Print the paste. Fire the board.
The machine on the homepage runs on one technical bet: low-temperature co-fired ceramic, laid down as paste and fired on your bench. This page is the reasoning behind it: the material, the machine architecture, why the same process prints the passives in, and the single problem the whole thing turns on.
An engineering brief for technical reviewers: reasoning and figures, not a finished spec.
FR-4 is the default. Ceramic is the ceiling.
Glass-epoxy (FR-4) is cheap and everywhere, and for most boards it's the right call. But it gives up on thermal conductivity, high-frequency loss, and hermetic sealing, which is exactly why RF front-ends, power modules, and aerospace and defense electronics (the hardware that has to survive the most extreme environments on Earth) move to ceramic substrates.
The catch has always been access. Co-fired ceramic lives behind specialist fabs with long lead times and high minimums: fine for a finished design, wrong for the early iteration loop. PrometheusPCB puts that substrate on the same bench you design at. Advanced manufacturing, desk-sized.
Firing temperature decides which metals you get.
Co-fired ceramic comes in two regimes, and the only real difference is heat. HTCC sinters at 1,400–1,600 °C, hot enough to rule out every high-conductivity metal, leaving refractory conductors like tungsten and molybdenum that run roughly an order of magnitude more resistive.
LTCC sinters at 850–900 °C. That window sits just below silver's melting point (961 °C), low enough to co-fire silver, copper, and gold, the conductors you actually want for RF and high-speed digital. Everything downstream follows from staying inside it.
| Property | LTCC | HTCC |
|---|---|---|
| Firing temperature | 850–900 °C | 1400–1600 °C |
| Usable conductors | Silver, copper, gold | Tungsten, molybdenum |
| Conductor resistivity | Bulk-metal class (low) | ~10× higher |
| Embedded passives | Yes: R, L, C in-layer | Limited |
| Best fit | RF / mmWave, high-speed digital | Rugged hermetic packages |
TABLE_01 · co-fired ceramic regimes
The deposition primitive is paste.
A fused-filament printer melts a thermoplastic strand. Co-fired ceramic isn't a thermoplastic; it's a ceramic-and-glass powder suspended in a binder, and the conductor is a separate metal-loaded paste laid down in the same build. Neither one feeds through an FFF hot end.
The correct primitive is direct ink writing, or robocasting: extrude a loaded paste through a fine nozzle under controlled rheology, place dielectric and conductor in the same pass, then co-fire the green part into a solid board. That fork in the road is what separates a real ceramic process from the FFF-based "PCB printer" projects that stall on materials.
The build log shows filament on our own bench, so it's worth being explicit. Today's hand-run coupons and test articles are FFF-printed from glass-filled filament, the fastest way to a fireable green body before the paste toolhead exists, and a deliberately conservative one. A filament body is the crudest green body this process will ever see: more binder to burn out, lower glass loading, more residual porosity than a paste we'd formulate. What we're testing is whether an LTCC-style board can be made in-house at all. If a circuit survives co-firing in that body, the test has proven the floor, not the ceiling. Moving to paste is an upgrade with decades of production LTCC behind it, not a new open question. The conductor already goes down as dispensed ink; the machine is, and stays, a paste process.
One melted strand
A single thermoplastic bead. Ceramic-and-glass powder won't behave as a filament, and there's no path to lay conductor in the same build.
Two pastes, one pass
Dielectric paste and metal-loaded conductor paste, placed in the same pass under controlled rheology, then co-fired into one ceramic board.
FIG_02 · deposition primitive · grey = dielectric, copper = conductor
The same nozzle that prints traces prints the passives.
This is the part that makes ceramic worth the trouble. Because direct ink writing already places more than one paste in a build, a resistor is a stripe of resistive paste between two vias; a capacitor is an electrode / dielectric / electrode stack; an inductor is a coil of the same conductor. All of them are formed inside the layers, in the same pass that lays the wiring: the headline capability LTCC has had in production for decades, pulled onto the bench.
That matters because passives dominate a bill of materials. Resistors and capacitors are routinely 70–90% of the part count on a real board. Print those in and the machine stops being a wiring printer and starts being a functional-electronics printer. What's left to place is the active silicon, plus the short list of tight-tolerance holdouts the caveat below names.
Resistive paste, in-layer
A length of carbon- or metal-oxide-loaded paste. Its resistance is set by geometry (width, length, thickness), so the CAD sets the value.
Electrode · dielectric · electrode
Two conductor plates around a high-k dielectric paste. Capacitance follows plate area and dielectric thickness.
Planar spiral coil
A spiral of the same conductor paste. Low inductance versus a wound part, but exactly what RF and some power rails need.
FIG_025 · the three passive primitives · copper = conductor, teal = dielectric
One materials problem. The whole architecture turns on it.
Fire a green board and both materials densify and shrink, but not by the same amount and not at the same rate. Control that co-firing shrinkage mismatch and the board comes out dimensionally true. Lose control of it and you get warpage, delamination, and broken traces.
Production LTCC fabs hold in-plane shrinkage to about ±0.2% using constrained-sintering techniques. Reproducing that kind of control in a desktop-scale machine is the central unsolved problem of this program. It has company on the bench, too: the build log tracks fired-part brittleness alongside it as a second measured materials problem.
Ceramic, open, and the passives printed in.
Capable, sealed, polymer
Multi-layer additive electronics exist today as six-figure closed appliances built around polymer and silver-ink. Capable machines, but not co-fired ceramic, and not open.
Real ceramic, off your bench
Co-fired ceramic made well, with the lead times, minimums, and queue that put it outside the early iteration loop where it'd help most.
A ceramic loop on your desk
A ceramic iteration loop a single team can afford, with the passives printed in, on an open, modular architecture you can build on instead of a sealed box.
Funded, founder-led, and building the first prototype.
PrometheusPCB is backed by the Texas Tech Innovation Hub JumpStart grant and NSF I-Corps customer-discovery funding, and is built by Aaron Wang (electrical systems + ME). The literature phase is behind us: heads-down building the first hand-run prototype and turning the open questions into bench experiments, while talking to the teams and labs who'd use it.
The seat still to fill is deep materials. If you work in ceramic processing, paste rheology, or constrained sintering, and the shrinkage problem above reads as a challenge rather than a dealbreaker, we should talk.