Building and Commissioning a Cleanroom

Class Is a Concentration, Air Change Rate Is a Judgement

The commonest mistake on a first cleanroom is to search for the ISO 7 air change rate and treat whatever comes back as a requirement. It is not one, because no ISO document sets it. The figures in circulation come from design guidance and accumulated practice, IEST-RP-CC012.3, Considerations in Cleanroom Design, being the usual origin, and they are the opening bid in a calculation whose real inputs are how much particulate the process throws off, how many people work in the space, how much heat the equipment rejects, and how the airflow is arranged. Two ISO 7 rooms serving different processes can legitimately be designed thirty air changes apart.

The flow arrangement matters more than the number. A mixed-flow room dilutes: filtered air enters through terminal modules scattered across the ceiling, mixes with what is already in the room, and leaves through low-level returns, so the classification is a dilution ratio and air changes per hour is a sensible way to describe it. A unidirectional room sweeps: the whole ceiling, or the whole ceiling over the critical zone, is filter face, and the air travels down across the work in a single direction and out. Describing a unidirectional zone in air changes per hour is arithmetically possible and practically useless, which is why those rooms are specified by velocity at the filter face instead.

General cleanroom practice for unidirectional flow sits in the region of 0.2 to 0.45 m/s, roughly 40 to 90 ft/min, and sterile pharmaceutical work is held to a tighter and higher band: EU GMP Annex 1, Manufacture of Sterile Medicinal Products, carries a guidance value near 0.36 to 0.54 m/s at the working position for grade A. Confirm the figure against the edition the project is actually being assessed under rather than carrying one across from the last job, because these values have moved between revisions and the tolerance around them has moved too.

The arithmetic for a mixed-flow room is short enough to do standing on the deck. Clear volume under the grid, multiplied by the target air changes, divided by sixty, gives the supply volume. Take a room 8.0 m by 5.0 m with 2.7 m clear beneath the ceiling: 108 m3, and at 60 air changes that is 6,480 m3/h, or about 3,810 CFM. Divide that by the total open face of the filter bank and you get face velocity, which in a mixed-flow room is not a cleanliness parameter at all. It is a check on whether the bank is large enough that the filters will not be screaming, will not be loading in six months, and will not be handing the fan a resistance it cannot carry. Eight modules at 1200 by 600 give 5.76 m2 of face, and the same 3,810 CFM across that comes out near 0.31 m/s, about 62 ft/min. Comfortable. Cut the bank to five modules to save money and the same air is doing 0.5 m/s through the media, which is a different filter, a different fan and a different noise complaint.

Setting Out a Ceiling That Is Mostly Filter

The grid is the drawing everything else negotiates with. Filter modules, luminaires, sprinkler heads, high-level returns, gas drops and the blank panels filling whatever is left all land on the same module, and once the grid is hung that module is fixed. Set it out from a datum you trust, which inside an existing shell is almost never a wall: old walls are out of square by more than a border panel can absorb, and a grid squared to one of them puts a taper across the room and leaves a filter module half in a border cell.

Filter coverage, not filter count, is what the class buys. An ISO 8 room may be satisfied with a handful of modules in a large ceiling; an ISO 6 room can run to a third or more of the ceiling as filter face; a unidirectional ISO 5 zone is effectively all of it. Distribute toward the process rather than evenly across the plan. A filter directly over the critical operation and a return low on the wall beyond it produces a sweep past the work; four filters spaced prettily on a symmetrical grid produce a nicely photographed ceiling and a recirculating eddy over the bench.

Two grid decisions are worth arguing about before the order goes in. The first is whether it is walkable, because service access from above is a different maintenance regime from changing filters on a scaffold inside a classified room, and a grid that is not rated for it will be walked on anyway by somebody at handover. The second is the seal at the filter seat: a compressible gasket is simple and serviceable, a gel seal costs more and holds a better scan result for longer, and swapping between them after the grid arrives is not a change you can make with a knife.

1. Establish the grid origin off a surveyed datum and a proven square, not off the nearest wall.
2. Run the module both ways and record the border width left at all four walls before anything is ordered.
3. Reserve cells for luminaires, sprinkler heads, detection and any high-level return, marking them on the same layout.
4. Place filter modules against the process, weighting coverage over the critical operation rather than spreading it evenly.
5. Check that every filter cell can be reached for scanning, and that the scan technician can stand where they need to.
6. Confirm in writing whether the grid is rated walkable, and if it is not, plan the access that replaces walking on it.

The Fan Has to Live With a Filter That Doubles

A HEPA filter is not a fixed resistance. It leaves the factory with a rated clean pressure drop and it is replaced at a rated final pressure drop, and on a typical terminal filter the final figure is roughly double the clean one. That is the range the air handling plant has to work across for the life of the filter, and a fan selected against the clean figure delivers design airflow on the day of commissioning and progressively less of it every month afterwards. The room does not announce this. It simply drifts out of classification while everybody assumes the filters are fine because nobody has changed them.

So the selection is made at the far end of the range and the plant is given the means to hold flow across it. Variable speed control, whether an EC motor or a drive, is the practical answer, ideally referencing the measured differential across the bank rather than a fixed speed somebody set once. Where the design uses fan filter units in the grid instead of a central plant, the same point applies unit by unit, and there are then a lot of small fans whose settings need recording somewhere other than the commissioning engineer's memory.

Protect the expensive filter with cheap ones. A terminal HEPA is a poor place to catch construction dust, lint and pollen, and a prefilter train ahead of it is what determines whether the terminals last five years or eighteen months. General ventilation prefilters are rated to ASHRAE Standard 52.2, Method of Testing General Ventilation Air-Cleaning Devices for Removal Efficiency by Particle Size, or to ISO 16890 outside North America, and the terminal filters themselves are classified under ISO 29463 or EN 1822, where an H14 grade is held to at least 99.995 per cent efficiency at the most penetrating particle size. Those two rating systems are not interchangeable and a filter quoted under one of them cannot be compared to a figure quoted under the other by arithmetic.

Keep construction dirt out of the ducts before any of this matters. Terminal filters go in late, after the dirty trades have left and after the system has been run on temporary media to flush the ductwork. Fitting the HEPAs early to protect them from theft, which happens, buys a bank that is half loaded before the room has ever been tested and a first scan result that condemns the filters rather than the programme that ruined them.

The Return Path Is the Half Nobody Draws

Supply gets the attention because it holds the filters. The return decides whether the room actually works. Air introduced at the ceiling and taken back at the ceiling short-circuits across the top of the space and leaves the work at bench height in a stagnant layer; the same air taken back through low-level chases in the wall panels is dragged down past the process, which is the entire point of the arrangement. Low-wall returns cost panel space and coordination with the electrical containment that wants the same cavity, and they are the difference between a room that classifies at-rest and one that classifies operational.

Size the return path for the volume the supply is actually pushing, remembering that a cleanroom deliberately runs supply-heavy: the excess is what leaves through the doors and the leakage and what holds the cascade up. Return chases behind panels, transfer grilles between the classified space and the gowning rooms, and the ducts running back to the air handler are all constrictions, and the velocity in each of them determines both the noise floor in a very quiet room and the pressure the fan has to find on top of the filters. Duct fabricated and sealed to the SMACNA HVAC Duct Construction Standards and leakage tested to the SMACNA HVAC Air Duct Leakage Test Manual keeps the return you drew and the return you get closer together.

Where the return runs through a raised floor or a ceiling plenum rather than in ductwork, that plenum is part of the classified system and inherits every obligation that goes with it: cleanable surfaces, sealed penetrations, no fibrous lining shedding into the airstream, and access for whoever has to clean it. A plenum return is cheaper to build and more expensive to ignore.

The Entry Sequence Is Part of the Room

People are the dominant particle source in almost every cleanroom, and the entry decides how much of that source gets in. It is a cascade of its own: a corridor, a gowning room held between the corridor and the room, sometimes a second gowning stage, and a stepover or bench line that physically separates the dirty side from the clean side so nobody crosses without changing footwear. Garment systems and the gowning sequence are covered in IEST-RP-CC003, Garment System Considerations for Cleanrooms and Other Controlled Environments; the room design accommodating them falls under ISO 14644-4.

Doors have to be interlocked, or at minimum sequenced by procedure, because two doors open at once in an airlock is the cascade defeated for as long as they stay that way. Interlocks bring their own obligations: an override for egress, coordination with the fire alarm strategy so the room cannot trap people, and a control panel somebody has to commission. None of that is HVAC work and all of it lands in the HVAC contractor's programme.

An air shower is a pass-through cabinet at the boundary that blows high-velocity filtered jets over a person before they enter, dislodging particulate from the garment surface and recirculating it back through the unit's own filter. It is common in semiconductor and electronics work and much less common in sterile pharmaceutical suites, where the value of blasting a gown at speed is contested. Where one is specified, treat it as a room in its own right within the cascade: it has a fan, a filter that loads, a noise output that a gowned person stands inside of, and two interlocked doors that must never be open together. Jet velocity, cycle duration and nozzle coverage angle come from the manufacturer's engineering data, because there is no standard that sets them.

The layout arithmetic is the part you can do yourself. Nozzles are arranged in columns across the chamber width and rows up its height, on opposing walls, so that the person is targeted from both sides at every height. A 1.0 m wide chamber with a 2.1 m coverage zone at the manufacturer's 0.3 m horizontal and 0.4 m vertical spacing works out at four columns by six rows, twenty-four nozzles per wall and forty-eight in total. Two things that arithmetic does not tell you: many units add ceiling nozzles or floor-level jets that sit outside the two-wall pattern, and every nozzle is a share of the unit's total airflow, so doubling the count without checking the fan gives a jet velocity that dislodges nothing. Noise is the other omission worth naming, because the exposure is short but genuinely loud, and it falls under OSHA 29 CFR 1910.95, Occupational Noise Exposure, in the United States and the Control of Noise at Work Regulations in the United Kingdom.