REVIEW OF SCIENTIFIC INSTRUMENTS

VOLUME 69, NUMBER 11

NOVEMBER 1998

Photodetector designs for low-noise, broadband, and high-power applications

Malcolm B. Gray,a) Daniel A. Shaddock, Charles C. Harb, and Hans-A. Bachor
Department of Physics, Faculty of Science, The Australian National University, Canberra, Australia 0200

Received 6 May 1998; accepted for publication 15 July 1998 We present design and performance details of three photodetector circuits that have been developed in the authors laboratory over the past eight years. These detectors have been optimized to meet the unique demands of experiments such as high power, high sensitivity interferometry, nonlinear optics, and laser noise measurements. The circuits are: a low-noise dc coupled dc 20 MHz general purpose detector, a low-noise broadband 151100 MHz detector capable of detecting 10 mW of light, and a high-power large dynamic range detector 30 kHz60 MHz capable of detecting up to 100 mW of light. We present bandwidth dynamic range and noise performance details for all three designs. In addition, we present detailed circuit schematics along with design and construction guidelines to enable assembly and use of these designs. 1998 American Institute of Physics. S0034-67489803910-0

I. INTRODUCTION

that can be readily performed to tailor the response to a particular experimental requirement.

Due to the unusual and demanding nature of many modern optical measurement techniques, it is often impossible to obtain commercial photodetectors that can optimally perform the required tasks. These tasks include: high power, high sensitivity, rf modulation interferometry, FM spectroscopy, cw frequency doubling, laser injection locking, frequency and intensity stabilization of solid state lasers, laser diode intensity squeezing, OPO squeezing, and electro-optic measurement and control. For the remainder of this article, we will refer to these tasks as quantum optics experiments. While there are numerous publications describing photodiode construction1,2 and performance,3,4 there is a genuine lack of published literature facilitating the implementation of these photodiodes into complete photodetector circuits for use in quantum optics experiments. Typically, photodetector circuit designs for these applications are developed in house and remain unpublished.57 Here we present detailed detector designs and performance data that we believe represents a signicant improvement on existing detectors. Over the past eight years, we have developed and rened three basic photodetector designs that have performed this broad range of experiments reliably and well. In this article we rst discuss pertinent characteristics of photodetectors and the requirements of typical quantum optics experiments. We then present our three basic detector designs. We present typical applications where these designs have proven useful. We also include a detailed discussion on the predicted performance and implicit design trade-offs. We then present schematic and construction details to a level such that an electronics technician, previously unassociated with optoelectronics, should be able to rapidly construct the circuits presented. The test details of the circuits are then presented with a nal section outlining possible variations
a

A. Quantum optics experimental requirements

Electronic mail: Malcolm.Gray@anu.edu.au 3755

While there are numerous commercial photodetectors readily available, the applicability of these devices is severely limited when considering quantum optics experiments. Typical quantum optics experiments require not only a high sensitivity detector but also one capable of detecting sufcient optical intensity so that shot noise is the dominant spectral feature at the detector output. This places stringent and unique dynamic range limits on the detector design. Many ber optics photodetectors and dedicated ampliers meet the high sensitivity requirement,8 however, very few are capable of detecting sufcient intensity to produce measurable shot noise at the detector output. For example, the HP ITA-12318 transimpedance amplier has excellent sensitivity of 2.8 pA/Hz (3.6 pW/Hz at 1064 nm), however, the amplier saturates at photocurrents of 300 A making it impossible for shot noise to dominate the photodetector output before the onset of saturation problems. Currently available photodetectors frequently cannot meet the demands of high-power, high efciency detection. For example, experiments with squeezed diode lasers require a high efciency photodetector capable of measuring 100 200 mW of light with a large signal bandwidth 1050 MHz. In addition, many quantum optics experiments need to accommodate large rf modulation components in the photocurrent. These modulation components typically arise from the need to control and measure complex nonlinear interferometric systems. It is therefore often required to make sensitive shot noise measurements in the presence of large deterministic and random spectral features. No commercial detectors are currently capable of performing these measurements.
1998 American Institute of Physics

0034-6748/98/69(11)/3755/8/$15.00

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

3756

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

B. Bandwidth

The basic bandwidth limiting mechanisms applicable to pin photodetectors are described in Ref. 9. For the photodetectors described below, the bandwidth was measured using an 830 nm Mitsubishi laser ML3401 with a home made intensity modulation circuit that provides a modulation bandwidth of 100 kHz1100 MHz. This provides a simple and convenient means of testing both InGaAs and Si detectors.
C. Noise performance

In order to quantify the noise performance of photodetectors, we use the following denition of noise equivalent power NEP.9 NEP is the fully amplitude modulated optical intensity required to generate a signal size equal to the noise spectral density recorded at the photodiode output when not illuminated. NEP therefore includes the effect of photodetector conversion efciency as well as the electronic noise and optoelectronic gain of the detector. In practice, we measure NEP by recording both the dark noise spectrum and the spectrum when illuminated with white light. This infers that the light source used is free from technical noise and exhibits only shot noise at the frequency range of interest. By taking the ratio of these two spectra, we can then determine NEP: NEP

FIG. 1. Circuit schematic for the low-noise general purpose detector.

P e I dc2 e W/Hz , P l P e

reection locking of FabryPerot interferometers, dc monitoring and intensity noise measurements of lasers, complex interferometer control, measurement of transfer functions of lasers, and injection locked lasers. A detailed schematic of the dc coupled transimpedance photodetector circuit is given in Fig. 1. The design relies on a standard transimpedance operational amplier10 op-amp front end stage using a Comlinear CLC420 op-amp. This front end stage functions by providing a virtual earth to sink all photocurrent. As the input impedance of the inverting input is very large 1 M, all photocurrent is forced through the feedback resistor R 1 with the consequent voltage drop dening the output voltage. Hence, the rst stage has a transfer function given by V out I pd R 1 . 2

where is the detector responsivity in A/W, P e and P l are the noise spectral densities at the detector output with no illumination and white light illumination, respectively, I dc is the illuminated dc photocurrent, and e is the electronic charge. Note that this measurement of NEP relies on either a measurement or estimation of the photodiode responsivity .
D. Dynamic range

The dynamic range of a photodetector is limited by a number of mechanisms the saturation or bleaching of the active region of the pin junction due to high intensity incident light, the maximum photocurrent the diode can endure and, the dynamic range of the amplier used in the detector. The result of all these dynamic range limiting mechanisms is that there will be an acceptable limit placed on all the following quantities: the maximum photocurrent and total optical power detected, the maximum optical intensity at the photodiode surface, the minimum dark noise spectral density at the photodiode output, and the maximum modulation components that can be accommodated on the optical input. As these four limits vary greatly among available photodiodes and photodetector circuits, it is necessary to carefully match experimental demands with available photodetector performance.
II. LOW-NOISE DC COUPLED GENERAL PURPOSE DETECTOR

This equation is valid for large op-amp loop gain frequencies below 20 MHz in practice and for output voltages less than 3.5 V CLC420 is a 5 V device with max output excursion of around 4 V. The second stage of the photodiode circuit provides a gain of 3.5 using a CLC430 in a standard op-amp circuit.10 As the CLC430 device is capable of around 12 V output swing, the gain matches the dynamic range of the second stage to that of the front end amplier. In addition, the second stage also provides a dual buffered output capable of driving two 50 loads.
A. Circuit limitations

The noise performance of this transimpedance conguration can be easily estimated based on published data.11 The total rst stage output noise is given by VN
out R 1

V n / R 1 2 I 2 n 4 KT / R 1 8

nV/Hz. 3

This photodetector nds a wide variety of uses in quantum optics experiments. Some typical applications include: rf

Here the square root term sums the equivalent input noise current terms in quadrature. The rst noise term V n / R 1 is the op-amp input noise voltage, the second term is the op-amp input noise current, while the third term is the thermal noise current produced by the feedback amplier. As can be seen, both current and voltage noise terms contribute equally in this circuit and thermal noise provides half the total noise. The second stage can be likewise analyzed and found to produce an equivalent input voltage noise of 5 nV/Hz. When added in quadrature with the front end noise and mul-

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

3757

FIG. 2. Frequency response of the low-noise general purpose detector. Vertical axis scale is arbitrary as absolute level depends upon incident optical power and modulation depth.

FIG. 3. NEP at 1064 nm for the low-noise general purpose detector.

C. Construction details

tiplied by the stage two gain 3.5, the nal output noise can be estimated at 30 nV/Hz. The detector NEP can then be estimated by NEP 30 nV 6 pW/Hz. 3.5R 1 4

The maximum output voltage swing obtainable from the second stage is approximately 12 V, this gives a maximum dynamic range of around 0.4 109 . At both low and high frequencies, this dynamic range will be reduced by: 1/ f noise below 5 kHz, op-amp loop dynamics above 20 MHz and large signal bandwidth of both stages in practice, this is limited by the harmonic distortion that is acceptable to the user. The maximum photocurrent for this circuit can easily be calculated as it is limited by the dc voltage swing of the front end amplier and also the stage two amplier due to the dynamic range matched gain at I maxVmax /R11.75 mA. At a wavelength of 1064 nm ( 0.77 A/W), this corresponds to a maximum input optical power of 2.2 mW.

In order to make high speed op-amp circuits reliable, stable, low noise and fast, it is necessary to carefully design the physical layout of the circuit board.12 Use a double sided printed circuit board with one side being the ground plane. We use surface mount components with low inductance earth connections made by simply putting a feedthrough to the ground plane. As several points of this circuit cannot be impedance matched, it is necessary to keep all nonmatched tracks as short as possible. This reduces parasitic capacitance and inductance to a minimum and ensures optimum frequency response. For instance, the output impedance of the photodiode is very large and cannot be easily impedance matched to the input op-amp impedance approximately zero due to the virtual earth. We therefore place the anode pad directly next to the op-amp input pin 2, IC1 and mount the diode on the ground plane. This limits lead length to 1.6 mm the thickness of the circuit board between anode and op-amp input. For high frequency operation, it is necessary to provide a low impedance path to ground for the diode cathode biased at 15 V. This is achieved by using a ceramic surface mount chip capacitor ( C 3 100 nF), and feed through to the ground plane, placed immediately adjacent to the cathode pad.
D. Variations

B. Circuit performance

Figure 2 plots the measured bandwidth of this circuit. As can be seen, the 3 dB bandwidth is about 20 MHz. In addition, the response at higher frequencies is seen to be well behaved, rolling off as 1/ f and free from any resonances. The noise performance for the CLC420-CLC430 circuit is given in Fig. 3. Between the high and low frequency limits, this design achieves a NEP7.5 pW/Hz. As the maximum detectable power is 2.2 mW, the maximum ratio, of shot noise given by 2 eI max to all other noise sources can be estimated by

2 eI max
NEP

12.5 dB.

While the circuit diagram shown in Fig. 1 uses an InGaAs photodiode Epitaxx ETX-500, spectral range 950 1600 nm, the diode can readily be replaced by a silicon diode for instance, a Hitachi HR8101 was also used to give operation over the spectral range 4001000 nm. Using the Hitachi HR8101 produces a frequency response identical to that of Fig. 2 however, the NEP is slightly worse (11.5 pW/ Hz at 830 nm). This is a fundamental effect due to the use of shorter wavelength radiation and not a diode design fault. Replacing the CLC420 op-amp with the faster CLC425 will extend the frequency response of the rst stage to around 100 MHz. However, the second stage performance

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

3758

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

coupled to a BNC connector output port. This ac coupling ensures that dc saturation effects, due to large photocurrents, are avoided. Power is supplied to the MMIC via the signal output terminal. R 3 and L 3 form a simple constant current source to drive the MMIC. In order to achieve maximum bandwidth and optimum noise performance the nominal supply current must be set to 30 mA for the INA-02170 MMIC, this implies R 3 240 for a 15 V supply.13

A. Circuit limitations

FIG. 4. Circuit schematic for the broadband low-noise detector.

will limit the bandwidth to around 60 MHz. Also, the noise performance of this circuit will degrade rapidly at frequencies above 20 MHz. For these reasons, we recommend use of our broadband detector see Sec. III at frequencies above 20 MHz rather than a high speed op-amp design.
III. BROADBAND LOW-NOISE DETECTOR

The noise performance of the broadband photodetector can be estimated based on the noise gure NF of the MMIC amplier 2 dB for INA-02170 MMIC. This value is derived for input and output impedances of 50 , however, we operate the MMIC with a severe impedance mismatch at the input terminal. This renders the following estimate quite inaccurate, however, it can be used as a relative comparison between different MMIC circuits. In order to estimate the total noise performance, we therefore determine the effective noise temperature of the rf input resistance.10 T R 290 10 NFdB/10 1 170 K. 6

The inferred electronic noise input current can then be calculated by I n

This photodetector is used in a large number of diverse quantum optics experiments where large bandwidth is required. Some typical applications include: broadband readout of optical squeezed state noise, measurement of the broadband response of complex interferometers, swept frequency measurement of FabryPerot interferometer properties and, high frequency reection locking of high nesse Fabry Perot interferometers for frequency stabilisation of lasers and FM spectroscopy. A broadband photodetector has been developed in order to measure quantum optical effects from around 15 to 1100 MHz. The detailed schematic of this circuit is shown in Fig. 4. The photocurrent is divided into two outputs via inductor L 4 . The dc and audio frequency photocurrent travels through the inductor to generate a small voltage across R 1 . In order to prevent signicant voltage drop across R 1 and a reduction in the reverse bias across the diode under large photocurrent operation, R 1 must be kept small 100 for I max 10 mA. The voltage generated across R 1 is then amplied by a positive gain op-amp single supply rail LM358 with the gain set by pot VR 1 . This enables the dc voltage output to be calibrated to optical power at the required optical wavelength. By using a LM358 op-amp operating with a single positive supply rail, we are able to power both the dc and rf circuits with a single 15 V supply. Alternatively, the rf photocurrent travels through the coupling capacitor C 6 and into the monolithic microwave integrated circuit MMIC,13 IC1. The rf input impedance of the MMIC converts the photocurrent to voltage which is then amplied by 30 dB within the MMIC. The MMIC therefore acts as an ac coupled transimpedance amplier with a nominal input impedance of 50 . The MMIC output is then ac

4 KT 13.7 pA/ Hz. R in

and the resulting NEP by NEP I n / 18 pW/ Hz. 8

The rf bandwidth of this circuit is limited at low frequencies by the inductor L 4 . This must be selected small enough in order to avoid inductor self-resonance effects and capacitive behavior at high frequencies. Once this inductor value is chosen, the low frequency behavior is partially determined. Selecting C 4 sets the resonant frequency of the series LC resonance ( L 4 C 4 ) and this produces a zero in the frequency response at low frequencies at 4 MHz in this case and denes the low frequency corner 15 MHz. As the zero pulls the frequency response down rapidly at frequencies approaching the resonance, the low frequency response rolls off very rapidly below the corner. The upper frequency limit is determined by the combined effects of the RC time constant and the drift transit time of the photodiode. For the diode used here either the Epitaxx ETX-300 or EG&G C30971E, the bandwidth into 50 is quoted as 700 MHz. As the rf input impedance of the INA-02170 MMIC is in fact close to 25 ,13 the RC time constant can be halved and the upper corner frequency can be extended to approximately 1100 MHz. The dynamic range limitations of this design are set by: the maximum photocurrent the photodiode can tolerate 10 mA and the tolerable harmonic distortion produced in the MMIC due to large modulation components. As the particular MMIC used here INA-02170 was selected for bandwidth and noise performance, the rf dynamic range is not

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

3759

causing the NEP to also rise. The minimum NEP of 15 pW/ Hz at 830 nm 9.5 pW/Hz at 1064 nm using ETX300 occurs at around 200 MHz. RF dynamic range performance is usually limited by the ability to maintain a reliable noise oor. As pointed out in Sec. I D, one effect of harmonic distortion within the MMIC is the production of a broadband noise oor several decibels above the predicted shot noise level. This can be a subtle effect and should be tested by referencing to a white light calibration if possible.
C. Construction details

FIG. 5. Frequency response of the broadband low-noise detector. Vertical position arbitrary.

particularly large13 third-order intercept. In Sec. III D, we present one technique for improving this dynamic range.

B. Circuit performance

Figure 5 plots the bandwidth of the photodetector from 1 to 1500 MHz. As can be seen, the photodetector rf output has a useful bandwidth from 15 to 1100 MHz. Within this range, the response is spectrally at to within 3 dB. At frequencies above 600 MHz, the laser diode drive circuit used to obtain this frequency response, contributes some 1 dB of the observed response deviations. The noise performance obtained by this photodetector EG&G C30971E, silicon diode is plotted in Fig. 6. As can be seen, optimum NEP is achieved across a broad range of frequencies from about 100 to 600 MHz. At frequencies below 100 MHz, the signal rolls off and the rising noise gure of the MMIC causes the NEP to rise. At frequencies above 600 MHz, the noise gure of the MMIC rises monotonically

In order to operate MMIC devices at frequencies around 1 GHz, a number of stringent physical layout rules should be followed14,15 in addition to those mentioned in Sec. II C. The rf signal tracks are designed to be a 50 microstrip transmission line see Fig. 4 and the BNC or SMA connector is soldered directly onto the output microstrip track. The MMIC is soldered directly onto both the input and output microstrip tracks. For stable operation of the MMIC, it is imperative that the two ground terminals be attached to the ground plane with an extremely low inductance connection. The manufacturers recommendation14 is to use a thin circuit board with three or more plated feedthroughs directly below the ground terminals. Due to ease of construction and availability of components, we use standard 1.6 mm ber glass circuit board. In order to reduce the ground inductance to an acceptable level, we use a small copper sheet approximately 5 mm wide that is soldered onto the ground plane, fed through a 5 mm slot, over the MMIC ground terminal and back through another 5 mm slot, to the ground plane. The MMIC ground terminal is then soldered onto this small copper sheet. While this technique is labor intensive and clearly not suitable for commercial production, it does allow successful MMIC circuits to be built with standard prototyping electronics facilities. It should be noted that if ground loop inductance is not sufciently small then the frequency response of the photodetector will start to peak at high frequencies around 800 MHz. If the inductance is still larger, then this peaking will develop into sustained oscillation of the MMIC. The photodiode must be placed very close to the MMIC input to minimise parasitic capacitance and inductance effects. We achieve this by placing the photodiode directly on the ground plane. Short leads 1.6 mm then connect the photodiode to both the MMIC input and the bias circuit.
D. Variations

FIG. 6. NEP at 830 nm for the broadband low-noise detector.

A number of different photodiodes have been successfully used with this broadband circuit. In addition to the ETX-300 and EG&G C30971E demonstrated here, we have used: ETX-500 with slower MMIC, Hitachi HR8102 Si and Hamamatsu S1721 Si. In order to optimize the noise performance and bandwidth of this circuit for these different photodiodes, the appropriate MMIC should also be selected. A brief summary of the MMIC and suitable photodiodes alternatives and their performance is given in Table I. Note that large dynamic range MMICs are available HP MSA-

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

3760

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

TABLE I. Comparison of different broadband photodetector circuit options and their performance. NEP Suitable (pW/ Hz) Bandwidth MHz photodiodes MMIC at 1064 nm 01170 8.5 3300 ETX-500 HR8102 S1721 ETX-300 C30971E ETX-100

Comments good stability G nd inductance not such an issue, lowest noise good NEP and bandwidth excellent frequency response stringent inductance requirements

02170 03170

9.5 11

151100 151500

1104, for example and would function well in this circuit. However, we have avoided these devices in order to optimize noise performance across the desired bandwidth. One approach that we have successfully used in order to increase the dynamic range of this detector circuit is to add a series LC circuit from the MMIC input to earth. By tuning the capacitance, this simple notch lter can be centered on the offending large spectral component. This approach provides up to 30 dB suppression in practice. The simple notch lter technique is particularly suited to reection locking FabryPerot interferometers. When locked to resonance, the modulation fundamental is minimized by the locking servo 75 MHz in our case while the second harmonic is maximized 150 MHz. It is typically this large second-harmonic component that causes harmonic distortion. As the second harmonic carries no useful information for reection locking, it can be readily suppressed by the notch lter. This approach has been successfully used at the authors laboratory to frequency stabilize a solid state laser to a FabryPerot interferometer. The frequency response for this particular detector is shown in Fig. 7. It should be noted that for this technique to be successful, the quality factor, Q of the notch lter must be sufciently large so that there is no signicant phase delay introduced at the fundamental frequency of interest. A similar approach in order to improve the dynamic range of this detector is to tune the existing LC series reso-

FIG. 8. Circuit schematic for the high-power large dynamic range detector.

nance ( L 4 C 4 ) to notch out low frequency spectral components. While this resonance is typically around 4 MHz, it can be modied by over a decade without introducing signicant problems at high frequencies. For example, we have tuned this zero and reduced R 1 to improve the Q to the relaxation frequency of our laser 500 kHz to avoid harmonic distortion and MMIC saturation at frequencies within our signal bandwidth above 15 MHz in this case.
IV. HIGH-POWER LARGE DYNAMIC RANGE DETECTOR

FIG. 7. Frequency response of the broadband low-noise detector with a 150 MHz notch lter at MMIC input. Vertical scale arbitrary.

This photodetector has found use in a number of demanding experiments including the measurement of highpower laser intensity noise, high-power shot noise limited interferometry, and quantum noise limited frequency stabilization of high-power solid state lasers. A detailed schematic is presented in Fig. 8. Photocurrent generated in a large area high-power photodiode EG&G 30641 or ETX-1000 is ac coupled into a high-speed opamp, IC1 CLC409 congured as a transimpedance amplier. Due to the virtual earth at the negative input, the low frequency corner occurs at 1/2 R 2 C 2 30 kHz. Above this frequency, photocurrent is coupled into the virtual earth pin 2, IC1 and through the feedback resistor to produce an output of V out R 1 I pd . The high frequency corner is primarily determined by the bandwidth of the CLC409 op-amp. The large input capacitance of high power photodiodes approximately 40 pF, produces an added pole in the open loop frequency response of standard voltage feedback op-amps16 at f p 1/(2 R 1 C pd). In order to stabilize the amplier under these conditions, it is necessary to add compensation, typically using a feedback capacitor in parallel with R 1 . This feedback capacitance must be of similar value to that of the photodiode for stable operation, leading to a closed-loop bandwidth of about f c 1/(2 R 1 C pd) 4 MHz. In order to expand this poor bandwidth, we make use of a current feedback op-amp; the CLC409. This op-amp topology increases

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

3761

FIG. 9. Frequency response of the high-power large dynamic range detector. Vertical scale arbitrary

FIG. 10. NEP at 1064 nm for the high-power large dynamic range detector.

the pole frequency to f p 1/(2 R 0 C pd), where R 0 is the input resistance of the current feedback op-amp,17 typically 50100 . This enables the bandwidth to be increased to around 60 MHz. In order to add compensation and ultimately to stop the op-amp oscillating with current feedback op-amps, the feedback resistor R 1 must be increased. While the optimum bandwidth of the CLC409 is obtained with R 1 250 , we need to increase this to about 1000 in order to compensate for the high frequency pole and obtain a at frequency response. The dc and audio frequency photocurrents travel through inductor L 1 and resistor R 2 . The voltage generated across R 2 is then amplied by a positive gain op-amp IC2 to provide a dc coupled output used for both aligning and dc power measurements. In order to operate with large photocurrents it is necessary to reduce R 2 to prevent signicant photodiode bias voltage drop. By restricting R 2 to 5.2 , we can accommodate up to 200 mA photocurrent with only 1 V drop in diode bias.

2 eI dc
NEP

20.5 dB.

Hence, while the circuit is quite noisy in absolute terms such as NEP, the relative noise performance for large optical intensities is excellent. In addition, this photodetector circuit produces very little harmonic distortion. For example, when detecting 100 mW of light, the photodetector can produce an rf output of 10 dBm with second- and thirdharmonic distortion less than 65 dBc . Hence, it is possible to measure both shot noise 20.5 dB above electronic and a large modulation component (10 dBm ) with a signal to noise ratio of 75 dB and less than 10 dB second- and thirdharmonic distortion measurable above shot noise assuming an analyzer RBW of 10 MHz. In practice, the dynamic range of the vast majority of quantum optics measurements undertaken with this photodetector are limited by the spectrum analyzer used and not the photodetector.
B. Circuit performance

A. Limitations

Unfortunately, the CLC409, and most current feedback ampliers, exhibits substantially more noise than standard voltage feedback devices. The output noise voltage for this circuit can be estimated in a similar way to that of the CLC420-CLC430 circuit above. When this calculation is performed, the resulting output noise is estimated at V N out 15 nV/Hz approximately two times the noise spectral density of the CLC420 circuit. The corresponding NEP estimate is also much larger; NEP20 pW/Hz at 1064 nm. The optical power handling ability is primarily limited by the electrical power that can be safely dissipated in the photodiode. For the two diodes used here, this limit corresponds to an input power of approximately 100 mW at 1064 nm giving I pd 80 mA with a 5 V bias applied to the photodiode. The maximum ratio, of shot noise to electronic noise can therefore be estimated at

Figure 9 presents the measured frequency response of the high-power photodetector obtained by using a tracking generator, our home made laser diode circuit and a spectrum analyzer. The 3 dB bandwidth is seen to extend from 30 kHz to 60 MHz. The NEP across this bandwidth is presented in Fig. 10. The NEP is seen to be approximately frequency independent across the range of 230 MHz. It should be noted that the NEP measured here is for the photodetector circuit using an ETX-1000 diode responsivity at 1064 nm is 0.77 A/W. If the EG&G diode is used, then the relatively poor responsivity of this diode at 1064 nm ( 0.6 A/W) causes the NEP to be correspondingly larger; NEP28 pW/ Hz at frequencies below 30 MHz. In other respects, the diodes are readily interchangeable. An indication of the dynamic range of this photodetector can be obtained from Fig. 11. Here we plot the intensity noise for a single frequency 700 mW Nd:YAG laser. Approximately 60 mW of light at 1064 nm is incident on the photodetector. The large spectral feature at 500 kHz is due to the relaxation

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

3762

Rev. Sci. Instrum., Vol. 69, No. 11, November 1998

Gray et al.

have not yet used this diode in the high-power circuit, it should enable the high-power performance obtained at 1064 nm to be replicated across the visible and near-IR spectral region 450950 nm. This would enable the efcient detection of bright squeezing in near-IR laser diodes typically 50100 mW.
ACKNOWLEDGMENTS

The authors thank Ping Koy Lam, Russell Koehne, Alex Rickter, Alex Eades, and David Cooper for helpful suggestions, ideas, and support in the development of the work herein reported on. This work was nancially supported by the Australian Research Council.
K. Vaccaro, S. M. Spaziani, E. A. Martin, and J. P. Lorenzo, IEEE J. Quantum Electron. 34, 991 1998. 2 Y. Shi, J. H. Zhao, J. Sarathy, H. Lee, and G. H. Olsen, IEEE Trans. Electron Devices 44, 2167 1997. 3 C. K. Sun, I. H. Tan, and J. E. Bowers, IEEE Photonics Technol. Lett. 10, 135 1998. 4 L. Y. Lin, M. C. Wu, T. Itoh, T. A. Vang, R. E. Muller, D. L. Sivco, and A. Y. Cho, IEEE Trans. Microwave Theory Tech. 45, 1320 1997. 5 N. Uehara, Internal GALILEO report SN: 2, Ginzton Laboratory, Stanford University, Stanford, CA, 1996 unpublished. 6 A. Dominjon and M. Yvert, Internal VIRGO report PJT94-014, AnnecyLAPP, France, 1994 unpublished. 7 R. Abbott, J. Mason and R. Savage, LIGO internal report T970052-00-D, CALTECH, Pasadena, CA, 1997 unpublished. 8 P. Fay, W. Wohlmuth, A. Mahajan, C. Caneau, S. Chandrasekhar, and I. Adesida, IEEE Photonics Technol. Lett. 10, 582 1998. 9 New Focus Application Note 1, Insight into High-Speed Detectors and High Frequency Techniques, Mountain View, CA, 1991. 10 P. Horowitz and W. Hill, The Art of Electronics, 2nd ed. Cambridge University Press, Cambridge, 1989. 11 Comlinear High-Speed Analogue and Mixed-Signal Data Book, National Semiconductor Corporation, Santa Clara, CA, 1997. 12 Comlinear Application Note OA-15, Frequent Faux Pas in Applying Wideband Current Feedback Ampliers, National Semiconductor Corporation, Santa Clara, CA, 1990. 13 Hewlett Packard Communication Components, GaAs and Silicon Products Designers Catalog, Hewlett Packard, Mountain View, CA, 1991. 14 Hewlett Packard Application Note S012, INA Series RFIC Ampliers, Hewlett Packard, Mountain View, CA, 1997. 15 Mini-Circuits Application Note, A handy how-to-use guide for MAR monolithic drop-in ampliers, Brooklyn, NY, 1992. 16 Apex Microtechnology Corporation Application Note 19, Stability for Power Operational Ampliers, Tuscon, AZ, 1993. 17 Comlinear Application Note AN, Current Feedback Ampliers, National Semiconductor Corporation, Santa Clara, CA, 1990.
1

FIG. 11. Laser intensity noise top trace, shot noise dashed trace, and electronic, or dark noise bottom trace for the high-power large dynamic range detector. Incident power80 mW at 1064 nm.

oscillation of the laser. Clearly, this detector is capable of recording this feature with no measurable harmonic distortion and still produce 15 dB shot noise/electronic noise.
C. Construction details

In addition to the circuit board layout guidelines mentioned in Sec. II C, the photodiode must be mounted in a high thermal conductivity heat sink. We mount the photodiode directly onto the circuit board ground plane and x a snug copper heat sink over the top of the diode. It is imperative that this heat sink be in intimate contact with the base of the photodiode for efcient heat removal.
D. Variations

One potential addition to this circuit is to mount the photodiode heat sink on a thermoelectric cooler in order to remove heat more efciently. The authors have not tried this technique as yet, however, this should enable greater power dissipated within the photodiode and may enable the photodetector to operate at higher optical powers. Another potential variation using this photodetector circuit is to use a Si photodiode Hamamatsu S1721. While we

Downloaded 02 May 2006 to 133.40.221.73. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp