Overshot water wheel
A few kilowatts, day and night, without eating — the first energy surplus large enough to support full-time specialists.
Build
Tier 3 · Medieval to early modern
Energy
Hard — the water works are most of the job
60–120 person-days including the leat
Crew 8–12
Home: Mechanical power, Level 3 · Mill, level 1 of 2 in its track. Serves 3 steps across 3 subjects.
Every build on this site has hazards that injure or kill — quicklime, carbon monoxide, molten metal, collapsing ground. Read the safety notice before starting, and read this build's own failure modes below.
What it is and why this shape
One healthy person sustains roughly 75 watts of useful work. A modest overshot wheel delivers around two and a half kilowatts, continuously, day and night, and it does not need feeding. That is the work of thirty people, and it is the first energy surplus large enough to support a class of people who make things instead of food.
Overshot, specifically. The three arrangements are not interchangeable:
- Undershot — the wheel sits in the current and is pushed. Simple, needs no head at all, and wastes most of the available energy.
- Breastshot — water enters at axle height. A reasonable compromise where head is limited.
- Overshot — water is delivered over the top and the wheel is turned by the weight of water in its buckets, not by the speed of the stream. Far more efficient, because you are harvesting the fall rather than the flow.
That difference is why the leat matters more than the wheel. Most of the labour in this build is the weir, the headrace and the launder that get water to the top of the wheel with its height intact. The wheel itself is carpentry.
Power is head × flow. A small stream with good fall beats a large flat one. Survey before you cut a single timber.
Elevation — head, wheel and races
Fig 1
Earth / subsoilTimberWater
Bucket geometry
Fig 2
TimberWater
Keyed parts
| 1 | Weir and intake | Raises the water level to feed the leat. Designed to be overtopped safely in flood, not to hold it back. |
| 2 | Leat / headrace on the contour | Carries water along the hillside, losing as little height as possible. Most of the earthwork is here. |
| 3 | Launder and control sluice | Delivers water onto the wheel just past top dead centre. The sluice is the throttle and the emergency stop. |
| 4 | Buckets | Shaped to fill quickly, hold water round the descent, and empty cleanly at the bottom. |
| 5 | Shrouds and soles | The side rings and the bucket floor. Together they form the vessel that holds the water. |
| 6 | Arms and axle | Compression spokes into a squared or cast axle. Everything is sized for shock, not average load. |
| 7 | Bearings | Hardwood, stone or metal in a housing that can be shimmed as it wears — because it will. |
| 8 | Pit wheel and gearing | Takes the drive off. A wheel turns far too slowly to be useful directly. |
| 9 | Tailrace | Takes water away fast. If the tailrace backs up, the wheel runs in water and loses most of its power. |
Specifications
| Wheel diameter | 3.0 m | Aim for roughly the available head minus the launder and clearance |
| Wheel width | 0.9 m | Sized to the flow. Width handles volume, diameter handles head |
| Total head | 3.2 m | Water surface in the leat to tailwater level |
| Flow | ≈ 40 L/s | Measure it in the driest month, not in spring |
| Theoretical power | ≈ 1.25 kW per m³/s per metre of head | P ≈ 9.81 × Q × H, in kW with Q in m³/s |
| Efficiency | 0.70–0.85 | Overshot wheels are genuinely efficient — comparable to many modern turbines |
| Shaft power | ≈ 2.5 kW | ≈ 3.4 horsepower. The sustained work of about 30 people |
| Rotation | 6–10 rpm | Slow. Gearing up is essential for almost every application |
| Rim speed | 1.5–2.5 m/s | Above this water is flung from the buckets before it does work |
| Bucket count | ≈ 36 | 260 mm pitch at the rim |
| Tailwater clearance | ≥ 150 mm | Below the lowest bucket. Flooded tailrace means severe power loss |
Bill of materials
| Material | Quantity | Notes and substitutes |
|---|---|---|
| Oak or elm for shrouds and arms | ≈ 2.5 m³ | Durable, and elm in particular tolerates constant wetting and drying. |
| Bucket boards | ≈ 3 m² | Consumable. Design them to be replaced individually without dismantling the wheel. |
| Axle | 1 × 3.5 m | Squared oak with iron gudgeons, or a forged iron shaft if you can make one. |
| Gudgeons and bearings | 2 sets | Iron journals in hardwood, stone or bronze housings. Must be shimmable as they wear. |
| Launder timber | ≈ 8 m | Sealed against leaks. Every litre lost here is power lost. |
| Sluice gate | 1 | Must shut fully and fast. This is your emergency stop, not a convenience. |
| Weir and leat materials | large | Stone, timber, puddled clay. This is most of the project by volume and by labour. |
| Pit wheel and pinion | 1 set | Wooden gear teeth on the large wheel, harder wood or iron pinion. Wooden teeth are replaceable — deliberately the sacrificial part. |
| Grease or tallow | ongoing | Bearings need lubrication continuously or they seize and burn out. |
Tools: Water level or levelling instrument · Saw, adze, auger, chisels · Compass and trammel for laying out the wheel · Forge for the ironwork · Barrows and digging tools for the leat
Construction
Survey the head and measure the flow
Before anything else, establish exactly how much fall you have and how much water. Use a water level — a long tube of water between two staffs — which is accurate over any distance and costs nothing. Measure flow by timing a floating object down a measured length and multiplying by the cross-section, or by damming and timing a container. Measure in the driest month you can wait for, because that is the flow the mill actually has.
✓ Check: Head and dry-season flow both measured twice, on different days, agreeing.
Set the wheel size from the survey
Wheel diameter should be close to the total head, less the launder depth and the tailwater clearance. Then set width from flow. Design the wheel around the water you measured — the commonest and most expensive mistake here is building the wheel first and then discovering the site cannot feed it.
✓ Check: Calculated power from P ≈ 9.81 × Q × H × 0.75 matches what the mill actually needs to do.
Build the weir and lay out the leat
The weir raises the water enough to start the leat; it is not a dam. Design it to be overtopped safely in flood. Then set out the leat on a very shallow, constant fall — typically 1 in 500 to 1 in 1000 — along the contour. Line or puddle it in porous ground: an unlined leat in sand can lose most of the flow before it arrives.
✓ Check: Water runs the full length of the leat at the design flow with no ponding and no scouring.
Build the pit, the bearings and the tailrace
Excavate the wheel pit and build the bearing supports on solid, independent foundations. Cut the tailrace generously and with fall — a restricted tailrace floods the bottom of the wheel and silently costs a large fraction of the power.
✓ Check: Tailrace carries the full design flow with the water surface well below the lowest bucket.
Lay out and assemble the wheel
Set out the shrouds with a trammel from a true centre and get them round and parallel. Fit the arms into the axle in compression. Assemble, then check for run-out by rotating it against a fixed pointer. An out-of-round wheel hammers its bearings and never stops doing so.
✓ Check: Run-out under 10 mm at the rim over a full revolution.
Fit the buckets
Sole boards angled about 30°, with a vent gap between sole and shroud so air escapes as water enters. Fix them so an individual bucket can be replaced without dismantling the wheel — they are the wear part and you will be doing this regularly.
✓ Check: Buckets fill to roughly two-thirds and empty completely before bottom dead centre.
Build the launder and sluice
The launder delivers water just past top dead centre, so it drops into the bucket rather than being thrown across it. The sluice must close fully and quickly — it is the only way to stop the wheel, and you will need it in a hurry at some point.
✓ Check: Sluice closes fully in a few seconds, from a position a person can reach safely while the wheel runs.
Commission on part flow
Open the sluice a little and let the wheel turn slowly. Watch the bearings for heat, watch for run-out, listen for knocking. Increase flow gradually over days. Grease bearings continuously from the start.
✓ Check: Bearings stay hand-warm, not hot, after an hour at full flow.
Gear it down to the work
At 6–10 rpm the wheel is useless directly. Take the drive through a pit wheel and pinion to get usable speeds, and use wooden teeth on the large wheel — they are the deliberate sacrificial element, replaceable individually, and they fail before anything expensive does.
✓ Check: Drive runs under load without teeth heating, chattering or riding up.
Plan for the flood
It will come and it will be bigger than you designed for. Have a way to divert the leat, lift or protect the wheel, and let flood water pass over the weir without taking the structure with it. Then inspect and repair afterwards — this is a routine maintenance event, not a disaster.
✓ Check: The mill has survived a full flood season with the leat intact.
Tolerances that matter
| Dimension | Target | Why it matters |
|---|---|---|
| Wheel run-out | < 10 mm at rim | An out-of-round wheel hammers its bearings on every revolution, continuously. |
| Leat fall | 1:500 to 1:1000 | Steeper scours the channel; flatter silts it up and the flow stops. |
| Tailwater clearance | ≥ 150 mm | A wheel running in its own tailwater loses a large fraction of its power to drag. |
| Delivery point | just past top dead centre | Earlier throws water over the wheel; later loses usable fall. |
| Bucket fill | ≈ two-thirds | Overfilled buckets spill on the way round, which is water lifted for nothing. |
| Bearing clearance | shimmable | Wooden and metal bearings wear continuously and must be taken up, not replaced wholesale. |
How this fails
Building the wheel before surveying the site. The classic and most expensive error: a fine wheel that the stream cannot fill, or one too small for the head available.
Sizing on spring flow. The mill then stops every summer, which is when you most want it running.
A restricted or silted tailrace. Power falls sharply and the cause is invisible — everything upstream looks correct.
No way to stop it. A wheel that cannot be shut down cannot be maintained safely, and someone will eventually be caught in the gearing. Fit the sluice and a brake before commissioning.
Unlined leat in porous ground. Most of the flow disappears into the hillside between weir and wheel.
Neglected bearings. They run hot, then seize, then burn out, sometimes taking the gudgeon with them.
Underestimating the flood. Design the weir to be overtopped. A weir built to resist a flood is a weir that gets destroyed by one.
Iron gear teeth against iron. Nothing is sacrificial, so a jam breaks something expensive. Wooden teeth on the large wheel are a designed weak point.
How you know it worked
VerificationThe wheel has run continuously through a full year including the dry month and a flood, driven at least two different machines under load, and the bearings have been taken up rather than replaced.
Scaling and variants
Smaller: a 1.5 m wheel on a 2 m head with 10 L/s gives roughly 150 W — modest, but enough to drive a grindstone, a bellows or a small lathe, and it can be built by two people in a fortnight.
Where head is scarce: use an undershot or breastshot wheel, or a horizontal (Norse) wheel, which is far simpler to build and needs no right-angle gearing because the shaft is already vertical. Both are less efficient and both are the right answer on a flat stream.
Where head is plentiful: a turbine extracts more from the same water and is far more compact, but it demands the machine tools and casting capability of a full tier 3 workshop. The wheel is what you build to get to a turbine.
Priority of uses, in order of return: grinding grain first — it is the single largest continuous labour in any settlement — then sawing timber, then driving the bellows for smelting, then trip hammers, then machine tools, then eventually a generator for electricity.