Multi-Cavity Mold Design: Balancing Runner Systems for Uniform Cavity Pressure

Cavity 1 is full. Cavity 8 is short. The runner is not “almost balanced.”

Uniform cavity pressure is the whole point of a multi-cavity silicone tool. You can machine eight identical impressions and still ship eight different parts if the rubber does not see the same pressure, temperature, and fill time in each hole. LSR cold-runner systems and HCR transfer/injection runners fail in different ways. RTV pour boxes with eight masters in one jacket are not runners at all; they are a gravity and air problem. Do not copy a thermoplastic fishbone into any of the three and call it done.

If you ignore balance, the defects are shorts, flash on the first cavities, cure mismatch, and a dimensional spread you will try to blame on ISO 3302-1.

What “balanced” means in rubber

Equal runner length on the CAD is not equal pressure at the gate. Balance is:

• Same fill time (or a known, tiny offset you designed).
• Same pressure when the cavity is full, so flash and density match.
• Same temperature at the gate, especially on LSR where the cold runner meets a hot cavity.
• Same cure at open, so cavity 8 is not green when cavity 1 is already tearing on eject.

LSR viscosity drops with shear. A short, fat path to cavity 1 and a long, thin path to cavity 8 will not “even out because silicone flows so well.” It will fill 1, pack 1, flash 1, and starve 8.

HCR is thicker and knits. Imbalance shows as incomplete fill and as hardness or density scatter after the press.

LSR: the cold runner is the job

Production LSR tools often use a cold runner (water-cooled) so the sprue stays uncured, feeding hot cavities. Balance that system like a manifold:

• Naturally balanced layouts (H, X, circular) beat a straight spine with eight takeoffs. The last takeoff on a spine is a different tool.
• Gate land and diameter must match. A worn gate on cavity 3 is a new runner.
• Valve gates, if you have them, need open time as part of balance, not only geometry.
• Cavity spacing for heat. A cluster that is thermally a blob will cure the center shots differently even if the fill was pretty.

Family molds (two different parts in one LSR tool) are unbalanced by intent unless you size drops for each volume. A 2 cm³ gasket next to a 15 cm³ grip is not “close enough.” Either dedicate cavities or accept a process window that is a tightrope.

Flash threshold on LSR is tiny. Overpacking the easy cavities to fill the hard one is how you drown the tool in flash and still have a short.

If you ignore thermal balance, the defect is a part that fills and then sticks in the hotter cavity.

HCR: the pot and the cull are the runner

Compression HCR is often one charge per cavity - balance is charge weight, not a runner. Transfer HCR uses a pot and runners; those runners cure and become scrap. Balance still matters, but you pay for every millimetre of runner in material.

Do not design a long, pretty HCR runner network as if it were LSR. Keep paths short, equal, and vented. Transfer imbalance shows up as knit lines and as flash at the pot.

RTV multi-cavity is not a runner conversation

Eight jewelry masters in one pour: tilt, last-fill vents, and a jacket that does not bulge. There is no injection pressure to balance. Vacuum helps. A “balanced sprue” carved in clay is a pour path. Useful. Not an LSR cold runner. If the production plan is 8-cavity LSR, the RTV eight-up only proves the part, not the manifold.

How to know you are not balanced

Shop checks that do not require a branded simulation suite (simulation helps; it is not a proprietary secret workflow):

1. Short-shot series. Stop early. Which cavity fills last? That is your restriction.
2. Flash map. Which cavity flashes first when you add a little shot size? That one is overfed.
3. Weight scatter of as-molded parts (same trim). More than your shrink study can swallow means pressure scatter.
4. Cavity-to-cavity dimensions on an F feature. ISO 3302-1 M2 on a 16-cavity unbalanced tool is a wish.
5. Cure / stick differences. Green or sticky in one hole only is heat or fill time, not “that silicone.”

Fix geometry or gate size before you “tune it in the process” forever. Process can hide a mild imbalance. It cannot hide cavity 8.

Cavitation versus class

M1 and even honest M2 want fewer cavities until shrink and balance are known. Duplicate steel after the first shots. Jumping to 16 cavities to hit a piece price is how all sixteen are wrong the same expensive way.

Aluminum prototype multi-cavity LSR can prove a runner idea. Wear and thermal drift will still move balance. Production balance lives in steel that you can still polish and recut.

FAQ

Why isn't equal runner length on the CAD a balanced LSR tool?

Balance is equal fill time, equal pressure at full, equal gate temperature, and equal cure at open. LSR viscosity drops with shear, so a fat short path packs and flashes while a long thin path starves. Matching branch drawings does not equalize pressure at the gate.

Can a small gasket share an LSR tool with a large grip?

Family molds are unbalanced by intent unless you size drops for each volume. A few cubic centimetres next to a large grip is not close enough. Dedicate cavities or accept a process window that is a tightrope.

How do I prove cavity 8 is the restriction without branded simulation?

Run a short-shot series to see who fills last, map which cavity flashes first when you add shot size, weigh as-molded parts after the same trim, and check F-dimension scatter. Green or sticky in one hole only is heat or fill time, not "that silicone."

Does an RTV eight-up prove an 8-cavity LSR manifold?

No. Eight masters in one pour are gravity, tilt, vents, and jacket bulge. There is no injection pressure to balance. The RTV eight-up can prove the part, not the cold runner.

Should I jump to 16 cavities to hold ISO 3302-1 M2?

Opposite. Honest M2 wants fewer cavities until shrink and balance are known, then duplicate steel. Sixteen wrong impressions is an expensive way to miss class. Fix geometry or gate size before you "tune it in the process" forever.