ISO 3302-1 for RTV Silicone Molds: A Drawing Class

If you're writing tolerances on a custom RTV silicone mold, silicone mold tolerance isn't machined-metal tolerance. Cite ISO 3302-1 (rubber products: dimensional tolerances) as a *drawing class* for rubber articles, pick a class that matches the face, and stop pasting ISO 2768 or a random ±0.05 mm on a 30A lip. ISO 3302-1 isn't another word for "precision," and it isn't a capability trophy. The master can be cut tighter than the dump will measure. Jacket sag, tin movement, and probe sink will eat a metal-style box.

This is a buyer's language guide for RTV cavities and the articles you cast in them. It doesn't invent a capability table. It doesn't claim a shop Cpk.

Rubber isn't machined metal: ISO 3302-1 as a starting language

ISO 3302-1 gives classes (commonly discussed as M1–M4 for molded products, with M1 the tighter commercial conversation and M3/M4 the wider) for linear dimensions of rubber articles. You pick a class per dimension or per drawing default. You don't pick "as good as steel." M2 is a common default conversation, not a promise.

Why metal defaults fail:

• Elastomers deflect under a CMM probe.
• Platinum RTV after a full cure typically moves about 0.1–0.3%; tin RTV keeps moving; heat-cured LSR and HCR *parts* shrink in a larger band (about 2–3% and 1.5–2.5%).
• Flash and pour seams are process scars, not machining burrs.
• Soft Shore 10A–20A skins ovalize in a fixture; a loaded tray can sag about 0.5–1.5 mm at the edge without a jacket.

ISO 3302-1 is a starting language, not a magic. Some customers still need a sealing-face note tighter than the default class, measured in a named fixture, at a named time after cure. Write that note. Don't apply it to every radius.

ASTM D3767 is another rubber dimensional practice some North American files cite. If you use it, use it on purpose. Don't mix two systems on one balloon.

Which dimensions belong in a tight class

Tight class belongs on:

• Pocket features that must mate or nest.
• Alignment keys on a split jacket.
• Logo depth only if the brand is the product and you will actually measure it.

Loose class belongs on:

• Cosmetic walls of the rubber block.
• Overall envelope of a floppy tray in free state.
• Jacket-adjacent dimensions unless sag is the defect.

A drawing that puts M1 (or a metal-like ±) on every number trains the shop to ignore all of them. Pick five datums. Fixture them. Live with the rest in a wider class.

Wall thickness of the *casting* is a special case. A thin chocolate membrane can be held as a fill-weight check, not a mill. A thick resin section can sink. Putting a metal-style ± on both without a dump story is how first article fails a number that never predicted function. If the wall is a seal, say so and test the seal. If the wall is cosmetic, give it a wide class.

Multi-cavity tools need a rule: every cavity meets the class, or the worst cavity is the tool. Photograph and weigh more than the hero pocket.

Shrink, wait time, and the number on the CMM

Measure after the process that the lot will see. Platinum RTV mold rubber: near-zero after a full cure; coupon the *dump*. Tin-cure continues to shrink, so a dimension taken on pour day isn't a dimension taken next month. Peroxide HCR that will be post-cured must be measured after post-cure — that's a *parts* rule on a different RFQ.

CMM vs optical: probes sink into rubber. Optical and vision methods still need a fixture that doesn't stretch the part. Contact measurements on 10A–20A are often a ritual. If you must use contact, specify tip, force, and wait. If that sounds like too much, you didn't need that tight class.

Master inspection (the pattern) can use metal or print practices. Don't paste those numbers onto the dump inspection. A 0.01 mm master miss can be real. A 0.01 mm miss on a floppy tray is often the clamp.

Drawing notes that survive a supplier change

Write:

• Process: platinum RTV or tin RTV mold-making (not "silicone").
• Shrink owner: shop or customer. Compensate the *cast* article; don't paste an LSR 2–3% class number onto the master.
• ISO 3302-1 class default, plus exceptions. Say it's a drawing class.
• Fixture or "free state" for floppy trays.
• Fill-weight check if pocket volume is the product.
• Time after cure when dimensions count (tin: also mold age).

China-origin and local shops can both hold a class if the class is rubber and the process is named. They can't both hold a copied metal box on an unfrozen 20A tray. The standard is the language. The process is the physics.

Master finish also interacts with measured size. Layer lines add texture depth. If you inspect only a CMM length and ignore sag, you aren't using a tolerance standard. You're using a single number as a personality test.

What high precision isn't

It isn't a brochure word. It isn't UL 94. It isn't FDA 21 CFR 177.2600. Those are other files. High precision isn't "our CNC is imported." It isn't ISO 3302-1 printed in bold. It's a jacket that controls sag, a shrink policy that is true, and a measurement method that doesn't crush the part.

If a supplier quotes "tolerance ±0.02 mm silicone" with no class, no Shore A, and no process, they're quoting a mill, not an elastomer. Send them ISO 3302-1 and the process sentence. Then you have a standard. Then you can argue first article.

For RTV *molds sold as products* (trays), decide whether you're tolerancing the outer jacket, the pocket volume, or the article that will be cast in them. Those three numbers aren't one ISO 3302-1 line. Pocket volume for chocolate or gummies is often a fill-weight check, not a CMM length. Write the check you will actually run on incoming inspection.

LSR and HCR *parts* share ISO 3302-1 as language. They don't share shrink or flash physics. Don't mix that inspection report with this tray.

Common mix-up: ISO 2768 on a 20A tray

A lot of RFQs treat a metal general-tolerance block as "high precision silicone." In the shop that usually means every number fails under a probe, so the shop ignores all of them. Use ISO 3302-1 classes. Metal practices belong on the master or on a steel *part* tool — different object.

What to ask when you're buying

• ISO 3302-1 class default, and which five dims are tighter?
• Are we tolerancing jacket, pocket, or the dump?
• Fill-weight vs CMM length for chocolate/gummy pockets?
• Platinum or tin, and at what age do we measure?
• Fixture or free state for the 10A–20A skin?

Worked example: 8-cavity bar, fill-weight not ±0.05 mm

A confectionery line wanted an 8-cavity bar, 120 x 60 x 12 mm, food-contact, about 500 pours a month, 38 g target per bar. The useful check was fill-weight on all eight pockets after a named wait at 23 °C, ISO 3302-1 as the linear default on the jacket keys, and a sag note on the outer row. A pasted ±0.05 mm on the floppy 10A–20A skin would have failed the clamp, not the cavity. A parallel LSR ±0.02 mm quote was a *part* mill story.

When this language does not apply

If you're inspecting steel LSR or HCR *part* tools, keep metal practices on the cavity and ISO 3302-1 on the article — different WYD-style RFQ. A one-off art casting in tin-cure doesn't need a library class; freeze a wait. Geometry can override quantity: a hollow chocolate on a mandrel is two measurement states (locked ID vs free face) even at 80 dumps.

Related reading: shrinkage factors, high-precision molds, quality inspection, custom RTV molds.

Tolerance block to copy onto the 2D

Copy this. A ± without these lines isn't a standard.

• Process: platinum or tin RTV
• ISO 3302-1 default class (often M2 as a conversation); exceptions listed
• Jacket / pocket / dump: which number is which
• Fill-weight method if volume is the product
• Fixture vs free state; measure at 23 °C after wait ____
• Shrink owner; no LSR 2–3% on the master

Send the STEP, the five dims that matter, and the dump compound. We'll mark which class belongs where and whether fill-weight beats a CMM box.

We would not recommend pasting ISO 2768 on a 20A lip

We would not recommend calling ISO 3302-1 "precision." We would not put M1 on every radius. We would not skip the jacket and then fail the outer row for sag. We would not raise Shore A first to "hold tighter numbers."

RUUIPON Mold Structure Decision: class the five dims that function; leave the floppy envelope wide. Flat back? open-back, still needs a jacket if 10A–20A. Deep undercuts that won't peel? split/core — tightness doesn't unlock a barb. Wet dump? mother mold before you tighten the class.

flowchart TD
  A[What are you tolerancing?] --> B{Jacket, pocket, or dump?}
  B -->|dump volume| C[Fill-weight at 23 C]
  B -->|keys / nest| D[ISO 3302-1 tighter class plus fixture]
  B -->|floppy envelope| E[Wide class or free state]
  A --> F{Metal ISO 2768 on rubber?}
  F -->|yes| G[Wrong language]
  A --> H{LSR or HCR part?}
  H -->|yes| I[Stop: different part RFQ]

Typical shop values

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

| Item | Typical shop value | | --- | --- | | Mix | Platinum 1:1 or 10:1; vacuum degas | | Cavity Shore | 10A–20A skin with jacket | | Platinum shrink after full cure | About 0.1–0.3%; tin keeps moving | | ISO 3302-1 | Drawing class; M2 is a common default conversation | | Loaded-tray edge sag without a jacket | About 0.5–1.5 mm | | Contrast: LSR / HCR *part* shrink | LSR about 2–3%; HCR about 1.5–2.5% — different RFQ |

FAQ

Can I put ISO 2768 on a silicone part drawing? You can, and you'll fight the CMM. ISO 2768 is for machined features. Use ISO 3302-1 classes for rubber articles, and metal practices only for the master or for steel cavity inspection on a *part* tool.

Which ISO 3302-1 class should I default to? Pick a default that matches duty, often a mid commercial class such as M2 as a conversation, then tighten only nest, keys, and location faces. Defaulting every dimension to M1 makes the drawing unmeasurable. The class isn't "precision."

Why did my tray shrink after I accepted first article? Tin-cure RTV movement after the measurement, or a dump measured warm. Measure after the wait the lot will see. Name that time on the drawing. Platinum after full cure is the small band; coupon the dump.

Do RTV trays and LSR parts share the same tolerance capability? They can share ISO 3302-1 as language. They don't share shrink or flash physics. A LIM shut-off and a poured jacket won't print the same number for the same CAD.

How should I measure a 20 Shore A tray without lying to myself? Fixture it the way it's used, prefer optical methods or fill-weight, and avoid tight contact CMM boxes. If the function is a dump, test the dump, not a floating overall length.