The deep-space power bus where the lighter option lost

What the work claims

This is a Phase A design-analysis paper, not a test report. Kimnach and Soltis, of NASA GRC, document the power management and distribution (PMAD) trade studies for a deep-space scientific spacecraft with a nominal 100 kWe electrical requirement under Project Prometheus's Nuclear Systems Program: a 20 kWe science payload and an ion propulsion load demanding up to 100 kWe.¹ They compare AC primary distribution (at 1000 Hz and 1500 Hz) against DC at various voltages, on efficiency, reliability, mass, thermal behavior, corona and radiation tolerance, and component technology readiness, and they recommend a dual-channel AC system fed by two 3-phase, 110 kVA alternators, each normally supplying 50 kWe but able to carry the full 100 kWe after a single failure.¹

The punchline, stated in the paper's own conclusions: the DC distribution channel comes in at approximately 218 kg versus 257 kg for AC, yet the AC topology is selected because its greater efficiency and lower control complexity outweigh the mass penalty.¹

How it works

The source is a pair of closed Brayton-cycle turbo-compressor/alternators mounted at one end of a 30 meter boom, with the PMAD bus and loads another 10 meters along. The architecture is dual everything: two independent sources, two buses, ten power processing units (PPUs) each driving two 10 kWe ion thrusters, a 20 kWe science and communications bus, and a 12 kWe auxiliary bus.¹ The alternator trade settled on a permanent-magnet synchronous machine rated 110 kVA, normally operated at 55 kVA, 400 Vrms, 1 kHz; a wound-field machine was rejected because a single-event upset that kills field current produces counter-torque and total loss of generation, a failure mode the PM machine does not have.¹ A parasitic load resistor bank, rated 125 kWe (25 kWe above the full load requirement), absorbs power whenever the thrusters throttle back, since a Brayton alternator cannot run unloaded.¹

The AC-versus-DC comparison is the core of the paper, and the harness numbers are its hardest data. For the 30 meter primary run sized to a 1% voltage drop per MIL-STD-975, the AC option uses twelve #6 AWG cables (four per phase, no neutral) weighing 55 kg, with I2R losses of 182 W at 50 kWe and 324 W at 100 kWe.¹ The DC option, at 400 V, needs eighteen #6 AWG cables (nine supply, nine return) at 83 kg, dissipating 198 W at 50 kWe but 783 W at 100 kWe.¹ The loss divergence at full power follows directly from the currents: 250 A on DC against 106.9 A per AC phase.¹

Voltage is the obvious escape from both mass and loss, and the paper quantifies it: the same primary harness drops from 55 kg at 400 Vrms to 22 kg at 1000 Vrms and about 2 kg at 10 kV.¹ It then declines to go there, for two stated reasons. Space-qualified switchgear of the period existed only to about 270 Vdc, and Paschen breakdown in the partial-pressure gas of low orbit plus potting and corona constraints argued for a conservative ceiling; the team selected 400 V nominal, 600 V maximum.¹ Frequency is the other lever. Transformer mass scales roughly as the inverse three-quarter power of frequency (the paper cites Schwarze, 1984), so the 600:400 V autotransformer designed for 1.5 kHz lands at about 37 kg, and the 11 kVA PPU transformers at 1 kHz at about 20 kg each.¹ Running the machine 1.5 times faster to hold 400 V at 100 kWe would have forced a 165 kVA design, so the study instead steps the bus to 600 V for the full-power mode (alternator points: 50 kWe at 400 V and 1000 Hz, 100 kWe at 600 V and 1500 Hz, 20 kWe at 388 V and 800 Hz, 10 kWe at 356 V and 920 Hz).¹

Two more trades close the loop. Active power-factor correction at every load could cut the primary harness from 55.4 kg at a 0.907 power factor to 43.6 kg fully corrected, but the paper judges the added electronics, in a radiation environment of roughly 4.3 Mrad gamma and 6x10^12 neutrons/cm2 over a 12-year mission behind 100 mil of aluminum, a substantial reliability risk for a 12 kg saving.¹ And on protection, AC wins decisively: the PPUs' self-commutating solid-state switches clear a grid fault within one cycle, AC switchgear uses inverse I-squared-t trips that a slow alternator can feed, while DC switchgear must actively current-limit, running hotter, with fuses ruled out as non-resettable.¹

The strongest case for it

The study's discipline is exemplary. Every comparison carries its measurement boundary: harness masses are for the 30 meter primary run alone with a stated 1% drop budget, the total system figures of 257 kg (AC) and 218 kg (DC) explicitly exclude the sources, switchgear mass models come from commercial off-the-shelf datasheets, and the authors flag that AC versus DC switchgear masses were not taken from optimum designs.¹ The voltage-ladder data point alone, 55 kg at 400 V against 2 kg at 10 kV with the component-availability reason for stopping where they did, is the kind of quantitative boundary that most architecture debates skip.¹

Just as strong is the refusal to launder component risk into the mass optimum. The PFC trade says, in effect, that 12 kg of copper is cheaper than the qualification burden of radiation-hardened active electronics at every load, and the corona constraint says the multi-kV harness is a paper optimum that the 2004 component base could not fly.¹ Those are the correct kind of reasons to reject a lighter design, and the paper states them openly.

Where a skeptic should push

Every number in this paper is analytical. There is no hardware, no thermal-vac test, no measured efficiency; the mass estimates are spreadsheet models fed by COTS datasheets and the efficiency comparisons are nominal design points.¹ That matters most for the 39 kg headline gap, which rests partly on switchgear mass approximations the authors themselves call non-optimum, and on the 270 Vdc component ceiling, which is a snapshot of 2004 technology, not a law of physics.¹

Second, the mission never flew. Project Prometheus was cancelled in 2005, so this architecture was never validated against integration reality, and the load of evidence is a single Phase A study plus the referenced power-conversion analysis for the Jupiter Icy Moons Orbiter.¹² Third, the thermal picture is asserted rather than modeled in this document: the 783 W of DC harness loss at 100 kWe, and the higher dissipation of DC switchgear, are stated as disadvantages but no thermal balance quantifies them.¹ Finally, the study's launch-cost framing (hardware mass valued at roughly $10,000 per kg, its stated circa-2004 figure) dates the economics; the mass-versus-complexity trade shifts as launch prices and component bases move.¹

Mass versus fault clearing in space power buses

The non-obvious implication: for high-power space buses, the mass-optimal architecture is routinely not the operable one, and the deciding variables live at the protection layer, not in the harness. Kimnach and Soltis's Tables 1 and 2 show DC losing the mass contest by 28 kg on cable alone yet dissolving that advantage at full power, where its 783 W of harness loss against AC's 324 W is a 460 W tax on exactly the propulsion mode where every watt is propellant.¹ Their Figure 3 adds the mechanism most designers underweight: transformer mass falling as frequency to the minus three-quarters means kilohertz-class AC distribution keeps magnetics small while giving passive, autotransformer-based voltage conversion that DC can only match with active converters at every load.¹

The genuine threat is that modern power electronics tempt teams to rerun this trade with the 2004 constraints deleted and reach the opposite answer for the wrong reason. Wide-bandgap switches now clear faults fast and hold off kilovolts, so a DC bus at multi-kV can be built lighter than anything in this paper; but the PFC lesson cuts the other way, because every active converter added for voltage conversion or power factor correction is also a radiation-qualification and single-point-failure liability in the 4.3 Mrad-class environment this paper budgets for.¹ The opportunity, grounded in the paper's own voltage ladder (55 kg at 400 V, 22 kg at 1000 V, 2 kg at 10 kV for the identical run¹), is that a kilovolt-class AC primary with kHz magnetics and passive step-down is a genuinely underexplored middle path for 100 kW-plus electric propulsion vehicles: most of the harness mass saving of high voltage, little of the DC converter count, and fault clearing the alternator can actually feed. The 39 kg that separated the two buses in 2004 is small; the architecture logic that produced the decision is what transfers.

The bottom line

Calibrated judgment: the study's internal logic is sound and its numbers, taken as Phase A estimates on stated assumptions, are self-consistent. High confidence that AC-with-autotransformers was the right choice given the 2004 component base and the fault-clearing requirement; high confidence in the qualitative lessons (voltage beats everything for harness mass, frequency buys magnetics mass, protection drives architecture); moderate confidence only in the specific masses, which rest on datasheet extrapolation and non-optimum switchgear models. What would confirm the claim is hardware: a kilohertz AC distribution demonstration at 50 to 100 kWe with measured loss and fault-clearing behavior. What would break it is a modern radiation-qualified high-voltage DC switchgear base that clears faults within a cycle without active current limiting, which would flip the trade the paper could not run in 2004.