sign in sign up
<<
Home , Laser diode

A GREATER MEASURE OF CONFIDENCE built-in potential or equilibrium potential) across an area called the (or layer). Reverse biasing the causes this potential to increase, and the depletion region to expand. Photons of sufficient en- ergy impinging on this region will create -hole pairs, which will separate due to the combined equilibrium potential and externally applied reverse bias. These charg- es will quickly drift away from the junction and be collected by the electrodes. Drift cur- rent generated by photon absorption in the —Choose depletion region is the main component of photocurrent. It also is the component with the fastest response time. and Use Wisely for Best Another component of photocurrent is called diffusion current. This component Results in Pulsed originates from charge carriers created by absorption of photons outside the depletion layer. While the majority of these charge Test Systems carriers will recombine in the neutral region, some of them may manage to diffuse slowly to the junction and make a contribution to the photocurrent. This contribution is not Helga Alexander, Keithley Instruments, Inc. welcome when a fast photodiode response is desired. Designing for Fast Response. For fast response time, a photodiode has to be de- Testing laser for fiberoptic communi- of these photodetectors involves a number of signed such that most of the photon absorp- cation systems requires a with compromises, but a fast response time often tion takes place in the depletion layer. Since a fast response. However, a 10–90% rise is the dominant concern. Nevertheless, the photons must penetrate the p-region to reach time is not the only important characteristic relationships between a photodetector’s re- the depletion layer, one way to achieve fast of a photodiode, particularly when it is used sponse time and other characteristics, such response is to make the p-region extremely for absolute optical power measurements in as device structure, external circuitry, and thin and to expand the depletion layer by pulsed mode. bias voltage, must be considered. applying as large a reverse bias as possible. Unfortunately, this method also increases Photodetectors for Photodetector Construction and the dark current and its associated noise con- Production Testing Performance tribution. Dark current is the small amount Given that 80% or more of a laser diode are the most commonly of charge flow that results when electron- module’s production costs are added during used detectors due to their low cost and packaging, it is imperative to start with only generally good performance. Like all semi- known-good laser diode chips for commu- conductor detectors, they make use of their nication modules. This requires laser diode internal photoelectric effect. When a photon testing at the bar or chip stage, using pulse is absorbed, it causes an electron to be raised techniques that prevent destructive self-heat- from the valence band to the conduction ing of uncooled devices. A pulsed current band, creating a charge-carrier pair (a hole source with a fast rise time is needed to drive in the valence band, electron in the conduc- the laser diode, and a photodetector with a tion band). When the device is connected to fast response time is needed to measure laser an external circuit, the resulting charge flow output accurately. is referred to as a photocurrent. Generally, PIN detectors are used, which The most basic photodiode has a simple can be based on , germanium, or III-V p-n junction structure. At the junction, diffu- compound technology, depending on appli- sion of into the p-type material and cation requirements. For pulsed light-cur- holes into the n-type material cause an op- Figure. 1. Representation of a PIN photodiode rent-voltage (LIV) testing, selection and use posing electric potential (sometimes called structure.

Photodetectors—Choose and Use Wisely for Best Results in Pulsed Laser Diode Test Systems April 2004 1 hole pairs are generated with the detector not exposed to light. Generation of dark cur- rent charge carriers is more likely to occur with the introduction of impurities in doped materials, such as those in a p-n junction photodiode. The dark current contribution increases as the depletion layer widens due to increases in reverse bias. Another way to control the width of the depletion region is to construct a p-i-n pho- todiode, where an intrinsic semiconductor layer is created between the p and n regions of the device (Figure 1). The undoped (or lightly doped) intrinsic layer of fixed width serves the same function as the depleted re- gion of a p-n junction. Since it lacks impurities that generate current carriers in the dark, photodiode sensitivity is improved [1]. More importantly, photon absorption is Figure. 2. Typical responsivity curve for Si, Ge, InGaAs concentrated in this region, thereby keeping undesirable diffusion current to a minimum. fast rise times to detect optical signals mod- possible in the production process. LIV tests The bias voltage, and thus the electric field, ulated at several gigahertz. Despite their also determine some of the optical and elec- is concentrated and virtually constant across speed, detectors of this type may lack the trical characteristics of the final packaged the absorption region. This greatly decreases characteristics needed for other applications, product. the average drift time of the optically gener- such as pulsed LIV testing. For LIV appli- An LIV curve is generated by applying ated carriers [2]. Hence, p-i-n photodiodes cations, the detector usually is expected to a range of drive currents to the laser diode are usually the detectors of choice when a make an absolute power measurement at var- and plotting voltage drop and light output fast response is required, as is the case in ious laser drive currents. Therefore, all the as a function of current (Figure 3). The test pulse testing of diode . light emitted by the laser must be captured, instrument is configured for a drive current Spectral Sensitivity. The bandgap of the either directly by the detector or by an in- sweep with small step increments and a si- detector material is the main factor that de- tegrating sphere that incorporates the detec- multaneous measurement of the laser diode’s termines the optical response of tor in the inner sphere wall. Since the beam forward voltage drop for each current step. a photodiode. Germanium (Ge) and indium emitted by a laser diode is highly divergent, The optical power output from the laser’s arsenide (InGaAs) photodiodes can intercepting all the light for absolute power rear and back facets are also measured at detect radiation ranging from approximately measurements requires a large area detector, each step, which requires calibrated photo- 800nm to 1700nm, making them well suited with a diameter of several millimeters. This detectors [3]. for testing telecom transmitter lasers. Sili- makes a fast fiberoptic receiver with a typi- An important characteristic obtained con (Si) is the detector material of choice cal diameter of 50µm unsuitable for use in from an LIV curve is the laser’s threshold for shorter , being sensitive from LIV testing. current. This is the injection current value 300nm to 1100nm. Although a particular There is another important difference in at which light output increases dramatically, photodiode can detect radiation over a range detector requirements for telecom and LIV signaling the onset of of wavelengths, its photocurrent output var- applications. A telecom photodiode detector (lasing action), as indicated by the knee of

ies considerably with wavelength for a given receives an optical pulse train and must de- the dL/dIF curve in Figure 3. The laser’s optical input power. This is known as the termine when the digital bits are on and off. slope efficiency (ΔL/ΔI) is another important spectral sensitivity (or responsivity) of the A fast response time helps ensure a low bit piece of information. It is desirable for the la- photodiode, and is dependent on detector error rate (BER). For absolute power meas- ser to produce a large increase in light output composition and device structure. Figure 2 urements in LIV testing, the detector must for a small increase in injection current. En- shows typical responsivities for Si, Ge and also be calibrated so that the photocurrent suring the laser’s output is linear and kink- InGaAs. Photodiodes based on all three generated by an individual optical pulse is free is another purpose of LIV testing. Kinks types of materials are readily available. related to the power input from the laser. are small bumps and abrupt, discontinuous changes in the slope of the LI curve, which Different Characteristics for LIV Testing are more easily seen on the dL/dI curve. Different Applications In laser diode production, LIV tests are Manufacturers base a laser’s maximum out- Detectors used as fiberoptic receivers in used to sort lasers for various applications put power rating on the linearity and absence the telecom industry must have extremely and to weed out defective devices as early as of kinks in the LI curve.

2 April 2004 Photodetectors—Choose and Use Wisely for Best Results in Pulsed Laser Diode Test Systems Figure. 3. “Smoothed” LIV curves, from originals obtained with a Figure. 4. Optical power output from a diode laser as a function of drive Keithley Model 2520 Pulsed Laser Diode Test System. current for pulsed and CW mode.

Unfortunately, both the threshold current • For any given optical power, the photo- tocurrents. and slope efficiency are influenced by tem- current produced by an optical pulse in- Of course, the rise time of the detector is perature changes caused by self-heating of put must be equal to that produced by a still an important consideration. This param- the device. Both quantities decrease with an CW input. eter is a measure of the time response of the increase in temperature. This makes thermal As can be seen from Figure 2, each de- photodiode to a light pulse input. It is defined control a necessity during diode laser test- tector material is sensitive over a clearly as the time required for the detector output to ing in (CW) mode. How- defined wavelength range. If the operating change from 10% to 90% of its “steady” or ever, early testing at the laser diode bar or wavelength of the laser under test is near the “settled” output level. As discussed earlier, chip stage usually precludes the use of active cutoff wavelength of a detector, a small shift telecom photodetectors are readily available cooling. in laser wavelength may cause a dramatic de- with response times in the nanosecond or Without active cooling, self-heating of crease in photocurrent. This can be avoided even picosecond range. However, as a pre- the laser chip can be avoided by performing by using a detector with a flatter responsivity requisite for absolute optical power measure- LIV sweeps in pulsed mode. Figure 4 illus- curve in the wavelength range of interest. ments, a detector’s photocurrent must reach trates the decrease in optical power during Another good rule of thumb is to choose 100% of its peak value before the end of each CW mode operation without active cooling, the smallest detector that will work for a par- pulse. In other words, its output current has as compared to pulsed mode. Note the rapid ticular application. Using a detector that is to be “settled.” If there is a slow diffusion decrease at higher CW drive currents. Pulse larger than necessary has severe disadvan- current component contributing to the total testing avoids this undesirable thermal ef- tages. An increase in area leads to a decrease photocurrent, a photodiode with a rise time fect, as long as the pulse duty cycle is kept in detector speed, due to an increase in junc- of a few nanoseconds could take several mi- low. Most pulse testing of laser diodes is tion capacitance. It also increases dark cur- croseconds to rise from 90% to its settled performed with pulse widths of 500ns to 1µs rent and cost. (100%) output value. For pulse widths of at a 0.1% duty cycle or less. This not only Nonetheless, a smaller detector makes 500ns or less, this would cause an inaccurate minimizes average power dissipation in the it more difficult to ensure that all the laser optical power measurement. device under test (DUT), it helps keep the radiation is intercepted and measured. A la- test cycle as short as possible. ser diode’s divergent beam requires detector Calibration Issues placement within a few millimeters of the Calibrating a photodetector or integrat- Detector Characteristics for Pulsed laser’s light emitting facet, which can create ing sphere/detector system means determin- Optical Power Measurements problems. For example, the close proximity ing the photocurrent the detector is expected Laser diode light output measurements of laser facet and detector aperture makes it to produce for a given optical input power at during LIV testing require a photodetector difficult to insert neutral density filters (of- a particular wavelength. These calibration with the following characteristics: ten required to avoid detector saturation). A constants, given in amperes/watt, can then • The photodiode must have adequate sensi- better solution is an integrating sphere with be used to convert photocurrent to optical tivity over the desired wavelength range. a built-in detector to capture diverging radia- power. • The active area must be sufficiently large tion. In this case, the size of the photodiode A calibration usually is performed over a to fully intercept a laser diode’s divergent is determined by the size of the integrating wide range of wavelengths, so the radiation beam. sphere, detector location in the sphere wall, source has traditionally been a monochro- • The rise time must be sufficiently short expected optical input power, and the ability mator and a halogen lamp with a spectrum for a pulse width of 500ns. of the instrumentation to measure low pho- that contains wavelengths from the visible to

Photodetectors—Choose and Use Wisely for Best Results in Pulsed Laser Diode Test Systems April 2004 3 Figure. 5. LI curves using the same diode laser and detector under three different pulse width conditions. Note that, in this case, shorter pulses cause the slope of the LI curve to be less steep because the photocurrent of each pulse was not settled.

the . The monochromator’s diffrac- pulse has settled Figure. 6. Pulse shapes of the same laser at various pulse widths, showing that the photocurrent value reported can be lower for shorter pulses than tion grating acts as a filter and allows only before the trailing for longer ones. certain narrow bands of wavelengths to pass edge of the optical through the output slit. The calibration is a input pulse occurs, the measured power will particular detector, source, and test setup comparison of the detector under test to an be representative of the actual optical power. may be longer than desirable. If a shorter NIST (National Institute of Standards and Figure 5 illustrates how a long settling time pulse width is necessary, the test engineer Technology) traceable reference detector. produces an incorrect slope in a laser diode’s can use data plots obtained with longer puls- Currently, calibration labs and NIST offer LI curve, i.e., an inaccurate power measure- es to make a few suppositions. For example, responsivity calibrations performed in CW ment. such plots can be used to infer the percentage mode only. NIST does not presently offer If a photodetector is intended to provide by which the power displayed is lower than it a detector responsivity calibration service absolute calibrated power measurements would have been if the pulse width were long performed with a pulsed source. Therefore, with a pulsed optical source, users should enough for the photocurrent to settle. How- analyzing the pulse performance of the pho- evaluate individual photocurrent pulse ever, this is not an ideal solution, and it un- todetector and test setup is usually required shapes with their own laser sources at differ- derscores the need to choose a detector that after CW calibration (discussed in the next ent pulse widths. This will show whether the gives the best possible results for a particular section). measured output power of each optical pulse application. is based on a “settled” photocurrent value Pulse Shape Considerations or if the measured value would actually be References The shape of the photodetector’s output higher if a longer pulse width were used (see 1. Hecht, Jeff. Understanding Fiber . current pulse is important for good LIV Figure 6). However, keep in mind that the Prentice-Hall, 1999. measurements. Ideally, the photocurrent photocurrent pulse is a result of both the 2. Donati, Silvano. Photodetectors: devic- pulse should have a perfectly flat top, rep- detector response and the shape of the input es, circuits, and applications. Prentice- resenting a “settled” value equal to the pho- pulse from the laser. If the laser has a slow Hall, 2000. tocurrent produced with an equivalent CW response to a short drive current pulse, this 3. DC Production Testing of Telecommuni- power input value. If the photocurrent is still could also be responsible for a longer rise cations Laser Diode Modules, Keithley rising at the end of the optical input pulse, time in the photocurrent output pulse. Application Note #2214. then the measured power will be lower than In some cases, the pulse width required the actual power. If the photocurrent output for an accurate power measurement from a

Specifications are subject to change without notice. All Keithley trademarks and trade names are the property of Keithley Instruments, Inc. All other trademarks and trade names are the property of their respective companies.

Keithley Instruments, Inc. 28775 Aurora Road • Cleveland, Ohio 44139 • 440-248-0400 • Fax: 440-248-6168 1-888-KEITHLEY (534-8453) • www.keithley.com

© Copyright 2004 Keithley Instruments, Inc. No. 2517 Printed in the U.S.A. 0304

4 April 2004 Photodetectors—Choose and Use Wisely for Best Results in Pulsed Laser Diode Test Systems