The problem is not the screw, it is the window
A torque specification for an M6 bolt has room to breathe. The bolt can be under-tightened by 20% and still hold; it can be over-tightened by 20% and still not yield. At M1 the same percentages collapse into a range narrower than the repeatability of a typical pneumatic driver.
The reason is geometric. Torque scales roughly with the cube of diameter, but friction scatter and driver repeatability do not scale down. A driver that holds ±5% at M6 is still ±5% at M1, and 5% of a 0.10 N·m target is 0.005 N·m — well inside the noise of a clutch that was designed for a much larger joint.
Calculate the window before you specify a torque
The target torque should sit between the minimum that produces the required clamp load and the maximum that the joint can survive. For a micro screw the upper bound is usually strip torque in the female thread, not the screw itself.
- Required clamp load: from the functional requirement (sealing, thermal contact, vibration resistance), not from a rule of thumb.
- Torque from preload: T = K × F × d, where K is the nut factor (roughly 0.2 for as-received steel, 0.13–0.16 for lubricated or coated).
- Upper bound: 60–70% of the measured strip torque in the actual mating material.
- Lower bound: the clamp load that still meets the function after relaxation over the service life.
Measure strip torque in the real material
Published strip torque tables assume a specific material and engagement length. They are a starting point, not a specification. An M1.4 screw that strips at 0.45 N·m in brass may strip at 0.28 N·m in a 2 mm aluminium wall, and at 0.6 N·m in a steel insert.
Run ten drives into the actual mating part, record the strip torque, and take the lower bound of the distribution. Design the target so that the upper tolerance of your driver is still below 70% of that value.
| Size | Typical strip torque, brass insert (N·m) | Typical strip torque, aluminium 2 mm wall (N·m) | Recommended target (N·m) |
|---|---|---|---|
| M0.8 | 0.10 | 0.06 | 0.04 |
| M1.0 | 0.18 | 0.11 | 0.07 |
| M1.2 | 0.26 | 0.16 | 0.11 |
| M1.4 | 0.36 | 0.22 | 0.15 |
| M1.6 | 0.50 | 0.31 | 0.22 |
| M2.0 | 0.90 | 0.55 | 0.40 |
Control the friction, or the torque means nothing
The nut factor K can vary from 0.12 to 0.25 depending on finish and lubrication. That is a two-to-one variation in clamp load for the same torque. If the finish changes between lots, the joint changes between lots, even though the torque setting never moved.
For critical micro joints, specify the finish system and the friction coefficient, and ask for torque-tension verification per lot. Zinc-nickel with a defined lubricant, or a PVD film with a specified µ, is more repeatable than an unspecified "zinc plated" callout.
Make the driver part of the specification
A torque specification is only meaningful if the tool can hold it. For M1 and below, use a current-controlled electric driver rather than a mechanical clutch, verify it with a torque transducer at the start of each shift, and check the first three parts of every lot.
If the process still produces scatter, the answer is usually to move up one screw size or to add a thread insert — not to tighten the torque tolerance further. At some point the joint is asking more of the fastener than the geometry can deliver.