Skip to content
←All work

Medical Devices

A diagnostic front end with a noise floor problem

An early-stage medical diagnostics company

recovered to bench

Noise floor

2

Board revisions

40

Units in pilot build

The problem

A benchtop prototype produced clean readings. The same circuit, once integrated into the product enclosure alongside the power supply and display, lost roughly a decade of usable resolution — enough to make the device useless for its intended assay.

What we did

  1. 01

    Reproduced the fault in a controlled setup to establish that the coupling was conducted through the supply rather than radiated.

  2. 02

    Redesigned the board with a separated analogue section, a quiet reference and a deliberate single-point return, rather than adding shielding after the fact.

  3. 03

    Moved the switching supply off the sensitive board entirely and specified a filtered feed.

  4. 04

    Built a functional test that measures the noise floor on every unit, so a regression cannot silently reach a customer.

Outcome

The integrated device recovered its bench performance, and the per-unit noise floor test became part of the standard build. The design change was documented as it happened, which kept the validation burden contained.

A worked example: a representative problem of this type, and the reasoning we would apply to it.

Next project

A controller built around a part that no longer exists

→