Power management for a thermionic-diode spacecraft

What the work claims

This is a design-study paper, reporting a Phase 1 trade effort rather than test data. Kimnach, of the Electrical Systems Development Branch at NASA Lewis, describes the power management and distribution (PMAD) system designed for the Air Force Phillips Laboratory's Integrated Solar Upper Stage (ISUS) Engine Ground Test Demonstration: a system to deliver 1000 W at 28 +/- 6 Vdc to the payload and spacecraft from thermionic-diode (TID) generation of at most 1070 W at a 2200 K emitter temperature.¹ The central design decision is the regulator: a Series Connected Boost Unit (SCBU), which processes only the difference between the sagging source voltage and the regulated bus, rather than the full load power.¹

How it works

ISUS was a bimodal solar-thermal upper stage: two concentrators heat a dual-cavity graphite core, with thermionic diodes encircling each cavity; the receiver-concentrator assembly doubles as the orbit-transfer engine and the on-orbit power plant. TID emitters run at about 2200 K peak and collectors at about 1000 K. Each cavity half is ringed by a string of 30 TIDs; each diode at 2200 K has a maximum cell voltage of 1.1 V and produces 35.7 W, giving a combined maximum of 1070 W and a string voltage of about 34 Vdc.¹

The source has two awkward properties. First, its voltage droops, both as the graphite core's thermal storage depletes (the requirement is that the TID output still hold 18 Vdc at the end of eclipse) and as individual diodes fail; the paper lists an explicit requirement that 40% of rated power remain deliverable if half the receiver is lost.¹ Second, it cannot be unloaded: if the loads draw less than the incoming thermal power, the core temperature climbs and the diodes overheat, so the PMAD must actively dump excess energy.¹

The trade study rejected the simple options. An unregulated bus minimizes weight (80.1 lb in the study's model, at 92.7% end-to-end efficiency) but only works if the source holds the load spec over all conditions, which a drooping thermal source does not; a capacitor cannot store enough energy, and a battery on the bus was judged to undermine the point of the TIDs.¹ Conventional buck, boost, and buck-boost regulators run at roughly 85 to 90% efficiency but process every watt. The SCBU instead floats in series with the source like an autotransformer and injects only the boost voltage needed to hold 28 V. Sized for the ISUS worst case (18 V source, 10 V boost, 85% assumed dc-dc converter efficiency), the internal converter processes at most 357 W, about 36% of the 1 kW load, and the unit runs at about 94% efficiency at maximum boost.¹ At the 18 V end-of-eclipse condition the regulated system holds roughly 91% efficiency versus 83% for the unregulated baseline, at a mass penalty of 10.2 lb; the complete PMAD including harnesses, the photovoltaic-and-battery subsystem for LEO and transfer, and control hardware totals 91.3 lb.¹

Two supporting mechanisms complete the design. A Power Shunt Module sinks up to 350 W to hold the energy-balance equation when payload demand falls, protecting the diodes from over-temperature.¹ And because the TIDs are thermally shielded until final orbit, a GaAs array with a nickel-hydrogen battery covers the transfer phase: the SCBU-regulated array is 413.7 W beginning-of-life at 2.06 m2, smaller than the 433 W, 2.25 m2 unregulated array, which mattered because only 2.2 m2 of vehicle surface was available and the unregulated array could not fully recharge the battery anyway.¹ The battery choice (15 Ah, about 21 lb, 52% depth of discharge over the 36-minute eclipse) was driven by what existed: the lighter 10 Ah cell would have forced 86% discharge cycles and was not available.¹

The strongest case for it

The architecture reasoning is genuinely good. Partial-power-processing regulation buys its efficiency advantage exactly when the source is weakest: at nominal voltage the converter barely works, and at the 18 V end-of-eclipse corner the regulated system still delivers about 91% against 83% for the passive baseline.¹ The topology is also single-fault tolerant by construction, since a failed SCBU leaves a direct electrical path around it, at the cost of losing regulation.¹ And the SCBU was not a paper invention: it came out of the Photovoltaic Regulator Kit Experiment (PRKE), which the paper cites as a flown 180 W unit built on a 60 W dc-dc converter, with the 1000 W ISUS version then in design as two paralleled 200 W converters.¹ The 96%-plus efficiency claim for low-boost operation is consistent with the paper's own sizing equation.¹

Where a skeptic should push

Every efficiency and mass figure in this paper is analytical: the 94% SCBU figure is computed from an assumed 85% dc-dc converter efficiency, the 91%-versus-83% system comparison comes from the paper's own model (its Figure 2), and no measured system performance appears anywhere in the document.¹ The 92.7% unregulated-baseline figure likewise rests on the modeled operating range.

The more telling evidence is programmatic. Written in 1996, the paper describes a ground demonstration already being descoped: only half a receiver (a RACCET) would be built in Lewis's Vacuum Tank 6, with the missing half of the TID string emulated by a diode simulator wired in series; the photovoltaic-and-battery subsystem was deleted from the demonstration entirely; and the start of testing had slipped to what the text calls "late early 1997."¹ In other words, the end-to-end PMAD-with-thermionic-source test the design deserved was itself partially a simulation. Emitter lifetime at 2200 K, the single hardest materials problem in thermionic conversion, is not addressed in this document at all. Treat the maturity honestly: a sound paper design, a heritage converter building block, and a ground test that, as described, would have qualified only part of the claim.

Partial power processing for drooping sources

The non-obvious implication is that this exact PMAD problem is coming back. A surface fission reactor with dynamic conversion, or any radioisotope system with a thermal engine, is electrically the same source Kimnach faced: a low-voltage, thermally inert generator whose output sags under load and which cannot be switched off. The lessons transfer directly. Size the regulator as a partial-power-processing converter from the start, because the mass-optimal unit processes only the droop headroom, here 36% of bus power.¹ And design the dump load as a first-class component: the Power Shunt Module exists because a thermal source at light load does not idle, it bakes, and the paper's energy-balance requirement means the shunt must be sized against the largest credible load drop, not against average conditions.¹

The genuine threat is organizational: modern spacecraft power engineering grew up around stiff photovoltaic-plus-battery buses, and the failure modes of a drooping source are foreign to that experience. A team that treats a reactor bus like a solar array bus will discover, in integration rather than on paper, that the bus collapses at the end of the eclipse analog and that the converter overheats its source exactly when the payload goes quiet. The opportunity, stated as inference from the paper's own sizing math, is that a converter which only processes the boost fraction stays small as bus power grows; with today's wide-bandgap switches, the SCBU-style partial-processing topology deserves a fresh look for kilowatt-class nuclear surface power buses, where processing 36% of the power instead of 100% is a compounding mass and efficiency saving.¹

The bottom line

Calibrated judgment: the regulator selection logic is correct and the efficiency arithmetic, taken on its stated assumptions, is sound; this is a competent engineering design study whose central insight survives. The maturity, however, is paper plus a descoped ground test plus one flown precursor building block, and none of the headline figures was measured on the system described. Confidence is high that partial-power-processing regulation is the right architecture for drooping thermal sources; it is moderate that the specific numbers would reproduce in hardware. What would confirm the claim is a system-level test of an SCBU-regulated bus against a real or high-fidelity simulated TID string across the full source-voltage range; what would break it is a converter efficiency materially below the assumed 85%, which the paper's own equation shows flows almost linearly into system efficiency.