Shrinkage: the 2% that ruins a tension frame

Shrinkage gets less attention than weight, width or brightness, and it is the figure most likely to produce a panel that cannot be installed. Heat is not optional in this process — dye-sublimation requires it, between roughly 180 °C and 205 °C depending on the material — so the fabric that comes out is always smaller than the fabric that went in. The question is only by how much, and whether you accounted for it before you cut.

How much does dye-sublimation fabric shrink?

Across Epic's six house fabrics, measured on the published test conditions:

Fabric	Length	Width	Tolerance	Conditions
Rio	0.9 %	0.7 %	< 2 %	190 °C, 0.65 m/min
Tesa	1.0 %	0.6 %	< 2 %	195–205 °C
Eos	1.0 %	0.3 %	< 2 %	205 °C, 0.65 m/min
Aureline	1.3 %	0.3 %	< 2 %	195–205 °C
Halo	2 % (single figure)		—	200 °C, 1 min
Lumex	2.7 %	2.3 %	< 3 %	190 °C

Every one of those is on spec. Lumex is not a defective fabric — it is a 3 % fabric performing at 2.7 %, and it carries the strongest breaking strength in the range by some margin at 921 N length and 984 N width. But a 3 % tolerance band and a 2 % band are different products to plan around, and the sheet is telling you so in the tolerance column rather than in the result.

Why does shrinkage matter more on a tension frame?

Because a silicone-edge graphic has no adjustment in it.

A grommeted banner that comes out 30 mm short gets taken up by the bungees and nobody notices. A pole-pocket backdrop has hem allowance to play with. A SEG panel has a bead sewn into a groove of a fixed dimension, and the fabric has to arrive at a size that lets it stretch into tension without either falling slack or refusing to seat.

The window is small. Most SEG systems want the finished graphic slightly undersized — commonly around 0.5 to 1 % — so the material comes up taut. Shrink less than planned and the panel sags; shrink more and the last corner will not go in, or it goes in and pulls the frame out of square. On a run of matched panels in a single wall, the failure is more visible still, because a 1 % difference between two adjacent panels reads as a crooked seam even when each one individually looks fine.

Do I compensate for shrinkage in the artwork or in the cut?

In the artwork, by scaling up before you print.

If the fabric shrinks 1.0 % in length, the file needs to be about 1.01 times the finished length so that what comes out is the size you wanted. Cutting oversized and trimming after is the alternative, and it wastes material and loses the bleed you wanted at the edge.

Two details catch people out. The first is that length and width shrink differently, and often very differently — Eos shrinks 1.0 % in length and 0.3 % in width, so a single uniform scale factor gets one axis right and the other wrong. The compensation has to be non-uniform, which means it has to be set per fabric and not as a shop-wide default.

The second is that the published figure is measured at stated conditions, and your conditions are probably not identical. A slower calender speed means more time at temperature, which means more shrinkage. Rio's 0.9 % is measured at 190 °C and 0.65 m/min. Run it hotter or slower and you will see more. The sheet gives you the starting point; your own test gives you the number you actually use.

How do I measure shrinkage on my own line?

Print a measured grid rather than guessing from the panel edge. Mark a 1000 mm square on the fabric before it goes through, in both axes, and measure the same marks after. A metre is long enough that a 0.5 % difference is 5 mm and readable with a tape; a 300 mm test square makes the same difference 1.5 mm and hides it inside your measurement error.

Do it on three samples from the same roll, not one, and do it again when you open a new lot. Shrinkage varies within a tolerance band across lots, and a compensation factor derived from one sample a year ago is a guess wearing a number's clothing.

Then write the factor down per fabric, per axis, per calender setting. That table is the single most useful piece of paper in a soft-signage shop, and no supplier can produce it for you — it is specific to your equipment.

Which fabric should I pick if dimensional stability is the priority?

Look at the tolerance band first and the result second. A fabric held to < 2 % will behave more predictably across lots than one held to < 3 %, even if a particular sample of the looser one measures lower today. Then look at the gap between the length and width figures: Eos at 1.0 % and 0.3 % needs a more carefully built compensation than Rio at 0.9 % and 0.7 %, which is close enough to uniform to be simple.

Every shrinkage figure above is from our own measured test data, with the test conditions named, because the conditions are half the number. The full sheets are on each product page — and if you want to run your own grid test before committing to a fabric, ask us for sample yardage.