Testing laser diodes for fiber optic communications systems requires photodiodes with fast response times. But 10 to 90 percent rise times are not the only consideration, particularly when photodiodes are used for absolute optical power measurements in pulsed mode. Given that packaging adds 80 percent or more to the cost of a laser diode module, it is imperative to ensure that only good laser diode chips enter communications modules. This requires testing laser diodes at the bar or chip stage with pulse techniques that prevent destructive self-heating of uncooled devices. These techniques drive the diodes using pulsed current sources with fast rise times. Photodetectors, generally PIN detectors with fast response times, accurately measure the lasers’ output. Photodiodes are the most commonly used detectors because of their low cost and good performance. Like all semiconductor detectors, they make use of an internal photoelectric effect. When a photon is absorbed, it raises an electron from the valence band to the conduction band, creating a charge-carrier pair — a hole in the valence band and an electron in the conduction band. If connected to an external circuit, the charge forms a photocurrent. For pulsed light-current-voltage testing, use of these photodetectors involves a number of compromises, but fast response or rise times are often the dominant concern. Rise time is a measure of how quickly a photodiode responds to a light-pulse input. It is defined as the time required for the detector output to change from 10 to 90 percent of its “steady,” or “settled,” output level. Light-current-voltage curves can reveal a laser’s threshold current, efficiency and maximum operating power. Output should be linear and kink-free. Kinks represent abrupt, discontinuous changes in the slope of the LI curve, which are more visible on the kink test curve. As important as rise time is, users must consider the relationships between a detector’s response time and its other characteristics, such as device structure, external circuitry and bias voltage. The most basic photodiode has a simple PN-junction structure, in which the diffusion of electrons into the P-type material and holes into the N-type material causes an opposing electric potential (or equilibrium potential) across the depletion region or layer. Reverse biasing the photodiode causes this potential to increase and the depletion region to expand. Photons that impinge on this region create electron-hole pairs that later separate because of the combined equilibrium potential and externally applied reverse bias. These charges quickly drift away from the junction and are collected by the electrodes. Drift current generated by photon absorption in the depletion region is the main component of photocurrent. It also has the fastest response time. Another component is the diffusion current, which originates from charge carriers created by the absorption of photons outside the depletion layer. Although most charge carriers recombine in the neutral region, some slowly diffuse to the junction and contribute to photocurrent, undermining response time. For a fast response time, a photodiode’s design must allow most photon absorption to take place in the depletion layer. One approach is to make the P-region extremely thin and to expand the depletion layer by applying as large a reverse bias as possible. Increasing the bias, however, increases the dark current and its associated noise contribution. Dark current is more pronounced in semiconductors with large numbers of impurity atoms, such as heavily doped PN-junction photodiodes. PIN photodiodes also help control the width of the depletion region. These devices incorporate an undoped (or lightly doped) intrinsic semiconductor layer between the p- and n-regions of the device. An intrinsic layer of fixed width serves the same function as the depleted region of a PN junction. Because it lacks doping impurities that can generate current carriers in the dark, it helps improve photodiode sensitivity. More importantly, by concentrating photon absorption in this region, it minimizes undesirable diffusion current. In PIN photodiodes, the intrinsic layer concentrates photon absorption and minimizes diffusion current to deliver a fast response. The bias voltage, and thus the electric field, is concentrated and virtually constant across the absorption region. This greatly decreases the average drift time of the optically generated carriers. Hence, PIN photodiodes are usually preferred for applications that require a fast response, as in pulse testing of diode lasers. PIN detectors are based on silicon, germanium or III-V compound technology, depending on the application requirements. The bandgap of the detector material determines the optical wavelength response of a photodiode. Germanium and indium gallium arsenide photodiodes detect radiation from approximately 800 to 1700 nm, making them suitable for testing telecom transmitter lasers. Silicon, sensitive from 300 to 1100 nm, is the preferred material for detecting shorter wavelengths. Although a particular photodiode can detect radiation over a range of wavelengths, its photocurrent output will vary considerably with wavelength for a given optical input power. Its spectral sensitivity or responsivity depends on detector composition and device structure. Depending on the application requirements, users select PIN detectors based on silicon, germanium or III-V compound technology. Each has a different spectral response, measured in amperes/watt of incident light vs. wavelength. Pulsed testing Detectors used as fiber optic receivers in the telecom industry must have extremely fast rise times to detect optical signals that are modulated at several gigahertz. Despite their speed, these detectors may lack the characteristics needed for applications in which the detector usually makes an absolute power measurement at various laser drive currents, such as pulsed light-current-voltage testing. Therefore, all the light emitted by the laser must be captured, either directly by the detector or by an integrating sphere that incorporates the detector in the inner sphere wall. Because laser diode beams are highly divergent, intercepting all the light requires a large-area detector with a diameter of several millimeters. This excludes fiber optic receivers, which typically have diameters of 50 μm. The fast response times of telecom photodiodes ensure a low bit error rate as they characterize digital bits in an optical pulse train. For absolute power measurements in light-current-voltage testing, however, the detectors must be calibrated to relate the photocurrent generated by an individual optical pulse to the power input from the laser. Light-current-voltage testing helps to determine which lasers are suitable for various applications and to weed out defective devices as early as possible in the production process. These tests also determine some of the optical and electrical characteristics of the final, packaged product. long settling time produces an incorrect slope in a laser diode’s photocurrent curve. These curves derive from the same laser diode and detector under three different optical pulse width conditions. For shorter pulses, the photodetector did not reach 100 percent of its output value before the end of the optical input pulse. Thus, the observed slope efficiency tends to fall off (magenta and yellow lines) from the actual efficiency (dark blue line), because the detector is not “seeing” the full amplitude of the input pulse. Applying a range of drive currents to the laser diode and plotting voltage drop and light output as a function of current generate a light-current-voltage curve. The test instrument performs a drive current sweep with small step increments and simultaneously measures the laser diode’s forward voltage drop for each current step. It also measures optical power output from the laser’s front and back facets at each step, which requires calibrated photodetectors. Threshold current increases and slope efficiency decreases with a rise in device temperature incurred by self-heating. Although this necessitates thermal control during testing in CW mode, testing at the bar or chip stage usually precludes the use of active cooling. Laser diode output power is temperature-dependent and drops because of self-heating unless cooled or operated in pulsed mode with a short duty cycle. Without active cooling, optical power decreases during CW-mode operation, particularly at higher drive currents. Pulse testing of laser diodes avoids this undesirable thermal effect if the pulse duty cycle is kept low. In the absence of active cooling, light-current-voltage sweeps in pulsed mode can avoid self-heating of the laser chip. Most pulse testing of laser diodes is performed with pulse widths of 0.5 to 1 μs at a 0.1-percent-or-lower duty cycle. This not only minimizes average power dissipation in the device under test, but also helps keep the test cycle as short as possible. If the operating wavelength of the laser under test is near the cutoff wavelength of a detector, a small shift in laser wavelength may cause a dramatic decrease in photocurrent. A detector with a flatter responsivity curve in the wavelength range of interest can help avoid this. Another good rule of thumb is to choose the smallest detector that will work for a particular application. Using detectors that are larger than necessary has severe disadvantages: Large-area detectors are slower because their junction capacitance is higher. A larger area also increases dark current and cost. Nevertheless, smaller detectors make it more difficult to intercept and measure all the laser radiation. A laser diode’s divergent beam requires detector placement within a few millimeters of the laser’s light-emitting facet, which can create problems. For example, the proximity of the laser’s facet to the detector’s aperture complicates the insertion of neutral density filters, which often are required to prevent detector saturation. One solution is integrating spheres with built-in detectors to capture diverging radiation. The size of the integrating sphere, the expected optical input power and the ability of the instrumentation to measure low photocurrents all determine the size of the photodiode. Photocurrent values can be lower for shorter pulses than for longer ones, as illustrated by photodetector output pulse shapes for various optical input pulse widths from the same laser. Such plots help estimate the degree of error in power measurements due to slow rise time (or shape) of the photodetector response. The rise time of the detector is still important. As a prerequisite for absolute optical power measurements, a detector’s photocurrent must reach 100 percent of its peak value before the end of each pulse. In other words, its output current must be settled. If there is a slow diffusion current component contributing to the total photocurrent, a photodiode with a rise time of a few nanoseconds could take several microseconds to rise from 90 percent to its settled output value. For pulse widths of 500 ns or less, this can cause inaccurate power measurements. Calibrating a photodetector or an integrating sphere/detector system means determining the detector’s expected photocurrent for a given optical input power at a particular wavelength. However, calibration usually is performed over a wide range of wavelengths and typically uses a monochromator and a halogen lamp emitting from the visible to the infrared. The monochromator’s diffraction grating filters narrow bands of wavelengths to pass through the output slit. Calibrating detectors generally requires comparison to a reference detector traceable to the National Institute of Standards and Technology (NIST). It provides constants measured in amperes per watt, which help translate photocurrent into optical power. However, calibration labs and NIST only offer responsivity calibrations performed in CW mode. Therefore, analyzing the pulse performance of the photodetector and the test setup requires another approach. The shape of the output current pulse is important for good light-current-voltage measurements. Ideally, the photocurrent pulse will have a flat top, representing a “settled” value equal to the photocurrent produced with an equivalent CW power input value. If the photocurrent still is rising at the end of the optical input pulse, the measured power will be lower than the actual power. If the photocurrent output pulse settles before the trailing edge of the optical input pulse, the measured power will be representative of the actual optical power. If a photodetector provides absolute calibrated power measurements with a pulsed optical source, users should evaluate individual photocurrent pulse shapes with their own laser sources at different pulse widths. This will show whether the measured output power of each optical pulse is based on a settled photocurrent value or if the measured value would be higher at longer pulse widths. However, keep in mind that photocurrent pulse is a result of both the detector response and the shape of the input pulse from the laser. If the laser has a slow response to a short drive current pulse, this could also result in a longer rise time in the photocurrent output pulse. In some cases, the pulse width required for an accurate power measurement from a particular detector, source and test setup may be longer than is desirable. If a shorter pulse width is necessary, the test engineer can use data plots obtained with longer pulses to make a few suppositions. For example, such plots can indicate the percentage by which the power displayed is lower than it would have been if the pulse width were long enough for the photocurrent to settle. This is not an ideal solution, however, and it underscores the need to choose a detector that gives the best possible results for a particular application. Meet the author Helga Alexander is an optical engineer with Keithley Instruments Inc. in Cleveland.