meltcalc

I assembled a spreadsheet and some models for predicting flow for different materials and hotend combinations so I could compare hotends and material flow predictions accurately. But I thought it might be useful for others, so I made a website for different visualizations for hotend and material comparisons.

I found most manufacturers tend to give wildly optimistic flow measurements for their hotends. Most of the calculators for max print speed for different hotends just seem to take manufacturer advertising numbers, or pluck them out of thin air.

Live at meltcalc.baconmilkshake.com.

79 hotends and 36 different base polymers for comparisons.

Max volumetric flow rate and print speed estimations with support for CHT/HF modifiers, different block types, and print temperatures.

Heater power requirement estimations.

Hotend price/performance calculations, including the price of the extender or high flow nozzle when you tick those options.

Hotend comparisons for total power that needs to be dissipated on the melt surface, and how long filament spends in the melt zone for a given flow rate.

Comparisons of estimated max flow rate and melt energy for different materials with a given hotend. This is basically melt index, which is very useful but is sadly not used for head to head material comparisons with a particular hotend very often.

A full description of the model and the design notes behind it is in theory.md. What follows is the short version.

Melting a cubic millimetre costs a fixed amount of energy, and a melt zone can only couple so much power into the filament per millimetre of heated length. Those two facts set a ceiling on volumetric flow, and everything on the site follows from it.

Melt energy. E = ρ (cp ΔT + h_f) per mm³, with cp averaged over the solid to melt interval and h_f the heat of fusion, which is zero for amorphous polymers. ΔT runs from the temperature the filament actually enters at, so a chamber or dryer rather than room temperature, up to the temperature the polymer has to reach to extrude. That is the melting point, or for amorphous polymers the lowest temperature at which they flow. The superheat from there up to the nozzle setpoint is counted separately, because it costs heater watts without changing the melt zone you need.

Flow ceiling. Q = q L / E, where L is the effective melt zone length and q is the power a millimetre of melt zone couples into the filament. There is no clean closed form for q, so it is calibrated on a standard nozzle running PLA sustaining 1.2 mm³/s per mm of melt zone. That works out to about 0.43 W/mm in copper, and every other material scales by the energy it demands to reach its own melting point. The calibration is editable on the site if your measurements disagree.

Effective melt zone. The database holds two lengths per hotend. One is the physical heated channel of a single bore, which is the number that describes the hardware. The other is what the model runs on, and it is entered by hand wherever a hotend does not behave like its dimensions: a multi bore block carries the total across all its bores, and a hotend with high flow geometry built into it carries what that geometry is worth rather than what it measures.

From there the model takes off 3.5 mm for the nozzle taper. Measured back from the tip that lands about halfway along the hex of a V6 nozzle, which is roughly where the bore starts narrowing to the orifice. Past that point there is little wall left against the filament and the pressure behaviour stops helping, so the length is there without melting much. It is a fixed deduction rather than a percentage because the taper is the same size whatever the block behind it is, and it is what brings long melt zones back in line, since they read optimistic against measurements otherwise.

Build options. An extender adds 8.5 mm of real heated length, so it counts against both figures. A CHT style nozzle adds no length but splits the bore into parallel channels, so the plastic meets more hot wall per millimetre. That buys roughly the same melting capacity, so the model counts it as an equivalent 8.5 mm against the effective figure only, and marks those hotends on the charts. Copper is the reference block; aluminium gives up 20% of the flow, brass and steel 30%.

Superheat. Running hotter raises q, but less than proportionally. It scales as (ΔT_set / ΔT_ref)^n, with n picked so that doubling a material's normal superheat gives 1.5x the flow. It caps at 2x, sits at 1 when you are at the material's own setpoint, and drops to zero at the melting point.

Heater power. Reported as P = Q E_setpoint / η at 30% efficiency, next to the smallest cartridge from the sizes people actually stock. Heater power is not treated as a second ceiling on flow, because a cartridge is the cheap swappable part and nobody is stuck with an undersized one.

Residence time. Melt zone volume over flow. The few hotends built for 2.85 mm filament hold 2.7x as much plastic per millimetre and are fed 2.7x slower, so the same melt zone gives them proportionally longer. On a multi bore block the flow and the volume both carry the bore count, so what comes out is the time one path sees rather than the whole hotend.

The backend models are all my own work and are based on first principles: melt zone length, specific heat capacity, heat of fusion, and some approximations for the effects of CHT nozzles and material selection for heat blocks.

For the frontend UI and graphs, the TypeScript and CSS was written with AI assistance if that matters to you.

There are a ton of caveats about heater placement and block geometry that this doesn't represent, so it can lead to flow being underestimated for some high performance hotends (Chube and Tricorn). I chose to go with the same base model for all hotends rather than bias results with a ton of correction factors.

I have not added any provision for the effects of nozzle diameter, though I would like to in the future. For higher flow rate hotends, larger nozzle diameters are almost certainly going to be needed.

There is no pressure drop or melt viscosity model, so a hotend that can melt a polymer may still fail to push it. Material properties are typical published values rather than brand specific measurements.

The material database also carries a practical flow factor, which is the share of the ceiling a polymer actually gets run at. It is shown as a stacked bar in the material views and is deliberately left out of the flow model. What holds PEEK to 20-40 mm/s is not heat transfer, it is interlayer bonding against a chamber well below its melting point, crystallisation, warping and viscosity. A longer melt zone fixes none of that, so folding it into the flow number would make the hotend comparison answer a question it isn't measuring.

The model is checked against published max flow tests at meltcalc.baconmilkshake.com/validation. As it stands that is 98 tests from 7 sources: the centre sits at 0.98x measured over predicted, R2 is 0.76, and about half the tests land within 25% of the prediction.

The per term results matter more than the overall number. The CHT credit comes out light, 1.69x measured against 1.47x modelled, which implies about 10.4 mm of equivalent melt zone rather than the 8.5 mm currently credited. The temperature term runs the other way and is softer than modelled, an exponent near 0.50 against 0.58. Nozzle diameter still has no term in the model and stock nozzles do not appear to need one, but CHT nozzles gain with diameter in a way the model does not capture.

More data would help, so if you have run a max flow test I would like it. Open an issue with:

• extruder
• hotend, and whether the nozzle is CHT/HF or a regular one
• nozzle diameter
• filament type and brand
• the max flow rate you measured

If you ran an Orca flow or temperature tower, the most useful pair of numbers is the rough flow where the surface first goes matte or starts looking slightly underextruded, and the flow where it obviously fails. Either one on its own is fine.

The project is open source, so if you want to contribute more hotends or materials that aren't already on the site, that is the most useful thing you can add.

data/hotend data.csv holds the hotends and data/materials.csv holds the thermal properties. The convention for melt zone length is that it is measured from the top of the hot part of the heatbreak or block down to the tip of the nozzle. Run pnpm data:update-db after editing either file, which regenerates src/lib/hotend-db.ts and src/lib/material-db.ts. Both generated files are committed. A malformed row is a build failure rather than a blank page.

See data/README.md for what the material numbers mean and how approximate they are. Prices go stale and melt zone lengths are measured or inferred from drawings, so corrections to either are welcome.

Install Node.js v24 and corepack:

npm i -g corepack@latest
corepack enable

Fork and clone the repository, then install the dependencies:

pnpm install

Run the development server:

pnpm dev

If you edited the CSVs, write them to the database first:

pnpm data:update-db

pnpm build produces the site and a small server that renders the link previews, and pnpm preview runs the built version.

Two licenses, split by what the thing is.

The website code is MIT. Fork it, ship it, sell it, no strings past keeping the copyright notice.

The model and the data are CC BY 4.0 — theory.md, the description of the model in this README, and the CSVs in data/. Use them anywhere, commercially or not, but credit meltcalc with a link. That is the part I care about, and it is the only condition on any of this.

The web UI was scaffolded from steppersimulator, whose own repository is CC BY-NC-SA 4.0. The permissive licensing here is by its author's explicit permission rather than anything the upstream license grants on its own.

data/validation.csv is the one part that is not mine to license. Those are other people's max flow measurements, gathered from public forum posts, vendor blogs and articles, and every row carries the source it came from in its Citation column. The CC BY grant covers the compilation, the cleanup and the analysis built on top of it, not the underlying measurements, and none of those sources have endorsed any of this. If you published one of these numbers and would rather it were not here, open an issue and I will remove the row.