This note explains why the power processing unit is the mass, cost and schedule bottleneck of electric propulsion, what changes at five kilowatts and above, where the merchant market stands, and how vidyut's discharge stage is designed to meet it. It is written for engineers and investors. Every number carries a bracketed reference; each reference is a link to the source.
A Hall-effect thruster ionises a gas, usually xenon and increasingly krypton or argon, in a magnetised annular channel, and accelerates the resulting ions electrostatically to produce thrust. The thruster body has no moving parts and a small number of electrodes. What it needs is precise, robust electrical power, and that is where the difficulty concentrates.
The power processing unit takes the spacecraft bus and generates every rail the thruster depends on: the anode (discharge) supply at several hundred volts and tens of amperes; the cathode heater and keeper that bring the neutraliser to emission and strike its plasma; the magnet supplies that hold the field shaping the discharge; the housekeeping rails that keep control alive first; and the control layer that sequences ignition, rides through arcs and thruster current oscillations, protects the unit, and reports health telemetry to the spacecraft. All of it in vacuum, across the thermal cycle, without disturbing the neighbouring subsystems.
Even in NASA's deliberately low-mass electric-propulsion architecture, the power unit is the heavier half of the system. Mass on a spacecraft is launch cost, so the unit that is easiest to overlook is the one that most decides the economics. [1]
The discharge supply processes roughly 95% of everything the PPU handles. Its efficiency sets the heat the spacecraft must reject, and its high-voltage survival through ignition and arcing is what qualification interrogates. [2]
The split below is representative of a Hall-thruster PPU. The heater, keeper, magnet and housekeeping rails together handle a few percent of the power; the discharge stage handles the rest. That is why the discharge stage is where efficiency, heat and high-voltage survival are decided, and why it is the one piece of vidyut hardware that runs on the bench today.
Sub-kilowatt units for small satellites are a solved and crowded market. The kilowatt-plus class, which does orbit raising for heavier buses and station keeping for large ones, is a different engineering problem in three specific ways.
Heat to reject scales with power and with inefficiency. At 5 kW out, a 90% stage dumps over half a kilowatt into the structure; a 98% stage, about a hundred watts. Reported peak efficiencies for resonant discharge supplies are 97.9% at 10 kW and above 97% at 20 kW. Our design target for the discharge stage is 98%. [5] [6] Try the efficiency cost tool ↓
At operating temperature both the thruster and the unit release gas. The local pressure around high-voltage conductors is therefore not the clean vacuum of a datasheet, and creepage, clearance and potting have to be designed for the degraded case, or the unit arcs when it is hottest.
Both are far cheaper than xenon and both have a higher first ionisation energy, so the discharge supply must operate across a wider and higher voltage range to keep utilisation up. A PPU used with a 5 kW-class thruster in 2025 tests ran 300 to 700 V. [7] [8] [9]
Every watt the stage does not deliver becomes heat; heat needs a path and a radiator, which are mass; mass is launch cost; and the same heat raises component temperature, which shortens life. Move the sliders. Every parameter is an estimate you can replace with your own. The model is illustrative, not a flight thermal analysis: it ignores view factors, heat sharing and the thruster's own load, and treats the hotspot rise as linear in loss. Open it full-page →
The largest public dataset on small electric-propulsion units in orbit comes from ENPULSION, which has flown more than two hundred field-emission thrusters and published its lessons. Almost every anomaly traces to the power unit rather than to the emitter: high-voltage sparking coupling into the low-voltage side and resetting the controller; commercial parts bought without lot control, so radiation test results did not transfer between production batches; command states, sent under schedule pressure, that could damage the unit. Their response was a redesign of the PPU around lot-controlled parts, spark resilience and fully autonomous firmware. [10] [11]
That is a different thruster class from ours, at a fraction of the power. The lesson transfers: electric propulsion is won or lost in the power electronics, and the cost of learning that in orbit is a redesign.
Most thruster makers build their own power electronics, and most of the companies that sell a standalone unit to third parties sell it in the sub-kilowatt class, bundled with their own thruster. A survey of the field in August 2026 gives the numbers below.
The kilowatt-class vendors, Thales Alenia Space Belgium, Airbus Crisa, SITAEL, Colorado Power Electronics and CisLunar Industries, are mostly aligned with their own primes in Europe, and the US vendors carry EAR 9A515 licensing friction for Indian buyers.
Two things have changed recently. First, India now has a high-power thruster of its own: ISRO's 300 mN, 5.4 kW stationary plasma thruster completed a 1,000 hour life test in March 2025 and is manifested for in-orbit validation on the TDS-01 technology demonstration satellite, for orbit raising to geostationary orbit. There is no domestic merchant vendor for its power electronics. [12] Second, the satellite buses that use kilowatt-class electric propulsion are getting heavier, not lighter. Starlink moved from 260 kg to around 800 kg to a planned 2,000 kg per satellite in three generations, and other constellation builders are following the same direction. At those masses electric propulsion does the orbit raising, and that is 5 kW class thrust. [13] [14]
vidyut designs one discharge stage and scales it. The first design target is the Class III stage, 5 to 12 kW, for the 300 to 600 mN thruster class that ISRO's thruster and its international peers occupy. The 1 to 2 kW and 50 to 500 W envelopes are planned derivatives of the same stage, sharing topology, control and protection. That is modularity rather than redesign, and it is how a small team covers three markets with one qualification story.
The founding team designed ISRO's own power processing units for its 300 mN and 75 mN stationary plasma thrusters, including ISRO's first zero-voltage zero-current switching (ZVZCS) module, at 6.5 kW and 90% efficiency. Soft switching moves the energy when there is nothing to slam: the transitions happen at zero voltage and zero current, which is what buys efficiency without an EMI penalty. Two of the three co-founders are named inventors on ISRO's Indian Patent 440340, “Method and Apparatus for Voltage Scaling in ZVZCS DC-DC Converter”; the patent belongs to ISRO. [17] A silicon ZVZCS discharge-stage breadboard is operating on the bench today, TRL 4 for that stage.
The engineering model is designed around gallium-nitride FET candidates with published radiation data, derated and screened for total-dose and single-event risk, with resonant soft switching. In representative converter mass budgets the magnetics dominate, and raising the switching frequency shrinks them when losses, EMI and thermal design are controlled. GaN's lower switching losses are the route to doing that at 5 kW while holding the efficiency that keeps the radiator small. No GaN hardware exists yet; device familiarisation with radiation-data-backed parts selection is the first milestone of the programme. [15]
The system design uses commercial parts under lot control and screening, with a process the team ran inside ISRO, instead of a fully hi-rel parts list. Part for part, commercial components cost about 100 times less than radiation-hardened equivalents and ship in weeks rather than quarters. They are not simply used everywhere because a commercial part carries no radiation guarantee and two lots of the same part number can behave differently, which is exactly what the operational record above found. Lot control, one lot code bought, held, screened and flown, is what turns a cheap part into a usable one. That process is the design asset and a candidate for vidyut's own patent filings once the engineering-model design freezes.
The power unit measures the thruster's voltages and currents on every cycle, which makes it the natural place to watch for degradation. ENPULSION traced extractor clogging through emitter impedance drift and used that signal to run a cleaning procedure on orbit. [11] The vidyut control spine is designed to expose the equivalent health metrics for a Hall thruster from the first engineering model, so that operators and the manufacturer learn from every hour of firing.
The programme runs in three six-month blocks, each with defined gates. The sequence is deliberately the same one a flight programme would use, compressed to what a focused team with the right heritage can execute.
Exit TRL 5 to 6 is the target for the coupled discharge stage only. No TRL is claimed for the full PPCU. Sequence and gates are published as they are passed.
Commercially, the programme does not wait for the coupled test. From month one we approach thruster and bus makers in parallel with engineering; year-one targets are a first order, a collaboration agreement, and the go-to-market groundwork, so the coupled test lands with buyers already engaged.
Each entry links to the source. Bracketed numbers in the text jump to the entry here.
Diagrams are representative, not a flight or ISRO design. “Flight-proven” and “qualified” in the sources describe those authors' own hardware. vidyut's operational hardware is the silicon discharge-stage breadboard, TRL 4 for that stage; GaN is the engineering-model design direction.