed trace with the detector being turned on or off. To make sure that the detector was in working condition, we connected its output to a DMM to measure the voltage drop across various load impedances. It turned out that we couldn't achieve more than a 1.5 V voltage drop which made us wonder whether the bias batteries were still good. We exchanged the batteries with known good ones (the old ones were indeed a little bit low), but still couldn't manage to measure a larger voltage drop. Either we didn't understand how to operate the device, or the device was not working properly, since we expected a larger maximum signal voltage with two internal 9 V batteries. Nevertheless, we tried to qualitatively measure the noise floor and the 1/f -noise. Since the network analyzer only operates out to 100 kHz, we measured the power level at this point. After increasing the optical power beyond a certain limit, the photoreceiver's noise floor seemed to exceed the network analyzer's noise floor, but the accuracy of the measurement wasn't high enough to identify the noise with the expected shot-noise floor. Even though the FFT analyzer didn't have a very high

10

Author: Carsten Langrock

AP304 Laser Laboratory Course, Lab 1

-70 -80 -90

Amplitude Noise (dBc/Hz)

-100 -110 -120 -130 -140 -150 -160 -170 -180 1 10 100 1k

1

2

3 4 5

10k

100k

1M

10M

Frequency (Hz)

Figure 5. Amplitude noise of argon-ion and DPSS lasers. i0 = 2.5 mA 1: Argon-ion laser noise. 2: DPSS laser noise. 3: Calculated shot-noise limit. 4: Photoreceiver noise floor. 5: System noise floor.

resolution at the low frequency end, we could certainly see the 1/f dependence of the noise. More careful measurements have to be made to accurately characterize the photoreceiver's noise properties. Unfortunately, the amount of time alloted for our group was not sufficient to study this part of the experiment in depth. To give a flavor for what results from more sophisticated amplitude noise measurements look like, I'd like to present some of the data that I've taken in the past. Fig. 5 illustrates nicely how a noise plot should be organized and what information it should contain. The purpose of this measurement was to compare several Ti:sapphire pump lasers according to their amplitude noise performance and to later correlate the Ti:sapphire's amplitude and phase noise with those results. To display the measurement system's sensitivity, the lowest graph in Fig. 5 is a

Author: Carsten Langrock

11

5. Conclusion

AP304 Laser Laboratory Course, Lab 1

plot of the system noise floor, which has been measured by terminating the system's signal port into 50. The next graph shows the photoreceiver's noise floor, which was measured with the photodiode being blocked completely. We can see that the photoreceiver's noise floor lies above the system's noise floor, just what you would want for a proper measurement setup. The other two graphs show actual amplitude noise data of two cw lasers, namely a DPSS laser (Coherent Verdi-V5) and a argon-ion laser (Coherent I-310) both running at 5 W output power. It is obvious that the modern DPSS laser performed much better almost over the entire frequency range covering 40 MHz. We can also see that the measurement didn't approach the theoretical shot-noise floor before approximately 2 MHz. It is therefore doubtful that we can measure the shot-noise floor at 100 kHz with the available equipment in the teaching lab.

5

Conclusion

In this lab experiment we had the chance to investigate the properties of various radiation detectors. We tried to explore the paramter space for each of them and to deduce their range of operation. It became apparent that each detector is inherently limted to produce valid data only in a particular setup beyond which it will not generate meaningful results. Therefore, the experimentalist has to choose the detector carefully according to the experiment's requirements. Special attention has to be paid to the range of linear operation, since nonlinear operation might introduce measurement errors which are hard to discover.

6

Acknowledgments

I would like to thank Sarah Katherine Braden and her group for providing the data for the PMT part of this experiment and Thomas Plettner for helping us with the experiment.

References

[1] Ryan P. Scott, Carsten Langrock, Brian H. Kolner. High-Dynamic-Range Laser Amplitude and Phase Noise Measurement Techniques. IEEE Journal on Selected Topics in Quantum Electronics, Vol. 7, No. 4, July/August 2001

12

Author: Carsten Langrock

Information

12 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate

1136778


You might also be interested in

BETA
K3 A 01_02 040501.pub
Columbia Engineering Bulletin 2011-2012
Microsoft Word - SURF Promo Book for 2012.docx
Microsoft Word - ELSC GU MTech syllabus 27-2-09