"Single Event Transients in 1.5V Voltage Regulators"
C. Poivey1, H. Kim1, C. Seidleck1, J. Karsh2, S. Pursley2, I. Kleyner2, R. Katz2, K. LaBel2
1 MEI, Technologies
As other bipolar analog devices, voltage regulators are known to be sensitive to single event transients (SET). However, as in typical applications large output capacitors are used to provide noise immunity, SET amplitude is generally small, less than 1V. These SET are a concern for low voltage applications. Over-voltages may cause destructive conditions. Under-voltage may cause functional interrupt and may also trigger electrical latchup conditions. Voltage regulators SET are critical in FPGA applications. For example, in the case of Actel FPGA RTAX-S family, core power supply voltage is 1.5V. Manufacturer specifies an absolute maximum rating of 1.6V and recommended operating conditions between 1.425V and 1.575V.
We tested two 1.5V low dropout voltage regulators for SET sensitivity, RHLP4913 from ST Microelectronics and MSK5900 from MS Kennedy.
Bias conditions of RHLP4913 are shown in Figure 1. The device was tested under different load conditions, output current and output capacitor. The device was also test without and with a RC filter, shown in red in the figure, made of a 0.1 ohm resistor, and a 200 mF capacitor. Input voltage is 3.3V +/-10%. Output Voltage is 1.5V.
Fig 1: RHLP4913 bias conditions
Test circuit contains a power supply for the input voltage, an electronic load for drawing current, and a digital scope for capturing any output anomalies. Once the, programmable output is present and the load conditions are set, the digital scope is set to trigger on and voltages that are above or below a predetermined threshold of 70 mV.
Figure 2 shows the SET cross section curves for different load conditions, low and high output current and low and high output capacitor, without filter. We can see in Figure 2 that output capacitor has little effect on SET cross section. However, output current does have an effect. The sensitivity is higher for high current loads.
Worst-case transients are shown in Figure 3. They were observed for the largest LET, the largest output current, and the lowest output capacitor. Larger output capacitor values reduce significantly the amplitude and duration of under-voltages. But still remain 200mV amplitude over-voltage transients that last for up to 2 ms, and short duration, 200 ns, bipolar transients of 300mV maximum amplitude.
The filter was effective to remove all long duration transients, but the short duration bipolar transients were not suppressed.
Figure 2: RHLP4913, SET cross section curve
Figure 3: RHLP4913 worst-case transients
MSK5900 was tested under similar conditions. Figure 4 shows the SET cross-section. We can see that MSK5900 is significantly less sensitive than RHLP4913. Maximum cross –section is about one order of magnitude lower and LET threshold is larger than 15 MeVcm2/mg, to compare with 2 MeVcm2/mg for RHLP4913. We can also see that MSK5900 is most sensitive for the lowest current load of 100 mA.
Figure 4: MSK5900 SET cross-section curve
Only one kind of transient was observed with MSK5900, 200 mV over-voltage transients with a worst-case duration of 4 ms. With the filter, the amplitude of transient is reduced to about 50 mV. Figure 5 shows typical transients with and without filter.
Figure 5: MSK 5900, typical SETs, without filter (top) and with filter (bottom)
The drawback of the serial resistance in the filter is a voltage dropout. Therefore, if the FPGA has different regimes, for example standby and high speed, the serial resistance may cause the output voltage to be out the specified limits recommended by ACTEL. We have investigated an inductive capacitive filter that does not cause any voltage dropout. Results will be presented in the final paper.
2006 MAPLD International Conference Home Page