Every mold has an ejection system and very few of them were designed. The ejector plate is usually laid out in the last week before the steel is cut, by copying the previous tool and moving the pins until they clear the cores. Nothing goes wrong until the trial, when the part sticks, or comes off with a white mark on a face the customer can see, or bows, or is levered out with a screwdriver while somebody says it will be fine at production speed. All four of those outcomes were decided by a drawing detail that cost nothing to change.
Ejection is a fight against four forces
A part does not simply sit in the mold and then get pushed out. It is held by four different effects, and only one of them is unavoidable. Knowing which one is holding your part is the whole diagnosis, and it is also the reason two parts that look identical on a drawing can need completely different ejection.
| Force | Where it comes from | What it looks like at the trial | What actually reduces it |
|---|---|---|---|
| Shrinkage grip | The part contracts onto the core as it cools | The part stays on the core and drags; the pins leave deep marks pushing it off | More draft, fewer tall core-side features, less packing |
| Friction and adhesion | Polished or textured faces in intimate contact with steel | It releases with a crack, then grips again further along | Draft, a release-friendly surface, a coating |
| Vacuum | The part seals against the core and a partial vacuum forms behind it | It holds until a pin punches it free, with a loud pop and a witness ring | An air path: core vent, pin clearance, air valve |
| Mechanical lock | An undercut cut for a snap, a clip or a thread | The part tears, or the tool is opened by hand and levered | Lifter, slide, collapsible core — or a change to the part |
Shrinkage is inherent: a molding compound that did not shrink would not be a molding compound. The other three are design choices, and all three are cheap to eliminate on a drawing and awkward to eliminate in steel.
There is a hierarchy in how they combine. Shrinkage grip sets the baseline force. Vacuum multiplies it, because the pin is now working against atmospheric pressure as well as friction. Friction and adhesion decide how that force is distributed, which is what produces marks. And an undercut overrides all of it: with a mechanical lock, no amount of draft or pin area will help, and the part will either tear or be pulled out of shape.
Shrinkage grip, and which half the part should stay on
The part shrinks onto whatever it is left on while it cools, and the amount of grip follows the shrinkage of the resin. Amorphous materials like ABS or PC shrink 0.4 to 0.8 % and grip lightly. Semi-crystalline materials like POM or PP shrink 1.5 to 2.5 % and grip hard, which is why a POM part is the classic case of a tool that will not let go.
The number that matters is not the shrinkage figure on the datasheet but the interference that has actually developed at the moment of ejection. In practice that is roughly 0.1 to 0.3 % of the dimension, because a good part of the shrinkage happens after the part leaves the tool. On a 150 mm deep core that still means 0.15 to 0.45 mm of radial squeeze over the entire contact surface, and friction is proportional to contact area.
Which half the part stays on is a design decision, not an accident. The part should stay on the ejector half, because that is the only half with a mechanism able to push it off. It stays there if the core side has more surface, more depth, more texture, or a lower temperature; it leaves for the cavity side if the cavity has the deeper features. When a part ends up in the wrong half, the mold still “works” — an operator reaches in with a screwdriver, or a robot gripper pulls it, and both of those damage the parting line over time. The drafting decisions that control this are part of the same review as the DFM checklist, and the surface side of it is covered in tolerances and surface finish.
How to bias the part onto the ejector half
Leaving it to chance is the usual approach, and it produces the usual result. There are four levers, in ascending order of how much they cost:
Cooling. The half that cools first holds the part, because the part contracts onto the colder steel. Running the core side around 5 °C cooler than the cavity is enough on many parts, and it costs nothing beyond a separate circuit.
Core-side features. Ribs, bosses and any depth on the core side give the part something to grip. A part that is nearly flat on the core side and deeply pocketed on the cavity side will stay in the cavity every time, and no ejector will reach it.
Surface. A texture or a coarser polish on the core side increases friction there. This is the cheapest lever on paper and the most dangerous in practice, because it also increases the force the pins need.
A small retention feature. A shallow groove or a very small undercut on the ejector side, 0.2 to 0.3 mm deep, will hold the part positively. It is a last resort, because it is an undercut and it will wear; but it is a legitimate one on a part that otherwise insists on staying in the cavity.
Why packing harder makes ejection harder
Higher cavity pressure packs the part more tightly onto the core and raises the residual stress frozen into it, so a process “improvement” that adds hold pressure can turn a free-releasing tool into one that sticks. The same applies to a longer hold time: once the gate has sealed, extra hold adds no weight, but it does keep the part pressed against the steel for longer while it cools.
The second counterintuitive point is that ejecting earlier is not automatically easier. A part ejected while still soft does not slide off; it stretches, whitens at the pins, and then warps as it cools unsupported. There is a window, and it is set by cooling, not by the ejector — which is why the first thing to check on a sticking tool is not the pins but the cooling balance.
How hard a given resin grips is fixed by its shrinkage, which varies by family rather than by brand: 0.4 to 0.8 % for amorphous materials and 1.5 to 2.5 % for semi-crystalline ones. The comparison is set out in plastic material selection, and it explains why a tool that releases POM with a tap will not release it with a nudge.
Draft angle: the cheapest line on the drawing
Draft is the only feature in a mold that costs nothing to add and cannot be added later. It is measured per side, from the direction of opening, and “1° draft” means each wall face is one degree off parallel to that direction.
The reason people underestimate it is that the numbers look generous until they are converted into clearance. On a wall 50 mm deep, 0.5° gives 50 × tan(0.5°) ≈ 0.44 mm of clearance per side. That is less than the shrink grip described above, which is precisely why a “half a degree is fine” wall sticks while a “one degree” wall does not.
| Situation | Minimum | Normal | Why |
|---|---|---|---|
| Smooth wall, under 20 mm deep | 0.5° per side | 1° | Short walls have little contact area, so friction is low |
| Smooth wall, over 50 mm deep | 1° | 1.5 to 2° | 0.5° over 50 mm gives only 0.44 mm of clearance per side, less than the shrink grip |
| Light texture, about 0.02 mm | 1.5° | 2° | Add roughly 1° for every 0.025 mm of texture depth |
| Heavy grain, 0.1 mm or more | 4° | 5° | The texture itself behaves as a shallow undercut |
| Ribs and bosses | 0.5° per side | 1° | They are deep, thin and surrounded by steel: the hardest features to release |
| Glass-filled resin | Add 0.5° | Add 1° | Higher friction against steel, and the surface is rougher |
| Shut-off and sealing faces | 0° | — | These faces are meant to close metal to metal; draft belongs on the release walls only |
Two rules make this practical. The first is to draft every wall that moves relative to steel, including the inside of a rib, the outside of a boss, and the sides of a small pocket — the features that get missed are always the small ones. The second is to add the texture allowance before the texture is applied, not after. A heavily grained surface etched to 0.1 mm is a forest of tiny undercuts, and a wall drafted at 1° for a polished finish will grip like a file once it is grained.
The draft review is worth doing on the 3D model rather than the drawing, with a draft analysis turned on. A wall that reads 1° in a section view is often 0° somewhere else because the parting direction changes across the part, and that single zero-draft patch is enough to hold the whole part.
How much force ejection actually needs
Ejection force is rarely calculated, and it is worth having an order of magnitude before the ejector plate is sized. The industrial rule of thumb is that ejection needs roughly 2 to 10 % of the clamp force used to mold the part. The spread is wide because it depends on shrinkage, draft, contact area and temperature, which is exactly why the number should be treated as a starting point rather than a design value.
A more useful bound comes from what the force is doing. Once the gate has sealed, cavity pressure at ejection is close to zero; the resistance is the friction of a shrunken part on steel. Contact area times a friction coefficient of roughly 0.2 to 0.4 against the residual pressure the steel is exerting gives a number in the same range as the rule of thumb, and it makes the two levers visible: reduce contact area (draft, shorter core-side features) or reduce the pressure (less packing).
The residual pressure is one of the few things a simulation can tell you before the tool exists. A packing analysis reports how much pressure is left in the cavity at the moment the part is ejected, and that figure is the load the pins are fighting; the same run also shows where the part is still warm and soft when it comes out. Reading that alongside the gate seal time is more useful than any rule of thumb, and the method is set out in mold flow analysis.
Size the mechanism, not the number
Whatever the estimate, the mechanism should be sized for two to three times it, because the worst case is a part ejected slightly too cold on a Monday morning with a new operator. That margin is not about the average cycle; it is about the cycle that breaks a pin.
The two weak points: pin buckling and a cocking plate
The mechanism has two weak points. The first is pin buckling: a long thin pin under compression behaves like a column, and the load it can take falls with the square of its length. A pin with a length-to-diameter ratio beyond roughly 40 is a pin that will eventually bend, and a thin pin in a deep core needs support close to its tip. The second is the ejector plate itself, which should run on guide bushings rather than on its own edges; an unguided plate cocks under load, and a cocked plate binds the pins on one side, which is how one half of a part gets deeper marks than the other.
Ejector pin layout: push where the part is stiff
The rule that covers most of it: push where the part is stiff, never where it is flat and free. A pin under a rib root transfers load into a wall that can take it. The same pin on a flat panel 30 mm from any support will dish the panel, and the dish will still be there when the part cools.
| Place the pin | Not here | Reason |
|---|---|---|
| Under a rib, at its root | On the thin web beside the rib | The rib is stiff and can take the load; the web will dish |
| Directly under a boss | Next to a boss, on the flat | A pin beside a boss cocks the boss and can crack it at the base |
| At a corner, set in from the edge | On the edge itself | A pin too close to the edge punches through or leaves a notch |
| Near the gate | At the far end of a long thin section | The gate area is the stiffest and most frozen part of the part |
| In a hidden or non-cosmetic area | On a class A face or a textured face | A pin breaks the grain and cannot be hidden afterwards |
| In a symmetric pattern about the centroid | All on one side | An unbalanced layout rotates the part off the core instead of pushing it |
Size and count follow from the same logic. A pin should be as large as the geometry allows, typically 3 to 8 mm on a technical part, and on a thin wall the pin diameter should be at least the wall thickness so it does not punch a hole. A large flat face is better served by a grid of smaller pins at 50 to 80 mm pitch than by four large ones at the corners, because the deflection between pins is what the customer sees.
Witness marks are a specification
Witness marks are a specification, not a by-product. On a hidden face, flush within about 0.05 mm is normal. On a visible face the pin is usually recessed 0.2 to 0.3 mm so the mark reads as a deliberate feature rather than a defect. What must not happen is a pin standing proud: that transfers the whole load onto one point and leaves a raised dot that will also mark the next part.
Multi-cavity and family tools
On a tool with several cavities, the ejector system has one extra requirement: every cavity has to release at the same moment, because the whole plate moves together. That is fine when the cavities are identical, and it becomes a problem when they are not. A family tool molding a housing and a small lid in the same shot has one part that grips hard and one that falls off, so the plate has to be designed for the stubborn one without over-driving the fragile one. In practice that means separate pin groups per part, or a spring-loaded pin that retracts when it meets no resistance.
The same logic applies across cavities of a multi-cavity tool after a repair. If one cavity has been polished or re-textured, its grip changes, and the part that used to release cleanly starts to drag. A tool that suddenly marks one cavity out of eight is usually telling you that cavity’s surface or draft has changed, not that the ejector has failed.
One useful side effect of pins: the clearance between a pin and its hole, typically 0.02 to 0.03 mm, is an air path. On a core that would otherwise seal, the pins are often the only vent the part has, which connects ejection directly to venting and explains why a part that sticks can also burn at the same spot.
Nine alternatives when pins are the wrong answer
Pins are the default because they are cheap, standard and adjustable. They are also the wrong answer in a surprising number of geometries, and the alternatives are not exotic: most of them are standard mold components.
| Method | Use it when | Its limitation |
|---|---|---|
| Ejector pins | There are stiff features to push against | Leaves marks; poor on a large thin flat face |
| Stripper plate | Thin-wall containers, open round or rectangular parts, no pin marks allowed | Needs a full perimeter to push on; adds stack height and cost |
| Sleeve ejection | Cylindrical bosses, tubes, long thin cores | Concentricity has to be held; sleeves wear |
| Blade ejection | Thin ribs and narrow deep slots | Fragile, and needs support very close to the tip |
| Air valve or air assist | Deep draws, large flat areas, parts that seal and hold | Needs clean dry air; leaves a small circular mark |
| Lifter | An internal undercut that can be released on the ejector stroke | Angle limited to about 15°; the sliding face wears |
| Collapsible core | An internal thread or a deep internal undercut | The most expensive option, and the most fragile in service |
| Slide, then eject | An external undercut | Tool cost, cycle time, and a parting line the customer will see |
| Runner ejection | Three-plate tools where parts drop free and the runner is ejected separately | Only applies to a three-plate layout |
The choice usually follows the wall thickness and the shape of the contact area. A thin-wall container has almost no stiff features to push on and a full perimeter to push with, so it gets a stripper. A tall cylindrical boss has a circumference, so it gets a sleeve. A deep draw with a sealed base has a vacuum problem, so it gets air. A part with an internal clip has a mechanism problem, so it gets a lifter — and the lifter angle is limited by the ejector stroke, not by preference.
Cost and lead time follow the choice rather than leading it: a stripper plate adds stack height and machining, a collapsible core adds weeks and a fragile mechanism, and both of those show up in the tool price and in the schedule long before they show up in a trial report.
Deep draw and thin wall: the two hard cases
Most parts sit between the two extremes and can be released with pins, a reasonable draft and a bit of care. The two ends of the range are where the standard solution fails, and they fail for opposite reasons: one has too much contact area, the other has too little stiffness.
Deep draw: the vacuum adds a hundred kilograms
As a deep part cools on a long core it contracts into a seal, and once sealed, the space behind it is at a lower pressure than the room. The pin is then not only fighting friction but atmospheric pressure over the sealed area: 0.1 N/mm² across a 100 × 100 mm base is 1,000 N, which is a hundred kilograms of extra force from the air alone.
That is why deep parts release with a pop and a witness ring, and why the fix is an air path rather than a bigger pin. A vent in the core, the natural clearance around the pins, or an air valve all work. A taper on the core wall helps the seal break earlier, and a slightly rougher core surface can help too, because it lets air creep along the interface. What does not help is adding pins: if the seal holds, five pins work against the same vacuum that one pin was fighting, and all five leave marks.
Thin wall: nowhere stiff to push
A 1 mm wall cannot take a point load without marking, and there is often nowhere stiff to push. The answers are a stripper plate, which spreads the load over the rim; many small pins rather than a few large ones; or ejection on the runner with the parts pulled by the sprue. Whichever is used, the ejection force has to be low, which brings the discussion back to draft: on thin-wall parts, draft is doing more work than the ejector is.
Thin-wall parts also distort differently. A thick part marks where the pin pushes; a thin part bows between the pins before it marks at all, so the complaint at the trial is often “the bottom is not flat” rather than “there is a mark”. Increasing the number of pins until the deflection disappears is usually cheaper than increasing their size, because the deflection is proportional to the span between supports.
Ejection marks and distortion: what the customer sees
Most ejection complaints are one of six patterns, and each has a different cause. Reading the pattern is faster than adjusting the process and hoping.
| What you see | What caused it | What to change |
|---|---|---|
| A white stress mark at the pin | Local yielding: too little contact area for the force needed | More pins, larger pins, or more draft so less force is required |
| A dish or dent opposite the pin | The pin is small relative to the span it is pushing | Spread the load: a grid of pins, or a stripper plate |
| The whole part bows | Unbalanced layout, or ejection before the part is stiff enough | Symmetrical layout about the centroid; longer cooling |
| A crack running from the pin | Pin too close to a corner, or the part is too cold and brittle | Move the pin inboard, raise the mold temperature, increase the radius |
| A stretched or oval boss | A pin pushing inside a thin cylindrical wall | Use a sleeve, so the load is spread round the full circumference |
| Drag marks down the side wall | Not enough draft, or a polished surface with no release | Add draft; check the polish direction and the mold temperature |
The common thread is that a mark is a load path made visible. Where the pin pushes, the material yields slightly, and whether that shows depends on how much force is concentrated on how little area. Adding pins, enlarging pins, and adding draft all do the same thing in different ways: they reduce the force per unit area. Changing the packing pressure or the cooling time does something else entirely, which is why process adjustments often move the mark without removing it.
Distortion is a different failure with the same origin. A part that bows after ejection was either pushed unevenly or released while still soft enough to creep. The first is a layout problem, the second is a cooling problem, and the cooling design usually has more influence than anything in the ejector stack.
Ejection and the cycle: the temptation to eject early
Ejection itself takes one to three seconds, which is small against a cycle of twenty or more. The temptation is to shorten the cooling time instead, eject while the part is still warm, and let it finish cooling outside the tool. On a thin-wall commodity part in PP that can work. On anything with a flat face, a tight tolerance or a visible surface it produces parts that bow, shrink past the tolerance, or mark at the pins, and the cost of the scrap is larger than the seconds saved.
What the ejection system contributes to cycle time is reliability rather than speed. A part that releases cleanly every shot can be handled by a robot on a fixed path; a part that sticks one shot in fifty needs a human watching, and a human watching costs more than the seconds. That reliability is worth designing for explicitly: ejection stroke at least the part depth plus 10 to 20 mm, a positive return so the pins cannot be left protruding, and a plate guided well enough that the pins do not bind.
Return pins: the failure that smashes cavities
Return pins and springs deserve a mention because they are the classic cause of a smashed cavity. If the ejector plate is returned only by springs, a stuck pin or a fatigued spring leaves a pin standing proud, and the next close drives it into the cavity. Pull rods, or a positive mechanical return, remove that failure mode entirely for very little cost.
The spring has a second failure mode that is easy to design out: a spring compressed to its solid height on every cycle loses force quickly, and one that has lost force will not return a plate that is binding. Specifying a spring with free length to spare, or replacing it with pull rods, is a decision made once and never revisited.
What fails after a hundred thousand shots
Ejection is a wear system, and the failures are predictable. Knowing them is useful because each one has a design decision behind it that costs almost nothing at build stage.
| Failure | Cause | How it shows up | Fix |
|---|---|---|---|
| Pin galling | Clearance too tight, soft pin, no lubrication | The pin squeals and the hole wears oval | Hardened or coated pin, correct clearance, polished hole |
| Pin breakage | Buckling: too long and too thin, or an unguided plate | Snaps at the plate, usually on a cold morning | Larger pin, guided ejector plate, support closer to the tip |
| Pin tip wear | Abrasive resin, glass or mineral filled | The pin gets shorter and the mark becomes a recess | Hardened or coated tip, or a larger contact area |
| Ejector plate cocking | Plate sliding on its own edges with no bushings | Pins bind on one side, marks get deeper on that side | Guide bushings on the ejector plate, hardened wear plates |
| Return spring fatigue | Springs compressed to solid height every cycle | The mold closes on protruding pins and damages the cavity | Longer springs, or pull rods instead of springs |
| Flash at a pin hole | Clearance opened up, or a worn pin | A thin ring of flash around every mark | Replace the pin, ream and bush the hole |
| Progressive sticking | Deposit or scale on the core, or texture polished smooth in places | Sticks after tens of thousands of shots, not at first | Clean on a schedule, restore the surface, review the release agent |
Two of those entries are worth designing against rather than repairing. Guide bushings on the ejector plate and a positive return are both small line items at build stage and both eliminate a whole failure class. The rest are maintenance: pins are consumables, and a tool that runs abrasive resin should be specified with hardened, coated pins from the start rather than discovering it after the marks go proud. The maintenance side of this is covered in mold maintenance, and the steel choice that decides how well a pin survives abrasion is in mold steel selection.
Ten things settled before the plate is drilled
1. Which half does the part stay on? Confirm it is the ejector half, and confirm it by geometry rather than by hope.
2. Is every release wall drafted, per side, including ribs and bosses? Check the 3D model with a draft analysis, not the drawing.
3. Has the texture allowance been added to the draft? Roughly 1° per 0.025 mm of grain depth, before the etch.
4. Are the pins on stiff features? Rib roots, bosses, corners and the gate area; balanced about the centroid.
5. Is the pin size sensible against the wall and the span? At least the wall thickness on a thin wall, and spread as a grid on a large flat face.
6. Is the witness mark location acceptable on the drawing? Flush on hidden faces, recessed on visible ones, never proud.
7. Is there an air path for a deep or sealed part? Core vent, pin clearance or an air valve.
8. Is the stroke at least the part depth plus 10 to 20 mm? And does the part clear the core before the robot enters?
9. Is the ejector plate guided and sized for two to three times the estimated force? With bushings, not on its own edges.
10. Is the return positive, not just spring-loaded? Pull rods or a mechanical return, so a stuck pin cannot be closed into the cavity.
Those ten answers belong on the tool drawing, not in a meeting. The related structural decisions — how many cavities, how the part is fed, how much force the press applies — are covered in cavity count, runner configuration and clamp force, and the point where all of it gets proven is the mold trial. The overview of how the whole chain fits together is in the injection molding process guide.
Frequently asked questions
How much draft angle do I really need?
One degree per side covers most smooth walls, with 0.5° as an absolute minimum on short shallow features and 1.5 to 2° on walls deeper than about 50 mm. Textured surfaces need more: roughly 1° extra for every 0.025 mm of grain depth, which puts a heavy leather grain at 4 to 5°. Ribs and bosses are the features that get missed and they are also the ones that grip hardest, so check them separately rather than trusting a drawing note that says “1° draft”.
Why does my part stick even though it has 2 degrees of draft?
Because draft is only one of four holding forces. The usual causes, in order: a vacuum behind a sealed deep section, an undercut that nobody noticed (often a texture or a tiny barb), over-packing that pressed the part onto the core, and the part being left on the cavity half where there is no ejection. The diagnostic is to look at how it releases: a pop and a ring means vacuum; a tear or a white line means an undercut; a heavy drag with no pop means friction and packing. Each points at a different fix, and none of them is “add more draft”.
Can I use air instead of ejector pins?
For some geometries, yes, and for deep draws it is often better than pins. An air valve or a burst of air through the core breaks the seal and lifts the part off, which avoids pin marks entirely and removes the vacuum component of the holding force. What it does not do is provide the mechanical push for a part with real shrink grip or any undercut, and it needs clean dry air at the press. Most deep-draw tools use both: air to break the seal, then a stripper or a few pins to move the part clear.
Why do ejector pins leave white marks?
A white mark is local yielding of the plastic: the stress at the pin exceeded what the material could take without deforming permanently, so the polymer chains were stretched and the area turned light. It is a stress concentration problem, not a surface problem. The fixes are to increase the contact area (bigger pins, more pins, or a stripper plate), to reduce the force needed (more draft, less packing), or to eject later so the part is stiff enough to take the load. Polishing the pin does not help, because the mark is inside the material.
How many ejector pins does a mold need?
Enough that no single pin carries a load the part cannot take, and enough that the push is balanced about the centroid. In practice that means at least one pin per stiff feature plus a pattern on any large flat area, which commonly lands between four and twelve for a part the size of a hand. The number matters less than the distribution: six well-placed pins on ribs will release a part that twelve pins scattered across a flat face will dish. A thin-wall container usually needs a stripper rather than any number of pins.
Can ejection cause warpage?
It can, in two ways. Ejecting a part that is still soft imposes a bending load on material that cannot yet resist it, and the resulting bow is permanent. And an unbalanced pin layout pushes one side first, rotating the part off the core instead of translating it, which twists it. But the larger share of warpage usually comes from cooling and shrinkage differences rather than from ejection, so a part that bows should be checked for a temperature difference across its two faces before the ejector layout is blamed. If the bow disappears when the cooling time is extended, ejection was not the cause.
What is a stripper plate and when is it better than pins?
A stripper plate is a plate that travels on the ejector stroke and pushes the part off the core around its whole perimeter, rather than at a few points. It is better whenever the part has a continuous rim to push on and no stiff internal features: thin-wall containers, buckets, caps, open boxes, and any part where a pin mark on the inside would be visible or would interfere with a seal. It costs more than pins because it needs an extra plate, a guided travel and a clearance around the core, and it needs the part to be able to slide off the core without catching on the stripper edge.
Why do ejector pins break?
Almost always in compression rather than in bending: a pin is a column, and a long thin one buckles. The load a pin can carry falls quickly as its length rises relative to its diameter, so a 3 mm pin running through a 200 mm stack is a design that will fail even if the calculated force looks modest. The other causes are an unguided ejector plate that cocks and side-loads the pins, a pin that binds because its hole has worn oval, and a part that sticks hard enough to exceed the estimate. Larger pins, a guided plate, and support close to the tip address all three.
