One BMS pressure alarm test often runs to forty pages of screenshots, and none of them shows the transmitter failing.
FDA's Computer Software Assurance guidance goes after exactly that. Classic CSV puts the effort into documentation: a pre-approved script, a screenshot per step, a signature per page. CSA moves it to intended use and risk, and tests hardest where a failure would hurt product or data.
Take the pressure alarm between a Grade C corridor and a Grade D change room. Under a scripted approach, the tester forces the value below setpoint, waits out the delay, captures the banner, acknowledges, captures the event log. Every step is written and evidenced in advance.
Under CSA, the first question is what the alarm is for. It protects the cascade, so it is high risk and still gets scripted tests on setpoint, delay and the record it writes. Then the tester spends time unscripted: pulling the transmitter signal, changing the setpoint without the right role, restarting the server mid-alarm. The record says what was tried and what happened, without a screenshot per click.
The guidance was written for medical device production and quality system software, so a pharma site applies it by analogy, usually alongside GAMP 5. We write the validation strategy in parallel with the design, which is where the test depth for a function like this gets decided.
Which BMS alarm has your thickest test pack and thinnest risk case?
Industries



