Yarn Hairiness: Read the Test Result and Find Its Processing Cause

A supplier reports lower yarn hairiness after a process change. The buyer's long-hair count has risen. Both results may be correct, because they may describe different parts of the yarn surface.

A useful investigation starts with the result's name, test basis and place in the process. Then it asks what changed in the fibre, yarn structure or handling. This sequence matters more than finding one impressive percentage.

Hairiness can be measured and controlled, but a single number cannot explain every surface fault. The examples below show how to compare readings, locate added hairs and judge whether a process trial has helped the fabric.

What a hairiness result describes

Staple yarn contains fibres whose ends or loops can project beyond the main yarn body. A test may respond to their combined length, count those beyond a chosen distance, or describe how the reading varies along the yarn. These are related questions, with different answers.

The yarn body is also a reference boundary. A loose strand seen in an image must be distinguished from the dense central structure. Where that boundary is placed can affect an image-based result.

H is commonly described in terms of the total length of projecting fibres relative to a one-centimetre yarn length. In a commercial optical system, it is a measured index based on the sensor response. It is not a technician's direct count of every fibre.

Haleem and Wang compared manual measurements with an Uster Tester 4 and a Zweigle G565. Their study found substantial differences between methods for the three yarns tested. Those were specific older instruments and yarns; the work does not provide a correction factor for every current tester. It does show why a method name belongs beside the number. See the original comparison study.

A lower H value need not mean fewer long hairs. A change could reduce many small projections while leaving the longer ones unchanged. The reverse is also possible. This is why a claim such as “hairiness improved” needs to say which feature improved.

Read the index before comparing values

The April 2026 Uster Tester 6 technical data distinguishes the OH hairiness sensor from the HL hair-length sensor. These are optional sensors, so the equipment name alone does not identify the available result.

Keep the symbol exactly as the report defines it. A result labelled S3 on another instrument needs its own method, threshold and reference length checked. Similar names do not establish identical measurement.

Do not calculate H from S3u by multiplying the count by 3 mm. Three millimetres is a threshold, not the actual length of every counted hair. The count also omits hairs below that threshold. A full length distribution would still need a compatible measurement model before it could be compared with an optical H result.

The distinction between H and sh also matters. A yarn may have an acceptable mean reading but severe short sections. Another may have the same mean with a much steadier surface. A buyer concerned about visual bands may care about that difference.

Put counts on the same yarn-length basis

Suppose one test finds 520 qualifying hairs over 200 m. A second finds 390 over 150 m. These are invented values used only to explain the calculation.

The common rate is:

Hairs per 100 m = counted hairs ÷ tested length in metres × 100

For the first test:

520 ÷ 200 × 100 = 260 hairs per 100 m

For the second:

390 ÷ 150 × 100 = 260 hairs per 100 m

The second raw count is lower because less yarn was examined. It is not evidence of a smoother yarn. This conversion makes the exposure equal, but it does not remove differences in sensor, speed, tension or threshold.

Pooling results needs the same care. Suppose two specimens produce 520 hairs over 200 m and 380 over 100 m. Their combined rate is:

(520 + 380) ÷ (200 + 100) × 100 = 300 hairs per 100 m

A simple average of their separate rates gives:

(260 + 380) ÷ 2 = 320 hairs per 100 m

These values answer different questions. The first gives equal weight to each metre of tested yarn. The second gives equal weight to each specimen's rate, despite unequal tested lengths. Neither should be presented without explaining the purpose.

If a purchasing decision concerns the average package, an equal-package analysis may be appropriate. If the question is the observed count across the full tested length, pool counts and lengths. Decide which population the result should represent before choosing the average.

Equal exposure does not mean equal coverage of the lot

Testing 400 m from one package and 100 m from each of four packages gives the same total length. The two plans cover the lot in different ways. The first gives a longer view of one package; the second includes more package origins. Neither plan automatically answers both questions well.

Choose package origins to match the suspected fault. For a position-related problem, include the suspect position and suitable comparison positions. For a lot comparison, spread the sample across the lot rather than selecting only packages nearest the laboratory. Keep a separate record of deliberate fault samples, since they are not a random picture of normal output.

Within a package, identify where the test length came from. An outer layer and a deeper layer may represent different production times or handling histories. If a complaint concerns intermittent sections, one short test near the surface may miss the fault.

A longer run through one sensor produces more measured yarn, but it does not create more independent parent packages. Report both the tested length and the number of parents. This makes the strength and limits of the comparison easier to judge without pretending that every centimetre is a separate production trial.

Check the measurement before blaming the process

A difference between laboratories may be a real yarn difference, a test difference, or both. Start by exchanging the method and result definitions. Ask for the instrument model, sensor, tested length and operating conditions.

ASTM D5647-07(2025) covers photoelectric hairiness testing but does not recommend it for commercial shipment acceptance. Its public guidance notes poor comparability between manufacturers' results and calls for checks of laboratory bias when differences matter. Test speed and tensioning arrangements can also influence results. The stated scope excludes novelty yarns.

That caution does not make the measurements useless. It changes how they should support a decision. A consistent in-house series may reveal a process change even when a numerical limit cannot be transferred to a different instrument.

For a dispute, use matched yarn from a well-mixed set of packages. Agree how specimens are assigned to the laboratories and how the results will be compared. Repeatedly sending only a visibly poor package to one laboratory would build selection bias into the investigation.

Keep conditioning on the agreed basis too. ISO 139 sets out standard atmospheres for textile conditioning and testing, along with an agreed alternative. Apply the relevant method and material requirements; do not assume that one fixed waiting time suits every yarn.

Check that the test specimens still represent the stated process stage. A sample labelled “ring yarn” may already have passed through winding, rewinding or another surface-changing operation. A test cannot locate the source of a difference if that history is missing.

Why fibre length deserves attention

A simple mass calculation explains why shorter fibres can change the opportunity for ends to reach the yarn surface. It does not predict a hairiness index.

Imagine two ideal samples, each containing 1 g of straight, uniform fibres. Both have a fineness of 2 dtex, meaning 2 g per 10,000 m. The first sample has a fibre length of 40 mm.

Mass of one fibre:

2 ÷ 10,000 × 0.040 = 0.000008 g

Number of fibres in 1 g:

1 ÷ 0.000008 = 125,000 fibres

With two ends per fibre, there are 250,000 fibre ends.

Now shorten every fibre to 32 mm while keeping the same fineness and total mass:

2 ÷ 10,000 × 0.032 = 0.0000064 g per fibre

1 ÷ 0.0000064 = 156,250 fibres

This gives 312,500 fibre ends, which is 25% more than before.

It does not give 25% more H or S3u. Most ends are embedded within the yarn. Their position, fibre arrangement and route through spinning affect which ends project. Real cotton also has a length distribution, rather than one length shared by every fibre.

The example isolates geometry. If fineness changes at the same time, the number of fibres per gram changes for another reason. Keep those effects separate. ApparelDiary's guide to yarn count systems helps distinguish the yarn's linear density from the fineness of its individual fibres.

When comparing lots, record the length distribution and short-fibre basis as well as the mean. A percentage by number is not the same as a percentage by mass. Two lots with equal average length can still have different amounts of very short material.

Do not confuse a short fibre with a short visible hair. The first describes the length of a whole fibre in the raw material or prepared sample. The second describes only the portion seen beyond the yarn body. A long fibre can leave a short end exposed, while much of it remains embedded. These measurements therefore belong on separate lines in the trial record.

Fibre properties do not act one at a time

Length is only one part of the investigation. Fineness, fibre type, shape, finish and mechanical behaviour belong in the material description. However, a change in several properties at once cannot establish which one caused the surface result.

Suppose a mill replaces one blend component and also changes yarn count, twist and spinning route. Lower hairiness in the new yarn may be useful, but that trial does not isolate the effect of the new fibre. It evaluates a whole product and process combination.

A Rieter study of mechanically recycled fibres compared selected tear fibres blended with virgin cotton. It reported changes in hairiness and abrasion as the blend and short-fibre content changed. Those variables moved together in the study, so its results are not a universal limit for every recycled material.

The same report's later fabric results are instructive. It found no worse pilling from the recycled fibres in the tested comparisons, despite higher hairiness. That finding limits any attempt to turn its yarn-surface trend into a general prediction of fabric life.

For a fibre trial, hold the yarn and processing conditions steady where practical. Where that is impossible, name the combined change. A useful conclusion can be narrow: this blend worked with this preparation route and these tested fabric conditions. It does not need to claim that one fibre shape or supply source is always superior.

Find where the extra hairs appear

The stage at which a sample is taken can be as important as the fibre lot. A yarn may leave spinning with one surface and reach the fabric machine with another.

Rust and Peykamian's winding study examined hairiness changes during winding, including tension and fibre transfer. Its public abstract reports increased hairiness after winding. That supports checking the winding stage, but it does not supply a suitable machine setting for every current yarn.

Take matched samples before and after the operation under investigation. Keep the test method fixed. Record the parent package so that a poor incoming yarn is not mistaken for a bad winding position.

If the added count is concentrated at one position, inspect its yarn path and component condition against the machine's approved setup. If all positions rise after a material change, broaden the investigation. Neither pattern alone proves a cause; it tells you where the next comparison will be most useful.

Avoid changing several settings and replacing several parts in one trial unless the aim is to assess that whole repair package. Otherwise, a successful result will not reveal which change mattered.

A matched winding trial with two valid percentages

Consider a teaching trial on four parent packages. Fresh adjacent lengths are assigned to reference testing and to two winding conditions. The control uses the existing approved arrangement. The trial uses a replacement guide of the specified type.

Test lengths, sensor settings and yarn conditions are held equal. The two routes are assigned in a balanced order to reduce a simple time-order bias. Each table value represents the same tested yarn length and is reported per 100 m.

The figures are hypothetical S3u results, not a recommended limit or a required sample size.

The control adds an average of:

630 − 420 = 210 hairs per 100 m

The trial adds:

510 − 420 = 90 hairs per 100 m

The trial's final count is lower by 120 hairs per 100 m. Relative to the control's final count, the reduction is:

120 ÷ 630 × 100 = 19.05%

Relative to the count added during winding, the reduction is:

(210 − 90) ÷ 210 × 100 = 57.14%

Both calculations are valid. Calling the second figure a 57.14% reduction in final hairiness would be wrong. It describes the reduction in the increase above the reference value.

Retain the package-level results. The four paired control-minus-trial differences are 120, 110, 120 and 130. Their consistency is encouraging in this invented example, but four parents do not establish a universal effect. More evidence is needed across normal material variation and operating periods.

The trial also leaves an average increase of 90. It has reduced added long hairs without removing them. That distinction may matter when a downstream process still has an unacceptable result.

The reference specimens also deserve scrutiny. They are fresh yarn segments from each parent, not the same physical lengths after repeated passage through a tester. If the incoming reference differs greatly between trial groups, compare the changes with care. A lower final count could partly reflect a smoother starting sample. Pairing by parent helps address that problem but does not remove all variation within a package.

Before extending the change, check other relevant outcomes. A lower hair count is not enough if the route adds breaks, damages yarn or creates an unstable package. Compare accepted output and fabric performance on the same material basis.

Look beyond the mean along the yarn

Suppose five equal-length blocks have mean H readings of 4.5, 4.5, 7.5, 4.5 and 4.5. Their overall mean is:

(4.5 + 4.5 + 7.5 + 4.5 + 4.5) ÷ 5 = 5.1

A second yarn stays at 5.1 in all five blocks. Its overall mean is identical, but the first yarn has a distinct high-hairiness section.

These invented block readings help explain the value of a lengthwise diagram. They are not a calculation of the instrument's sh result. Changing the length over which readings are averaged changes what variation remains visible.

A historical Uster core-yarn application report describes a periodic hairiness fault that was not clear in the mass signal. Its investigation linked the pattern to the particular roving traverse arrangement. The case supports examining the hairiness trace in its own right. It does not justify disabling a machine function on another line.

When a peak repeats, compare its spacing with possible process cycles. When it occurs only once, examine handling and sample history. Keep the original trace so that smoothing or a large average does not hide the feature being investigated.

For repeated production monitoring, establish a stable reference period before judging a shift. The practical guide to statistical process control explains why process variation and specification limits serve different purposes. A small numerical change is not automatically evidence that a repair worked.

Use images to answer a defined question

A microscope can help show whether the projections are short ends, long loose fibres or loops. Images are most useful when the question is narrow: does a suspect section contain a different kind of projection from its matched reference?

Keep scale, lighting, background and selection rules consistent. Include more than one field. Choosing only the worst image from one sample and the neatest from another creates a persuasive picture without a fair comparison.

Telli's 2023 investigation of image-based hairiness measurement examined 13 ring yarns with different materials and colours. Relationships between image results and Uster readings depended on the group being compared. The study shows why a useful correlation in one set is not proof that two methods agree for every yarn.

Use images alongside the defined test results. If an image method is introduced for routine decisions, establish its repeatability and relationship to the existing method with representative samples. A clear photograph alone does not establish a transferable numerical limit.

Confirm the result in the intended fabric

A surface measurement is an input to a product decision. It is not a direct measurement of fabric pilling, shade appearance or service life.

For example, ISO 12945-2:2020 describes fabric pilling, fuzzing and matting assessment using the modified Martindale method. Its subject is the fabric response under that method, rather than a conversion from yarn H to a fabric grade.

Use matched fabric constructions and finishing conditions when comparing yarn trials. Otherwise, a fabric change can conceal or exaggerate the effect attributed to hairiness. Keep the assessment suited to the intended use and the buyer's agreed method.

A process change may be accepted when it improves the relevant yarn measure and the fabric result, with acceptable production behaviour. If only the yarn number improves, say so. That is a useful finding, but it is a smaller finding.

Turn the investigation into a clear decision

A useful hairiness report lets another reader follow the comparison without guessing. It identifies the yarn and process stage, names the instrument result, states the tested length and records the relevant test conditions.

Then it separates observation from explanation. “Long-hair counts rose after winding” is an observation. “The guide caused the rise” is a proposed cause until a controlled comparison supports it.

Finally, state the result on the right basis. Give the final level, the added amount where relevant, and the package or lengthwise spread. Link acceptance to the agreed fabric use. This makes the test a practical way to improve yarn processing while keeping its limits visible.

References

1. Uster Tester 6 technical data, April 2026.

2. ASTM D5647-07(2025): photoelectric hairiness testing, public scope and significance.

3. Haleem and Wang: comparison of manual and commercial hairiness measurements.

4. Telli: investigation of image processing for yarn hairiness, 2023.

5. Rust and Peykamian: hairiness changes during winding, 1992.

6. Uster: hairiness variations in core yarns, historical application report.

7. Rieter: recycling in the staple-fibre spinning process.

8. ISO 139: standard atmospheres for textile conditioning and testing.

9. ISO 12945-2:2020: modified Martindale pilling, fuzzing and matting method.