TPU is a block copolymer that behaves like rubber without being one. Along each chain, rigid urethane-rich segments alternate with long flexible polyol segments. The rigid blocks cluster together into hard domains that act as physical crosslinks — they hold the material's shape the way sulphur bridges hold vulcanised rubber together — except that heat melts them apart instead of burning the part. That is the whole trick: a genuinely elastic material that can still be extruded through a nozzle.
Shore 95A puts it near the hard end of the rubber scale, roughly the feel of a skateboard wheel or a heavy-duty castor. It is the grade most people should start with, and the grade most people should stay on.
The polyester-versus-polyether choice nobody mentions
The soft segment is either a polyester polyol or a polyether polyol, and filament listings almost never say which. It matters for anything that lives a hard life.
- Polyester TPU has better abrasion resistance and stands up to oils and fuels. It is also vulnerable to hydrolysis: warm, wet service breaks the ester links and the part gradually turns sticky and weak.
- Polyether TPU shrugs off water and stays flexible at low temperature, at the cost of some abrasion and oil resistance.
A gasket for a damp pump housing wants the polyether. A wear pad on a workshop tool wants the polyester. If the technical data sheet does not say, that is itself information about how much the brand knows about its own product.
Why the extruder is the whole problem
Between the drive gear and the melt zone, the filament is a slender column under compression. Push a column hard enough and it does not compress, it buckles sideways — the same behaviour as a metre rule pressed end-on. A rigid filament has enough stiffness to resist; TPU does not, so anywhere the filament path has an unsupported gap, it kinks, coils and jams.
Every piece of practical TPU advice follows from that one mechanic:
- Direct drive, not a Bowden tube. A long Bowden path is a very long unsupported column, and it is why TPU has a reputation for being impossible on some machines.
- Close the gaps. Any place the filament crosses air inside the extruder is where it will escape. Extruders sold as "flexible-capable" mostly differ in having no such gap.
- Slow down. Feed force rises with speed, so 20 to 35 mm/s is the realistic band. This is not caution, it is the buckling limit.
- Minimal retraction. Long retractions stretch the elastic filament rather than pulling melt back, and the extruder then loses track of where the material is.
What slowing down actually costs
Take a phone case printed flat: 12 mm tall, 0.2 mm layers, three walls, 15% infill, a 0.4 mm nozzle, about 35 cm³ of plastic over an 11,000 mm² footprint. At the 25 mm/s TPU tolerates, that is 6.4 hours. Run the identical job at the 50 mm/s a rigid filament would happily take and it finishes in 3.2 hours. Both figures come from the print time estimator, and the estimator carries a ±30% error band because a prism is a crude stand-in for a real model.
The three extra hours are the real price of flexible filament, and they land on machine time rather than on material. If you cost prints for other people, a TPU job priced on grams alone is priced wrong.
Hardness is not the only softness control
This is the thing that most changes how people design in TPU. A part's perceived stiffness comes as much from its geometry as from the Shore number on the spool. Two walls and 10% gyroid infill in 95A feels dramatically softer than four walls and 40% infill in the same filament. Shore hardness describes the bulk material; the printed part is a lattice.
The practical consequence: try wall count and infill before buying a softer spool. A great many people order 85A because their 95A case felt rigid, when the case was simply printed nearly solid.
What TPU 95A is genuinely bad at
- Bridges and steep overhangs. The extrudate stays soft and sags, and support material welds itself to a flexible surface far more stubbornly than to a rigid one.
- Sharp dimensional work. The part relaxes elastically after printing, so a hole measured straight off the plate and the same hole a day later can differ. Print a test piece rather than trusting a scale factor.
- Anything that needs to be glued reliably. Cyanoacrylate holds on TPU better than on most plastics, but a flexing joint peels a rigid glue line apart eventually.
- Wet filament. Water in TPU boils in the melt and produces a furry, bubbled surface. Four to six hours at 60 °C fixes it, and an open spool in a humid room re-ruins it in a weekend.
- Long-term sunlight if it is unpigmented. Aromatic TPU — the common, cheap kind — yellows under UV. The part usually still works; it stops looking new.
What people make from it, and what they should not
Phone and camera cases, drone landing feet, RC tyres, machine feet and anti-vibration pads, gaskets and blanking plugs, cable strain reliefs, watch straps, and belts for light-duty drives. Its abrasion resistance is genuinely outstanding — better than most cast rubbers — which is why it survives roles that eat softer materials.
The misuse worth naming is skin-contact and wearable parts assumed to be inert. A filament's polymer may be benign while its pigments, plasticiser residues and processing aids are unknown, and a printed surface is porous. For a strap worn daily, that is worth a moment's thought rather than none. Where a softer feel is the actual goal, the 95A against 85A comparison covers what changes and what breaks.