Sie sind auf Seite 1von 6



Radiation Damage in Power MOSFET Optocouplers

Allan H. Johnston, Fellow, IEEE, and Tetsuo F. Miyahira
AbstractRadiation damage is investigated in optocouplers with power MOSFET output stages. They differ from conventional optocouplers, incorporating an intermediate photovoltaic chip to allow the MOSFET gate voltage to be controlled by the light-emitting diode. These optocouplers are sensitive to ionization as well as displacement damage, and can fail catastrophically from degradation in either the power MOSFET or the LED. Radiation testing must take both mechanisms into account. Index TermsMOSFET, optocoupler, proton damage, space radiation effects.

II. BASIC CONSIDERATIONS Light-emitting diodes are one of the key components of optocouplers. Several types of LEDs are used by various optocoupler manufacturers, ranging in wavelength from 660 to approximately 900 nm. The technology and wavelength of the LEDs used within optocouplers are usually not specied by the manufacturer. The sensitivity of LEDs to proton damage varies more than three orders of magnitude, depending on the manufacturing technology. The most sensitive are amphoterically doped devices [9]. Their optical power decreases by a factor of two or [2]. At the other exmore at treme, high-speed double-heterojunction LEDs with thin active regions are not signicantly degraded until uences exceed ap. Thus, the rst concern for optoproximately couplers in space applications is to determine the type of LED. Minority carrier lifetime has been shown to correlate with displacement damage sensitivity. It is possible to measure lifetime from the optical power of discrete LEDs [10], but usually not for optocouplers. An electrical measurementreverse-recovery timecan be used to determine lifetime (and the approximate sensitivity of LEDs to radiation damage) when direct access to the output of the LED is not possible [9]. The other important component is the output stage, which imposes two difculties. First is the highly nonlinear manner in which optocouplers with complex output stages are affected by the optical power of the LED. The second difculty is that of interpreting the specic electrical conditions used by the manufacturer for the wider range of electrical conditions in which the device is likely to be used. This is particularly difcult for optocouplers with power MOSFET outputs because the specications apply to a very narrow range of conditions: a forward current of 5 to 10 mA for the LED, and a xed pulse width30 ms with a 1 A loadfor switching conditions for the Avago HSSR7110. The leakage current is also included in the specication. The devices can be applied over a wide range of conditions. For example, large leakage currents may occur at the output after irradiation that would not affect switching applications for currents on the order of 1 A, but would be critical for those involving currents of a few milliamps. Design details are not provided in the specication sheet for optocouplers with power MOSFETs, only the overall performance. HSSR-7110 optocouplers were disassembled in order to determine how the LED was coupled to the MOSFET. Detailed examination showed that an intermediate element (a separate chip), containing 10 series-connected photodiodes, was used as a photovoltaic source to develop the positive gate voltage required to turn on the MOSFET. The photovoltaic source is powered by light from the LED when it is turned on, and located directly beneath the LED (which is on a bridge over the top), as shown in Fig. 1. The output stage is a conventional power

I. INTRODUCTION PTOCOUPLERS are among the most sensitive devices to space radiation. Some are degraded by proton uences [1][5], which can occur during a as low as single solar are in space, while others are extremely sensitive to transients from heavy ions and protons [5][8]. Most studies of permanent damage have concentrated on basic optocouplers with simple phototransistors (or slightly more complex variants), which allow detailed measurements of the output stage characteristics. This makes it possible to distinguish the various mechanisms that contribute to overall degradation of the current transfer ratio: the decrease in LED optical power, photoresponse, transistor gain, and, in some cases, darkening of lenses or coupling media. Not all optocouplers are as straightforward as those with basic phototransistor outputs, making it more difcult to determine the dominant failure mechanisms. This paper discusses radiation damage in optocouplers with power MOSFET output stages, using proton and gamma irradiation to investigate the mechanisms. An earlier paper considered the effects of neutron irradiation on power MOSFET optocouplers that examined the effects of neutron activation on their characteristics [9]. The maximum neutron uence in that study was which was too low to determine the sensitivity of the LEDs to displacement damage. The devices used in this work include three types of optocouplers with power output stages, the HSSR-7110, -7111, and -7112. The main difference between them is that the HSSR-7112 requires a lower LED current5 mAcompared to 10 mA for the other two devices. All were manufactured by Avago Semiconductor (formerly Agilent).

Manuscript received September 22, 2006; revised May 15, 2007. The research in this paper was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration (NASA). The authors are with the Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109 USA (e-mail:; Digital Object Identier 10.1109/TNS.2007.903172

0018-9499/$25.00 2007 IEEE



Fig. 1. Physical diagram of the HSSR-7110 optocoupler. The LED is on a separate assembly that is placed over the internal photovoltaic source. Fig. 3. Output (drain) current vs. LED current for a 7111 optocoupler after proton irradiation.

Fig. 2. Functional schematic of the HSSR-7110 optocoupler with power MOSFET output stage.

MOSFET. The internal node between the photovoltaic source and the gates of the two power MOSFETs is a oating node that is not directly accessible. A simplied schematic of one section of the dual device is shown in Fig. 2. The power MOSFETs are discrete devices, with no provision for photocurrent collection, relying on the voltage from the photovoltaic chip for turn on. The response time of the overall circuit is very slow because several milliseconds are required for the photovoltaic source to develop a sufciently high voltage to switch the power MOSFET. Before radiation testing was done, the wavelength of the LED (884 nm) was measured on a partially disassembled device with a spectrometer. The reverse-recovery time was also measured, with a value of 2.1 . This is more an order of magnitude higher than the reverse-recovery time of 700 nm (GaAsP) and 870 nm (heterojunction AlGaAs) LEDs used in other optocouplers from that manufacturer [9]. It is in the same range as that of amphoterically doped LEDs, indicating that the HSSR-7110 will be far more susceptible to radiation damage compared to other optocouplers from that manufacturer. III. EXPERIMENTAL PROCEDURE For the initial characterization, devices were irradiated without bias (all pins grounded) using 51-MeV protons at the University of California, Davis, cyclotron. DC measurements were made using an Agilent 4156 parameter analyzer. Switching measurements were made using a conventional pulse generator and digital oscilloscope. For the power MOSFET

optocouplerswhich have a very slow response timethe switching measurements were made using a single pulse in order to minimize injection-enhanced annealing in the LED. For output currents below 100 mA the transfer characteristics are very steep, mainly because the internal voltage across the photodiode stack rises very rapidly with current. A diagnostic measurement for LED degradation in the HSSR-7110 was made by measuring the input LED current required to produce a current of 40 mA at the power MOSFET output. This effectively determines the threshold condition for turning on the MOS output transistor, and is roughly analogous to current transfer ratio in a conventional optocoupler. The LED current where the power MOSFET begins to turn on is much lower than the input current required in the device specications for high-current switching. This diagnostic measurement provides a way to determine how the LED is degraded in packaged assemblies that is more straightforward compared to the normal electrical parameters. gamma Subsequent diagnostic tests were done using a ray source. The devices for those tests were repackaged, removing the LED assembly, and adding an additional bond wire to the internal pad that is used to connect the (oating) output of the photovoltaic chip to the gate of the MOSFETs. This additional connection allowed direct measurement of the MOSFET characteristics as well as the output characteristics of the photovoltaic source. Although the internal LED was removed for the special diagnostic tests, it was possible to use an external LED to determine how the photovoltaic chip responded after it was degraded by the gamma rays. That same technique was used on devices that had been irradiated with protons for diagnostic tests. IV. RADIATION DAMAGE OF OPTOCOUPLER ASSEMBLIES A. Proton Damage The effect of proton radiation damage on the turn-on threshold can be more easily evaluated at low drain current. The nominal value of LED input current to produce a 1 mA output current was about 100 , 50 times lower than the 5 mA value required in the specication sheet for the 7112. Fig. 3 shows



Fig. 4. Comparison of proton damage in LEDs used in the HSSR-7110 with other LED technologies used by Avago Semiconductor.

Fig. 5. The effect of radiation damage on switching waveforms for the HSSR7110 optocoupler with an input LED current of 10 mA.

how the output drain current of a complete HSSR-7111 optocoupler (with the internal LED) is affected by proton damage, using a logarithmic scale to show the effects of leakage current as well as characteristics at higher drain current. For these measurements, the drain current was limited to 40 mA. The device in this gure was irradiated without bias (all pins at ground). The threshold current increases with proton uence because the LED light output is reduced. However, at high radiation levels the leakage current at the output increases because of a competing mechanism, the negative threshold shift of the internal power MOSFET. Thus, although the LED current required for switching has increased to about 0.37 mA after , the large increase in leakage a uence of current that takes place because of the negative threshold shift of the power MOSFET results in very large leakage currents. In this regime the output can no longer be turned off because of the second mechanism. The device in Fig. 3 was irradiated without bias. The leakage current mechanism will take place at much lower radiation levels for devices that are irradiated under bias because MOS transistors are more sensitive to radiation when a positive voltage is applied to the gate. In typical applications of these optocouplers, the gate voltage is only positive when the LED is turned on, biasing the photodiode stack. Consequently, in switching applications the duty cycle of the LED will be important in determining how leakage current is affected in circuit applications. A comparison of LED degradation of the HSSR-7110 with LEDs at two different wavelengths used in other Avago products is shown in Fig. 4. Those measurements assume that changes in the MOSFET and photovoltaic chip can be ignored at low uence, and that degradation in that regime is dominated by the decrease in light output from the LED. Damage was evaluated using the approach developed by Rose and Barnes, with for amphoterically doped LEDs, and for heterojunction LEDs [9], [11]. For the HSSR-7110, the slope is very close to 1, validating the assumption that (at low proton uences) the change in the transfer characteristics is dominated

by degradation in the LED. As expected from the reverse-recovery measurements, the HSSR-7110 was far more sensitive to radiation damage than the other optocouplers, which use different LED technologies. Switching time also degrades after irradiation. The effect of radiation damage on switching waveforms, conforming to the conditions specied by the manufacturer, are shown for two the switching different uences in Fig. 5. At time was degraded signicantly (the pre-irradiation values were 0.2 ms), but the device was still capable of switching the 1-A load. At the next uence the 30-ms pulse width was only capable of driving the power MOSFET to about 70% of the required load. It is evident from the waveform that the maximum output current will depend on pulse width once we reach the point where the specied pulse width condition is no longer capable of driving the output to saturation, and that radiation tests that are limited to the specic conditions in the specication will not necessarily be applicable to the broader range of applications that can be used for this type of device. Unit-to-unit variability, temperature, and aging also affect the interpretation of radiation data. LED degradation in typical devices is masked because of the very conservative way in which the device specications are set. The LED input current (before irradiation) with an output . This current of 50 mA was very low, typically about 150 is nearly two orders of magnitude lower than the LED forward current specied for the 1-A switching condition, 10 mA. Thus, under the specied conditions for switching, the LED is driven far higher than the minimum current required, providing a great deal of extra margin. As a result, radiation tests on the overall device show less degradation than would be expected from a conventional optocoupler that uses such sensitive LEDs. The large difference between the minimum conditions to affect the power MOSFET and the overdriven conditions for switching effectively mask the high sensitivity of the LED to radiation damage. Note however that the specications do not require such extreme margins. Even though typical devices begin to turn on



Fig. 6. Effect of total dose degradation on the transfer characteristics of an HSSR-7111 optocoupler (unbiased irradiation).

Fig. 7. Effect of total dose on a power MOSFET from an HSSR7111 optocoupler, irradiated with 5.5 V on the gate.

at LED currents on the order of 100 , the threshold current could be much higher and still allow a device to meet the overall requirements. We did not test a sufcient number of devices to determine how much variability occurs in the transfer characteristic margins. Adding a screen for such margins at lower output currents is an obvious way to reduce the risk of getting a device with lower internal margin that would fail at much lower radiation levels. B. Total Dose Effects on Optocouplers Radiation degradation in power MOSFET optocouplers is also affected by changes in the threshold voltage of the power MOSFET and degradation in the photovoltaic chip. Although unbiased irradiation is often used for conventional optocouplers, degradation of the power MOSFET and PV chip will be greater when bias is applied during irradiation. The effect of total dose degradation on the transfer characteristics of an unbiased device is shown in Fig. 6. In contrast to proton degradation (see Fig. 3), the LED current threshold decreases with radiation. This is due to the decrease in MOSFET threshold voltage, not LED degradation. Note that the leakage current is very high after a total dose of 25 krad(Si), which was also observed during the proton tests. In contrast to the proton tests, the LED current required to switch the power MOSFET decreased during irradiation when tests were done with gamma rays. The reason for this is that ionizing radiation produces very little degradation in LEDs, causing the transfer characteristics to be dominated by the negative threshold shift of the power MOSFET. V. DIAGNOSTIC TESTS WITH SPECIAL COMPONENTS Diagnostic tests were done on several parts that were modied by removing the LED assembly and the optical coupling compound, and bonding an additional lead to the internal connection between the output of the photovoltaic chip and the gate of the power MOSFET. LEDs are relatively unaffected by ionizing radiation (the non-ionizing energy loss for gamma rays is about two orders of magnitude lower than that of 50-MeV protons). Thus, to rst order we can assume that the LED output does not change

when gamma radiation is used, and that the net effect on the optocoupler is dominated by the change in threshold voltage of the MOSFET. However, MOSFET degradation depends on the gate bias during irradiation, so that we expect more degradation when the MOSFET is on Consequently, total dose diagnostic tests were done using biased and unbiased conditions. Fig. 7 shows the MOSFET characteristics before and after total dose radiation for a typical power MOSFET from the HSSR-7111. The irradiations were done by applying 5.5 V to the gate, which is the nominal output voltage of the photovoltaic chip. There is some stretch out in the characteristics at higher levels, indicating some contribution from interface traps, but the slope is nearly constant at low total dose levels. The threshold shift in this example is 0.095 V/krad(Si). A similar test with the gate at ground produced smaller threshold 0.04 V/krad(Si). Nearly identical shifts, approximately threshold shifts [ 0.091 V/krad(Si) for biased devices] were observed in unpublished JPL data for commercial n-channel MOSFETs. Thus, the data for the MOSFETs in the optocoupler are very similar to total dose data for other commercial power MOSFETs. Results from McGarrity show that this corresponds to a gate oxide thickness of 700 angstroms, assuming 50% hole trapping efciency [12], in close agreement with these results. Test of the internal photovoltaic chip were also done, using a special test xture that positioned an external LED over the device. Gamma rays had very little effect on the photovoltaic voltage, as shown in Fig. 8. Those tests were done using an external LEDwith a wavelength of 880 nmto evaluate the performance of the photovoltaic chip. The maximum voltage, which was about 6.2 V before irradiation, decreased by approximately 70 mV after irradiation to 30 krad(Si). Although we did not do proton tests on devices that were disassembled, the samples that were used for the proton tests were disassembled after they were returned from the proton facility. Devices that were irradiated to a uence of had a photovoltaic voltage that was 0.79 V lower than that of typical devices that had not been irradiated, a decrease of about 13%. This indicates that changes in the photovoltaic chip are a second-order effect compared to the other degradation mechanisms in the optocoupler.



Fig. 9. Switching speed for two HSSR7110 optocouplers after proton degradation at two different temperatures. Fig. 8. Effect of total dose on the output voltage of the internal photovoltaic chip, using an external LED.

VI. DISCUSSION Unlike conventional optocouplers where displacement damage is the dominant factor for both the LED and the other internal electronics, the characteristics of the internal power MOSFET in MOSFET optocouplers causes these devices to be highly sensitive to total dose damage. The most important mechanism is the negative threshold shift, which allows the device to switch at much lower voltages that in turn require lower light output from the internal LED. This produces the opposite effect on overall optocoupler performance compared to the dominant effect with proton damage. There are several important consequences. First, total dose tests are required in addition to displacement damage for these devices, even though they use internal LEDs that are highly sensitive to displacement damage effects. The total dose tests have to be done with different bias conditions (essentially with and without forward bias on the LED) because that determines the gate voltage on the internal power MOSFET, which affects total dose degradation. Second, output leakage current will become very large when the gate threshold shift is high enough to cause the device to conduct when the gate voltage is near zero. Leakage currents 100 mA were observed in total dose tests of some devices at 30 krad(Si) which essentially mean that the device can no longer be turned off, and that the LED no longer controls the output state (except for very high output currents). Similar leakage current effects can occur with protons. Third, the sensitivity of the LED and the power MOSFET to temperature also affect the results. Fig. 9 shows the dependence of switching speed on proton uence for two HSSR-7110 optocouplers at two different temperatures. Measurements at 50 cause the switching speed to be more than 20% higher compared to room temperature. The partially offsetting effects of displacement damage and total dose damage in MOSFET optocouplers make it particularly difcult to interpret radiation test data for space applications of these devices. For many spacecraft, energetic electronswhich produce less displacement damage than compa-

rable total dose levels from protonsare a signicant part of the environment. It will be necessary to separate the displacement damage and ionizing damage contributions of the nal environment in order to determine the overall effect on these types of optocouplers. This will be far more involved than using simple non-ionizing energy loss factors because the damage mechanisms are partially compensating. Finally, these optocouplers are hybrid devices, which often have more variability between devices compared to conventional microelectronics. Previous work has shown that proton radiation damage in optocouplers with 870 nm LEDs varies more than a factor of ve [13], and therefore we expect a wider range in radiation sensitivity for the LEDs within the MOSFET optocouplers compared to LEDs fabricated with heterostructures. We also have to be concerned about the initial threshold voltage of the power MOSFETs. If the pre-irradiation threshold voltage is lower than average, then the leakage current mechanism will occur at lower radiation levels. Initial ) ranged from 1.07 to threshold voltage values (for 1.46 V, a signicant range. Temperature also affects radiation damage. LED output decreases almost exactly 1% per . The threshold voltage of the . The latter factor power MOSFETs changes 6.5 mV per causes the leakage current to be about an order of magnitude once higher at 40 compared to the leakage current at 20 the power MOSFET has shifted to the point that it is clearly in the depletion mode. VII. CONCLUSIONS This paper has discussed radiation degradation in optocouplers with power MOSFET output stages, which use an internal photovoltaic chip to provide the gate voltage required to switch the power MOSFET when the LED is turned on. When sufcient degradation occurs in the LED, the output stage will no longer saturate. Unlike other optocouplers from Avago Semiconductor, these devices use amphoterically doped LEDs that are strongly which degraded at proton uences of about causes them to degrade at much lower levels than other optocouplers from that manufacturer.



However, total dose degradation in the power MOSFET allows it to turn on with lower voltages, making these devices unusually sensitive to total dose effects compared to other types of optocouplers. Total dose effects in the MOSFET partially compensate the lower light output of the LED (from displacement damage), producing the opposite effect in a radiation environment that is dominated by electrons rather than protons. Consequently the net effect in a real space environment depends on degradation of the LED, and threshold voltage shift in the power MOSFET (degradation of the photovoltaic source is a second-order effect). The presence of two distinctly different mechanisms makes it necessary to do displacement damage as well as total dose tests on these devices. It also requires more involved analyses of effects from specic radiation environments in order to evaluate their performance in space. The temperature sensitivity of the power MOSFET and the LED are strong enough to reduce the failure level in typical applications where the optocoupler operates at 40 to 50 . The temperature sensitivity and potentially larger unit-to-unit variability in radiation damage of the different internal components have to be taken into account when these devices are used in space environments. REFERENCES
[1] H. Lischka, H. Henschel, O. Kohn, W. Lennartz, and H. Schmidt, Radiation effects in light-emitting diodes, photodiodes and optocouplers, in Proc. 2nd European Conf. Radiation and its Effects on Components and Systems, St. Malo, France, Sep. 1993, pp. 226231. [2] B. G. Rax, C. I. Lee, A. H. Johnston, and C. E. Barnes, Total dose and proton damage in optocouplers, IEEE Trans. Nucl. Sci., vol. 43, no. 6, pp. 31673173, Dec. 1996.

[3] K. A. LaBel, S. D. Kniffen, R. A. Reed, H. S. Kim, J. L. Wert, D. L. Oberg, E. Normand, A. H. Johnston, G. K. Lum, R. Koga, S. Crain, J. R. Schwank, G. L. Hash, S. Buchner, J. Mann, L. Simkins, M. DOrdine, C. A. Marshall, M. V. OBryan, C. M. Seidlick, L. X. Nguyen, M. A. Carts, R. L. Ladbury, and J. W. Howard, A compendium of recent optocoupler radiation test data, in Proc. IEEE Radiation Effects Data Workshop, Reno, NV, Jul. 2000, pp. 123146. [4] R. Germanicus, L. Dusseau, F. Saigne, S. Barde, R. Ecoffet, O. Mion, R. Calvel, J. Fesquet, and J. Gasiot, Analysis of the proton-induced permanent degradation in an optocoupler, IEEE Trans. Nucl. Sci., vol. 49, no. 3, pp. 14211425, Jun. 2002. [5] R. A. Reed, P. W. Marshall, A. H. Johnston, J. L. Barth, C. J. Marshall, K. A. LaBel, M. DOrdine, H. S. Kim, and M. A. Carts, Emerging optocoupler issues with energetic particle-induced transients and permanent radiation degradation, IEEE Trans. Nucl. Sci., vol. 45, no. 6, pp. 28332841, Dec. 1998. [6] A. H. Johnston, T. Miyahira, G. M. Swift, S. M. Guertin, and L. D. Edmonds, Angular and energy dependence of proton upset in optocouplers, IEEE Trans. Nucl. Sci., vol. 46, no. 6, pp. 13351341, Dec. 1999. [7] R. A. Reed, C. Poivey, P. W. Marshall, K. L. LaBel, C. J. Marshall, S. Kniffen, J. L. Barth, and C. Seidlick, Assessing the impact of the space radiation environment on parametric degradation and single-event transients in optocouplers, IEEE Trans. Nucl. Sci., vol. 48, no. 6, pp. 22022209, Dec. 2001. [8] P. J. McMarr, M. E. Nelson, H. Hughes, and K. J. Delikat, 14-MeV neutron and cobalt-60 gamma testing of a power MOSFET optocoupler, IEEE Trans. Nucl. Sci., vol. 50, no. 6, pp. 20302037, Dec. 2003. [9] A. H. Johnston and T. F. Miyahira, Hardness assurance methods for radiation degradation of optocouplers, IEEE Trans. Nucl. Sci., vol. 52, no. 6, pp. 26492655, Dec. 2005. [10] A. L. Barry, A. J. Houdayer, P. F. Hinrichsen, W. G. Letourneau, and J. Vincent, The energy dependence of lifetime damage constants in GaAs LEDs for 1500 MeV proton, IEEE Trans. Nucl. Sci., vol. 42, no. 6, pp. 21042107, Dec. 1995. [11] B. H. Rose and C. E. Barnes, Proton damage effects on light-emitting diodes, J. Appl. Phys., vol. 53, no. 3, pp. 17721780, 1982. [12] J. M. McGarrity, Considerations for hardening MOS devices and circuits for low radiation doses, IEEE Trans. Nucl. Sci., vol. NS-27, no. 6, pp. 17391744, Dec. 1980. [13] T. F. Miyahira and A. H. Johnston, Trends in optocoupler radiation degradation, IEEE Trans. Nucl. Sci., vol. 49, no. 6, pp. 28682873, Dec. 2002.