Designing Custom RTV Silicone Molds for Depositing Lines

If you're putting a custom RTV silicone mold on a depositing line (or a chocolate wheel), design starts at the master and the jacket — not after a robot vendor visits a tray only a person can peel. The line needs consistent presentation: jacket nests or pins, dump orientation that repeats, fill height, no undercut that needs two thumbs every cycle, and flash that doesn't vary by cavity. If those are missing, you've bought a depositor and a payroll.

This isn't a hymn to lights-out factories. Many 2026 gummy cells still mix a depositor with a human demold. That's fine. Design the mold so the mix is a choice, not a rescue. LSR valve-gate take-out is a different cell on a parts OEM. If the RFQ is full of EOAT and cold-runners, you're on the wrong product line.

Jacket presentation on the line

Decide how the tray sits: pin nests, pocket in a fixture, or "we'll tape it." Then design:

• A locate that repeats every dump (no flipping in the fixture)
• Dump orientation vs gravity and the depositor nozzles
• Fill height and pour-head so end cavities aren't starved
• A grasp or peel zone that isn't a show face or a logo
• A place to park the tray for a vision check (flash, fill) without stretching the skin

Low-durometer RTV will sag in a floppy glove. Design the jacket for sag, or cool enough to handle, or both. Optical A-range faces stick to pads; a specified duller grip on the jacket rim can be the handling feature.

Silicone's stretch lets you skip some splits. That stretch is the enemy of a line if the dump hangs on a core and needs a peel path. Engineering should call which undercuts are stretch-off and which need a keyed split. Stretch-off that only works with two human hands isn't automated.

Flash and mixed cavities as line killers

Variable flash changes the silhouette a camera sees and the thickness a demold jig expects. It also means a human with a scalpel is still in the cell. Design keys for the flash spec, then maintain them. Automation without PM is a week of success.

Multi-cavity trays that mix flash by cavity will teach the line the average and fail the outlier. Balance and recast are depositing projects. Cycle-time claims should include demold, not only deposit. A 12-second dump with a 20-second peel isn't a 12-second cell. Put take-out time on the same sheet as dump, or the sheet is fiction.

Common mix-up: "add a robot" to a glove with no nests

A lot of RFQs treat a production tray as something an integrator can "just pick up." In the shop that usually means there is no grasp zone, no jacket pin, and no flash spec. China cluster programs often automate later, after volume is real. That's rational. What isn't rational is pouring a 24-cavity skin with no locate and then asking the line to invent one. Put nests on the first production jacket.

Inserts, two-shot resin, and downstream print should see the same orientation the jacket provides. Dropping dumps into a bin to "sort later" throws away the only free orientation you had.

What to ask when you're buying

• Intended take-out: depositor + human peel, jig, or future robot?
• Jacket locate: pins, nests, fixture envelope from the integrator?
• Max flash at gripper / vision window; peel keep-outs?
• Dump orientation and fill height on the drawing?
• If year one is manual and year three might be a cell, is the grasp zone already on the jacket?

If the first year is operator and the third year might be a robot, still add the nest. Recutting a show face later to make a pad land is an ECO you can avoid.

When depositing-line design is the wrong brief

Hand-poured resin jewelry at 40 casts doesn't need pin nests. LSR part take-out (valve gates, EOAT crash volume) is a parts-OEM conversation. Geometry can override volume: sticky high-undercut gummies may stay on a smaller tray because peel, not pocket count, is the cycle.

Related: high-volume dumping, engineering before the pour, gummy trays.

Worked example

A 24-cavity gummy bear, 18 mm, food-contact platinum, fiberglass jacket with two dowel nests for a depositing line, tens of thousands of dumps in season. Layout froze dump-up orientation, kept stretch-off ears in the split discussion, and put a dull jacket rim for handling. A shop that offered "we'll add a robot later" without nests was declined. Flash at the split was the thing that actually kept a person in the cell until they recast two nicked keys.

Line-design note block to put on the drawing

• Take-out: depositor / operator / jig
• Jacket locate (pins, nests) + fixture envelope
• Grasp keep-out and peel path
• Max flash at vision window
• Dump orientation and fill height
• Do not oil or spray unless listed (contamination and gripper slip)

Send the fixture envelope (or a photo of the depositor nest) and monthly dumps. We'll design the jacket locate and peel path on the first production skin — not after the first sticky dump.

We would not recommend adding a robot to a glove with no nests

We would not recommend pouring a 24-cavity skin with no locate and asking the integrator to invent one. We would not treat stretch-off that needs two thumbs as automated. We would not copy a gummy nest RFQ onto a hollow chocolate with mandrels. We would not oil or spray a food tray to help a gripper if that spray was not on the first dump.

RUUIPON Mold Structure Decision: pin nests or a fixture pocket on the *first* production jacket. Soft Shore 10A–20A? jacket for sag, or the outer row goes 3–6% heavy. Undercuts that cannot peel? keyed split before the cell. Year-one manual, year-three maybe a robot? still add the grasp zone now.

flowchart TD
  A[Line] --> B{Jacket locate: pins / nests?}
  B -->|we'll tape it| C["Stop: not a depositing mold"]
  B -->|yes| D{Peel needs two hands?}
  D -->|yes| E[Add split or change undercut]
  D -->|no| F[Dump orientation + fill height + flash spec]
  F --> G{Sag 0.5-1.5 mm on outer row?}
  G -->|yes| H[Stiffer jacket / fewer cavities]
  G -->|no| I[Rate vs peel time on the same sheet]

Typical shop values

Caption these as typical shop values, not a named lot.

| Item | Typical shop value | | --- | --- | | Depositing cavity count | 12–24 pockets is a common gummy/chocolate band | | Cavity Shore | 10A–20A platinum RTV, fiberglass jacket with dowel nests | | Mix | Platinum 1:1 or 10:1 by grade; vacuum degas until the foam collapses | | Platinum shrink after full cure | About 0.1–0.3% | | Loaded-tray edge sag without nests / jacket | About 0.5–1.5 mm | | Outer-row gummy vs center if the loaded tray sags | Edge often 3–6% heavy | | Cycle sheet | Include peel/take-out; a 12 s dump with a 20 s peel is not a 12 s cell |

FAQ

Can I automate a tray that still needs hand deflash? You can automate dump and still trim. You shouldn't pretend the labor is gone. If trim is the cost you wanted to kill, fix the keys and the split first. A robot presenting a flashy gummy to a scalpel is a conveyor, not a cell.

Do more cavities make robot take-out easier? They can raise rate if the jacket locates and flash is even. They also add pockets that die first. On low volume they may be a monument. On high dump counts they're part of the same decision as the fixture. Quote both as a system.

How soft is too soft for vacuum handling on RTV dumps? It depends on wall, temperature, and sag, not only Shore A. Very low-durometer membranes wrinkle and mark. Jacket rims, tabs, or a mechanical fork may beat vacuum. Prove it on first dumps at demold temperature, not on a room-temperature souvenir.

Should vision inspect every cavity? If flash or fill is a launch risk, yes, at least at SOP start. Vision trained on cavity 1 will fail cavity 6 if balance is off. Keep cavity identity in the image. Vision isn't a substitute for key PM.

Is a chocolate wheel the same design problem as a gummy depositor? Same jacket-locate idea, different temperature and demold. Don't copy a gummy nest RFQ onto a hollow chocolate with mandrels and expect the same peel. Name the line.