Ferguson's Mechanical Paradox is a fixed-axis epicyclic gear arrangement where three loose gears with slightly different tooth counts ride on a common arbor and mesh with the same driving pinion, producing forward rotation, reverse rotation, and apparent standstill from a single input. James Ferguson built the first one in 1764 to demonstrate planetary motion in his orrery. The purpose is to show how a 1-tooth difference between meshing gears completely reverses output direction. The outcome is a tabletop demonstration that still trips up engineering students 260 years later.
Ferguson's Mechanical Paradox Interactive Calculator
Vary the fixed sun and three loose gear tooth counts to see which gears creep forward, stand still, or reverse per carrier revolution.
Equation Used
The calculator applies Ferguson's tooth-count differential to each loose gear. A loose gear matching the fixed sun gives zero output, one fewer tooth gives forward creep, and one extra tooth gives reverse creep.
FIRGELLI Automations - Interactive Mechanism Calculators.
- Fixed sun gear is grounded.
- All loose gears are driven by the same carrier pinion.
- Ideal tooth geometry with no backlash or arbor slop.
- Positive output ratio indicates forward creep; negative indicates reverse creep.
Inside the Ferguson's Mechanical Paradox
Take three thin gears, stack them loose on a common arbor, and let that arbor swing around a fixed central wheel on a rotating carrier. The same pinion on the carrier meshes with all three loose gears at the same time. Here is the trick — the three loose gears have tooth counts of, say, 19, 20, and 21, while the central fixed wheel has 20 teeth. As the carrier rotates, the pinion rolls around the central 20-tooth wheel and drives all three loose gears together. The 20-tooth loose gear matches the fixed wheel exactly, so it stays still in space. The 19-tooth gear creeps forward. The 21-tooth gear creeps backward. One input, three different outputs, all from a 1-tooth difference per stage.
The mechanical paradox is in that word "paradox" — visually nothing about the three gears looks different, yet they behave in opposite ways. The principle is pure tooth-count differential. If the loose gear has the same number of teeth as the fixed sun wheel, the loose gear's angular position relative to ground is locked. Add a tooth and it lags. Remove a tooth and it leads. The ratio you actually see at the loose gear is (Zfixed − Zloose) / Zloose per carrier revolution, which is why a 1-tooth offset on a 20-tooth gear gives you 5% output per carrier turn.
What goes wrong is almost always backlash and arbor slop. If the loose gears wobble on the arbor by more than about 0.1 mm radial, the meshes hunt and you lose the clean stationary behaviour on the matched gear — it twitches forward and back instead of holding still. Tooth-count errors are not forgiving either. Cut a 19-tooth gear with one tooth slightly thick and it will bind against the pinion at one point per revolution. Ferguson's original wood-and-brass build held this together with hand-fitted bushings; modern reproductions need pinion-to-gear centre distance held to within 0.05 mm or the demonstration loses its punch.
Key Components
- Fixed Central Sun Wheel: A stationary gear (typically 20 teeth) rigidly grounded to the frame. The carrier rotates around it. Its tooth count is the reference all three loose gears compare against — match it exactly and the loose gear stands still, deviate by 1 tooth and the loose gear rotates.
- Rotating Carrier Arm: The driven input. It carries the pinion shaft and swings it around the fixed sun wheel. One revolution of the carrier produces small fractional rotations of each loose gear depending on its tooth-count offset. Carrier RPM is typically 5-30 RPM in demonstration builds.
- Common Pinion: A single pinion (often 8-12 teeth) mounted on the carrier that simultaneously meshes with all three stacked loose gears AND the fixed sun wheel. Centre distance from pinion to loose-gear stack must be held within 0.05 mm or backlash dominates the output.
- Three Loose Gears on Common Arbor: Three thin gears with tooth counts straddling the fixed sun wheel — typically Z−1, Z, and Z+1. They ride free on a common arbor on the carrier. Their differential motion against the fixed sun wheel is what produces forward, stationary, and reverse outputs from one input.
- Output Pointers or Dials: Pointers fixed to the top of each loose gear so the observer can see forward, stationary, and reverse motion side by side. In Ferguson's original orrery these were planet position indicators showing Mercury, Venus, and Earth at scaled orbital rates.
Where the Ferguson's Mechanical Paradox Is Used
The Ferguson's Mechanical Paradox lives in two worlds — historical orreries and modern teaching kits. It rarely shows up in production machinery because you can get the same differential output from cleaner planetary gearboxes, but for showing students how tooth-count math drives output direction it is unbeatable. Every mechanism museum that runs gear demonstrations has one.
- Astronomical Instruments: James Ferguson's original 1764 orrery used the paradox to drive Mercury, Venus, Earth, and Mars indicators at correct relative orbital rates from a single hand crank. The Royal Society of Arts in London still holds Ferguson's drawings.
- Engineering Education: MIT's gear-train teaching collection and the Cornell Reuleaux Kinematic Models collection both include Ferguson Paradox demonstrators built in the 1880s by Voigt of Berlin. Students see the 1-tooth-difference principle without any math.
- Science Museum Exhibits: The Science Museum in London and the Smithsonian both display working Ferguson Paradox models. Visitors crank a handle and watch three identical-looking gears rotate forward, stop, and reverse at once.
- Horology and Clockwork: Specialty astronomical clocks — for example the long-case orrery clocks made by John Pearson in the 1820s — use a paradox stage to drive multiple celestial indicators off one arbor without needing separate gear trains for each planet.
- Mechanism Design Research: Researchers at the University of Cambridge Whipple Museum have used Ferguson-style paradox stages as a benchtop reference to validate epicyclic gear-ratio derivation software, because the small tooth-count offset isolates the differential term cleanly.
- Hobbyist Maker Builds: 3D-printed Ferguson Paradox kits sold by Thingiverse contributors and on Etsy run with carrier RPMs of 10-20 RPM and demonstrate the principle with FDM-printed gears at module 1.5.
The Formula Behind the Ferguson's Mechanical Paradox
The output rotation of each loose gear, per revolution of the carrier, is set by the difference between the fixed sun-wheel tooth count and the loose gear's tooth count. Push the offset to zero and the loose gear holds still — that is the magic stationary output. At a 1-tooth offset on a 20-tooth gear, you get 5% rotation per carrier turn, which is the demonstration sweet spot — slow enough to show direction clearly, fast enough that observers do not lose patience. Push the offset above 3 teeth and the demo loses impact because the gears now visibly spin at obviously different speeds and the paradox feel disappears. Below 1 tooth offset is impossible (gears need integer teeth), so 1 tooth is both the minimum and the most dramatic.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| ωloose | Angular velocity of the loose gear relative to ground | rad/s | RPM |
| ωcarrier | Angular velocity of the carrier arm (the input) | rad/s | RPM |
| Zfixed | Tooth count of the fixed central sun wheel | teeth | teeth |
| Zloose | Tooth count of the loose gear under analysis | teeth | teeth |
Worked Example: Ferguson's Mechanical Paradox in a museum exhibit Ferguson Paradox demonstrator
You are sizing the three loose gears on a Ferguson Paradox demonstrator for the Whipple Museum's hands-on gear gallery in Cambridge. The visitor cranks a handle that drives the carrier at a nominal 12 RPM through a 5:1 reduction. The fixed sun wheel has 20 teeth at module 1.5. You want one pointer to creep forward, one to hold dead still, and one to creep backward → all visible in a 30-second visitor interaction. You pick loose gear tooth counts of 19, 20, and 21 teeth.
Given
- ωcarrier = 12 RPM
- Zfixed = 20 teeth
- Zloose,A = 19 teeth
- Zloose,B = 20 teeth
- Zloose,C = 21 teeth
Solution
Step 1 — at the nominal 12 RPM carrier speed, compute the output of the 19-tooth (forward) loose gear:
Step 2 — compute the output of the 20-tooth (stationary) loose gear. The tooth counts match exactly, so the differential term is zero:
Step 3 — compute the output of the 21-tooth (reverse) loose gear:
Step 4 — at the low end of typical visitor cranking, around 5 RPM carrier, the 19-tooth pointer creeps at 0.26 RPM (one full pointer revolution every 3.8 minutes). At that pace it looks dead still to a casual observer — the museum staff complained the paradox failed to register in test runs. At the high end of cranking, 30 RPM carrier, the forward pointer hits 1.58 RPM and the demonstration becomes a spinning blur instead of a paradox. The 12 RPM nominal sits in the middle — you see the forward pointer make about a quarter turn in 30 seconds, the back pointer doing roughly the same in reverse, and the middle pointer rock-solid still. That contrast is what sells the paradox.
Result
At 12 RPM carrier input, the three pointers run at +0. 632 RPM, 0 RPM, and −0.571 RPM respectively. That gives the visitor about 19° of forward pointer rotation in a 30-second interaction — slow enough to look mysterious, fast enough that the direction is unambiguous. At 5 RPM the demonstration falls flat (only 8° of pointer travel in the same window), and at 30 RPM the forward and reverse motion blurs into general spinning. If your stationary pointer twitches instead of holding still, check three things in order: pinion-to-loose-gear centre distance (must hold to 0.05 mm or backlash hunts the mesh), arbor radial play on the loose-gear stack (more than 0.1 mm and the gears wobble axially and lose tooth contact briefly each revolution), and tooth-count verification (it is shockingly common to find a vendor sent you 20-tooth gears stamped as 21).
Ferguson's Mechanical Paradox vs Alternatives
The Ferguson Paradox is a teaching tool first and a power-transmission element a distant second. Compare it to the gearboxes that actually do the same job in production — a standard planetary gearbox or a harmonic drive — and you see why it never made it out of the demonstration cabinet.
| Property | Ferguson's Mechanical Paradox | Standard Planetary Gearbox | Harmonic Drive |
|---|---|---|---|
| Typical input speed | 5-30 RPM (demo only) | 100-6000 RPM | 500-6000 RPM |
| Reduction ratio per stage | 1:19 to 1:21 (intentionally tiny differential) | 3:1 to 10:1 | 30:1 to 320:1 |
| Load capacity | < 0.5 Nm (decorative) | 10-10,000 Nm | 1-3,000 Nm |
| Backlash sensitivity | Critical — 0.05 mm centre distance | Moderate — 5-15 arcmin typical | Very low — under 1 arcmin |
| Manufacturing complexity | Low (hobby-grade) | Medium | High (flexspline manufacture) |
| Primary application fit | Education, orreries, museums | Industrial actuation, robotics | Precision robotics, aerospace |
| Cost (representative) | $50-300 kit | $200-3000 | $800-8000 |
Frequently Asked Questions About Ferguson's Mechanical Paradox
This is almost always a profile or pitch error in one of the two matched gears. The math says zero output only if both gears have identical tooth count AND identical effective pitch diameter. If your 20-tooth loose gear was cut on a different hob than the 20-tooth fixed wheel, even a 0.02 mm pitch-diameter mismatch produces a small rolling error per carrier revolution that integrates into visible drift over 30 seconds.
Quick check — pull the loose gear, swap it with the fixed sun wheel, and see if the drift reverses direction. If it does, your gears are not a matched pair. Cut both stationary-pair gears from the same blank stack on the same setup if you can.
You can, but you lose the paradox effect. The whole demonstration depends on the three loose gears looking visually identical to the observer. At 1 tooth difference on a 20-tooth gear the diameters differ by less than 5% and the eye cannot tell them apart. At 3 teeth difference on a 20-tooth gear the diameters differ by 15% and a sharp observer immediately spots that one gear is bigger than the others — the trick is gone.
If you want faster pointer motion, raise carrier RPM instead of widening the tooth offset. 1 tooth offset at 20 RPM carrier gives the same pointer rate as 2 tooth offset at 10 RPM, but only the first one preserves the visual paradox.
Yes — when engineers say "the mechanical paradox" in a gearing context they almost always mean Ferguson's specific 1764 arrangement of three loose gears with neighbouring tooth counts on a common arbor. There are other so-called paradoxes in mechanism theory (the geometric paradox in linkages, for example) but in the gear-train world the term is reserved for Ferguson's design.
Three causes in order of likelihood. First, FDM layer lines on tooth flanks create high spots that hit one tooth on the pinion harder than the others — sand the gear flanks lightly with 400 grit or print on side with a 0.1 mm layer height. Second, the arbor hole prints undersized on most consumer FDM machines by 0.2-0.3 mm; ream it to nominal or design 0.4 mm clearance into the CAD. Third, your fixed sun wheel may be slightly off-centre on its mounting boss — a 0.3 mm eccentricity produces a once-per-revolution tight spot that feels exactly like a gear-tooth error.
Planetary gearbox, no contest, unless your goal is specifically the visual paradox effect. A planetary gearbox gives you cleaner reduction (10:1 to 100:1 in a single stage), much lower backlash, and far better load capacity. The Ferguson arrangement is a teaching device — its only advantage is that you can show forward, stop, and reverse from one input simultaneously. If you only need one output direction at one ratio, the paradox stage is just an inefficient planetary with extra parts.
For an orrery where you want multiple planet indicators at proportional rates from one drive, the paradox stage is genuinely elegant — but you would still typically pair it with a downstream reduction stage to get the actual orbital rate ratios.
The 19-tooth and 21-tooth outputs are not symmetric — they should not run at the same speed. The forward output runs at carrier × (1/19) = 5.26% of carrier, and the reverse runs at carrier × (1/21) = 4.76% of carrier. The reverse is always about 10% slower than the forward in a 19/20/21 stack. If yours is closer to symmetric, you probably have your 19 and 21 gears swapped on the arbor — check the tooth count by counting twice, because at module 1.5 the difference between a 19 and 21 tooth gear is only about 3 mm in outside diameter and easy to miss visually.
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
