Protocols copied from before 2015 still calculate a 95% upper confidence limit for rooms with few sampling locations.
The calculation came from the 1999 edition of ISO 14644-1. For a room with two to nine sampling locations, the tester took the mean of the location averages, worked out a standard deviation and a 95% upper confidence limit (UCL) with a t-distribution factor, and compared that with the class limit.
ISO 14644-1:2015 removed the step. The number of sampling locations now comes from Table A.1, by room area, and each location is judged alone: the average concentration at every location has to meet the class limit.
The statistics are still there, inside the table. Table A.1 is built so that, if every location passes, there is at least 95% confidence that at least 90% of the room area is within the class limit.
A protocol that cites the 2015 edition and keeps the UCL formula mixes two methods. It can fail a room in which every location passes and one of them reads much higher than the rest, a failure the current standard does not define. It also tends to keep the old number of locations, the part that should have changed.
EU GMP Annex 1 (2022) points to ISO 14644 Part 1 for the minimum number of sampling locations and their positioning, so the template is worth opening before the next requalification.
Industries




