1. Introduction
Pulse width modulators (PWMs) are commonly used in switching mode power supply systems, on which the single event effect (SEE) is of growing concern for spaceflight. To fully understand the SEE mechanisms, PWMs were tested by broad ion beams [1-6], SEE cross-section as a function of heavy-ion linear energy transfer (LET) or proton energy were measured to provide information needed for predicting SEE error rates in space. However, due to the random nature of the individual ion strikes, the ion beam test doesn’t provide the temporal and spatial information to understand the SEE mechanisms.
Subsequently, studies on SEE of PWMs were performed with laser beams [7-9] for detailed analysis of SEE mechanisms [10-14]. A. M. Chugg et al. [7] revealed some details of PWM SEE behavior by injecting laser pulses on few positions, such as delayed onset of latch-up near the single-event latch-up threshold. Y Ren et al. [8,9] showed that single event transient (SET) pulses on the internal reference voltage circuit might interfere with the protection circuits and cause system reset, and reported the effect of temperature-induced quiescent operating point shift on SET sensitivity. Other studies focused on the correlation of heavy ion and laser testing on PWMs [15-16] or the relationship between proton and laser testing data on PWMs [17].
However, reports on temporal characteristic analysis of SEE in PWMs are rarely seen. V. Pougetet al. [18] showed that the SEE sensitive areas of a portion of the AD7821 ADC changed with time after the conversion start, indicating the temporal characteristic analysis would be required to comprehensively characterize the SEE performance.
In this work, we obtained the SEE signatures for different injection times through the output cycle in different modes of oscilloscope trigger. A time-domain error-identification method was used in the temporal characteristic analysis of SEE.
2. Experimental details
A current-mode PWM implemented in bipolar technology was investigated in this work. Top package was removed for the laser beams to irradiate the active silicon regions. In addition, there is only one layer of metal interconnection in the chip and the routing is not dense, so some fraction of the laser may bypass the metal and generate SEEs.
The extracted circuitry has been mapped back onto the physical die layout and the block diagram. Fig.1a shows a photomicrograph of the device and the main blocks, which include the OSC(oscillator) block, the Error Amp(error amplifier) block, the UVLO(under voltage lock out) block, the VREF(reference voltage) Good Logic block, the Current Sense Comparator block, the SR Latch block and the T Latch block.
The device under test (DUT) was irradiated on a pico-second neodymium-YAG laser system PL2210 (Fig.1b) at the Northwest Institute of Nuclear Technology. The system delivers laser pulses centered at 1064 nm (1.2eV) with pulse width of 21.8ps and 1 kHz repetition rate. The single photon energy is greater than the silicon band-gap (1.12eV) so that an electron in the valence band can be excited to the conduction band, and leaves a hole there. The strong laser beams were attenuated and passed through a wave plate-polarizer combination to precisely control the incident pulse energy. With a 100× microscope objective, the laser beam was focused on the device in a 1.4μm Gaussian spot (FWHM).
The laser pulses were injected on the bipolar transistors of different circuit blocks. Relevant signals were captured for each SEE event using a digital oscilloscope. They include the reference voltage (VREF), the device output (OUTPUT), the oscillator input signal named RT/CT, and the error amplifier output (COMP).The incident laser energy was 50nJ. Making an accurate determination of the fraction of incident light reaching the active silicon regions is necessary for quantitative measurements, but this is out of the scope of this paper.
A time-domain error-identification method should be used in the temporal characteristic analysis of SEE. Here we use reference signals to decide the phase relationship between the injection times and the output cycle. The choice of reference signals for different circuit blocks will be discussed below.
3. Results and Discussion
3.1. The OSC block
For the OSC block, the typical SEE signatures after each laser shot were both a disturbance in the oscillator input signal and a phase shift in OUTPUT. As can be seen in the block diagram of the PWM (Fig.2), the pulse width modulation is achieved by both SR latch and T latch. The OSC block provides the oscillator signal to control the S input of SR latch and the clock signal of T latch. The S input of SR latch can hardly change the latch output when the R input is disabled. Therefore, SEE for the OSC block may be primarily due to the changes in the clock signal of T latch, which is related to the oscillator input signal. In view of the above analysis, the reference signal of the OSC block can be the oscillator input signal named RT/CT.
According to the change of the oscillator input signal after each laser shot, the phase relationship between the injection time and the output cycle could be determined. The results for a laser shot occurring on a high level and a low level of OUTPUT are illustrated in Fig.3. Fig.4 provides the timing waveforms. With the laser shot on a high level of OUTPUT, the clock signal of T latch that controls the falling edge of OUTPUT was delayed; with the laser shot on a low level of OUTPUT, the clock signal of T latch controlling the rising edge of OUTPUT was delayed.
3.2. The Error Amp block
For the Error Amp block, the typical SEE signatures after each laser shot were both a voltage transient in the error amplifier output and a change in OUTPUT. This block can reset the SR latch by pulling the error amplifier output below a voltage, which forces high level output of the current sense comparator (Fig.2). Therefore, SEE for this block may be explained that the negative transient in the error amplifier output affected OUTPUT by resetting the SR latch temporarily. So, the reference signal of the block can be the error amplifier output, COMP.
According to the change of the COMP after each laser shot, the phase relationship between the injection time and the output cycle was observed (Fig.5a). With the laser shot on a rising edge of OUTPUT, there was a missing pulse. For other injection times, there might be a change in duty cycle (Fig.5b) or even no errors (for example, the laser shot on a low level of OUTPUT).
Fig.6 provides the timing waveforms for different resetting times through the output cycle. It can be seen that the error was more serious when the resetting occurs on a rising edge of OUTPUT (blue waveforms), which means that the SEE behavior may be more serious when the laser shot occurs on a rising edge of OUTPUT.
3.3. The UVLO block and the VREF Good Logic block
For the UVLO block and the VREF Good Logic block, all the measured signals might be affected by a laser shot, because the blocks have the function of providing signals to other blocks. Two typical SEE signatures were shown in Fig.7.
For Type 1 SEE signatures (Fig.7a), OUTPUT was decided by the oscillator signal and the error amplifier output. There was also a disturbance in the reference voltage, contributing to the voltage transient in the error amplifier output. The disturbance in the oscillator signal affected the clock signal of T latch, and the negative transient in the error amplifier output reset the SR latch temporarily.
For Type 2 SEE signatures (Fig.7b), OUTPUT was decided by the oscillator signal and the reference voltage. The disturbance in the oscillator signal affected the clock signal of T latch, and the disturbance in the reference voltage raised CURRENT SENSE pin (Fig.2) above 1V, resetting the SR latch temporarily.
The temporal characteristics of SEE in the two blocks depend on the SEE response of some related blocks.
3.4. Other circuit blocks
For the current sense comparator block, and the T and SR latch blocks, the typical SEE signature observed after each laser shot is shown in Fig.8. Except for OUTPUT, there was no obvious change in the measured signals. This behavior can be explained by the fact that an error in these blocks can hardly propagate into the front blocks for the open-loop test.
Although there is no reference signal for these circuit blocks, the phase relationship between the injection times and the output cycle can be indicated to some extent. For example, with a laser shot on a rising edge of OUTPUT (Fig.8a), there was a phase shift; with a laser shot on a high level of OUTPUT (Fig.8b), there was a transient in the high level of OUTPUT. It seems that the SEE behavior is more serious when the laser shot occurs on a rising edge of OUTPUT.
3.5. Discussion
In the experiments, we found that some SEE sensitive locations on the PWM die typically only produced an SEE response for one-in-tens of laser pulses delivered. This was reported by Chugg et al[2],who attributed the phenomenon to enhanced sensitivity during output switching. According to our experiment results, there might be no errors when the laser shot occurs on the low level of OUTPUT for the Error Amp block. This means that the SEE sensitivity might be only exhibited during transitory periods of the output cycle for some locations. In addition, it should be noted that the error signatures for different injection times through the output cycle were sometimes obtained by setting different trigger modes. For example, Figs.3a and 3b were obtained by using positive and negative pulse width triggering, respectively. This is a possible indication that some SEE behaviors might be neglected by using a single trigger mode.
To see the layout, the microchip die should be illuminated with a white light source directed through the microscope objectives. In our experiment, when the transistor TI9 in the schematic of the UVLO block (Fig.9) was illuminated with a white light source directed through the 100× microscope objective, the response waveforms for low and high illuminator brightness were observed (Fig.10). Some other transistors/diodes circled in Fig.9 were also sensitive to the illumination. It can be seen that the normal function of the device was influenced and the influence was more serious for higher illuminator brightness. This phenomenon may be due to some light-dependent materials introduced in the semiconductor manufacturing process of these components. Fortunately, the device comes to normal when the illumination was turned off. Therefore, it is necessary to turn off the illumination during SEE measurements in some cases.
4. Conclusion
By injecting laser pulses in different circuit blocks, insights into the nature and origin of SEE are provided. SEEs for the OSC block are primarily due to the changes in the clock signal of T latch. SEEs for the Error Amp block can be explained that the negative transient in the error amplifier output affects OUTPUT by resetting the SR latch temporarily. SEEs for the UVLO block and the VREF Good Logic block result from the changes in the clock signal of T latch as well as the resetting of the SR latch. According to the above analysis, mitigation approaches are required for T latch and SR latch. Circuit hardening solutions can be added to mitigate the single event transients in the clock pin of T latch and the reset pin of SR latch. In addition, the reference signals used to decide the phase relationship between the injection times and the output cycle are selected for different circuit blocks through the mechanism analysis.
The SEE signatures for different injection times through the output cycle are compared and it is shown that the SEE behavior is more serious when the laser shot occurs on a rising edge of OUTPUT for the Error Amp block, the current sense comparator block, the T latch block and the SR latch block. In view of the result, SEE error rate may decline when the PWM works at lower frequency. Frequency dependence of SEE sensitivity of this device can be considered in heavy-ion beam testing.
Single event effects in pulse width modulation controllers
. IEEE Trans. Nucl. Sci.43,2968-2973 (1996).doi: 10.1109/23.556893The SEU in pulse width modulation controller with soft start and shutdown circuits
.Single event effects testing of the Linfinity SG1525A pulse width modulator controller
.Total dose and single event testing of a hardened single-ended current mode PWM controller
.Development of measurement system for single event effect on pulse width modulator
. Atomic Energy Science and Technology 48, 717-722 (2014). doi: 10.7538/yzk.2014.48.s0.0717Experimental research of heavy ion and proton induced single event effects for a Bi-CMOS technology DC/DC converter
. Journal of Semiconductors 36, 115010-1-115010-5 (2015). doi: 10.1088/1674-4926/36/11/115010Laser simulation of single event effects in pulse width modulators
. IEEE Trans. Nucl. Sci.52, 2487-2494(2005). doi: 10.1109/TNS.2005.860721Single-event effects analysis of a pulse width modulator IC in a DC/DC converter
. Springer Journal of Electronic Testing: Theory and Applications 28,877-883 (2012) doi: 10.1007/s10836-012-5338-8The effect of temperature-induced quiescent operating point shift on single-event transient sensitivity in analog/mixed-signal circuits
. Springer Journal of Electronic Testing: Theory and Applications 30, 377-382 (2014). doi: 10.1007/s10836-014-5448-6Pulsed-laser testing for single-event effects investigations
. IEEE Trans. Nucl. Sci. 60,1852-1875 (2013).doi: 10.1109/TNS.2013.2255312Phase-dependent single-event sensitivity analysis of high-speed A/MS circuits extracted from asynchronous measurements
. IEEE Trans. Nucl. Sci.58, 1066-1071 (2011). doi: 10.1109/TNS.2011.2125989Time-resolved scanning of integrated circuits with a pulsed laser: application to transient fault injection in an ADC
. IEEE Trans. Nucl. Sci.53, 1227-1231 (2004). doi: 10.1109/TIM.2004.831488Pulsed-laser testing methodology for single event transients in linear devices
. IEEE Trans. Nucl. Sci. 51, 3716-3722 (2004). doi: 10.1109/TNS.2004.839263Comparison of SET’s in bipolar linear circuits generated with an ion microbeam, laser light and circuit simulation
. IEEE Trans. Nucl. Sci.49, 3163-3170 (2002). doi: 10.1109/TNS.2002.805346Correlation of heavy-ion and laser testing on a DC/DC PWM controller
. Springer Journal of Electronic Testing: Theory and Applications 29, 609-616 (2013). doi: 10.1007/s10836-013-5379-7Experimental study on single event transient of current mode PWM controller
. Nuclear Techniques 37, 120401-1-120401-6 (2014). doi: 10.11889/j.0253-3219.2014.hjs.37.120401Single-event transient measurements on a DC/DC pulse width modulator using heavy ion, proton, and pulsed laser
. Springer Journal of Electronic Testing: Theory and Applications 30, 149-154 (2014). doi: 10.1007/s10836-013-5431-7Time-resolved scanning of integrated circuits with a pulsed laser: application to transient fault injection in an ADC
. IEEE Trans. Instr. and Meas.53, 1227-1231 (2004). doi: 10.1109/TIM.2004.831488