A single differential gauge across several AHU filter stages cannot say which stage loaded, so teams change everything or guess. A gauge per stage, marked with the commissioning reading and change limit and kept on the calibration list, allows each stage to be changed on condition. BMS transmitters add a trend.
Annex 1 initial validation of an aseptic process needs at least 3 consecutive successful simulations covering all working shifts, then periodic runs twice a year per process, line and shift. The cleanroom must hold its qualified state throughout. Agree an HVAC freeze date with QA, and list in advance which jobs are like-for-like and which need an impact assessment.
21 CFR 11.10(e) requires secure, computer-generated, time-stamped audit trails. For an environmental monitoring system, that means the software logs setpoint, alarm-limit and user changes itself, nobody can purge entries, every clock follows one time source, and the requirements sit in the URS before a platform is chosen.
A pre-filter changed at 150 Pa on a 0–1,000 Pa gauge sits in the bottom 15% of the dial with about 20 Pa of uncertainty. A 0–250 Pa gauge puts it at 60% with 5 Pa. Range each stage's gauge and transmitter to its own change point.
GAMP 5 Second Edition (2022) scales validation effort to risk to patient safety, product quality and data integrity, and adds agile and cloud guidance. On a GMP BMS, test alarms, audit trail and data transfer deeply. A quick check: count low-risk functions that still carry full test scripts.
When a supply AHU trips with extract fans running and dampers open, a Grade B room loses its 10–15 Pa margin and pulls air from the lower grade. Tight-shutoff dampers with a tested leakage class, defined fail positions, interlocks, proven position and a supply-first restart keep the cascade, and a deliberate fan trip at FAT proves it.
A paper room-pressure log with every reading in the same pen at the end of the shift fails ALCOA+ before any number is checked. Walking one sheet against the nine attributes, from attributable initials to enduring ink, shows that most data-integrity failures are fixed by redesigning the form.
21 CFR 11.10(e) says record changes shall not obscure previously recorded information. A SCADA log that stores only the new setpoint fails on that phrase. Three tests before OQ: read the trail after a change, try to edit it as administrator, and compare its retention with the batch records.
When cooling and reheat valves open together at 20 ± 0.5 °C, the cause is usually a zero gap between the two control sequences. Check the trend first. Where the product data supports 20 ± 2 °C, cooling above 21 °C and reheat below 19 °C let the room drift between with both valves shut.
A fixed-speed fan filter unit loses airflow as its HEPA filter loads. Set at 0.45 m/s, it has 20% to lose before the 0.36 m/s bottom of the Annex 1 guidance range. Constant-airflow EC units hold velocity if they are selected at the final pressure drop, and their speed trend doubles as a filter-loading record.
Measuring every room against one reference stops errors stacking but leaves room-pair differences at up to ±2 Pa with ±1 Pa sensors. Room-to-room sensors show airflow direction directly but add errors along the chain. Record one set, compute the other, and state the error budget.
Individual logins make records attributable, but user lists drift as people move and leave. A periodic access review exports users and roles from the system, checks them against HR's leavers and movers, justifies every admin or generic account and is signed with its corrections, the record an inspector will ask for.