Shrinking a 10,000 m³/h duct from 700 × 500 mm to 600 × 400 mm raises velocity from about 8 to 12 m/s and velocity pressure from 38 to 80 Pa, so every fitting costs about twice. Set separate limits for mains and branches in the design basis.
A recovery test with a 20-minute limit, traced to a 15-minute URS requirement, passes and leaves the requirement unproven. Reading a trace matrix from each test back to its requirement finds these weak links and orphan tests. Pick three signed results at random and trace them back.
If a Grade C area grows from 800 to 950 m² between concept and basic design and the estimate is carried forward, the design has 19% more cleanroom than the budget covers, found at tender. Write gate criteria before the meeting: design basis frozen, URS traced, estimate re-run, FMEA risks owned, a schedule the team can staff.
A 10,000 m³/h fan at 1,500 Pa and 0.6 overall efficiency draws about 6.9 kW, around 61 MWh a year in a cleanroom. Every 100 Pa removed at design stage saves about 0.46 kW for the life of the plant. A pressure budget per AHU in the design basis is where it starts.
Bicarbonate-buffered media sit in incubators at around 5% CO2, and losing the supply raises the pH. Choose cylinders or bulk early, use a two-bank manifold whose changeover alarm reaches someone at night, and put the regulator, point-of-use filter and cleanroom penetration on the drawings.
Air in a ceiling void has never passed a HEPA filter. If the void sits above room pressure, leaks bring it down past light frames and HEPA collars. Keep one void at or below the lowest room under it, seal duct joints, and measure room-to-void pressure at commissioning.
Most of a cleanroom HVAC contract is paid before the turnover documents exist. Split the retention into two parts, released on acceptance of the turnover package and of the documents, and size the documentation part on what it would cost the owner to finish as-builts, calibrations and retests without the contractor.
After an over-budget cleanroom HVAC tender, three savings recur: an EN 1886 L2 casing instead of L1, fewer room pressure transmitters, and deleted door drop-seals. Each can reappear as an OQ failure in airflow, alarm testing or pressure differentials, so check every line against its URS requirement and test before accepting it.
ISO 14644-1 allows a room to be classified as-built: installation complete, services running, no equipment, materials or people. On a new suite it shows whether the HVAC, filters and envelope can reach the class alone. It says nothing about machines or people, so the GMP classification is still done at rest and in operation.
SEMI F47 expects semiconductor tools to ride through sags to 50% for 200 ms, 70% for 0.5 s and 80% for 1 s. Fan drives, contactors and controls are often bought to no such curve, so the tool stays up while the cleanroom loses airflow. Specify the support systems to the same curve.
In unidirectional airflow at 0.45 m/s, a 28.3 L/min particle counter needs an inlet about 37 mm across to sample isokinetically, and a 100 L/min counter about 69 mm. An undersized inlet draws air in too fast, large particles miss it by inertia, and the ≥5 µm count reads low.
Micro-environments hold ±2% RH at the tool and let the room run a looser band. A 2 m³ enclosure at 0.4 m/s over 1 m² replaces its air every 5 seconds. Steady leakage is the harder case: with the room at 45% ± 5% RH, size the enclosure for 50% RH at the highest room temperature.