This is the technical companion to our piece on how rocket mass heaters work and whether you should build one. That article covers the decision, the code and insurance problem, and the safety case. This one assumes you have made the decision and want the numbers.
Nearly every failed rocket mass heater we have seen or read about failed for one of four reasons: the cross-sectional area was not consistent through the system, the heat riser was too short or not insulated, the duct run was too long, or the mass was built before the core was tested. All four are avoidable arithmetic. What follows is that arithmetic, with the caveat that a heater is a combustion appliance in a building, and the safety sections of the companion article are not optional reading.
The one number everything else obeys
A rocket mass heater is designed around a single figure: the cross-sectional area of the gas path, universally shortened to CSA. Pick it first, then derive everything else from it.
The rule is that the CSA stays essentially constant from the top of the feed tube, through the burn tunnel, up the riser, over the barrel gap, and out through the duct to the chimney. The gas is a single moving column. Pinch it anywhere and the whole system loses draft; balloon it anywhere and the gas slows, drops its ash, and stalls. The two permitted exceptions are the manifold, which is deliberately generous, and the chimney, which may be equal or slightly larger. Everything else matches.
Systems are named by the diameter of their duct. The three that get built are 6, 7, and 8 inch.
| System | Round CSA | Square equivalent | Typical use |
|---|---|---|---|
| 6 inch | 28.3 sq in | 5.3 × 5.3 in | Small, tight spaces. The largest size generally recommended for a batch box feeding a piped bench. |
| 7 inch | 38.5 sq in | 6.2 × 6.2 in | The middle ground, and often the right answer. |
| 8 inch | 50.3 sq in | 7.1 × 7.1 in | Larger buildings, longer duct runs, bigger fuel. The common workshop and barn size. |
The arithmetic is just the area of a circle, A = πr². An 8 inch duct has a radius of 4 inches, so π × 16 = 50.3 square inches. To build that same area as a square firebox you take the square root: √50.3 = 7.1 inches on a side. In practice many builders round the square opening up slightly, to 7.5 inches for an 8 inch system, on the reasoning that a square channel has more wall friction than a round one. Rounding up a little is safe. Rounding down is not.
Bigger is not better. An oversized system in a small building is the most common design error after a short riser. It overheats the space, it needs more fuel than you want to process, and it is harder to keep drafting. Size to the building, and if you are between sizes, take the smaller one and make the mass larger.
The J-tube core, dimensioned
Three notes that matter more than the exact ratio you pick.
The riser is insulated, never massive. This is the difference between a rocket core and a fireplace. Wrap it in ceramic fibre blanket or build it from a perlite-clay mix. If you build the riser out of solid brick, it will suck heat out of the gas, the column will not be buoyant, the draft will collapse, and the thing will smoke back at you. People rebuild entire heaters over this one mistake.
The feed tube cannot be arbitrarily deep. The feed is in a tug of war with the riser: the riser’s buoyancy pulls gas sideways down the tunnel, while the feed tube is itself a vertical chimney trying to pull flame upward. Keep the feed short enough that the riser always wins. For an 8 inch system, 16 inches is roughly the practical limit, and going deeper to burn bigger wood is exactly how people end up with flames licking out of the feed. If you want to burn bigger fuel with less tending, the answer is not a deeper feed tube. It is a batch box.
The burn tunnel floor should be flat and the ceiling can taper down slightly toward the riser. Some builders drop the tunnel ceiling to about 12.5 inches for an 8 inch system, which accelerates the gas into the riser.
The barrel gap, which is where most first builds go wrong
Gas leaves the riser, hits the inside top of the barrel, turns 180 degrees, and flows down the annular space between the riser and the barrel wall. Both of those passages have to satisfy the CSA rule, and the top gap is the one that usually gets pinched.
The duct run, and how long is too long
The horizontal duct buried in the bench is where the heat gets harvested, and it is also the thing consuming your draft. Every foot of horizontal run and every turn costs you pressure that only the chimney can supply.
Working limits, and treat these as ceilings rather than targets:
| System | Practical maximum horizontal run | Notes |
|---|---|---|
| 6 inch | about 25 to 30 ft | Less if you have more than two 180 degree turns. |
| 7 inch | about 35 to 40 ft | |
| 8 inch | about 50 ft | The usual figure quoted, and rarely worth approaching. |
Two design rules go with the length. Slope the duct slightly upward from the manifold toward the chimney, something on the order of a quarter inch per foot, so the gas is always moving slightly uphill toward the stack and any condensate drains back to a cleanout rather than pooling in the middle of your bench. And put a cleanout at every change of direction, accessible from outside the finished cob. You will not remember where they are in three years, so photograph the duct layout before you bury it.
There is a temperature floor as well as a length ceiling. If the exhaust arrives at the chimney below roughly 100 to 120 degrees, water vapor starts condensing inside the stack and draft goes soft. Extracting the last few degrees of heat is a false economy that costs you the draft that makes the whole appliance work.
What overshooting actually feels like, from a build that did
We can be specific about this because we got it wrong. On our own build we got ambitious: we ran the tubing too far and we built too much mass while we were at it. Both errors push the same direction, and neither announced itself until the thing was finished and buried.
The symptom was not poor heat output. It was that the system would not establish draft from cold on its own. A long duct and a large cold mass together present the fire with more resistance than a cold chimney can overcome, so at the moment of lighting there is not yet enough hot gas anywhere in the system to start the siphon that the whole design depends on. It would smoke back rather than catch.
What fixed it was an inline fan in the pipe to start the draw, and we want to be clear about how well this worked, because the purist answer to a too-long duct is “build it shorter” and that answer is useless to anyone who has already buried the thing in two tons of cob. The fan made a huge difference. It turned a heater that would smoke back at us into one that lit reliably every time. Once the core was hot and the riser was doing its job, the fan came off and the system self-drafted for the rest of the burn.
If you have an outlet anywhere near the heater, this is a perfectly reasonable engineering choice and not a confession of failure. It is cheap, it is reversible, it fixes the problem completely, and it takes about an hour to fit. Getting the duct length right the first time is better. A fan is what you reach for when better is no longer available, and there is nothing wrong with reaching for it.
Why cold starting is hard for everyone, not just for overbuilt systems
Draft is produced by the density difference between the hot gas column in your chimney and the cold air outside it. Run the numbers for a 15 foot stack on a 30 degree day and the problem becomes obvious:
| Flue gas temperature | Natural draft available | What is happening |
|---|---|---|
| 30°F (dead cold) | 0.000 in w.c. | No fire yet. There is literally no draft, in any system. |
| 100°F | 0.030 in w.c. | Just lit. Barely anything to work with. |
| 150°F | 0.047 in w.c. | Warming. Marginal against a long duct. |
| 300°F | 0.085 in w.c. | Establishing. The siphon is taking hold. |
| 500°F | 0.117 in w.c. | Running. Plenty of reserve. |
At the moment you strike the match, the draft available to you is exactly zero, and it stays small for the first several minutes. Every rocket mass heater passes through that window. A short duct on a tall chimney gets through it on the strength of a handful of burning newspaper. A long duct feeding a cold mass does not, because the resistance never fell to zero even though the driving pressure did. A startup fan is simply a way of supplying the pressure that physics has not produced yet, and switching it off once physics catches up.
Sizing the fan: the two numbers that matter
Airflow. Take the duct cross-sectional area and multiply by the gas velocity you want. At startup you only need slow, steady movement, on the order of 180 feet per minute, roughly 3 feet per second:
| System | Duct area | Startup flow (180 fpm) | Running flow (360 fpm) |
|---|---|---|---|
| 6 inch | 0.196 sq ft | about 35 CFM | about 71 CFM |
| 7 inch | 0.267 sq ft | about 48 CFM | about 96 CFM |
| 8 inch | 0.349 sq ft | about 63 CFM | about 126 CFM |
So the airflow you actually need is small. Somewhere between 35 and 70 CFM will start any of these systems, and 100 CFM is generous.
Static pressure, which is the number people ignore and the reason cheap fans disappoint. You need to replace the missing natural draft, call it 0.05 to 0.12 inches of water column from the table above, and additionally overcome the friction of a long duct run with several turns, which at these low velocities adds roughly another 0.05 in w.c. for a run near the length limit. Budget for about 0.15 to 0.2 inches of water column.
Now the trap. Inline duct fans are advertised at their free air rating, measured against zero resistance, which is a condition that exists nowhere in your building. A fan sold as “240 CFM” may deliver a small fraction of that at 0.2 in w.c., and many inexpensive axial inline fans stall almost completely somewhere between 0.1 and 0.2. Find the manufacturer’s fan curve and read the CFM at 0.2 in w.c. rather than the number on the box. If the curve is not published, that is informative in itself. As a rule of thumb, a mixed-flow or small centrifugal inline fan holds pressure far better than a cheap axial one, and a modest mixed-flow fan will beat a much larger axial fan at the only operating point you care about.
Temperature rating, which is the one that will bite you. At startup the gas is cool and any fan survives. If the same fan sits in the flue path once the system is running hot it will fail, and a plastic-bladed HVAC duct fan in a 400 degree flue is a fire risk rather than an inconvenience. Three ways to handle it, in order of preference: mount the fan at the coolest point in the system, which is right before the chimney after the gas has given up its heat to the mass; use a fan rated for flue-gas temperatures; or fit it into a removable tee with a cap, so it is only in the gas path during startup and is physically out of the system the rest of the time.
Position it to pull, not to push. This one is safety rather than performance. A fan pushing into the duct pressurises everything downstream of it, and a pressurised duct with any leak pushes combustion gas out into the room rather than drawing room air in. A fan near the chimney end that pulls keeps the whole duct under slight negative pressure, which is the condition you want. Either way the duct has to be genuinely leak-tight, and carbon monoxide alarms are not negotiable, which is the same conclusion the safety section of the companion article reaches by a different route.
The power draw is trivial, so do not let the off-grid argument scare you off. A small inline fan pulls somewhere around 25 to 45 watts, and you only run it for the first ten or fifteen minutes of a burn. That is 6 to 11 watt-hours per start, which is well under a third of an ordinary 10,000 mAh USB power bank, and nothing at all to a modest battery station. If your reason for building this heater is warmth during an outage, a startup fan does not meaningfully compromise that, provided you have thought about it in advance and own the battery.
The non-electric version of the same fix is a bypass damper, and if you are still at the design stage it is the better answer because it costs nothing to run and cannot fail. A short path from the barrel straight to the chimney lets you get the stack hot before you ask the gas to travel the length of a cold bench. Build the bypass and you will probably never need the fan. Build both and cold mornings stop being interesting.
The other half of our fix was not in the heater at all. We insulated the building’s outside walls, which is what finally made the oversized mass behave like an asset instead of a heat sink pointed at the outdoors. After that the bench would carry heat for about half the night, which is what this size of system should do. The arithmetic for why that worked is in the next section, and it is the same point in a different form: the mass was never the problem.
Sizing the mass: the calculation almost nobody does
This is the part where a build stops being folklore. The question is not “how big should the bench be”, it is “how much heat can this bench hold, and how fast does my building lose heat”. Those two numbers together tell you how often you will be lighting fires, and they are the difference between a heater you load once a day and a chore.
Step one: the heat capacity of the mass. Cob runs about 110 pounds per cubic foot with a specific heat around 0.2 BTU per pound per degree Fahrenheit. Multiply those and you get a useful constant:
about 22 BTU stored per cubic foot of cob, per degree F of temperature rise
Take a bench 14 feet long, 30 inches wide, and 20 inches tall. That is 14 × 2.5 × 1.67 = 58 cubic feet. Its heat capacity is 58 × 22 = 1,280 BTU per degree F. Its weight is 58 × 110 = 6,400 pounds, which is 3.2 tons and which we will come back to.
Step two: usable storage. The mass does not heat evenly. The cob against the duct gets very hot while the bench surface stays at a temperature you can sit on, so what you want is the average rise across the whole mass, which in a well-charged bench is somewhere around 40 to 60 degrees above room temperature. At 50 degrees:
1,280 BTU/°F × 50°F = about 64,000 BTU stored
Step three: how long that lasts. Divide by your building’s heat loss rate, and this is where the answer becomes personal:
| Building | Heat loss | 64,000 BTU lasts |
|---|---|---|
| Tight, well-insulated 600 sq ft shop | 4,000 BTU/hr | 16 hours |
| Average insulated 900 sq ft space | 9,000 BTU/hr | 7 hours |
| Drafty uninsulated barn | 20,000 BTU/hr | 3 hours |
Our own build is the worked example of that row. An oversized mass in a building with uninsulated walls is not a bigger battery, it is a bigger radiator aimed at the weather: the extra thermal capacity we had gone to the trouble of building simply gave the outdoors more surface to pull from. We did not fix it by shrinking the bench, which would have meant demolishing it. We fixed it by insulating the outside walls, and the same mass that had been disappointing then held heat for roughly half the night. Nothing about the heater changed. The number in the right-hand column did.
That table is the most important thing on this page. The same heater is a once-a-day appliance in one building and an all-day job in another, and the variable is not the heater. Every claim you read about a rocket mass heater burning once a day and staying warm until tomorrow is quietly assuming a small, tight, well-insulated building. If your envelope is leaky, insulation is a cheaper and faster win than a bigger bench, and it should come first. Our guide to insulation techniques is the honest first step for anyone whose number looks like the bottom row.
Step four: how much wood charges it. Seasoned wood at around 20 percent moisture delivers roughly 6,400 BTU per pound. The mass captures maybe half to sixty percent of the fire’s output, since the barrel is radiating directly into the room the whole time and some heat goes up the stack. So to put 64,000 BTU into the bench you need to burn on the order of 110,000 to 130,000 BTU gross, which is:
120,000 ÷ 6,400 = roughly 19 pounds of dry wood
An armload. That matches what people actually report loading into a batch box for an evening burn, which is a good sign that the arithmetic is describing something real.
Floor loading: do this calculation before anything else
The bench above weighs 6,400 pounds and sits on a footprint of 14 feet by 2.5 feet, which is 35 square feet.
6,400 lb ÷ 35 sq ft = 183 pounds per square foot
Residential floors are commonly designed for a live load of 40 pounds per square foot. You are proposing to put more than four times that on one strip of floor, concentrated along a line.
On a concrete slab this is a non-event and you can stop worrying. Over a framed floor and a crawlspace or basement it is a structural question with only three real answers: build on grade instead, add posts and footings under the bench line to carry it to the ground, or have an engineer size the reinforcement. It is not a question to answer by feel, and it is much cheaper to answer before you have three tons of cob in the room. This is also the calculation an insurer or a building official will ask about first, so having it written down is useful for more than one reason.
Materials
Almost none of this needs to be bought new.
Core: firebrick for the burn tunnel and feed, since ordinary brick spalls at rocket temperatures. Fire clay mortar, not portland cement, which fails with heat cycling.
Riser: a ceramic fibre blanket wrapped form, or a perlite and clay slip mix, or a purpose-made refractory tube. The requirement is insulating, lightweight, and able to survive 2,000 degrees.
Barrel: a 55 gallon steel drum, unlined and burned out first. Food-grade drums are ideal. Burn off any paint or residue outdoors before it comes near the building.
Manifold: brick or a welded steel transition, sized generously as above.
Duct: standard steel stovepipe or ducting of the system diameter. It is entombed in cob, so it is not serviceable, which is an argument for the heavier gauge.
Mass: cob, which is clay, sand, and chopped straw. Urbanite, cut stone, or brick can bulk out the interior. Whatever you use, it needs to be in good thermal contact with the duct, so no voids.
Chimney: proper insulated class A chimney where it passes through anything combustible, with correct clearances and roof height. This is the one component where salvage is a bad idea and code compliance is non-negotiable.
The build sequence, and the step people skip
Order matters here more than in most projects, because two thirds of the way through it becomes very difficult to change your mind.
1. Answer the floor loading and permitting questions. Both are cheap now and ruinous later.
2. Lay the foundation and the base. Non-combustible, level, and sized for the whole footprint including the bench.
3. Build the core dry, without mortar. Stack the firebrick for the feed, tunnel and riser exactly as designed but assembled dry so it can be taken apart.
4. Set the barrel and run temporary duct to the chimney, and then burn it. This is the step people skip and it is the most valuable one in the entire build. Fire it repeatedly. Listen for the rocket sound. Watch the chimney: you want heat shimmer, not smoke. Check that it drafts from cold without pushing smoke back at you. Measure exhaust temperature at the far end. If any of it is wrong, you take a dry stack apart and adjust the riser height or the gap in an afternoon. If you have already cobbed it in, you own a two ton mistake. Do not let anyone talk you out of this step because the weather is turning.
5. Only now, mortar the core and build the permanent manifold and duct run. Photograph everything, with a tape measure in the frame, before it disappears.
6. Pressure-check the duct for leaks before burying it. A smoke test at the cleanouts will find bad joints while you can still reach them.
7. Build the mass in lifts, letting each layer dry. Do not trap moisture inside a thick monolithic pour.
8. Dry the mass out slowly with small fires over days or weeks. Rushing this steams the water inside the cob and cracks it. Small cracks are cosmetic and normal, and cob is trivially repairable.
9. Commission it. Carbon monoxide alarms in place and tested, a flue thermometer installed, the bypass operating, and someone present through the first several full burns.
Living with it
Ash accumulates in the manifold and at the turns, and the system tells you when: burns get lazier, the rocket sound softens, startup gets fussier. Once a season through the cleanouts is a reasonable rhythm, more if you burn hard.
Burn dry wood, and take that more seriously than you would with a conventional stove. The entire clean-burn argument for this design depends on reaching combustion temperature quickly, and wet wood spends the first part of every burn boiling water instead. Small diameter, well seasoned, stored under cover.
Expect cracks in the cob and ignore the cosmetic ones. Expect the first cold start of the season to be the most awkward of the year, since you are warming several thousand pounds of masonry and a cold chimney at the same time. And expect that the thing will not behave like any heater you have owned, because you are no longer operating a fire, you are charging a battery.
If you would rather not tend it: the batch box
Everything above describes the classic J-tube, which is the design most people build first because it is the simplest to get right. The modern refinement is the batch box, developed and documented by Peter van den Berg, which replaces the vertical feed tube with a proper firebox and door. You load it full of splits, light it once, and it burns as a single batch for roughly 45 minutes to an hour with no tending. It is generally cleaner and needs less babysitting than a J-tube, at the cost of a more demanding geometry with tighter tolerances and a secondary air channel that has to be built correctly.
If the tending is what worries you about this project, do not solve it by enlarging a J-tube feed. Build a batch box instead, and take the dimensions from batchrocket.eu, which publishes the full drawings and dimensional tables free. The rest of this article, the CSA rule, the barrel gap, the duct limits, the mass and floor calculations, applies unchanged.
The permies rocket mass heater forums remain the best place to check a design before you build it, and the community there is unusually willing to tell someone their numbers are wrong. Post your dimensions before you buy brick. It is the cheapest review you will ever get, and our companion article on how these heaters work and whether to build one covers the code, insurance and safety ground you should settle first.
