RTV Silicone Mold Design Mistakes That Force Recasts

If you're reviewing a custom RTV silicone mold, most recasts are drawing mistakes, not pour mistakes. The usual set: treating elongation as a draft waiver on a future *part* tool, forgetting that shrink is process-specific, parking a logo on a shut, writing Shore A as if it were a compound, and freezing CAD before a split or a pour direction exists. Each one is cheap to fix in STEP. Each one is a second master or a second pour in the shop.

This is a diagnostic list for RTV design reviews. It isn't a gallery of every possible flash defect on an LSR steel tool. Those *part* tools are a different WYD-style quote.

Treating elongation as a draft waiver

Silicone stretches. That isn't a license for a barb that will tear a 10A–20A lip. Platinum or tin RTV cavities can peel with little *mold* draft because the rubber flexes. Copying that behavior onto an LSR steel tool is the mistake — and that conversion is a different RFQ. If the RTV prototype was a glove mold of a figurine, the production conversion still needs metal-tool draft and a strip direction.

Soft grades (low Shore A) tear more easily on a barb even though they elongate more. Don't use 10A as an excuse for a deeper undercut. That's how a letter serif or a sealing bead comes back in two pieces. Texture makes the same mistake worse: grain on a master is a field of tiny undercuts. If you add grain after a zero-draft review, you have a new part.

Shop order: add a fillet so the feature strips; test a shallow undercut on the grade; split the mold or pick a core; or change the article. Don't skip to "it will peel" because a video showed a glove mold of a miniature.

Forgetting shrink is process-specific

Many platinum RTV mold rubbers shrink about 0.1–0.3% after a full cure. Tin-cure RTV continues to shrink in storage. Heat-cured LSR and HCR *parts* shrink in a larger band (LSR often about 2–3%, HCR about 1.5–2.5%). If two people each think the other applied compensation, the article is systematically off, and the CMM won't save you.

Mistakes that look like "the shop cannot hold tolerance":

• LSR shrink copied onto an RTV master.
• RTV near-zero shrink copied onto LIM steel (wrong object).
• CAD already compensated *and* the pourer compensated again.
• Tin mold measured on pour day and treated as a library dimension next season.

Write the policy on the 2D. One owner. Measure first article against that policy, not against a metal ISO 2768 fantasy. Rubber articles are often dimensioned with ISO 3302-1 classes; that's a drawing class, not another word for "precision." Using a machined-metal default is a related drawing mistake.

Putting the logo on the wrong face

On RTV, a molded-in mark lives on the master. It needs depth, draft on serifs, and a home that isn't the shut. Logos on the parting get sheared. Logos on a stretchy undercut fill once and tear. Logos that are a 0.1 mm hairline will copy in a soft RTV peel and starve later if someone treats the same art as HCR compression — which is a different tool.

Pad-print isn't molded-in. Drawing a print as engraved master wastes finishing time and still needs a print process. Decide. Recessed marks in the pocket hide wear from knives and washers. Raised marks on a bakeware face take abuse.

The other wrong face is the sealing lip. Texture and logos there leak or print into the mating part. Hide the ego.

Specifying Shore A as if it were a material

50A platinum LSR, 50A peroxide HCR, and 50A tin RTV aren't one rubber. Tear, compression set, extractables, and cure chemistry differ. Food-contact (FDA 21 CFR 177.2600, LFGB) lives on a grade and a test, not on a durometer. "Medical 50A" isn't ISO 10993.

Design mistakes follow: a drawing that says 20A because the CAD engineer liked the flex, with no process, then a shop quotes HCR that can't fill the membrane. Or a buyer copies a craft RTV TDS onto an LSR RFQ. For this site, name platinum RTV or tin RTV, then pick Shore A inside that family — often 10A–20A for a food or gummy skin with a jacket behind it.

Color is the same class of mistake. A Pantone on a title block isn't a pigment package, and some pigments move cure or food-contact status. Call the grade, then the color.

Freezing CAD before the split and pour exist

A pretty article with no strip direction, no parting, no pour seam, and no jacket callout isn't frozen. It's a sculpture. Shops that pour it anyway will pick a split you hate. Shops that wait will look slow. Both are reactions to the same design hole.

Before freeze:

• Parting visible on a screenshot.
• RTV pour seam; keys if it splits; core pull path if it's hollow.
• Jacket / mother mold: yes/no and material.
• Master finish: CNC / post-cured SLA plus barrier / supplied sample.
• Logo depth and which face.
• Shrink owner.

Manufacturers in China or anywhere else will pour the sculpture if the PO says hurry. That isn't a cultural mystery. It's an incentive. Stop the incentive in the design review, not in the complaint after first article. A marked split screenshot takes minutes. A recast takes the calendar you already promised to sales.

Related misses that show up in the same review: missing radii at the root of a rib, walls of rubber "as thin as possible," and a wet dump on a 10A–20A skin with no jacket. Those aren't exotic. They're the difference between a DFM mark-up and a silent pour.

The recast you're trying to avoid is almost always one of these five. Find it on the PDF while the A/B is still in the kit.

Common mix-up: one DFM checklist for RTV, LSR, and HCR

A lot of RFQs treat "silicone DFM" as one PDF. In the shop that usually means LSR draft and gate notes on a chocolate tray, or an RTV glove's zero draft copied onto steel. RTV peels. LSR injects into a heated cavity with a cold-runner. HCR compresses a millable charge. Draft, shrink, flash, and gates don't transfer.

What to ask when you're buying

• Is the process on the title block platinum RTV or tin RTV (not "silicone")?
• Where is the split, and are keys drawn, not "shop to advise"?
• Jacket / mother mold: yes/no, material, and which Shore skin sits against the master?
• Who applies shrink, and is the CAD already compensated?
• Logo depth and which face; undercuts: peel, split, or core?

Worked example: 40 mm lollipop letter, negative draft

A confectionery buyer needed a 40 mm lollipop with a letter that had negative draft, about 800 pours a month, food-contact, single-cavity first tool then a 6-cavity sheet. We'd move the split off the letter face, put CNC bosses on the jacket, and keep a high-tear platinum skin against the master so the serif wouldn't tear on peel. Copying the zero-draft RTV peel onto a future LSR *part* tool would have been a recut of steel; that stays a separate conversation.

When this diagnostic does not apply

If production is molded LSR or HCR *parts*, don't freeze an RTV glove as the production toolpath. Open a parts RFQ with strip direction and runner notes — different WYD-style quote. A tiny shallow one-piece tray can skip keys. Geometry can override quantity: one undercut-heavy 80 mm figurine at 30 resin casts still needs a multi-piece jacket.

Related reading: design before the pour, how molds are made, custom logo on the master, custom RTV molds.

Design-review pack to copy onto the 2D

Copy this onto the freeze:

• Process: platinum RTV / tin RTV
• Shrink owner (shop or customer); CAD compensated yes/no
• Parting screenshot; key style; core pull path
• Jacket material and Shore of the skin
• Master finish: CNC / post-cured SLA plus barrier / supplied sample
• Logo: raised or recessed in the pocket, depth, face
• Freeze date and revision letter

Send the STEP, a marked split screenshot, and the dump material. We'll mark keys, jacket, and whether the undercut is a peel, a split, or a core before anyone mixes A/B.

We would not recommend freezing zero draft because a glove mold peeled

We would not recommend copying RTV peel rules onto an LSR part tool. We would not raise Shore A first to hide a missing split. We would not leave shrink with two owners. We would not park a logo on the shut and call it "branded."

RUUIPON Mold Structure Decision: flat back? open-back. Deep undercuts that will not peel? two-part, plug, or inner core. Soft 10A–20A or large wet casting? mother mold. Don't skip the jacket to "save design time."

flowchart TD
  A[Title block says platinum or tin RTV?] -->|no| Z[Wrong DFM pack]
  A -->|yes| B{Can it peel without tearing 10A-20A?}
  B -->|yes, shallow| C[One-piece plus jacket if soft]
  B -->|no| D[Keyed split or inner core]
  D --> E[Freeze keys, pour direction, logo face]
  C --> E

Typical shop values

Caption: Typical shop values (not a named lot).

| Item | Typical shop value | | --- | --- | | Cavity Shore | 10A–20A skin | | Mix | Platinum 1:1 or 10:1; vacuum degas | | Platinum shrink | About 0.1–0.3%; do not paste LSR 2–3% onto the master | | Loaded-tray edge sag without a jacket | About 0.5–1.5 mm | | HCR *part* contrast | About 1.5–2.5% on a different drawing | | 40 mm lollipop letter, negative draft | Single cavity first, then 6-cavity sheet at about 800 pours/month |

FAQ

Why did my zero-draft RTV peel tear if silicone is stretchy? Elongation isn't infinite, and a 10A–20A lip tears on barbs and on serifs with no fillet. Add a split, a core, or a fillet. Don't cite a glove-mold video. If you meant an LSR *part*, that's a different RFQ with metal-tool draft.

Who should apply the shrink factor? One named owner, shop or customer, written on the 2D. Dual compensation and an LSR percent on a platinum master are the usual dimensional misses.

Can I add the logo after the mold is poured? Only as a recast against a refinished master, or as pad-print. Engraving a finished RTV face isn't a shop process. Put the mark on the master in the first CAM.

Is "50 Shore A food grade" enough on a drawing? No. Name platinum RTV or tin RTV, plus the test path if you need FDA 21 CFR 177.2600 or LFGB. Durometer is a point on a curve, not a grade.

What's the cheapest design review that prevents a recast? A 30-minute call on parting, shrink owner, process, and logo face, with a marked screenshot. Skipping it to "save time" is how you buy the pour twice.