Single-Event Transients in Voltage Regulators

Abstract

Single-event transients are investigated for two voltage regulator circuits that are widely used in space. A circuit-level model is developed that can be used to determine how transients are affected by different circuit application conditions. Internal protection circuits-which are affected by load as well as internal thermal effects-can also be triggered from heavy ions, causing dropouts or shutdown ranging from milliseconds to seconds. Although conventional output transients can be reduced by adding load capacitance, that approach is ineffective for dropouts from protection circuitry

Full text

Page 1

Copyright © 2006 IEEE. Reprinted from IEEE TRANSACTIONS ON NUCLEAR SCIENCE


This material is posted here with permission of the IEEE. Such permission of the IEEE does not in any
way imply IEEE endorsement of any of Jet Propulsion Laboratory's products or services. Internal or
personal use of this material is permitted. However, permission to reprint/republish this material for
advertising or promotional purposes or for creating new collective works for resale or redistribution must
be obtained from the IEEE by writing to pubs-permissions@ieee.org.



By choosing to view this document, you agree to all provisions of the copyright laws protecting it. 

Page 2

IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 53, NO. 6, DECEMBER 20063455
Single-Event Transients in Voltage Regulators
A. H. Johnston, Fellow, IEEE, T. F. Miyahira, F. Irom, and J. S. Laird
Abstract—Single-event transients are investigated for two
voltage regulator circuits that are widely used in space. A cir-
cuit-level model is developed that can be used to determine how
transients are affected by different circuit application conditions.
Internal protection circuits—which are affected by load as well as
internal thermal effects—can also be triggered from heavy ions,
causing dropouts or shutdown ranging from milliseconds to sec-
onds. Although conventional output transients can be reduced by
adding load capacitance, that approach is ineffective for dropouts
from protection circuitry.
Index Terms—Radiation testing, single-event transient, voltage
regulator.
I. INTRODUCTION
D
are strongly affected by circuit application conditions, making
it difficult to extend tests with a specific circuit configuration
to other applications [1]–[9]. This paper discusses SETs in
linear voltage regulators, using two widely used devices as
test vehicles. These regulators can be used over a very wide
range of input voltages, with input/output voltage differences
up to 40 V. The wide range of supply voltages and variation in
output loading conditions are critical parameters in determining
SETs for these devices, but increase the complexity of radiation
testing and data evaluation.
In studying SETs, the first step is to establish the criteria
for the amplitude and duration of output voltage response. For
voltage regulators, transients with small amplitude or short du-
ration are often of secondary importance, but may induce un-
acceptable noise for critical circuits. Transients with large am-
plitude can produce voltage conditions that may damage other
circuits, or produce noise-like effects that disrupt normal circuit
functions. We have to keep in mind that the SETs from these
regulators can take place even with a relatively large capacitor
presentatthedeviceoutput.Thelowoutputimpedanceandhigh
current compliance of the regulator basically overcome the ex-
pected filtering effect of bypass capacitors—up to the value of
capacitance used during SET measurements—producing tran-
sients in a network that is often assumed to be immune from
such effects when a first-order analysis is done by circuit de-
signers.
EALING with single-event transients (SETs) has proven
to be a complex problem. Transients in linear devices
Manuscript received July 14, 2006; revised September 6, 2006 This work
was carried out at the Jet Propulsion Laboratory, California Institute of Tech-
nology, under contract with the National Aeronautics and Space Administration
(NASA).
The authors are with the Jet Propulsion Laboratory, California Insti-
tute of Technology, Pasadena, CA 91109, USA (e-mail: allan.h.john-
ston@jpl.nasa.gov; tetsuo.f.miyahira@jpl.nasa.gov;
gov; jamie.laird@jpl.nasa.gov).
Digital Object Identifier 10.1109/TNS.2006.886215
farokh.irom@jpl.nasa.
Fig. 1. Simplified diagram of the LM137 voltage regulator.
ThispaperevaluatesSETsintworegulatorcircuitsthatcanbe
used over a wide range of voltage and load conditions. Circuit-
levelmodelsaredevelopedthatcanbeusedtoextendtestresults
to a wider range of application conditions.
II. DEVICES SELECTED FOR TESTING
A. Basic Properties
Two voltage regulators were selected for testing, the LM117
a positive voltage regulator, and the LM137, a negative voltage
regulator with similar properties. Both regulators can deliver
currents up to approximately 1 A (in the TO-39 package ver-
sions used in the study), with output voltages from 1.25 to 40
V. This allows a great deal of design flexibility, but it makes it
farmoredifficulttoestablishconditionsforradiationtestingthat
can be applied to the wide range of possible application condi-
tions.
The circuit design of the two regulators is very similar. They
contain an internal bandgap reference voltage and use a floating
currentsourcethatallowstheoutputvoltagetobeadjusted—rel-
ative to the bandgap reference—with an external resistor net-
work. Internal current limiting and temperature sensing circuits
are included to improve reliability [10], [11].
An example of the basic circuit configuration is shown in the
simplified diagram of Fig. 1 for the LM137 negative voltage
regulator. An internal bandgap reference in series with the ad-
justment pin and an external resistor network (not shown) de-
termines the output voltage. The circuit has an internal current
limiter that will shut down the regulator if the current through
0018-9499/$20.00 © 2006 IEEE
Authorized licensed use limited to: Jet Propulsion Laboratory. Downloaded on March 19, 2009 at 11:58 from IEEE Xplore. Restrictions apply.

Page 3

3456 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 53, NO. 6, DECEMBER 2006
Fig. 2. Details of current shutdown circuit for the two regulators.
the 0.04 ohm resistor exceeds approximately 1 A. Although ini-
tial sensing of a high current condition is done with an npn tran-
sistor, turn on of the lateral pnp transistor ultimately produces
the shutdown condition.
Although large output capacitancesare often used for voltage
regulators,theLM117doesnotrequireacapacitorattheoutput.
A 1
F capacitor is recommended at the output of the LM137
to increase stability and improve transient response, but lower
values can be used, particularly for low load currents. A wide
rangeofcapacitorvalues,loadcurrentsandinput/outputvoltage
are used in applications of these devices, all of which affect the
SET response. One of the motivations for this paper is a specific
application in flight hardware where the load capacitance of the
LM117 was only 0.1
F.
B. Current and Thermal Protection
The current protection circuit is more involved than shown
in the simplified diagram of Fig. 1. An additional transistor is
interposed between the sense resistor and the npn transistor to
reduce the temperature sensitivity of the current limit value. In
order to increase application flexibility the current limit is au-
tomatically reduced when the input-output voltage exceeds ap-
proximately 13 V by a zener diode network, shown in Fig. 2
[11]. Current through the zener diodes switches out part of the
multi-emitter transistor, reducing the current threshold from ap-
proximately 1.2 A at low voltage to about 300 mA for high
values of input-output voltage.
This has important consequences for radiation testing be-
causecurrentshutdowncanalsobetriggeredbyheavyions.The
sensitivity of the circuit to shutdown is substantially different
when the device is tested with high power supply voltages,
where the current limit is reduced through the automatic com-
pensation provided by the zener diode/multi-emitter network.
Radiation tests must be done over a range of input/output
voltage conditions in order to determine the sensitivity of the
shutdown circuit to heavy ions.
TABLE I
PROPERTIES OF IONS USED FOR TESTING
III. EXPERIMENTAL PROCEDURE
Heavy-ion tests were done at Brookhaven National Labora-
tory (in vacuum), and at Texas A&M (in air). The ions used for
testing are listed in Table I. The package lids were removed be-
fore testing. All tests were done using normal incidence. Addi-
tional diagnostic tests were done using
(also in vacuum).
Most tests were done with an output voltage of 5 V (or
V for the LM137) using appropriate resistor combinations at
the adjustment pin. The output load current was 20 mA, which
was well above the minimum start-up current required, but low
enough to avoid overheating the device during the tests when
the devices are in vacuum. Reed relays at the output were used
to change the capacitive load. Four different capacitance values
were used: 0.1, 0.32, 1.1 and 4.7 F. A Tektronix TDS784 os-
cilloscope was used to measure the waveforms, storing them
on an external computer for later analysis. The oscilloscope
trigger was set for a differential voltage of 0.2 V, capturing all
waveforms thatexceeded that value during eachrun. The output
voltage that will affect other devices is application dependent,
and is usually much larger than the 0.2 V threshold condition.
The set of waveforms could be analyzed afterwards for spe-
cific values of output voltage and time duration that are crit-
ical for specific applications, which is usually determined by a
worst-case analysis.
fission fragments
IV. TEST RESULTS
A. Output Transients
LM117 Positive Regulator: Positive- and negative-going
transients can occur in the LM117. Negative-going transients
(with the exception of those caused by shutdown and thermal
protection) have very short duration, typically
are quickly corrected by the emitter-follower output circuit.
Therefore the main emphasis of this paper is on positive-going
transients that not only persist for longer times, but can poten-
tially destroy other circuits if they are too large.
SETs from the LM117 depended strongly on capacitive
loading. Transients with output amplitude less than approxi-
mately 0.3 V are within the active operating range of the control
amplifier, and persist for time intervals that are of the same
s, and
Authorized licensed use limited to: Jet Propulsion Laboratory. Downloaded on March 19, 2009 at 11:58 from IEEE Xplore. Restrictions apply.

Page 4

JOHNSTON et al.: SINGLE-EVENT TRANSIENTS IN VOLTAGE REGULATORS3457
Fig. 3. Generalized waveform of positive-going transients for two different ca-
pacitive load conditions. Recovery does not begin until the waveform is close
to the initial voltage value.
Fig. 4. Cross section for SETs in the LM117 voltage regulator for various
values of load capacitance.
order as the normal electrical transient response (
Transients with higher amplitude go beyond the linear control
range,resultinginanextendedtimeresponsebecausetheoutput
stage is asymmetric, designedto source current, notto absorb it.
The general nature of those positive-going transients is shown
in the diagram of Fig. 3. The dashed line corresponds to a load
capacitor that is twice that of the waveform in the solid line. An
impulse-like positive response occurs for approximately 70 ns
that is higher for the condition of reduced load capacitance.
Recovery of the output waveform depends on the load cur-
rentandoutputcapacitance,withnearlyconstantslopethatcon-
tinues until the output voltage is within the linear range of the
control amplifier. At that point the voltage starts to recover with
adampedoscillatorybehaviorthatpersistsforseveralmicrosec-
onds. The oscillatory period is somewhat longer with higher
load capacitance. The amplitude is lower when tests are done
with a higher load capacitance, but the time interval for the ini-
tial“chargingperiod”isthesame.Themaineffectofhigherload
capacitance is to extend the linear recovery period, as shown.
A triggering threshold of 0.2 V was used to capture transients
during various test sequences. Fig. 4 shows the dependence of
the cross section on LET for transients above 200 mV. No tran-
sientswereobservedduringthetestsatBrookhaven—whichhas
ions with much shorter range—with the highest value of load
capacitance (4.7
F), but transients were observed for the 4.7
F condition during tests at Texas A&M, which has ions with
much longer range.
s).
Fig. 5. Maximum value of output transient observed for experimental runs
where several hundred events are captured.
The maximum value of the transient voltage increased with
LET; transients
V occurred even with a load capacitance
of 1.1
F for LET values of 26.6 MeV - cm
The maximum peak transient voltage that was observed during
a test sequence where
to 400 total events were observed
is shown in Fig.5. This is a usefulway to compare experimental
results because often the main concern is whether the transient
voltage will exceed maximum allowable voltage ranges for cir-
cuits that are powered by the voltage regulator. The maximum
amplitudeislowerforhighvaluesofloadcapacitance,asshown.
The difference between the input and output voltage was 15 V
for these measurements. The threshold LET was very low, well
below the approximate 15 MeV-cm
from protons are expected. Although we did not include proton
tests in our study, upset from high-energy protons is likely to
be important for these devices, particularly when they are used
with low output load capacitance.
LM137 Negative Voltage Regulator: Although the LM137
is a negative-voltage regulator, it uses a Darlington npn output
stage that is very similar to that of the LM117 with the input
and output terminals interchanged. Negative-going transients
were observed for the LM137 that closely parallel the posi-
tive transients from the LM117. The LM137 is less stable than
the LM117, and the damped oscillatory behavior during the re-
covery period persisted for about twice as long for the LM137
compared to the LM117.
The SET cross section for a load capacitance of 0.32
shown in Fig. 6, comparing results for two different threshold
conditions. The cross section is about a factor of five lower for
transients
Vcomparedtothelowerthresholdvoltagecon-
dition (0.2 V) that was used during the individual runs. The sat-
uration cross section and threshold LET of the LM137 are very
similar to that of the LM117 (Fig. 4), even though circuit design
details are somewhat different for the two devices.
and above.
value where upsets
F is
B. SEU Effects in Protection Circuits
Althoughtheprotectioncircuitsinthetworegulatorsarevery
similar,mostoftheexperimentalworkonSEUeffectswasdone
on the LM137, using a laboratory
Cf fission source. After
Authorized licensed use limited to: Jet Propulsion Laboratory. Downloaded on March 19, 2009 at 11:58 from IEEE Xplore. Restrictions apply.

Page 5

3458 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 53, NO. 6, DECEMBER 2006
Fig. 6. SET cross section for the LM137 negative regulator for two different
threshold conditions.
correction for passive over-layers, the range of the fission frag-
ments is approximately 10 m for “heavy” fragments and 12.5
m for “light” fragments. The depth of the n-isolation wells in
these circuits is approximately 12 m. One way to assign an ef-
fective LET for
ions is to integrate over the ion range,
yielding an effective LET of 21 MeV-cm
ments and 24.5 MeV-cm
for the light group. This is appli-
cable to npn and lateral pnp transistors where charge collection
in the substrate is cut off by the buried layer, but not for sub-
strate transistors where charge collection can extend beyond the
isolation well into the substrate. Fortunately, substrate transis-
tors are not directly involved in the current shutdown circuitry
(see Figs. 1 and 2), and we expect that the tests with
a reasonable simulation of heavy ion effects once the correction
is made for ion range.
Current Protection: Tests of the current overload protection
circuitweredoneundervariousloadconditions,usingathermo-
electric cooler to limit the temperature rise in the device when
it was operated at high load current. The LM137 regulator was
in a vacuum chamber for those experiments. Device tempera-
ture was monitored with a thermocouple. When the load current
was increased to approximately 550 mA the
could trigger the current protection circuit. A typical waveform
is shown in Fig. 7. A dropout occurs at the output with widths
from about 0.5 to 1.4 ms when the current protection circuit is
tripped.
The current protection SEU mode does not depend on output
capacitance. The cross section, measured with
cm . The area of the collector of the npn transistor in the
currentlimitcircuitis
anism where that transistor is turned on by the ion. A rough es-
timate of the event rate for galactic cosmic rays in space can be
made by assuming that the LET threshold is 10 MeV-cm
and that the saturation cross section is the same as the cross
section measured with
Cf. The resulting estimate is an upset
probability of 0.5% per year (at solar maximum).
It was only possible to trigger current limiting when the regu-
lator was drivinga relativelyhighload current. Thus,this mech-
anismisunlikelytobeimportantatmoderateloads,eventhough
it is important for applications with high loads. A SPICE anal-
ysis of the circuit was done to calculate the effect of different
for heavy frag-
are
fragments
Cf, was
cm ,consistentwiththemech-
,
Fig. 7. Output voltage of the LM137 when the internal current protection cir-
cuit is triggered by californium ions.
Fig. 8. SPICE analysis showing how the base-emitter voltage of the npn tran-
sistor in the current limit circuit of the LM137 is affected by output current.
load currents on the threshold conditions for this mechanism.
Those results are shown in Fig. 8, assuming an internal shunt
resistor of 0.04 ohms. The ordinate shows how the base-emitter
voltageofthenpntransistorinthecurrentlimitstage(seeFig.2)
is affected by load current.
Thermal Protection: A clever circuit technique is used in
both regulators that will shut down the regulator if a sudden
thermal overload occurs. This is incorporated into the current
protection circuit by locating the two transistors in the current
protectioncircuitindifferentlocationsonthedie.Onetransistor
is close to the large-area output transistor, where most of the
power is dissipated, and the other is at the opposite edge of the
die. A large difference in localized temperature will activate the
shutdown circuit within a few milliseconds [11].
Thermal shutdown could also be triggered by heavy ions.
Testsofthethermalprotectioncircuitweredonebyallowingthe
devicetoheatuptotemperaturesabove
with
Cf. When the ions were not present the device operated
normally. Thermal shutdown caused the device to turn off for
much longer periods—on the order of seconds—compared to
,andirradiatingit
Authorized licensed use limited to: Jet Propulsion Laboratory. Downloaded on March 19, 2009 at 11:58 from IEEE Xplore. Restrictions apply.