Contact

Professional · Internship

Analog Signal Conditioning Internship

A one-week summer internship with Professor Antonio Selmo on analog signal conditioning for transducers, studying and measuring the modules of a universal conditioning board used by an ESA laboratory.

This summer I spent a week in Professor Antonio Selmo's laboratory for an internship on analog signal conditioning for transducers: everything that sits between a sensor and the analog-to-digital converter, and has to turn a small, noisy electrical signal into a clean voltage that still means something. The thread running through the week was a universal transducer-conditioning board Professor Selmo designed, which is used today by a European Space Agency laboratory in Padova.

The modules, one at a time

Before looking at the board as a whole, I worked on its building blocks as standalone circuits. Each module — op-amp gain stages, the instrumentation-amplifier front end, the filters, the excitation for resistive sensors — was built on its own, on a breadboard or a perfboard, and measured on the bench before moving to the next.

The theory came alongside: how a transducer's transfer law has to stay invertible, how to map a sensor's range onto the converter's input without saturating an intermediate stage, why a remote Pt100 needs a four-wire Kelvin connection, how interference couples into long cables, and how the ideal op-amp model gives way to offset, bias current, noise and finite bandwidth. Building each block by hand made that theory concrete very quickly.

Measuring the filters

One of the modules I tested was the filter stage. Professor Selmo had a board with three fifth-order low-pass filters built on it — one Chebyshev, one Butterworth and one Bessel — all with their cut-off around 1 kHz. I measured the three of them to understand how they behave, both in the frequency domain and in the time domain.

In frequency, I swept each filter from 100 Hz to 10 kHz and recorded its gain and phase on the oscilloscope's Bode plot. Chebyshev cuts off the sharpest and Bessel the gentlest: at 2 kHz they attenuate by about −39 dB and −13 dB, with Butterworth in between.

In time, I fed them a square wave and compared input and output. Chebyshev rings and Butterworth overshoots, while Bessel adds the shortest delay and keeps the shape of the signal almost intact. In a measurement chain, where a transient has to come out undistorted, that is why the Bessel filter is the one used on the conditioning board. The full set of captures is in the report.

Op-amp stages and four-wire resistance

The op-amp stages were measured the same way, with a function generator on the input and both signals on the scope: an inverting amplifier at 800 Hz, sine responses at different frequencies, and an integrator turning a square wave into a triangle whose amplitude grows as the frequency drops.

For resistive sensors like the Pt100, I set up a four-wire measurement: a decade resistance box standing in for the sensor, with the excitation current and the sensed voltage on separate pairs of leads into an HP 3456A, so the lead resistance drops out of the result. With the box set to 50 Ω, the meter read 50.05 Ω.

The report

At the end of the internship, Professor Selmo and I wrote a report that brings the week together. The theoretical content is his; my part was studying those principles, measuring the board's modules and documenting the design philosophy behind it.

The report follows the measurement chain in order — transducer theory, range mapping, excitation and loading, input integrity and interference, filtering, op-amp and instrumentation-amplifier foundations, gain and calibration, and output stages — with the bench measurements above built into the filter chapter. Its final chapter applies all of it to the ESA board: the selectable inputs and excitation, the filter banks, the gain and offset stages and the output buffer, down to the populated board configured for a four-wire Pt100 channel.

Documents & links

Let’s build
something together.

[email protected]