What voltage does a plasma actuator need, is plasma cooling safe, and where DBD plasma actuators genuinely fall short: eight objections, answered fairly.
The objection is almost never "I do not believe the physics". It is "I cannot put a kilovolt supply inside a product I have to certify, service and insure." Every serious evaluation of plasma actuators for electronics cooling ends in a list of reasons to say no. This is that list: physics, mitigation, and what is left.
What voltage does a plasma actuator need? A surface dielectric barrier discharge actuator needs a few kilovolts to tens of kilovolts of alternating drive at kilohertz frequencies, but it draws only microamps to a few milliamps, so the hazard is set by current and available energy, not by the voltage figure alone.
Published work brackets that range: 14 to 22 kV at 1 to 1.5 kHz for wire electrodes and 7.5 kV peak to peak at 23 kHz for vortex generators [1], and 13.4 kV peak to peak at 5 kHz elsewhere [2]. The mechanism is covered in how DBD plasma actuators work.
Electrical injury is caused by current through tissue, not open-circuit voltage. OSHA training material puts the perception threshold near 1 mA, the let-go range at 9 to 30 mA for men and 6 to 25 mA for women, with death possible at 50 to 150 mA [3]. An actuator sized for local cooling sits far below that: the case cited below drew 15 µA [4].
That does not clear a safety file. IEC 62368-1 bounds accessible energy source class ES1 at roughly 30 V rms and 60 V dc [5], so the kilovolt node has to be enclosed, everything reachable kept at ES1, and fault energy limited.
What remains: creepage and clearance, potting, high-voltage interconnect, single-fault behaviour and a service policy that stops a technician probing a live actuator. Not exotic, but work you would not otherwise do.
It does. Non-thermal discharge dissociates oxygen and some recombines as ozone: quantity depends on waveform, geometry and duty cycle, but in air it is never zero.
The OSHA permissible exposure limit is 0.1 ppm (0.2 mg/m³) as an 8-hour time weighted average and the NIOSH limit is a ceiling of 0.1 ppm at any time [6][7]. Product thresholds are tighter: California caps indoor air cleaning devices at 0.050 ppm [8] and the FDA caps indoor medical devices at 0.05 ppm [9]. Waveform and duty cycle cut generation, ozone decays back to oxygen with published half-lives in air of 20 to 1524 minutes [10], and catalysis handles the rest.
What remains: in a sealed enclosure with no purge path, generation and decay settle at an equilibrium you cannot calculate reliably. It has to be measured, not modelled. More in ionic wind reliability and ozone explained.
The discharge is filamentary. Fast current pulses repeat at the drive frequency, so emission is broadband rather than a single tone, the driver adds high dV/dt switching noise, and actuator wiring is an efficient antenna at chassis lengths. Under 47 CFR 15.109 a Class A device at 10 metres must stay under 90 µV/m from 30 to 88 MHz and 300 µV/m above 960 MHz [11]. Shielding, short leads, filtering and separation are the fixes.
What remains: proximity. If the actuator must sit centimetres from a low-noise amplifier, a precision ADC front end or a coherent optical receiver, filtering may not save you: one of the two has to move.
A charged surface attracts particulate. Deposits change surface conductivity and permittivity, and conductive or hygroscopic dust creates preferential discharge paths. Filtered and sealed enclosures largely remove this, and there is no bearing or impeller to clog.
What remains: in dirty industrial or unfiltered outdoor air it is a genuine limitation that degrades quietly.
Water alters ion mobility and surface conductivity, so performance is not humidity-independent. In tests at 70 per cent relative humidity with water sprayed onto the surface, thrust after 5 seconds stood at 44 per cent of the dry value, falling to 38 per cent at the heaviest water loading [12]. The discharge then dries its own surface: thrust reached 84 per cent of dry thrust after 80 seconds, and almost fully recovered after about 290 seconds at lower drive [12].
What remains: the transient loss is large and condensation stops it until the surface clears. In a sealed edge AI enclosure this rarely bites; in uncontrolled ambient it is a design case you must state.
Continuous partial discharge on a polymer surface causes chain scission, erosion and pitting, then puncture. This is the dominant failure mode. In one comparative study, Kapton-based actuators were aborted after half an hour and two and a half hours; a painted electrode on glass survived six hours but its power consumption fell by 60 per cent; screen-printed copper ran ten continuous hours and held within plus or minus 3 per cent after a two-hour run-in; alumina ceramic and quartz glass were most stable [13].
What remains: the largest genuine gap. Public continuous-run durability is reported in hours; a line card expects tens of thousands. No dataset at that duration is published, and any supplier claiming one should be asked to show it. Our work sits under research.
Correct, and it is a ceiling, not a problem to be engineered away: ionic wind gives metres per second, not tens, and electrical-to-fluid conversion is around 1 to 2 per cent [4]. The win is local: superimposed on an existing 0.3 m/s bulk airflow, an ionic wind gave roughly a factor-of-two increase in the average heat transfer coefficient and about 20 K of extra cooling, drawing under 100 mW (67 mW at 15 µA) [4].
What remains: if the problem is total heat removal from a box, a fan is still correct. See ionic wind versus fans.
A new active component brings a driver to qualify, EMC re-testing, a safety file update, ozone measurement, a life programme and a second-source question. For a design that already meets its thermal target with a fan, that arithmetic does not close.
What remains: the cost is front-loaded and unavoidable. It pays only where a fan cannot go: sealed enclosures, form factors thinner than any bearing, contamination-sensitive optics, and the last 10 to 20 K of a boundary layer no bulk airflow reaches.
| Objection | Physics | Mitigation | What is left |
|---|---|---|---|
| Voltage | kV AC drive, microamp current [3] | Enclose the HV node, ES1 outside [5] | Creepage, potting, service policy |
| Ozone | Discharge splits O₂ in air | Waveform, duty cycle, catalysis [10] | Sealed boxes need measurement |
| EMI | Broadband filamentary pulses | Shielding, short leads, filtering [11] | Cannot sit beside an LNA |
| Dust | Charged surface attracts particulate | Filtered or sealed enclosure | Dirty air still degrades it |
| Humidity | Water changes ion mobility | The surface dries itself [12] | Condensation stops it |
| Ageing | Partial discharge erodes polymer | Ceramic dielectric, printed copper [13] | No data at product duration |
| Bulk flow | 1 to 2 per cent conversion [4] | Use locally, over existing flow | Fans win on volume |
| Cost | New driver and qualification path | Only where a fan cannot go | Fan-served designs never recover it |
Five of the eight are ordinary engineering work: voltage containment, EMI control, dust exclusion, humidity envelope and integration cost. Expensive, boring, and not a reason to reject the technology.
Three are not. Dielectric lifetime at product duration is not publicly demonstrated. Equilibrium ozone in a sealed enclosure must be measured case by case. And an actuator centimetres from a sensitive analogue or RF receiver may simply not be possible. If any of those is a hard requirement, the answer today is no, and you should hear it now, not at qualification. Wider context sits in ionic wind and solid-state cooling; to test an open question, talk to us with the enclosure, not the concept.
What voltage does a plasma actuator need? A surface DBD actuator typically needs a few kilovolts to tens of kilovolts of alternating drive at kilohertz frequencies. Published examples include 14 to 22 kV at 1 to 1.5 kHz and 13.4 kV peak to peak at 5 kHz. Current draw is in the microamp to milliamp range.
Is plasma cooling safe to put near people? Electrical injury tracks current, not voltage: OSHA training material places the perception threshold near 1 mA and the let-go range at 9 to 30 mA, while a DBD cooling actuator draws microamps. The discharge is also non-thermal. The high-voltage node must still be enclosed and treated as inaccessible under IEC 62368-1.
How much ozone does a plasma actuator produce and is that regulated? Some ozone is unavoidable in air. The OSHA limit is 0.1 ppm as an 8-hour average and the NIOSH ceiling is 0.1 ppm at any time. Product-level thresholds are stricter: California requires indoor air cleaning devices below 0.050 ppm and the FDA requires indoor medical devices at or below 0.05 ppm.
How long does a DBD plasma actuator last? Published continuous-run data is measured in hours, not years. Comparative testing showed Kapton-based actuators failing within half an hour to two and a half hours, while screen-printed copper actuators held to within plus or minus 3 per cent over ten hours after a run-in period. Lifetime at product duration is still an open question.
Can a plasma actuator replace a fan? Not for bulk airflow. Electrical-to-fluid conversion is around 1 to 2 per cent and induced velocities are metres per second. It is used to break up a boundary layer where no fan can be placed, where it has produced roughly a doubling of the heat transfer coefficient on top of an existing 0.3 m/s flow.