A cap screw is sold by two numbers, diameter and length, and every dimension on its data sheet is a function of the first. Head diameter, head height, socket size — pick M8 and they are settled before you decide whether you want 20 mm or 80 mm. The thread length is the one column where that habit produces a wrong answer, because it is settled by the diameter too, and it does not follow the length at all.
Most of the screws here do not even have the column
Of the 1,614 screws in this database, 506 carry a thread length and 1,108 carry none. That split is not scattered rows — it is whole standards. The four metric tables that print the column are DIN 912, ISO 4762, DIN 7991 and ISO 10642, 436 rows between them. The other 16 standards have no unthreaded shank to describe: 512 headless set screws, which are thread from end to end by construction, plus countersunk (150), cheese (150), pan (150), button (73) and low socket (73) heads.
At one diameter it never changes
Take M8 in DIN 912. Ten lengths are stocked, and the thread length is 28 mm on all ten:
| Length L (mm) | Thread length b (mm) |
|---|---|
| 12 | 28 |
| 16 | 28 |
| 20 | 28 |
| 25 | 28 |
| 30 | 28 |
| 35 | 28 |
| 40 | 28 |
| 50 | 28 |
| 60 | 28 |
| 80 | 28 |
A M8 × 12 is thread from end to end, because 28 mm of thread does not fit on a 12 mm screw. A M8 × 30 has 28 mm of thread and 2 mm of plain shank. A M8 × 80 has the same 28 mm of thread and 52 mm of plain shank. The extra 50 mm you paid for bought shank.
The rule behind the column is 2d + 12
Across all 436 metric rows the thread length is the diameter doubled plus 12 millimetres — fifteen diameters, no exceptions:
| d (mm) | b (mm) | b ÷ d |
|---|---|---|
| 2 | 16 | 8.0 |
| 2.5 | 17 | 6.8 |
| 3 | 18 | 6.0 |
| 4 | 20 | 5.0 |
| 5 | 22 | 4.4 |
| 6 | 24 | 4.0 |
| 8 | 28 | 3.5 |
| 10 | 32 | 3.2 |
| 12 | 36 | 3.0 |
| 14 | 40 | 2.9 |
| 16 | 44 | 2.8 |
| 18 | 48 | 2.7 |
| 20 | 52 | 2.6 |
| 22 | 56 | 2.5 |
| 24 | 60 | 2.5 |
The constant term is what makes the ratio fall. At M2 the thread is 8.0 times the diameter; at M24 it is 2.5 times. Small screws are almost entirely thread, large ones are mostly shank, and the crossover is not a design choice anyone made per size — it is the same 12 mm added to every one of them.
Half of the short ones are all thread anyway
The table prints b on every row, including rows where it cannot be a length. 216 of the 436 metric rows have b larger than L — a thread longer than the screw — because b is the figure for the size, not a measurement of that part. Another 12 sit exactly at b, and 208 have a shank.
So b is a floor, and the column header says so: DIN 912 labels it "Thread length b (min)". For a screw shorter than b, the whole screw is thread and the printed number tells you nothing about the part in your hand. For a screw longer than b, you get at least b.
What that costs when you need engagement
Suppose a joint needs 40 mm of thread engagement. No M8 in these tables will give it: the guaranteed thread is 28 mm and no length changes that. The smallest metric cap screw whose b reaches 40 mm is M14, at 40 mm.
The alternative is to leave the cap-screw tables altogether. The families with no b column are thread over their whole length, so there L and the thread are the same number — a DIN 84 cheese head screw 40 mm long is 40 mm of thread, at a fraction of the strength of a cap screw. That trade is the real choice: engagement, or clamp load.
Thread engagement itself is covered in nut height and thread engagement, which works from the nut side of the same joint.
The inch table is a short list, not a formula
ASME B18.3 prints the same column over 10 diameters, but it is a ladder of fractions rather than a formula. 10 diameters produce only 8 thread lengths:
| d (in) | b (in) |
|---|---|
| 0.112 | 0.750 |
| 0.138 | 0.875 |
| 0.164 | 1.000 |
| 0.190 | 1.000 |
| 0.250 | 1.000 |
| 0.313 | 1.125 |
| 0.375 | 1.250 |
| 0.500 | 1.500 |
| 0.625 | 1.750 |
| 0.750 | 2.000 |
The first 4 steps are eighths of an inch, 0.750 to 1.250; the next 3 are quarters, 1.250 to 2.000. And 3 diameters share the single value 1.000 in. A formula cannot do that — 12 mm on top of twice the diameter gives fifteen different answers for fifteen sizes, where the inch table gives eight for ten.
The unit belongs to the table, not to the column
On a metric row b is millimetres. On an inch row it is inches, and the number is small enough to read as millimetres by mistake: 1.000 means 25.4 mm, not 1 mm. At that size the two systems land close — a 1/4-20 cap screw (6.35 mm) guarantees 1.000 in of thread, 25.4 mm, while the metric M6 guarantees 24 mm — but the printed figures differ by a factor of twenty-five, and nothing on the row says so except the unit.
Where this stops
The table gives one figure per diameter and calls it a minimum. It cannot tell you the thread length of a particular screw on a particular drawing — for that you measure the part, or you buy to a drawing that states it. What the table can do is stop the assumption that length buys thread, which is the assumption that quietly ruins an engagement calculation.
What to do with it
- Size the thread from the diameter, not from the length: at M8 the number is 28 mm and it will be 28 mm on every length.
- If the joint needs more engagement than 2d + 12 gives, change the diameter or change the family. Buying a longer screw is not one of the options.
- Check the unit on the row before comparing a metric and an inch screw. The column looks identical and the two numbers are not.