Crimp height: what it tells you and how tightly to control it

Crimp height tells you whether the tool closed the terminal to its set-up position on this wire. It is a quick, non-destructive signal of compression that you can chart all shift. Hold it to the terminal maker's tolerance, which on TE's general table is ±0.05 mm for wire over 0.5 mm² and up to 6 mm², and ±0.02 mm on the finest wire,¹ and react to shifts inside the tolerance as well as to readings outside it. A height inside tolerance does not prove a good crimp, because it cannot see cut strands, insulation in the barrel or the wrong crimper.

What the number stands for

Height stands in for compression. TE's tooling paper defines an area index: total area after crimping divided by the area of wire plus barrel before crimping, with 80% as a typical design point. It shows the minimum and maximum cross-sections that one terminal specification allows, a variation of ±3.5%.² At a fixed crimp width the index moves in proportion to height. On TE's example crimp, 0.068 in (1.727 mm) high with a tolerance of ±0.002 in (±0.051 mm), the height tolerance alone is ±2.94%. That is the same order as the ±3.5% TE shows for one terminal, though the two are separate examples, so take it only as a sense of scale.

The same paper works out that a 2% change in crimp width costs 35% of that height tolerance,² so changing the crimper can move compression with nobody touching the height adjustment.

Resistance can rise on either side of the window, as the harness manufacturing guide explains, and Molex sets its specification as a balance between electrical and mechanical performance across stranding, coatings and platings.³

How tight is tight enough?

Use the maker's figure for your terminal and wire. TE's general guideline steps the tolerance by wire size.¹ The other columns are calculated.

"Uses" is the instrument's error as a share of half the tolerance. Molex's crimp micrometer quotes 0.001 mm resolution and ±0.002 mm accuracy,⁴ and Mitutoyo's 342 series 0.001 mm and a maximum permissible error of ±0.003 mm.⁵ Cp 1.33 is the level Molex calls acceptable for most applications,³ and your customer may ask for more. Sigma is width ÷ (6 × Cp).

These are TE's general values for its open barrel terminals, and a product specification overrides them.¹ One TE 3.3 mm pitch receptacle specifies ±0.03 mm on an applicator even at 0.51 to 0.53 mm², where the general table would allow ±0.05 mm. Its hand tool table for the same wire gives 0.90 mm ±0.05 mm against 0.93 mm ±0.03 mm, so read the row for your tooling.⁶

Against the tolerance, the instrument's error is the small part. The larger unknown is what happens when a person places a point on a formed crimp, and neither maker publishes a repeatability figure for that. Molex recommends a gauge capability study.³ The table sizes the possible results, taking a range of gauge standard deviations as an illustration.

Each cell is 6 × gauge sd ÷ tolerance width. As Minitab reports the AIAG bands, under 10% is acceptable, 10 to 30% depends on the application and the costs, and over 30% is unacceptable.⁷ At ±0.02 mm a gauge sd of 0.002 mm already sits on the 30% line.

Measurement noise also lowers the capability you report, which Molex warns about for micro-terminals.³ A process truly at Cp 1.33 reads 1.14 through a gauge sd of 0.003 mm at ±0.02 mm, and 1.29 at ±0.05 mm. Target accuracy gets squeezed too. A process sigma of 0.004 mm at ±0.02 mm leaves only 0.004 mm off target before Cpk falls below 1.33, about the size of the micrometer's own error. At ±0.05 mm the same process has 0.034 mm of room.

Five pieces cannot give you a Cpk

Molex asks for at least five readings to confirm a set-up and at least 25 for capability, and says itself that set-up samples have limited statistical value.³ NIST says the usual interval needs at least 25 pieces and that capability studies generally need 100.⁸ A reading of 1.33 is worth this much at each size.

The intervals use NIST's approximation, Cpk ± z × √(1/(9n) + Cpk²/(2(n - 1))), with z of 1.96 for the range and 1.645 for the lower bound.⁸ Below 25 pieces that is outside what NIST calls valid, so treat those rows as rough. As a cross-check, the exact chi-square interval on Cp gives 0.46 to 2.22 at 5 pieces and 0.96 to 1.70 at 25. The two probability columns are simulated: 400,000 runs per cell, centred normal process, independent pieces. In the same simulation the lower bound held about 97% of the time even at 5 pieces, so here it errs on the cautious side.

So 25 pieces reading 1.33 tells you, with 95% confidence, only that the process is above 1.00. A five-piece reading tells you the average is about where you put it, which is what a set-up check needs. Do not write a Cpk on it.

Short-run shops cannot run a 25-piece study on every set-up, and Molex says so. Its answer is to establish capability before the tool goes into production, then keep one control chart per tool and wire size across colour, length and set-up changes.³ Twenty set-ups at five readings each give you 100 readings and show how far set-ups differ from one another. Treat the result as a long-run figure. If set-ups vary as much as pieces do, the 95% range on a reading of 1.33 is about 1.07 to 1.61 in simulation, wider than the 100-piece row.

The chart also gives you a working answer to "how tightly?". Molex's limit for samples of five is the average of the means ± 0.577 × the average range.³ With an invented average range of 0.010 mm, that is ±0.006 mm around the average, well inside a ±0.05 mm tolerance. At that range the total sd is about 0.004 mm, so a gauge sd of 0.003 mm would be about half the variance on the chart. Run the gauge study before you chase the chart. The tolerance tells you what to reject, and the chart limits tell you when to go and look.

A reading outside tolerance is different from a shift inside it. Stop the tool and hold everything back to the last good reading, as in the pull test containment steps. Then look for the cause.

If your terminal and wire have no published height, ask the terminal maker. Do not interpolate: one TE sheet's rows for 0.31 to 0.38 mm² and 0.51 to 0.53 mm² differ by 0.13 mm, and a 0.5 mm² wire is in neither.⁶

A set-up routine

This is condensed from Molex's 21-step procedure for its own tooling,³ which has the detail. Where your press or applicator maker says otherwise, follow them. Disconnect power and refit guards around every adjustment.

1. Check that the crimper and anvil are clean, unworn and right for this terminal and wire. Your target comes from the product specification or applicator log sheet.¹
2. Install the tooling, cycle it by hand, and check alignment and feed.
3. Crimp samples under power and set, in this order, the cut-off tab and bell mouth, the conductor brush and the insulation position.
4. Loosen the insulation crimp, because adjusting it can change the conductor crimp height slightly.
5. Crimp at least five and measure each. Put the blade across the top of the crimp, mid-barrel and away from the bell mouth, and bring the point onto the bottom radius, not onto the extrusions. Use the ratchet where there is one and do not overtighten the point.⁴⁵ Adjust until five are on target.
6. Pull test the conductor crimp, then set the insulation crimp, crimp again and measure the conductor height again.
7. Record the target, the readings, the tool, the terminal reel lot and the wire lot.

During the run, Molex says to check height every 250 to 500 parts, and to record the point on the chart before you adjust, so the chart shows the process and not your corrections.³ That interval is 50 to 100 minutes at 300 parts an hour, 12 to 25 minutes at 1,200 an hour and 5 to 10 minutes at 3,000 an hour. TE adds a check for each batch and after each change of terminal reel, wire, crimp tool or setting, and asks for regular microsections from the first crimp on.¹

When the reading moves, what to look at first

The causes are from Molex's troubleshooting tables and control chart notes.³ The order of checks is an editorial suggestion, quickest first.

Why a height inside tolerance does not prove a good crimp

TE calls crimp height "the overriding quality characteristic", and in the same document asks for microsections, pull force checks and, on semi-automatic and automatic machines, preferably crimp force monitoring.¹ Height is the check you can afford to run often, and the others cover what it cannot see.

On a press, the closing point is set mechanically. Molex defines shut height as a distance at bottom dead centre, lists a changed shut height among the causes of off-target height, and lists cut or missing strands as a source of scatter.³ The inference is that a press closes to about the same place whether or not every strand is there, and the reading moves only a little. Presses do flex under load,⁹ so a crimp with fewer strands may read slightly low, but we found no published figure.

TE's area index can size what that gap hides. Hold the crimped area fixed, assume the barrel has half the area of the wire (an invented ratio), and take out 2 of 19 strands. The index goes from 80 to 86.0, or 85.3 to 87.0 for barrel ratios of 0.7 and 0.3, well beyond the ±3.5% TE shows for one terminal. To get back to 80 the height would have to fall by 7.0%, which is 0.121 mm on a 1.727 mm crimp, more than twice the ±0.051 mm tolerance. In this illustration, the micrometer can read on target while the copper sees a crimp 0.12 mm too high.

You can test this on your own wire. Crimp sets with one, two and three strands removed and measure them. Schleuniger suggests the same exercise for setting up a crimp force monitor.⁹ It shows how much your chart would notice on that terminal.

A monitor has limits as well. TE asks for it to flag a loss of more than 9% of the conductor section through cut strands, and insulation over more than half the crimp length.¹ Worked through, a monitor set to that rule should flag 1 of 7 strands or 2 of 19, and need not flag 1 of 19 or 5 of 65. Whether it can depends on the application. Schleuniger, a monitor supplier, says companies usually find the best headroom, the gap in peak force between a crimp with wire and one without, is 30 to 40%, and that harder terminals shrink it. Ask for that figure before you rely on the monitor. The same article says it often finds monitors switched off on shop visits.⁹ Molex wants the operator to fan each bundle and check it by eye as standard practice,³ and a look at the brush is the quick way to notice strands that did not make it into the barrel.

Notes

1. TE Connectivity, Application Specification 114-18022, Rev K, 2011, general guidelines for contacts with open crimp barrels, copy hosted by Farnell. This is Rev K, the copy we could open. Check TE for a later revision. Written first for the automotive industry; it says it can be used for other industries. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

2. TE Connectivity, Crimp Tooling: Where Form Meets Function, 2011, white paper, copy hosted by ChipDip. A tool maker's paper, used here for its definitions and worked examples and not for its product claims. ↩ ↩²

3. Molex, Quality Crimping Handbook, order no. 63800-0029, revision C, 2009, copy hosted by DigiKey. Written for Molex open barrel terminals and tooling. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹²

4. Molex, Crimp Height Micrometer 63800-2850, application tooling specification, copy hosted by ChipDip. ↩ ↩²

5. Mitutoyo, Crimp Height Micrometers, Series 342, catalogue page, copy hosted by a distributor. ↩ ↩²

6. TE Connectivity, Application Specification 114-106094, Rev D, copy hosted by Farnell. ↩ ↩²

7. Minitab, More on how to interpret gage output. Software vendor's blog, reporting the AIAG guidelines. The AIAG manual itself was not read. ↩

8. NIST/SEMATECH, e-Handbook of Statistical Methods, section 6.1.6, What is process capability?. US government publication. It assumes normally distributed data. ↩ ↩²

9. G. Garcia-Cota, Schleuniger, Crimp Force Monitoring: The Recipe for Success, supplier article. ↩ ↩² ↩³