Roger Shawyer’s EmDrive: Engineering The “Impossible Drive”
The EmDrive was called “impossible” because it promised propulsion without propellant – but the longer the debate ran, the more impossibility shifted from theory to practice. The hard part wasn’t drawing a cone on a whiteboard—it was engineering the device and a measurement protocol capable of distinguishing micro-thrust from heat, wiring, drift, and electromagnetic side effects.
The “impossible” part of the EmDrive was proving it worked
The EmDrive became famous because it offered a seductive proposition: thrust without propellant. Microwaves go in, thrust comes out, and the spacecraft doesn’t have to throw anything overboard. If that were true at useful levels, it would change mission design—no propellant budgets, no tank fractions, no “you’re done when the xenon’s gone.” It would be an engine that turns electricity directly into momentum, continuously, for as long as you can supply power and dump waste heat.
The phrase “impossible drive” stuck because the claim looks, at first glance, like a violation of conservation laws. A sealed cavity pushing itself through space feels like pulling on your own bootstraps. That immediate conflict between “astonishing capability” and “textbook physics” turned the EMDrive into a cultural event—half engineering rumor, half physics argument, half internet tribal war.
But the deeper story is that the EmDrive lives right where experiments are easiest to fool and hardest to interpret: at micronewton-level forces, inside hardware that heats, expands, flexes, and carries high-frequency currents. In that regime, the experiment isn’t just measuring the claim—the experiment creates the conditions where false positives can look like discoveries. You can be honest, careful, credentialed, and still get a “thrust signal” that vanishes when you change a mount, reroute a cable, or improve thermal symmetry.
Roger Shawyer’s APEC talk and Q&A make this tension explicit. He insists the EmDrive is not mystical: it’s an engineering device with strict geometric and RF requirements. His critics insist the burden is not storytelling but replication: if a propellantless thrust exists, it must survive the best artifact suppression that modern thrust metrology can produce. That clash—engineering claim versus measurement reality—is the backbone of the EmDrive narrative.
What Roger Shawyer says the core principle actually is
At the headline level, Shawyer’s EmDrive is a tapered microwave resonant cavity—often drawn as a truncated cone (a frustum). You inject microwave power through a coupler, establish a strong resonant field pattern in the cavity, and (in Shawyer’s model) the electromagnetic forces on the cavity walls do not perfectly cancel because the cavity’s geometry changes the field behavior from one end to the other.
Shawyer’s key insistence is that popular explanations are wrong in the way they’re wrong. He rejects the cartoon of photons ricocheting like pellets in a tin can. In his telling, you must treat the cavity as a waveguide/resonator problem—an “electrical machine” where fields and induced currents exchange momentum with the conducting structure. The cavity walls are not passive bumpers; they are part of the electromagnetic system that stores and redirects energy and momentum.
“So what is EmDrive? Well, it’s a new propulsion technology which is a propellantless technology — not reactionless, but propellantless.” — Roger Shawyer
A core ingredient in his reasoning is waveguide behavior—especially the idea that the effective wave propagation characteristics depend on cross-section. In a tapered waveguide, the group/guide velocity can vary with geometry. Shawyer argues that this affects how electromagnetic momentum and force distribute at the two ends of the cavity. The result, he claims, is a net axial force that points toward the larger end of the cone.
This is also where Shawyer becomes very “engineer” and very specific: he argues geometry is not cosmetic. He claims that flat endplates and certain cavity choices introduce phase errors and wavefront problems that prevent the internal field regime he believes is required. His patent language, likewise, emphasizes shaped end features and control concepts—suggesting that, in his view, the EmDrive is not “a cone plus microwaves,” but a tuned resonant system with strict design and control constraints.
What kind of thrust is the EmDrive “supposed” to produce?
The EmDrive’s claimed performance is part of what makes it so polarizing. If it were merely at “photon rocket” levels—tiny thrust equal to the momentum carried by emitted radiation—it would be physically uncontroversial but practically unexciting. What drew attention were claims and reports at levels far above simple radiation pressure expectations.
Shawyer has described generations of development where increasing Q-factor (how sharply the cavity resonates) and better control yield dramatic improvements. In his APEC material, he discusses thrust-to-power figures moving from early demonstrators into numbers that—if real—would be competitive with conventional electric propulsion on a per-kilowatt basis, while eliminating propellant consumption. He also points to other reported results (including overseas efforts) as suggestive support.
NASA’s Eagleworks work put a different number into the public bloodstream: on the order of a millinewton per kilowatt. That’s far smaller than Shawyer’s most ambitious figures, but still wildly larger than pure photon thrust per unit power. It landed in the uncanny valley: too big to ignore, too small to trust without extraordinary care, and right in the zone where experimental artifacts can masquerade as propulsion.
The thrust question therefore becomes a fork: either the EmDrive effect is real and scales in some predictable way with power, Q, geometry, and mode structure—or the “thrust” is an emergent behavior of the test stand itself, correlated with power, heating, and wiring in ways that look convincing until you design the experiment to kill those correlations. The performance story is inseparable from the measurement story.
“Does the rocket equation apply?” and the myth of “no top speed”
Shawyer has argued that the rocket equation “doesn’t apply” to the EmDrive, and in a narrow sense he’s pointing to something true: the rocket equation is fundamentally a bookkeeping law for vehicles that accelerate by throwing reaction mass out the back. If you don’t expend propellant mass, you don’t face the same exponential penalty for large delta-v that chemical rockets do.
But “not governed by the rocket equation” is not the same as “no limits.” A propellantless thruster trades propellant limits for power and thermal limits. If you must supply kilowatts (or megawatts) for months or years, you have to carry power generation, energy storage, conversion electronics, and radiators. Heat rejection becomes a first-class design constraint. As thrust stays low, acceleration stays low, and missions become “continuous thrust spirals” rather than impulsive burns.
There’s also a conceptual pitfall: some popular coverage treated “no propellant” as “free momentum.” In reality, propellantless thrust is entirely compatible with known physics if momentum is carried away by something—photons (in a laser/antenna), reflected sunlight (in a sail), or beamed power momentum exchange. Those are reaction engines too; the “reaction mass” is radiation.
So the clean way to phrase the EmDrive’s “top speed” implication is this: if it worked as claimed, the EmDrive would not remove physics limits—it would change the constraint landscape. Instead of “how much propellant do you have,” the governing questions become “how much sustained electrical power can you field,” “how efficiently can you convert it,” and “how much waste heat can you reject for how long.”
EmDrive vs ion drives: what would actually be different?
Ion drives and Hall thrusters are the real-world benchmark because they already do what EmDrive enthusiasts want: high-efficiency, long-duration propulsion powered by electricity. They are slow but steady, with thrusts in the millinewton to newton class depending on power level and design, and with flight heritage to prove they’re more than lab curiosities.
If an EMDrive-like device produced thrust without propellant at even modest fractions of ion-drive thrust-per-kilowatt, the “killer feature” wouldn’t necessarily be raw thrust. It would be endurance and mass fraction: no xenon tanks, no propellant depletion, no mission-ending fuel exhaustion for stationkeeping and attitude control. Even a small but genuine propellantless thrust could matter if it’s reliable and continuous.
But conventional electric propulsion has an advantage the EmDrive has never earned: its thrust is straightforward to measure and reproduce. You can model the ion current, exhaust velocity, thrust, and efficiency with high confidence, and independent labs can repeat your performance tests. With EmDrive-style claims, the measurements have tended to live too close to the noise floor of complex apparatus, making every comparison feel conditional: “assuming the thrust is real” becomes a permanent footnote.
There’s a second uncomfortable point: if the EmDrive effect collapses toward pure radiation pressure, it becomes a photon thruster in disguise—technically real, but far weaker per watt than even modest electric propulsion. That’s why the debate keeps circling back to a single practical question: does the EmDrive produce thrust significantly above the photon-thrust baseline, in a way that survives the best artifact suppression?
Engineering constraints from mainstream physics: the conservation-law design brief
From an engineering standpoint, mainstream physics sets the design constraints: if thrust is real, you must identify the momentum exchange path and then measure it in a way that survives the ugliest failure modes of micro-force experiments. The conservation objection isn’t just a philosophical “no”—it’s an engineering requirement. A closed cavity can’t produce net external momentum unless something outside the box participates, whether that “something” is emitted radiation, coupling to external fields, or an interaction your model can specify and your lab can test.
That constraint is why the photon-thrust benchmark is so useful as a design reference. If all you’re doing is leaking radiation or behaving like a quiet photon rocket, your thrust-per-power will be capped at a very small value. Any claim far above that is either a real new mechanism—or a systematic error dressed up as a breakthrough. Engineers don’t need to settle the metaphysics to use this benchmark: it’s a practical yardstick for what magnitude and scaling behavior a credible measurement should reveal.
This is where the best mainstream criticisms actually help the story: they translate into test demands. If someone claims millinewton-per-kilowatt propellantless thrust, then the experiment must demonstrate (a) repeatable scaling with power and resonance, (b) robust null configurations, (c) reversibility with orientation or device configuration, and (d) immunity to thermal drift, cable forces, and electromagnetic coupling that can fake exactly those signatures. Skeptics are not only saying “it can’t be true”; they’re saying “here’s what you must engineer into the test before we should believe you.”
So mainstream skepticism doesn’t end the EmDrive story—it tightens it. It forces the EmDrive discussion into an engineering brief: define the system boundary, specify the momentum bookkeeping, predict the scaling laws, and then build an apparatus that cannot be fooled by heat, wiring, and drift. In that framing, “physics skepticism” becomes “engineering discipline.”
Dissenting opinions: how the physics might work — as testable engineering signatures
These dissenting ideas matter only insofar as they produce testable engineering signatures. If the EmDrive is not a Newton-violator, then it must be exchanging momentum with something—either by emitting momentum-carrying radiation, by coupling to external fields, or by exploiting a vacuum/inertial effect that has predictable dependencies. The value of dissent is not reassurance; it’s that it offers concrete predictions you can design experiments around.
One family of “it doesn’t violate Newton” explanations reframes the EmDrive as having an exhaust—photons. The paired-photon / hidden-exhaust concept says the cavity may emit radiation in a way that isn’t captured by simplistic intuition, providing an equal-and-opposite momentum flow without an obvious plume. If that’s true, the engineering signature is clear: there should be measurable anisotropic radiation leakage correlated with thrust, and the momentum flux implied by that leakage must match the force. This is a solvable measurement problem, not a philosophical argument.
Etkin’s dissenting view takes a different route: he argues the system isn’t meaningfully closed because it cannot be isolated from broader field interactions, and proposes an “energodynamic” account involving vortical field structures and directed forces while preserving momentum conservation via an expanded system boundary. Whether one buys it or not, it suggests different engineering discriminators: dependence on orientation, environment, shielding, and other boundary-condition manipulations beyond what a sealed, purely internal EM stress model would predict.
A third family of proposals leans on vacuum structure or modified inertia ideas: if the mechanism involves vacuum fluctuations, Unruh-like effects, or inertia modification, it should predict specific scaling with frequency, cavity dimensions, modulation, Q-factor, and operating modes—and those predictions must survive the artifact suppression that destroys ordinary false positives. In practice, “dissenting physics” becomes useful when it tells you what knob to turn and what curve you should get back.
Here’s the engineering takeaway: dissenting theories only matter if they help you design a discriminating experiment. Measure directional radiation leakage and momentum flux. Test environmental coupling by changing shielding, cable routing, magnetic conditions, and boundary configurations. Look for scaling laws with frequency, Q, mode, and modulation that persist when the apparatus is engineered to be boring. If the EmDrive is real, it should survive that gauntlet—and if it isn’t, these same tests will show you exactly where the illusion lives.
Eagleworks: the “positive signal” era (and what they thought they were seeing)
NASA’s Eagleworks work matters because it was an institutional, instrumented attempt to measure an anomalous thrust signature rather than an anecdotal “it moved” claim. Their torsion pendulum approach was designed for very small forces, with careful calibration methods and a measurement chain intended to discriminate thrust-like deflections from background drift.
A key element of their public claim was repeatability across configurations: forward, reverse, and null tests were used to argue that a real effect tracked the device configuration rather than arbitrary drift. Their most-cited result, reported in a peer-reviewed venue, described thrust-to-power behavior on the order of a millinewton per kilowatt in vacuum. That number was large enough compared to pure photon thrust to trigger renewed theoretical speculation, and small enough to remain deeply vulnerable to systematic error.
“The test campaign included a null thrust test effort to identify any mundane sources of impulsive thrust, however none were indicated. Thrust data from forward, reverse, and null suggests that the system is consistently performing with a thrust to power ratio of 1.2 ± 0.1 mN/kW.” — NASA Eagleworks
Eagleworks also implicitly acknowledged the problem they were up against. When you power a resonant cavity, you create thermal gradients and mechanical changes. Over time, even tiny shifts in geometry can change resonance and couple into the thrust-balance baseline. Eagleworks explicitly discussed thermal behavior and the need for improved experimental approaches that would be more immune to center-of-gravity shifts from thermal expansion.
In hindsight, Eagleworks’ greatest contribution may not be “proof of EmDrive,” but a high-profile demonstration that careful labs can see thrust-like signatures in these systems—followed by a roadmap of how much harder the experiment must become before the community should believe those signatures as propulsion rather than as instrument response.
The Measurement Problem: when your “thruster” is smaller than your experiment’s bad habits
The EmDrive lives in a regime where your apparatus has more ways to lie than you have ways to check it—unless you design the test stand like a paranoid engineer. At micro-Newton levels, thermal drift can dominate: heating causes expansion, expansion shifts center-of-mass, and the balance “moves” even if the device is producing zero net force. Over long runs, creep and relaxation in mounts can create slow baseline shifts that mimic thrust curves.
Then come electromagnetic couplings. Feedthrough currents, cable routing, partial shielding, and ground loops can all generate Lorentz forces when currents flow in Earth’s magnetic field or in local fields near the apparatus. A cable can act like a tiny motor. A connector can heat unevenly and create a mechanical bias. RF leakage can couple into nearby structures and create spurious forces. At the scale of claimed EmDrive effects, none of this is “minor.”
Shawyer’s own engineering critique targets a different but related trap: you might not even be driving the cavity the way you think you are. He argues that single-port tuning based on reflected power can “tune the loop, not the cavity,” producing impressive-looking RF metrics without actually loading the cavity volume correctly. One practical implication he emphasizes is thermal: if the input hardware heats strongly while the cavity endplates do not, you may be dissipating power in the feed system rather than establishing the resonant field regime your theory assumes.
This is why the decisive EmDrive experiment is not “build a cavity.” It’s “build a test that is boring.” A boring test is one where cable forces are neutered, thermal gradients are symmetrical or compensated, resonance control doesn’t correlate with mechanical drift, and null configurations behave exactly as predicted. In this regime, the apparatus is not a neutral observer; it’s an active participant—and the only way to trust it is to engineer out its personality.
TU Dresden / SpaceDrive: the “artifact elimination” era (and why the signal disappeared)
TU Dresden’s SpaceDrive effort is widely viewed as a turning point because it treated the EmDrive not as a breakthrough to validate, but as an artifact problem to solve. Their framing was explicit: high-power RF hardware creates numerous ways to generate false-positive thrust readings, and some of those false positives can even survive naïve consistency checks if the artifact itself changes with orientation or configuration.
“Measuring a thruster with a significant thermal and mechanical load as well as high electric currents, such as those required to operate a microwave amplifier, can create numerous artefacts that produce false-positive thrust values” — TU Dresden / SpaceDrive team
Their engineering response was to redesign the measurement ecosystem. Among the most telling choices was removing feedthrough interactions by self-powering the device with a battery pack—a direct attempt to eliminate forces and couplings that ride along cables and vacuum chamber interfaces. They also built and iterated thrust balances with the specific goal of suppressing thermal drift and mechanical baseline shifts that can masquerade as thrust.
The result was stark: across tested frequency bands and resonant conditions, they reported no thrust signal above their instrument’s noise floor and argued that any anomalous effect was constrained to below classical radiation-force equivalence—effectively pushing any surviving “EmDrive thrust” down toward the photon-thrust regime. In story terms, this doesn’t just say “we didn’t see it.” It says “we engineered the usual suspects out, and what remained was indistinguishable from nothing.”
This is where the EmDrive controversy becomes sharply defined. If you believe the Dresden results, the earlier positive signals were artifacts. If you believe Shawyer’s rebuttals, Dresden did not test the correct geometry and/or the correct RF operating regime—and therefore their null result is not decisive for his version of the device. Either way, the burden shifts: the next credible pro-EmDrive experiment must match Shawyer’s defined design constraints and survive Dresden-grade artifact suppression.
Conclusions and takeaways: what the EmDrive leaves behind
If you weigh the public record by experimental rigor, the trajectory is clear: early work produced thrust-like signatures; later, more artifact-hardened experiments drove those signatures down into the noise. That doesn’t prove that no RF cavity could ever produce propellantless thrust—but it does mean the strongest tests to date have not supported a thrust effect above classical radiation-pressure baselines.
The best pro-EmDrive argument still standing is not “the physics is solved.” It’s “the replications didn’t test the right device.” Shawyer’s position—curved endplates, strict tolerances, correct coupling, correct mode control, and the rejection of flat-plate implementations—is coherent as an engineering claim. But engineering claims become scientific claims only when independent teams can build the same device, operate it in the same regime, and measure the same effect under conservative, artifact-resistant protocols.
So what would actually settle it? A pre-registered design and protocol, transparent cavity geometry and RF characterization, battery-powered operation in vacuum, rigorous null configurations, orientation reversals, thermal symmetry checks, magnetic coupling controls, and full raw data release—ideally replicated by multiple labs. That’s not bureaucracy; it’s the price of credibility at micro-Newton scale.
The key nuance is this: TU Dresden’s null result is best read as a high-quality upper bound for the configuration and regime they tested, and as a demonstration of how convincingly false positives can arise—and disappear—once you engineer the usual suspects out of the apparatus. But it does not, by itself, settle Shawyer’s contention that many tests were not representative of a properly built, properly coupled, high-Q EmDrive operating in the regime he specifies. The EmDrive is worth further attention only in one form: a faithful Shawyer-spec build coupled to Dresden-grade artifact suppression and transparent data release—either producing a repeatable scaling law that survives nulls and reversals, or tightening the bounds until the claim becomes engineering-irrelevant.
References
APEC 4/3, Part #1 — Roger Shawyer — EmDrive (YouTube)
APEC 4/3, Part #2 — Roger Shawyer — EmDrive — Q&A Session (YouTube)
Futurism — “New Experiment Will Test EmDrive That Breaks the Laws of Physics” (Jun 6, 2019)
Google Patents — GB2493361A “A high Q microwave radiation thruster”
UK Patent Application (PDF) — GB2493361A (official patent PDF mirror)
V.A. Etkin (PDF) — “Energodynamic theory of the Shawyer’s engine”