Design freeze decides the route: a manifold with an ECO freeze record can take hard tooling, an unfrozen one stays on printing. The date the 3D model stops changing is what matters, not the day the drawing is issued. Every route cost after that point follows from it.
A design freeze is the point where the model and the drawing stop moving, so the cavity, the gating and the fixture can be built once. Before that point, one change shifts the runner centre, the flange face and the port position together. On a 2.5 mm wall manifold those three features are linked, so a tool cut early carries the error into every part it makes.
The freeze is a process decision, not a paperwork step. A frozen model fixes the datum scheme for machining and the allowance budget for vacuum investment casting. An open model moves both, and the second move is the costly one.
The three routes absorb the same manifold geometry change at very different speeds.
Printed metal: re-slice and reprint, about 1 day, no tool touched.
Printed wax pattern: about 2 days for a new pattern, then the normal 12-day shell and pour cycle.
Hard tooling: the cavity is recut in about 6 weeks for a 4-cavity tool, plus the first article.
That gap is what the freeze is worth in days. One day of change is affordable during development, and 2 days is tolerable. Six weeks is not, and a tool cut against an open model pays the cycle twice.
Printing carries an unfrozen design because nothing permanent is cut. Printed metal gives a fit check on port alignment and bracket clearance within about 4 days, which is fast enough to iterate twice inside two weeks. It does not prove casting behaviour, so it answers geometry questions only.
A precision wax pattern is the middle choice. The pattern changes in about 2 days, and the casting that follows runs the same 7-layer shell and 1600 °C pour that production will use. For an unfrozen design that still needs cast data, this is the shortest path to it.
Hard tooling carries a frozen design. A 4-cavity tool holds one geometry across 300 parts or more, and the price per part reflects that. Cutting it before the freeze turns a design decision into a scrapped cavity.
An engineering change order (ECO) record, not a verbal agreement. The record names the model revision, lists the changes it closes, and states the date the geometry was locked. A tool released without it has no baseline for the first article.
The freeze record also sets the allowance. This manifold keeps 1.5 mm on the six runner faces and the turbo flange for a casting route, and 1.0 mm for a printed part. Both numbers assume a frozen geometry, so a late change sends the allowance back through review before the first cut.
Design freeze expires when the requirement changes, not when time passes. A new heat cycle, a different turbo flange interface or a tighter flatness target reopens the geometry, and the route is checked again. The freeze record is a trigger to re-examine, not a permanent approval.
Two limits hold even after a freeze. A frozen design still loses money below 300 parts per batch, because the tool is not amortised, and it still cannot hold a wall under 2.0 mm on a casting route. A freeze makes tooling possible; it does not make tooling sensible.
Flatness shows whether the decision held. The machined faces run to 0.05 mm and are checked on a coordinate measuring machine (CMM), with the casting tolerance at CT6 under ISO 8062-3. Dimensional control at this step is what catches a change that was never frozen.
Send the model revision and the ECO status with the request, and the route recommendation comes back with the change cost for each option: 1 day, 2 days or 6 weeks.