The Scientific Hobbyist

Your kiln knows the temperature. It doesn't know where the glass is.

That distinction is the difference between annealing and hoping.

You already know what thermal shock is. You know the annealing range for COE 33 borosilicate — 515°C to 565°C — and you understand why the ramp through that range has to be controlled. You've read the thermal expansion coefficient data. You have opinions about which glass formulations are more forgiving and which are not.

So I'm not going to explain annealing to you.

What I want to talk about is a measurement problem that almost every annealing kiln in the accessible price range has — and that almost no one in the lampworking equipment space has bothered to document, because documenting it means admitting that their product can't solve it.

A standard annealing kiln has one thermocouple. It is mounted in the kiln wall or ceiling. It measures the temperature at that point.

It does not measure the temperature of the glass.

During the annealing range — 515°C to 565°C for COE 33, the critical window where molecular stress can still redistribute before it's locked in — the temperature differential between the probe position and the glass surface ranges from approximately 30°C to 80°C, depending on load, position in the work volume, and kiln geometry. That differential varies by session, by how much glass is in the kiln, and by where you place it.

A fixed annealing ramp — the programme you set before you start — is calculated against the probe temperature, not the glass temperature. If your probe is reading 540°C and your glass is at 480°C, your ramp is advancing through what your controller thinks is the annealing range while your glass hasn't entered it yet.

The fixed ramp problem compounds when you're running multiple pieces of different mass. A single bead of 4g and a sculptural element of 22g have different thermal mass — they heat and cool at different rates. A fixed ramp calibrated for the average of those two pieces is wrong for both of them.

This is not a marginal error. For a 22g piece in a kiln with a ramp calibrated for a 4g bead, the heavy piece may not reach the annealing range at all during the hold period. The stress doesn't equalise. It locks in. The fracture happens six hours later when you're handling the finished piece — or six days later when your customer is.

The Ignis Algorithm is not a marketing term for a smarter fixed ramp. It is a responsive annealing controller trained on thermal expansion data for 47 borosilicate and soft glass formulations.

At session start, it takes three inputs: the glass type you select from the library, the estimated mass from the infrared sensor at the kiln entrance (±15% accuracy), and the opening temperature differential across the seven-zone array at the moment of glass introduction. From these it calculates the opening phase — the most critical transition, where the glass is moving from working temperature toward the annealing range.

During the annealing hold, it monitors the differential across all seven zones in real time. If glass in zone 3 is cooling faster than glass in zone 6, the algorithm extends the hold time for the zones showing faster cooling — adjusting the ramp to match the actual thermal state of the actual glass, not a generic assumption about what glass of that type and approximate mass should be doing.

The curve it runs for three 8g beads is different from the curve it runs for one 4g bead and one 20g sculptural element. This is not marketing. It is thermodynamics applied correctly.

The Ignis-7 glass library is published in full at ignis7.com/glass-library. Each formulation entry includes: COE, annealing point, strain point, softening point, and notes on colour chemistry interactions (cadmium, cobalt, silver-bearing glasses) that affect annealing behaviour.

The library is the result of Dr. Mara Chen's materials science research combined with field documentation from the Ignis beta user group. It is updated when new formulations are documented. When you select a glass type, you are giving the algorithm the specific thermal parameters it needs to optimise the curve for that glass's actual behaviour — not a generic borosilicate approximation.

The Ignis-7 manifold controls propane and oxygen flow to a resolution of 0.01 LPM via CNC-machined needle valves. Both gas lines have integrated flash arrestors. The pressure gauges read to ±0.5 PSI accuracy.

The practical implication: your oxygen:propane ratio at a given pressure is a documented parameter. You can record it, reproduce it, and share it. The flame character that produced a particular result in a particular glass type is no longer a tacit memory — it is a session data point exportable via USB-A.

A hobbyist needle valve operates at a resolution of approximately 0.1–0.5 LPM. The flame character varies by valve position time, temperature effects on the valve seat, and upstream pressure variation. It is not a reproducible variable. The Ignis-7 manifold makes it one.

Gas regulators: not included — regulator specifications must be matched to your specific gas supplier and tank configuration. We specify CGA 510 for propane, CGA 540 for oxygen, and provide a verified supplier list at ignis7.com/setup.

The torch: not included — the manifold is torch-agnostic. Compatibility verified for the 12 most common torch models at ignis7.com/compatibility.

Without the Ignis Algorithm. Without the safety integration. Without the glass library. Without session data logging. Without the assembler signature and calibration certificate.

The session data export is USB-A. Format is CSV. The full parameter log — gas ratios, pressures, zone temperatures, annealing curve — is yours to analyse.

Ignis-7 Digital Lampworking Station — $4,500 direct. Full technical specification, testing data, glass library, and alternative-sourcing calculation at ignis7.com/technical. Assembled in Portland, Oregon.