High Bandwidth Piezoelectric Amplifier

The problem

The wind tunnel setup drives vertical oscillations of vortex generators using a commercial Thorlabs piezoelectric actuator through a Thorlabs amplification device. That device is built for precision positioning in a laboratory, not for repeated dynamic operation under aerodynamic load, and it kept failing under off-axis loading.

The job was to replace the amplification stage while keeping the existing actuator, so the constraint set was fixed by hardware we were not going to change.

The commercial structure it replaced managed about 9.4× amplification, but at the cost of force, and 0.4 mm of amplified displacement in wind tunnel testing.

What I did

I pursued a rhombus-based compliant structure, following the amplification-ratio modelling in Ling et al. A rhombus converts the stack's small, stiff axial extension into a larger transverse displacement, with no sliding joints to wear or bind.

I read a good deal about flexure hinges while working out the geometry, and then deliberately left them out — the simpler structure was easier to make and easier to reason about, and I could not justify the added complexity from what I understood at the time.

I ran an FEA sweep on the arm thickness, trading amplification against first resonance — the two requirements pull in opposite directions, since the compliance that buys displacement also drops the natural frequency. The geometry I selected predicted 11.8× amplification with a first mode at 912 Hz, against a target near 1 kHz.

The geometry went through two generations of hardware. PLA prints came first, cheap enough to iterate on and enough to check that the stack seated properly. The tested article was cut from 6061 aluminium on the water jet — I produced the drawings and submitted the work order; the machine shop cut it.

Set screws at each end provide uni-axial pre-compression, which maximises force transfer into the stack and reduces the risk of fracturing the piezoelectric ceramic.

How it turned out

The prototype measured 253 µm maximum displacement — 5.6× amplification, against 11.8× predicted. It reached a blocking force of 132 N against roughly 100 N for the device it replaces, a 32% improvement, which is enough to drive the PLA vortex generators with only minor changes to the housing.

So the design did what it was brought in to do — more force, and a structure that does not break under off-axis load — and came in well under the predicted amplification.

Plotting the amplifier's stiffness line against the actuator's own force-displacement characteristic puts the operating point at 42.1 µm and 61.0 N, a spring constant of 1.45 N/µm. That is what the stack delivers into this structure rather than what it delivers unloaded.

I should be straight about what that 5.6× is and is not. It is one measurement, from one test, that I never repeated. I do not know whether the design genuinely amplifies less than predicted or whether that run was bad data, because I did not go back and check. Two weeks after that test I was in finals, then the summer internship started, and by the time the new school year came round I no longer had the time for the lab. The project stopped there rather than concluding, and nobody has picked it up since.

I would rather say that than offer an explanation for a gap I never actually investigated.

What I'd do next

The first thing is not a design change, it is repeating the measurement. One data point is not a result, and everything downstream — whether the FEA needs revisiting, whether the interface is losing displacement, whether the geometry is simply wrong — depends on knowing that 5.6× is real.

After that: a swept-sine test to find the actual first resonance, which was never measured either, and wind tunnel validation of the assembled housing.