Pulsed Characterization of a UV LED for Pulsed Power Applications on a Silicon Carbide Photoconductive Semiconductor Switch

Transcription

1 Pulsed Characterization of a UV LED for Pulsed Power Applications on a Silicon Carbide Photoconductive Semiconductor Switch Nicholas Wilson, Daniel Mauch, Vincent Meyers, Shannon Feathers, James Dickens, and Andreas Neuber Center for Pulsed Power and Power Electronics Texas Tech University Lubbock, Texas Abstract The electrical and optical characteristics of a highpower UV LED (365 nm wavelength) were evaluated under pulsed operating conditions at current amplitudes several orders of magnitude beyond the LED s manufacturer specifications. Geared towards triggering of photoconductive semiconductor switches (PCSS) for pulsed power applications, measurements were made over varying pulse widths (25 ns 100 µs), current (0 A 250 A), and repetition rates (single shot 5 MHz). The LED forward voltage was observed to increase linearly with increasing current (~3.5 V to 53 V) and decrease with increasing pulse widths. The peak optical power observed was >30 W and a maximum system efficiency of 23% was achieved. The evaluated LED and auxiliary hardware were successfully used as the optical trigger source for a 4H-SiC PCSS. The lowest measured on-resistance of the SiC was approximately 67 kω. I. INTRODUCTION Owing to its wide bandgap (3.23 ev), Silicon Carbide (SiC) is an excellent material for photoconductive semiconductor switches (PCSS). In general, PCSSs are in certain aspects superior to other types of power switches; they typically exhibit high speed, reasonably long lifetimes, negligible jitter, and, if fabricated from certain materials, high electric field strength. [1][2] If a PCSS is to be triggered intrinsically, a light source with a higher photonic energy than the semiconductor s bandgap must be used (< 380 nm for 4H-SiC). Significantly limiting the portability and deployability of PCSSs is the requirement of short, high intensity pulses, historically only available from pulsed laser systems, excimer sources, or dielectric barrier discharge systems, all large, mechanically delicate, and energetically expensive. Previously, lasers with wavelengths ranging from 307 to 380 nm (4.04 ev to 3.26 ev) and intensities on the order of > 10 µj/cm 2 have been used to successfully trigger a SiC PCSS. [1][3][4] It was found that optimal photocurrent efficiencies for a 20 kv SiC switch were achieved with the wavelength range of 360 nm to 370 nm [3] with optimal energies between 20 µj/cm 2 and 50 µj/cm 2. In recent years, LEDs with a power output of a few watts with a wavelength of 365 nm became commercially available. These devices were primarily intended for operation in continuous mode for curing and sterilization. [1] However, when operated in pulsed mode, a substantially higher level of power is emitted from the device as will be discussed in the following. LEDs used as a trigger source for SiC PCSSs offer several advantages over lasers owing to their compactness, mechanical resilience, and potentially higher efficiency, all at substantially lower cost. Hence, LEDs may be a viable alternative for SiC PCSS trigger sources. Previous experiments utilizing a UV LED similar to the presented work were focused on triggering a SiC PCSS. At most, this earlier LED experiment was able to output 4.1 W of optical power, where around 16% of the light is incident on the conduction region. The lowest on-resistance of the SiC PCSS was about 158 kω. [4]. The design and characteristics of this PCSS are discussed in Bragg et. al [5]. The radial symmetry of this design allowed for a more efficient transmission of light in the conduction region of the switch. Utilizing an improved driver board design and a nextgeneration LED, it was possible to push the experimentally observed output power to about 35 W. Improved light coupling yielded a minimum on-resistance of about 67 kω for the SiC PCSS. II. EXPERIMENTAL SETUP The setup is comprised of the LED driver board which is activated by a pulse generator, see Fig. 1 for a basic diagram. The pulse generator triggers a MOSFET driver, which then triggers a low-side, power MOSFET as the main switch. Two DC supply voltages were needed for the board: one for the MOSFET driver at 10 V, and the other for the LED variable from 6.5 to 250 VDC.

2 energy sensor manufactured by Thorlabs (ES111C) placed above the LED as illustrated in Figure 1. Using the pyroelectric energy sensor, several measurements of the optical energy on target were recorded. The first round of experiments involved placing the sensor (active area: 0.95 cm 2 ) at different distances in front of the LED (using no lens or reflector) and recording the energy readings, see Table 1. Pulse peak power was calculated from pulse duration and total energy. Due to the angular distribution of the emitted light (cf. Fig. 2A), only a fraction of the emitted light is able to be measured at a given instant and position. Therefore, in order to determine the total optical power, the fractionally received power is divided by the factor h( ) defined in Eq. 1. : Figure 1: Simplified circuit of LED driver board, DC Supply Voltage ranging from 6.5 VDC to 250 VDC; R L current limiting resistor 1 Ω to 0.5 Ω;10 VDC supply for MOSFET Driver The printed circuit board was designed to minimize the amount of parasitic inductance. Specifically, a strip line geometry with a large ground-plane was employed throughout where possible; any large loops were avoided. Due to the significant average power flow in burst mode, up to 200 W (2 kw peak power), wide copper planes (> 5 mm), 34 µm thick, were used instead of narrow traces (~1 mm). To characterize the single pulse LED response, the LED supply voltage was swept from 6.5 VDC to 250 VDC, the upper limit established by the chosen voltage rating of the energy storage capacitors. The current used for the LED is sourced from the 700 F capacitor bank on the board. This 700 µf capacitor bank was comprised of complimentary combinations of capacitors ranging from 0.01 µf to 100 µf in order to source the wide-band frequencies of the fast square current pulses. R L, the current limiting resistor, ranged from 0.5 Ω to 1 Ω, with the lower end being used in later experiments for higher current through the LED. In addition to single shot characterization, the LED was tested with repetition rates ranging from 2 khz to 5 MHz. Lastly, the LED was used to trigger the SiC PCSS under varying LED supply voltages. To cope with the wide radiation angle of the LED for triggering the PCSS, an optical coupling system was developed to maximize the light intensity delivered onto the PCSS s narrow (8.5 mm by 1 mm) rectangular gap. Due to the circular emission profile of the LED, efficiently delivering high light intensity on target is challenging. III. MEASUREMENTS To measure the amount of current traveling through the LED, the differential voltage across R L, a low inductance carbon composite resistor, was recorded. Dividing the differential voltage by the measured R L yielded the current. The optical power of the UV LED was measured by using an EOT Silicon PIN photodiode (ET-2030), and a pyroelectric / (1) From Eq. 1, is the fraction of the optical power being measured, is the angle of the sensor position relative to the normal direction of the LED, and is the function that fits the light intensity pattern from Fig. 2A. Finally, is the distance of the sensor from the LED. Note that the result of Eq. 1 is only valid for the sensor dimension << d. Further note that ranges from 0 to /2 instead of the typical since the LED emits in only 180, and is the azimuthal angle. Table 1: Pyroelectric energy measurements (0.95 cm 2 sensor area); no lens, no reflector Mode Single Single Burst Distance (mm) DC Supply Voltage (V) Rep. Rate (khz) none none 100 # of pulses Pulse width (µs) 1,000 1, Avg. energy during pulse (µj) (on sensor) STDV (nj) Intensity during pulse (W / cm 2 ) Capture percent: 3.89% 1.35% 68% Total UV Power during pulse (W) Expected power (from datasheet [6], W) N/A

3 Figure 2A: LED angular intensity distribution utilizing a photodiode sensor placed 7.62 cm from LED (sensor area to distance ratio equivalent to 0.3 degree angle) ; no lens, no reflector. IV. FABRICATION & EQUIPMENT The parts used in this experiment were chosen because they were designed for both high power and speed applications. The system is able to source 220 A and has a rise time of approximately 15 ns. A. LED The LED used in the experiment was from LED Engin, model LZ4-04UV00. It is composed of 4 dies which were connected in a parallel configuration. The manufacturerspecified wavelength ranged between 365 nm and 370 nm [6]. The optical spectrum of the LED was recorded utilizing a spectrograph with a spectral resolution much smaller than the observed spectral width, see Fig. 2B. In the device data sheet, the minimum forward voltage from the LED is stated to be between 14 V and 18 V. The absolute maximum ratings for current, both DC and pulsed, is stated to be 1,000 ma. At 1,000 ma, the LED is rated at a radiant flux of 4.6 W [6]. Please note that the datasheet values were taken with the LED die connected in series instead of parallel, for this paper, the dies on the LED were wired parallel. The LED performed up to values one order magnitude higher than the stated values in the datasheet. B. MOSFET Driver and MOSFET The MOSFET driver (IXRFD630) and MOSFET (DE N44A) are both from IXYSRF, which are both high speed, power devices. Both devices have a rise and fall time of approximately 5 ns. The MOSFET can withstand up to 1,800 W and has an ESR of 13 mω.[7][8] C. PCSS The PCSS used is a previously developed lateral-geometry switch where anode and cathode are located on the same face of the device, see Fig. 3 [9]. Ideally, the UV light, represented by the blue arrows in Fig. 3, needs to fill the gap between the anode and cathode, roughly 8.5 mm by 1 mm, generating free charge carriers in the otherwise insulating material. Figure 3: Front and top views of PCSS, not to scale [9] D. PCSS Circuit Figure 2B: Spectral content of UV LED radiation under pulsed operation. Nothing else observed between 180 nm and 800 nm. A simplified version of the PCSS circuit is represented in Fig. 4. Again, the blue arrows indicate UV light necessary to excite the PCSS. The V charge in this experiment was set to 2 kv. An equivalent resistance of about 11 Ω was used for R load. The load current, I load, was recorded using a 0.49 Ω T&M Research SBNC-5-5 current viewing resistor (CVR).

4 w Figure 4: PCSS simplified circuit V. SINGLE SHOT LED CHARACTERIZATION RESULTS The first series of experiments consisted of testing the UV LED with different pulse widths and current levels. The current was controlled by a combination of changing the voltage and load resistance, R L from Fig. 1. Multiple LEDs were tested. Typical forward voltage drop may be inferred from Fig. 5. One observes that as the current increases, the forward voltage across the LED increases linearly, and as the pulse width increases the pulse averaged forward voltage decreases. Figure 5: Pulse averaged current (for different pulse widths) vs forward voltage, jumps in traces result from replacing LED Figures 6A and 6B illustrate the spectrally integrated LED light output at 50 ns and 500 ns, respectively with a 100 V supply. The 50 ns pulse width has a more noticeable imperfection in the rising edge for the on state of the LED, which is mostly attributed to the current rise time, see Fig. 6A. Please note that the delay between the current and the LED turning on is attributed the cable length to the photodiode (~1 m). The roughly 50 ns rise is still visible in the 500 ns pulse width, but is less noticeable because of the time scale. It takes just over 50 ns for the LED to reach its peak optical power output. Figure 6A (Top): Spectrally integrated LED light output and current at 50 ns pulse width; Figure 6B (Bottom): Same as (A) with 500 ns pulse width From previously triggering the PCSS with a UV laser system, it is obvious that the LED must be pushed to its power limits in order to achieve a noticeable drop in PCSS resistance during the LED trigger pulse. That is, the optical power emitted from LED needs to optimized and focused on target with high optical efficiency. The former is achieved by pushing extreme current through the LED, I LED from Fig. 1. The latter is achieved by effectively coupling the emitted optical power through the use of a parabolic reflector, discussed later. Increasing the pulsed current amplitude revealed diminishing returns in optical power, see Fig. 7. Noticeable results diminish after about 100 A. When the LED is subjected to high currents approaching ~100 A, the LED lifetime begins to degrade rapidly. The power dissipation increases drastically as current increases, see Fig. 7. Similar to the in Fig. 7, the amount of electric power dissipated by the LED begins to at about 150 A or around 10 kw. Power efficiency, defined as the ratio of optical power to electrical power, begins to diminish at the higher currents, see Fig. 7. Since the optical power is tens of watts and electrical power is thousands of watts, power efficiency easily falls to less than 1% at higher currents. Even though electric power can reach high numbers, the amount of energy in a single pulse reaches a maximum of around 5 mj owing to the short pulse duration.

5 LED will still operate at much higher duty cycles (cf. Figs. 11 and 12), as high as 50%, but will not be able to continue this rate for more than a few microseconds. As seen in Figure 8A and 8B, the waveforms from the LED s optical power are consistent throughout the duration of the burst operation; however, the optical power does increase slightly near the end of the operation, see Fig. 8A and 8B. Throughout testing, the LED exhibited consistent output waveforms in burst operation. Figure 8A shows a 1 MHz repetition rate with a 50 ns pulse width at 100 V supply voltage. Figure 8B is a close up of the waveforms from Figure 8A. Figure 7: Top: Pulse averaged optical power vs current, I LED, (pulse averaged current); Center: electrical power; Bottom: Electrical efficiency defined as optical power / electrical power for different pulse widths. Jumps in traces result from replacing LED between experiments. Figure 8A: Optical power of LED with a 1 MHz repetition rate, 50 ns pulse width, 100 V supply voltage, 5% duty cycle VI. REPETITION RATE LED CHARACTERIZATION RESULTS Testing the LED in burst mode at high repetition rate proved to be more difficult than the single shot operation. In brief, the capacitor bank needs sufficient capacitance to support the number of pulses in the burst the LED makes during its operation. The high currents the LED was subjected to caused failures due to increased thermal loading for long burst durations when compared with single shot operation at the same pulsed current. The LED began to noticeably degrade at currents >100 A with repetition rates >100 khz. For continuous operation, the maximum duty cycle at which the LED may be safely operated is estimated from: (2) Where is the duty cycle, i.e. the ratio of LED pulse width to the pulse period, is the maximum allowed junction temperature (130 C [6]), is the ambient temperature, is the power dissipated in the device (80 A results in a drop of 20 V across the device), and is the thermal resistance of the device (1.1 K/W [6]). For instance, with 80 A, the maximum duty cycle for the device is 6%. In burst mode, the Figure 8B: Detail of waveform in 8A VII. LED OPTICAL COUPLING To trigger the PCSS most effectively, the UV light must fill the gap between the anode and cathode, see Fig. 3. The more power and energy delivered to this gap by the LED, the lower

6 the on-state resistance of the PCSS. To accomplish this effectively, an optical system must be able to deliver a high percentage of the LED s overall light output onto this target. The system consists of a parabolic reflector followed by 2 plano-convex lenses with focal lengths of 60 mm and 75 mm. A. Parabolic Reflector The first item used is a parabolic reflector which will ideally collimate the light radiating from the LED. The parabolic reflector is made from aluminum which reflects roughly 90% of 365 nm light and is coated with silicon dioxide which also allows transmission of roughly 90% of the same wavelength of light. The parabolic reflector is 62.9 mm long with an opening of 50.8 mm (diameter). From the equation stated above, if the LED were an ideal point source placed at the focal point of the parabolic reflector, around 84% will be collimated and could ultimately be focused onto a target. The LED, however, is not a point source; it has a light emitting surface that is 2.2 mm X 2.2 mm. With this considered, one can safely assume that roughly 70% of the light can be focused onto the energy sensor. The 84% calculation arises from the light rays that do not get reflected by the reflector. Any rays that fall within +/ will not be reflected, see Fig. 9. If the rays are not reflected, they are too close for the lens to be effective. These rejection angles are determined by the parabolic reflector s dimensions. B. Plano-Convex Lenses Immediately following the parabolic reflector, two planoconvex lenses are placed, see Fig. 9. Their focal lengths are 60 mm and 75 mm. These lenses combine to form an optical system with a focal length of ~35 mm. The light incident on the lenses from the parabolic reflector is now significantly focused at around 33 mm from the lenses. This may be problematic for triggering the SiC PCSS because of the large angle incident on the sample, which can be as high as 54. Table 2: Pyroelectric energy sensor readings, with optical setup, energy readings reflect average from 30 repetitions. Virtually all of the focused light fell on the 0.95 cm 2 target area. DC Supply Voltage (V) Rep. Rate (khz) none none # of pulses Pulse width (µs) Avg. energy (µj) (on sensor) Avg. power during pulse (W) (on sensor) Intensity during pulse (W / cm 2 ) To simulate the gap on the lateral geometry PCSS, a mask was created to be placed over the energy sensor with the same dimensions as the gap in the PCSS. The mask had a gap that was 8.5 mm by 1 mm. The amount of light making it on target is now substantially less, see Table 3. However, the pulsed intensity has increased by 100% and 55% for the conditions listed in the first and second columns of Tables 2 and 3, respectively. This is simply due to slit capturing the brightest portion of the beam cross section. Please note that the energy levels for 50 V and 100 V were too low for an accurate reading with the available equipment. Table 3: Same as above, except with 8.5 mm X 1 mm mask DC Supply Voltage (V) Rep. Rate (khz) none none # of pulses 1 1 Pulse width (µs) Avg. energy during pulse (µj) (on sensor) Avg. power during pulse (W) (on sensor) Intensity during pulse (W / cm 2 ) Figure 9: Plano-convex lens setup, 2 plano-convex lenses placed about 5 mm apart, ~35 mm effective focal length VIII. PCSS RESULTS Using the optical setup from Fig. 9, the LED s performance on triggering the PCSS, cf. Fig. 4, was measured in single shot operation, see Fig. 10. Several tests were conducted measuring the amount of energy delivered to a target, the energy sensor in this case, see Table 2. Virtually all of the focused light fell on the 0.95 cm 2 target area on the energy sensor.

7 Figure 10: Single shot PCSS results, I Load current. V Charge = 2000 V, LED supply voltage = 50 V The rise time of the switch is approximately 200 ns. The PCSS setup was designed to reduce the amount of parasitic capacitance, by using large gaps (5.5 cm) between the high voltage and ground plates. The minimum estimated on resistance of the PCSS is around 67 kω. One advantage of using the LED is that the system is ultracompact in addition to having a very fast and consistent repetition rate. With the system, repetition rates up to 5 MHz have been realized on the PCSS (50% duty cycle), see Fig. 11. Repetition rates of 1 MHz and 100 khz were also examined on the PCSS, Figs. 12 and 13. The PCSS has a standard deviation of < 0.2% from pulse to pulse. The LED s pulse width in Figs is 100 ns. Figure 12: Repetition excitation of PCSS, I Load current, V charge = 2000 V, LED supply voltage = 100 V, 10% duty cycle Figure 13: Same as above except with 1% duty cycle, last two repetitions (100 khz repetition rate) Roughly 10% of the total light emitted by from LED makes it on the conductive region of the PCSS because of the geometry of the PCSS. A 40 times higher output than currently possible would be required, ~1 kw of optical power, to drive currents as high as 30 A (~50 Ω on-resistance) through the bulk PCSS, assuming 100% of the LED s optical power could be focused on the gap. Figure 117: Repetition excitation of PCSS, I Load current, V charge = 2000 V, LED supply voltage = 50 V, 50% duty cycle IX. SUMMARY It is possible to use a UV LED to trigger a PCSS. The onstate resistance of the PCSS, however, is still high. There is room to improve the switch performance of the PCSS triggered by the LED. Future work will include different geometry of PCSSs. Due to the circular shape of the focused beam, a

8 different type of geometry switch, such as one with a round target area would allow for a more efficient transmission of light to the target. ACKNOWLEDGMENT This material is based upon work supported by the Air Force Office of Scientific Research under award number FA REFERENCES [1] S. Doǧan et al., "4H SiC photoconductive switching devices for use in high-power applications," Applied Physics Letters, vol. 82, no. 18, pp , May [2] Y. Sun, M. Yang, C. Song, H. Guo, and S. Jiang, "Experimental research on silicon carbide photoconductive semiconductor switch," Microwave and Optical Technology Letters, vol. 57, no. 8, pp , May [3] C. James, C. Hettler, and J. Dickens, "Design and evaluation of a compact silicon carbide Photoconductive semiconductor switch," IEEE Transactions on Electron Devices, vol. 58, no. 2, pp , Feb [4] D. Mauch, C. Hettler, W. W. Sullivan, A. A. Neuber, and J. Dickens, "Evaluation of a pulsed ultraviolet light-emitting Diode for triggering Photoconductive semiconductor switches," IEEE Transactions on Plasma Science, vol. 43, no. 7, pp , Jul [5] J.-W. B. Bragg, W. W. Sullivan, D. Mauch, A. A. Neuber, and J. C. Dickens, All solid-state high power microwave source with high repetition frequency, Review of Scientific Instruments, vol. 84, no. 5, [6] LED Engin, 365nm UV LED Gen 2 Emitter, LZ4-04UV00 datasheet, March 17, [7] IXYSRF. 30 A Low-Side RF MOSFET Driver, IXRFD630 datasheet, [8] IXYSRF. DE N44A RF Power MOSFET, DE N44A datasheet, [9] D. Mauch, W. Sullivan, A. Bullick, A. Neuber, and J. Dickens, "Performance and characterization of a 20 kv, contact face illuminated, silicon carbide photoconductive semiconductor switch for pulsed power applications," th IEEE Pulsed Power Conference (PPC), Jun

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24