// Showcase

Measuring diaphragm thickness: a feasibility study instead of a machine purchase

Ceramic diaphragms for pressure sensors are ground to a defined thickness.

01Starting point

Ceramic diaphragms for pressure sensors are ground to a defined thickness. That thickness sets the pressure range of the finished sensor, and because the grinding process scatters wider than the classes it is supposed to produce, every single part has to be measured and sorted. The requirement was ±1.5 µm at 3σ, without contact, at five measuring points, in a cycle time under 3 s. The two narrowest classes are just 20 µm wide. And the open question was not only which sensor to measure with, but what to measure at all: the thickness itself, or the deflection of the diaphragm under pressure.

02Approach

We worked out both before anyone designed a machine. The capacitive variant failed on its own geometry: the five-point requirement shrinks the electrode to 2 mm, the capacitance to between 0.09 and 0.26 pF, and the sensitivity to 0.08 % per µm, the error propagation lands at ±47 µm instead of ±0.5 µm. Not an optimisation problem, simply the wrong method at this part size. The chromatic confocal measurement came close to specification, and we then took it apart in six gauge R&R runs: three parts, both faces, three appraisers, ten measuring points, three repeats.

03Result

The result was unambiguous and not the expected one. Repeatability in five of the six runs lies between 0.020 and 0.032 µm: the sensor is many times better than it needs to be. Reproducibility lies between 0.11 and 0.77 µm, up to thirty-six times larger. Against a 10 µm tolerance band the measurement consumes between 6.7 % and 51.3 %, with an identical sensor throughout. That settled where the money belongs: not in a better measuring head, but in the fixture and in repeatably reaching the five measuring points. As an aside, the ground face and the raw sintered face turned out to be indistinguishable in measurement behaviour.

What's inside

A feasibility study covering two sensing principles with full error propagation, a surface analysis from an accredited laboratory over 2 × 2 mm² areas at 50 µm line spacing, six gauge R&R evaluations with their raw data, and two fully worked machine concepts: thickness at five points, or deflection under a test pressure of up to 4 bar, each with schedule, gates and cost. Plus the comparison that carries the decision: the deflection variant needs a 5 s cycle instead of 3 s, 26 weeks instead of 20, and roughly 1.9 times the outlay.

And what happened next

What was ordered was the first of two project phases: development of the measuring head for the thickness route, ten weeks, with the single aim of demonstrating accuracy, repeatability, reproducibility and cycle time before a machine is built around it. Worth keeping is the side finding from the surface analysis: what counts as "thickness" depends on the measuring spot, filtered over two points the same profile yields a 3 µm peak, filtered over twenty points a 20 µm one. That question belongs in the specification, not in the instrument.

All showcases

Get in touch

We are ready for new challenges

You bring the challenge. We develop the solution. From the first idea through concept and development to the validated machine, we guide your project.