Power and Batteries

Energy storage, power semiconductors and radioisotope sources used by planetary and orbital robotics programs, with the qualification campaign or the flight result that establishes each one.

One cell type behind a planetary robot has a published qualification campaign: the lithium-ion cell of the Mars Exploration Rover battery [1]. Everything else in the class is characterized rather than qualified. Power semiconductors appear in radiation compendia as screening results at a named beam [2][3][4], and the radioisotope generators are published either as a design value or as a procurement requirement, never as a measured output [7].

Cells qualified for a planetary robot

Part	Manufacturer	Used by	Source
10 Ah lithium-ion cell	Lithion/Yardney	spirit, opportunity	[1]
8-cell Rover Battery Assembly Unit	Lithion/Yardney	spirit, opportunity	[1]

Ratings and qualification results

• 10 Ah lithium-ion cell, Lithion/Yardney: eight cells per Rover Battery Assembly Unit, cycled 3.0 to 4.1 V [1]. Ratings: the battery had to hold 24 to 36 V, deliver 220 Wh at launch and at least 283 Wh per sol at 0 C on the surface, take 270 cycles at 50 percent depth of discharge or 90 sols, and pass multiple 30 A pulses of 50 ms over the -20 to +30 C range. Those are the requirements the design was written against, quoted without the power system analysis behind them, and no unit demonstrated them as a set [1]. Qualification: at JPL, C/5 discharge at 1.60 A with the cell charged at the discharge temperature delivered over 123 Wh/kg at 23 C and over 97 Wh/kg at -20 C, the cold case being over 82 percent of the room-temperature capacity [1]. Charging warm and discharging cold gave more capacity in every case. High-current pulse capability for pyrotechnic events was demonstrated at 0 C as 30 A pulses of 50 ms held above 3.0 V at 100, 75 and 50 percent state of charge, in a rapid succession of about 42 pulses at 21 to 30 A. Below 50 percent state of charge the end-of-discharge voltage falls further, which is why the battery went into entry fully charged [1].
• 8-cell Rover Battery Assembly Unit, Lithion/Yardney: a flight-configuration engineering battery on the bench [1]. Ratings: eight cells in series, manually balanced by resistive discharge on the ground unit where the flight unit had active cell balancing. Qualification: permanent capacity loss after simulated cruise plus 90 sols was under 4 percent, the figure showing 10.026 Ah falling to 9.690 Ah, or 3.4 percent [1]. Cruise was simulated by storage on the bus at 30.40 V and 10 C at 70 percent state of charge, and the capacity check ran at 20 C, C/5 charge and discharge, 32.40 V charge and 24.0 V cut-off. The loss reads worse when checked at -20 C, which the authors attribute to impedance [1].

What the campaign does not establish

Every number above is a bench measurement on prototype or engineering hardware. No telemetered capacity, voltage or impedance from Spirit or Opportunity appears in the source, so the flight battery is reached by analogy with a ground article [1]. Three figures need reading with that in mind.

• Cycle life. Over 900 cycles at 100 percent depth of discharge with over 80 percent of initial capacity retained, at 23 C at C/5 with a C/100 taper cut-off, is one named cell, YL094, on a test still running when the paper was written [1]. It bounds the life from below and carries no spread, so it is not a qualified life for the type.
• Impedance growth. Cells rose 30 to 40 percent over cruise plus 90 sols by current interrupt at 0 C, 5 A for one minute with the impedance computed an hour later [1]. No absolute impedance is given, and the authors state the derived value depends strongly on the discharge current and duration used, so the number describes the trend rather than the part.
• The thermal profile was wrong in the direction that matters. The sol profile was built for a 0 to -20 C range and the rovers ran warmer, by an amount the paper does not quantify [1]. One battery was tested, without the flight balancing.

A separate 8-cell 25 Ah lander battery, MSP01 hardware rather than MER, lost 0.6 percent of capacity permanently over 11 months of simulated cruise storage, 29.085 Ah falling to 28.900 Ah, stored connected to the bus at 10 C and float charged at 3.875 V per cell to hold about 70 percent state of charge [1].

Power electronics characterized against radiation

Screening results against a named beam and facility. None of these parts carries a published flight qualification result on a planetary robot.

Part	Manufacturer	Facility	Source
SGF15E100 GaN HEMT	SSDI	Texas A and M cyclotron	[2]
SI9112DY switch-mode controller	Vishay Siliconix	Texas A and M cyclotron	[3]
EN5322QI 2 A buck regulator	Altera	Texas A and M cyclotron	[4]
TPS7A3001-EP linear regulator	Texas Instruments	Texas A and M cyclotron	[4]
HSYE-117RH voltage regulator	Intersil	Goddard gamma irradiator	[2]
MIC4427 MOSFET gate driver	Microchip	Goddard gamma irradiator	[2]
EPC2019 eGaN FET	EPC	Ohio State reactor	[2]

Ratings and campaign results

• SGF15E100 GaN HEMT, SSDI. Heavy ion at the Texas A and M cyclotron, 15 MeV/u tune, gate-source voltage 0 [2]. Qualification: last pass 300 V and first fail 350 V drain-source at normal incidence with silver at surface LET 42 MeV-cm2/mg [2]. The threshold is not a scalar. At 45 degree tilt with copper at LET 20 MeV-cm2/mg in silicon, the part survived 500 V with the ion track perpendicular to the two-dimensional electron gas channel and burned out at 500 V with the track parallel to it [2]. A burnout voltage quoted without the track orientation does not bound the part.
• SI9112DY switch-mode controller, Vishay Siliconix. Heavy ion at the Texas A and M cyclotron. Qualification: single-event latchup LET threshold 23 to 27 MeV-cm2/mg at 25 C, falling to 11 to 19 MeV-cm2/mg at 85 C [3]. A part screened at room temperature has been screened at its easiest condition.
• EN5322QI 2 A synchronous buck regulator, Altera, heavy ion at the Texas A and M cyclotron [4]. Qualification: destructive latchup at LET 42.8 MeV-cm2/mg at room temperature after a fluence of 7.5e6 ions/cm2, with a permanent supply current increase; at elevated temperature the part failed at 4.2e6 ions/cm2 [4].
• TPS7A3001-EP linear regulator, Texas Instruments. Heavy ion at the Texas A and M cyclotron. Qualification: no latchup to LET 42.8 MeV-cm2/mg, which was the highest LET available in that campaign [4]. The companion TPS7A4700 low-dropout regulator returned the same bound.
• HSYE-117RH voltage regulator, Intersil. Cobalt-60 at the Goddard gamma irradiator, low dose rate. Qualification: line and load regulation out of specification at 75 krad(Si) [2].
• MIC4427 MOSFET gate driver, Microchip. Cobalt-60 at the Goddard gamma irradiator, low dose rate. Qualification: functional failure at 20 krad(Si) in one of eight devices, with rise time out of specification at the same level [2].
• EPC2019 eGaN FET, EPC. Neutron displacement damage at the Ohio State reactor with cobalt-60 dose. Qualification: on-resistance increase held under 10 milliohm through 4.3e14 n/cm2 at 1 MeV equivalent plus 1.2 Mrad(Si) [2].

The compendium entries are two or three parts each, and the authors warn that behavior varies with application and between lots [3]. Several of the latchup results are one-sided bounds set by the highest LET the beam could reach, 42.8 MeV-cm2/mg, which is below the 75 to 85 MeV-cm2/mg customarily used for heavy-ion qualification [4].

A total ionizing dose number belongs to a dose rate and a bias

The regulator and gate driver figures above are low-dose-rate results, and the distinction is not bookkeeping. Bipolar and BiCMOS parts can degrade further at low dose rate than at high, which MIL-STD-883 Method 1019 handles with a Condition D default rate of 10 mrad(Si)/s [8]. One JPL campaign ran the Texas Instruments DAC5675A-SP 14-bit DAC at three rates under cobalt-60: a manufacturer capability of 150 krad(Si) taken at 2.53 rad(Si)/s with five parts all in specification, a last pass level of 30 krad(Si) when re-run static biased at 10 mrad(Si)/s with two of five parts failing functionally after 35 krad and a third at 40 krad, and 150 krad(Si) again at 150 mrad(Si)/s [5]. The effect appears between 150 and 10 mrad(Si)/s rather than scaling with rate throughout, and the authors do not claim 30 krad(Si) as the threshold, because the low tail of the distribution may extend below it on five parts [5]. Europa Clipper set its own low dose rate at about 45 mrad(Si)/s, derived from the mission dose profile rather than from the Condition D default, because a Condition D campaign to 300 krad(Si) would have taken a year [8].

Bias state moves the answer as far as dose rate does. The NVE IL815T giant magnetoresistive digital isolator degrades parametrically and functionally at 100 krad(Si) at 100 mrad(Si)/s and 5 V with sync pulled high and output enable grounded, with supply currents and timing out of specification; both pins pulled to 5 V is the worst case, and the recommendation is to fly with both grounded [5]. The same dependence decides a part’s usable dose in the Europa Clipper campaign: the ST Micro RHR61 operational amplifier failed functionally at 62 krad(Si) when biased as a comparator with a large input differential, because internal bias current grows with that differential [8]. The part’s dose capability is therefore a property of the board around it. Every result in that campaign is one wafer lot, one bias circuit and room temperature [5].

Radioisotope power

The Multi-Mission Radioisotope Thermoelectric Generator on curiosity and perseverance dissipates 1830 W of waste heat at beginning of life alongside about 110 W electrical [6]. Both figures are design values carried into a flight thermal analysis rather than a measured output, and the waste heat is a thermal design input before it is a power one. The same 110 W appears in the program literature with no source, temperature or fin-root condition attached [7].

Nothing comparable is published for the Next-Generation RTG. What the literature gives is the procurement envelope posted in Idaho National Laboratory sources-sought notice INL-18-012: up to 500 W electrical for the largest variant at 16 GPHS Step-2 modules, internal hot junction below 1100 degC, sink at -269 degC, fueled system mass below 60 kg, 17 year design life, 22 to 34 VDC flight envelope, fin root 50 to 200 degC, and a primary mode above 50 Hz [7]. These are requirements a bidder has to meet. No hardware existed when they were posted, and the document is marked pre-decisional [7]. The enhanced MMRTG is at the same stage: at least 77 W at the end of a 17 year design life is a project requirement, and the roughly 600 degC skutterudite hot-side figure is a design operating point from a life-prediction model, with no couple life test result reported [7].

Regulators, drivers and references screened at a named beam

A second group of power parts from the same compendia, each a single-event screen rather than a qualification [3][8].

Part	Manufacturer	Facility	Source
HS-117RH voltage regulator	Intersil	Texas A and M cyclotron	[3]
HS-4423BRH MOSFET driver	Intersil	Texas A and M cyclotron	[3]
RH-117H voltage regulator	Linear Technology	Texas A and M cyclotron	[3]
LS2803R3S DC/DC converter	International Rectifier	Texas A and M cyclotron	[3]
JANTX2N6790 power MOSFET	International Rectifier	Texas A and M cyclotron	[8]
2N7480 power MOSFET	Microsemi	Texas A and M cyclotron	[15]
SG1524BJ PWM controller	Microsemi	Texas A and M cyclotron	[8]
CGH40120F, CGH40180PP GaN HEMT	Cree	Texas A and M cyclotron	[8]
OP497 quad op amp	Analog Devices	Texas A and M cyclotron	[8]
LM111 voltage comparator	National Semiconductor	Texas A and M cyclotron	[8]
LTC2904ITS8 supply monitor	Linear Technology	Heavy ion, facility unnamed	[4]
LTC6655BHLS8 voltage reference	Linear Technology	Heavy ion, facility unnamed	[4]
LMP2022 op amp	Texas Instruments	Heavy ion, facility unnamed	[4]
OLS500SB optocoupler	Skyworks	UC Davis cyclotron	[2]
HSSR-7111 opto isolator	Broadcom	63 MeV protons	[16]

Ratings and campaign results

• HS-117RH voltage regulator, Intersil. Radiation-hardened adjustable positive regulator. Ratings: the hardened counterpart of the HSYE-117RH gamma result above. Qualification: positive single-event transient threshold below 4 MeV-cm2/mg [3]. A hardened part that transients at 4 MeV-cm2/mg is hardened against dose, not against ions.
• HS-4423BRH MOSFET driver, Intersil. Radiation-hardened dual MOSFET gate driver. Ratings: hardened process. Qualification: dynamic transient threshold below 84.6 MeV-cm2/mg with a 5 V 1 kHz square wave input and an average transient width of 23 microseconds, and no transient at all with the inputs held static at 0 or 5 V [3]. The part is only sensitive while it is switching, so a static bias screen would have passed it.
• RH-117H voltage regulator, Linear Technology. Radiation-hardened adjustable regulator. Ratings: bipolar. Qualification: positive transient threshold below 2.7 MeV-cm2/mg [3].
• LS2803R3S DC/DC converter, International Rectifier. Hybrid point-of-load converter. Ratings: full input voltage range and load tested. Qualification: transients under 20 mV at 51.5 MeV-cm2/mg with no latchup [3].
• JANTX2N6790 power MOSFET, International Rectifier. Discrete n-channel power MOSFET. Ratings: gate-source held at 0 V for the test [8]. Qualification: gate rupture and burnout onset at 130 V drain-source under krypton at 37 MeV-cm2/mg, falling to 70 V under xenon at 50 [8]. The safe operating voltage nearly halves between two ions in the same campaign, so a single-ion result is not a derating.
• 2N7480 power MOSFET, Microsemi [15]. A 60 V n-channel part rated 22 A, on the RH2 hardened process. Ratings: three samples tested. Qualification: gate rupture and burnout safe operating area measured with undegraded silver at a surface transfer of 41 MeV-cm2/mg at normal incidence and room temperature, gate-source swept from 0 to -16 V, published as a curve rather than a threshold [15]. The part performed better than earlier fabrication lines of the same design and agreed with the manufacturer’s own data.
• SG1524BJ PWM controller, Microsemi. Switching regulator controller. Ratings: worst-case bias at 25 C. Qualification: no latchup, gate rupture or burnout above 84.6 MeV-cm2/mg over 1e7 ions/cm2 of 15 MeV/amu gold, with duty-cycle shifts and transients observed [8].
• CGH40120F, CGH40180PP GaN HEMT, Cree. Commercial radio-frequency gallium nitride high electron mobility transistors. Ratings: not qualified for space. Qualification: no gate rupture and no burnout observed under heavy ion [8]. The compendium reports no threshold, so the entry is a clean screen and not a bound.
• OP497 quad op amp, Analog Devices. Precision bipolar quad operational amplifier. Ratings: supplies of plus and minus 5, 10 and 15 V. Qualification: transient threshold 1 MeV-cm2/mg with a saturated cross section of 1e-2 cm2, the same at all three supplies [8].
• LM111 voltage comparator, National Semiconductor. Commercial bipolar comparator. Ratings: worst case is the highest supply, 15 V, and a large differential input of 4.75 V because of its emitter-follower input [10]. Qualification: transient threshold below 28 MeV-cm2/mg with a saturated cross section about 2e-3 cm2 at 83 [8]. Worst-case saturated cross section is about 1e-3 cm2/device with a threshold under 3 MeV-cm2/mg, falling to 8e-6 cm2/device at 20 mV overdrive and a 5 V supply, and raising the part to 76 C did not move it [10]. About one part in three from a single wafer lot showed anomalously large input bias current degradation under gamma, a bimodal distribution the compendium answers by recommending larger lot acceptance samples [10].
• LTC2904ITS8 supply monitor, Linear Technology. Dual supply voltage monitor. Ratings: commercial. Qualification: no latchup above 42.8 MeV-cm2/mg, which was the highest transfer available in that campaign [4].
• LTC6655BHLS8 voltage reference, Linear Technology. Precision series voltage reference. Ratings: commercial. Qualification: destructive latchup observed [4]. The campaign publishes the outcome without a threshold, so the part is disqualified rather than characterized.
• LMP2022 op amp, Texas Instruments. Commercial precision operational amplifier. Ratings: commercial. Qualification: no latchup above 42.8 MeV-cm2/mg [4].
• OLS500SB optocoupler, Skyworks. Optically coupled isolator. Ratings: parameters measured against specification. Qualification: all parameters in specification to 2e11 protons/cm2 of 64 MeV protons at the UC Davis Crocker Nuclear Laboratory [2].
• HSSR-7111 opto isolator, Broadcom. Solid state relay optocoupler. Ratings: degradation depends on the forward voltage the part is run at. Qualification: displacement damage degradation grows beyond 30 krad(Si) under 63 MeV protons [16].

A destructive event that was not what it was called

The Analog Devices RH3845 synchronous step-down controller was carried into the Europa Clipper power subsystem, and its destructive single-event failure was first recorded as burnout in the part. Testing at the Lawrence Berkeley 88-inch cyclotron with the input and switch-node pins tied together at 41 V or above gave a destructive threshold under 20 MeV-cm2/mg and a saturated cross section near 7e-5 cm2/device [10]. Decoupling those pins moved the average onset to 44 V over five flight-lot devices at 38 MeV-cm2/mg, with almost no change between -5 C and room temperature, which is the evidence against burnout: burnout is strongly temperature dependent. The mechanism is transient-induced cross conduction in the external synchronous FETs, and raising the in-line gate drive resistance from 2 to 10 ohms removed the destructive failure entirely up to 100 V at 79 MeV-cm2/mg [10]. The fix is a board-level gate drive choice, not a part derating.

Two further results on the same part bound its use. A non-destructive functional interrupt starts at 38 MeV-cm2/mg, in which the controller sticks in a linear state with the switch node near 0 V and the output rising to about 8 V against a 5 V nominal; a pre-load current of 10 mA eliminated it over 1e7 ions/cm2 at 79 MeV-cm2/mg on five devices [10]. A separate cross-conduction event has an onset transfer of 0.1 MeV-cm2/mg, low enough that the authors flag the device as potentially proton sensitive, with a calculated galactic cosmic ray rate of 0.3 events per device-year at 1 AU that is a prediction and not a Europa figure. Under low dose rate the part stayed functional to 300 krad(Si), the only non-conformance being bias supply current slightly out of specification from 250 krad(Si) [10].

Parts qualified against a 300 krad total dose requirement

The Europa Clipper vault sees 2.7 Mrad(Si) unshielded and 150 krad(Si) behind 500 mils of aluminum, and parts are required to hold 300 krad(Si) at a radiation design factor of two [9]. All results below are cobalt-60 at the JPL low-dose-rate room irradiator, at a mission-derived 45 mrad(Si)/s.

Part	Manufacturer	Result	Source
HS-117EH voltage regulator	Renesas	Above 300 krad(Si)	[9]
ISL70003ASEH buck regulator	Renesas	Above 300 krad(Si)	[9]
ISL74422BRH MOSFET driver	Renesas	Above 300 krad(Si)	[9]
HS-508BEH analog multiplexer	Renesas	Above 300 krad(Si)	[9]
RH1013MW dual op amp	Analog Devices	Above 300 krad(Si)	[9]
JANSF2N2857 NPN transistor	Semicoa	Above 300 krad(Si)	[9]
HS-OP470AEH quad op amp	Renesas	Parametric at 100 krad(Si)	[9]
IS-139ASEH quad comparator	Renesas	Parametric at 200 krad(Si)	[9]
HS-303AEH analog switch	Renesas	Parametric at 200 krad(Si)	[9]
JANSF2N2907A PNP transistor	Semicoa	Parametric at 200 krad(Si)	[9]
RHR61 rail-to-rail op amp	ST Micro	Parametric at 50 krad(Si)	[9]

Ratings and campaign results

• HS-117EH voltage regulator, Renesas. A 40 V positive adjustable regulator [9]. Ratings: biased at 15 V for the test. Qualification: above 300 krad(Si) at 42 mrad(Si)/s with no parametric or functional failure, against a source-control drawing that guarantees only 50 krad at low dose rate [9]. The drawing understates the part by a factor of six, which is what the campaign was run to find out.
• ISL70003ASEH buck regulator, Renesas. A 9 A synchronous buck point-of-load regulator [9]. Ratings: tested biased at 13.2 V and unbiased. Qualification: above 300 krad(Si) with no parametric or functional failure, again against a 50 krad drawing guarantee [9].
• ISL74422BRH MOSFET driver, Renesas. Radiation-hardened gate driver [9]. Ratings: 15 V, biased and grounded. Qualification: above 300 krad(Si) [9].
• HS-508BEH analog multiplexer, Renesas. Eight-channel analog multiplexer [9]. Ratings: plus and minus 15 V, biased and grounded. Qualification: above 300 krad(Si) [9].
• RH1013MW dual op amp, Analog Devices. Dual precision operational amplifier. Ratings: biased and unbiased. Qualification: above 300 krad(Si) at low dose rate with no parametric or functional failures [9].
• JANSF2N2857 NPN transistor, Semicoa. Discrete NPN bipolar transistor. Ratings: biased and unbiased at 45 mrad(Si)/s. Qualification: above 300 krad(Si) with no failures [9].
• HS-OP470AEH quad op amp, Renesas. Quad precision operational amplifier. Ratings: unbiased is the worst case for this part. Qualification: bias current and open-loop gain out of specification at 100 krad(Si), with no functional failure to 300 krad [9].
• IS-139ASEH quad comparator, Renesas. Radiation-hardened quad voltage comparator. Ratings: dielectrically isolated bipolar with triple-redundant comparators and hardened majority voting [11]. Qualification: bias current and open-loop gain out of specification at 200 krad(Si), open-loop gain saturating at 12 dB from 260 krad, no functional failure to 300 krad [9]. Its output transients disappear above 5.8 mV of differential overdrive, tested to 83.9 MeV-cm2/mg with gold; below that, transients about 2 microseconds long appear [11].
• HS-303AEH analog switch, Renesas. Dual single-pole double-throw analog switch. Ratings: plus and minus 15 V. Qualification: turn-on and turn-off time out of specification at 200 krad(Si), and functional failure unbiased between 220 and 250 krad(Si) with no recovery after a 48 hour anneal at 50 C [9].
• JANSF2N2907A PNP transistor, Semicoa. Discrete PNP bipolar transistor. Ratings: biased at 12 V. Qualification: saturation voltage out of specification at 200 krad(Si), functional beyond 300 krad [9].
• RHR61 rail-to-rail op amp, ST Micro. Radiation-hardened operational amplifier. Ratings: the failure level depends on how the part is wired. Qualification: supply current out of specification at 50 krad(Si) at 100 mrad(Si)/s in a generic op amp bias, functional failure at 62 krad(Si) in a comparator bias configuration and at 150 krad(Si) as a voltage follower, both at about 40 rad(Si)/s [9]. There is no recovery after a 168 hour room-temperature anneal, and biased parts were worse than unbiased.

Three of the parts above are guaranteed 50 krad(Si) by their source control drawings and measured above 300 [9]. The gap is the reason the campaign exists, and it runs in one direction only for these lots: a drawing limit is a floor the vendor will defend, not an estimate of the part.

A cold-survivable motor controller module

The Cold Survivable Distributed Motor Controller is a 0.3 kg, 10 by 10 by 3 cm module driving one 3 A brushless DC motor with two resolver channels, a brake driver and point-of-load regulation from a 28 V bus [14]. Its stated survival range is -180 to +125 C in storage without survival heating, against a -55 to +85 C operating range, which is a design goal with no thermal cycling result reported. The gallium nitride buck converter at its point of load holds regulation at 2 A until 256 krad(Si), where the output drifts out of specification [14]. Adding three tantalum shield elements is credited with about another 100 krad in simulation, and the resulting 356 krad(Si) figure is an engineering expectation: no irradiation was performed at that level [14].

Cells and solar cells with a flight or life record

• LO-26S SX-2 lithium sulfur dioxide D cells, Duracell [12]. Ninety-two battery packages flown on the Long Duration Exposure Facility for 69 months. Ratings: primary cells powering experiments to their designed loads throughout. Qualification: about 30 percent of initial capacity lost in flight against 11 percent on ground controls, which the analysis attributes to storage temperature rather than to the space environment, the controls never exceeding 40 F while the flight batteries cycled between 41 and 95 F [12]. Every LiSO2 battery met its load, so the capacity difference is a shelf-life result and not a failure.
• AO171 SAMPLE silicon solar cells, built at Marshall Space Flight Center. Silicon cells on Kapton polyimide substrates, flown on the same facility. Ratings: seventeen cells measured after retrieval against Lockheed pre-flight data. Qualification: maximum power point degraded 6.5 percent on average with a standard deviation of 1.75, and cell-to-cell interconnects survived the thermal cycling with no debonding from the contact pads [11]. Two of seven modules were lost from the spacecraft outright because atomic oxygen eroded the polyimide substrate holding them, so the failure was the carrier and not the cell [11].
• 7 Ah lithium-ion cell cycled inverted, Yardney [13]. Screening article for the Mars Exploration Rover battery orientation decision, one sample. Ratings: seven amp-hour cells, not the ten amp-hour flight design. Qualification: one cell failed after about 630 cycles at 100 percent depth of discharge inverted, and destructive physical analysis found no detectable cause [13]. The accelerated follow-up showed no orientation effect over about 1000 cycles at 40 C or 300 days of float at 25 C, and a cell failed after 115 days of float at 55 C with rising self-discharge, a temperature outside any rover condition. The result nonetheless set the flight rule, which is to fly the battery inverted during launch only and upright on the surface.

Two more analog parts, and a fabrication line change

• MAX4595 analog switch, Maxim. Ratings: characterized by a Weibull fit rather than a single threshold [17]. Qualification: single event transient threshold 32.1 MeV-cm2/mg with a limiting cross section of 2e-3 cm2, Weibull shape 2.5 and width 71, at the Texas A and M cyclotron [17]. The predicted transient rate is 1.55e-4 per device-day in geostationary orbit and 4.80e-1 on the October 1989 worst day, a spread of three and a half orders of magnitude on the same part.
• RHR64 quad rail-to-rail op amp, ST Micro. Ratings: the datasheet rating is 100 krad(Si) at high dose rate [5]. Qualification: parametric failure of supply current and input offset voltage at 100 krad(Si) at 100 mrad(Si)/s in a voltage follower, with severe failures at 200 krad, and functional failure at 62 krad(Si) in a comparator bias against 150 krad(Si) as a voltage follower, both at about 40 rad(Si)/s [5]. There was no recovery after a 168 hour anneal. The part meets its datasheet number in one bias configuration and fails at two thirds of it in another.

Moving a part number to a different fabrication line moves its single-event response, and not always in the same direction. Twenty-five devices each of two International Rectifier power MOSFETs from the Temecula line were tested against prior data on the same part numbers from El Segundo [18]. The IRHN57133SE, a 130 V 20 A device in an SMD-0.5 package, degraded, taking events at lower drain-to-source voltage under krypton at 49 and 59.8 MeV-cm2/mg with the beam tuned to the surface transfer used in the manufacturer’s earlier work. The IRHN57250SE, a 200 V 31 A device in SMD-1, moved the other way and improved: inducing any event with bromine at 41.3 or krypton at 27.1 MeV-cm2/mg required biasing outside the manufacturer’s absolute maximum ratings, so the defensible statement is immunity to about 37 MeV-cm2/mg within ratings, while gold at 84 does produce events inside them [18]. The authors state that the statistics do not support attributing either change to the line move.