Designing a cleanroom inside a data center: ISO 8 assembly areas and high-density test rooms
A worked case study of an ISO 8 assembly cleanroom and a high-density test room inside a data hall: particle control, rack heat, acoustics, humidity, and pressure.
What this means
A cleanroom inside a data center has to satisfy two sets of rules in one space. ASHRAE recommends that data halls themselves be kept clean to ISO 14644-1 Class 8, so an assembly room built there targets ISO 8 or better while server chassis are open. In the worked case, a 2,400 square foot assembly room is governed by particle control at 20 air changes per hour, while a 600 square foot test room holding 180 kW of racks needs about 22,750 cfm for cooling alone, about 190 air changes per hour. That airflow, not particle control, drives the test room's pressure, acoustic, and filtration decisions.
Equipment and model context
- A hypothetical ISO 8 server assembly room and an adjoining high-density test room built inside an operating data hall
- The layout, loads, and selections form an illustrative case study, not a completed project
This is an illustrative case study built from published guidance to show how the requirements interact. It does not describe a real project, and every load and selection is a stated assumption. Fire protection follows NFPA 75 and the local code, and the IT equipment maker's inlet limits apply to the test racks.
What this covers
- How ASHRAE's ISO 8 recommendation for data halls sets the starting point for an assembly cleanroom.
- A worked assembly room governed by air changes and a worked test room governed by rack heat.
- How to arrange pressure between a data hall, an assembly room, and a test room with sharply different airflows.
- Where acoustics, humidity, filtration, and fire protection change the design.
What changes the result
- Rack power and server temperature rise, which set test room airflow.
- People, packaging, and open chassis in the assembly room, which set its particle release.
- Whether cooling airflow stays inside the test room or crosses its boundary.
- Server fan noise at full test load, which sets acoustic treatment and worker exposure.
Why a cleanroom ends up inside a data hall
Servers and accelerators are assembled, integrated, repaired, and burned in close to where they will run. Opening a chassis exposes boards, connectors, and heat sinks to whatever settles on them, and ASHRAE's data center contamination guideline explains why that matters: settled dust with a low deliquescent relative humidity absorbs moisture and can become conductive.
The same guideline recommends that data centers be kept clean to ISO 14644-1 Class 8, and says that level can be reached by continuously filtering room air with MERV 8 filters and filtering air entering the data center with MERV 11 or, preferably, MERV 13 filters. A data hall built that way is already an ISO 8 space at rest. What it lacks for assembly work is control in operation, with people, packaging, and open equipment inside, and that is what the cleanroom adds.
The case study and its basis of design
The assembly room is 40 by 60 feet with a 12 foot ceiling, 28,800 cubic feet, for twelve technicians at benches. Test benches and burn-in carts draw an assumed 25 kW. The room is classified ISO 8 in operation and held at 72°F (22°C), inside the ASHRAE recommended inlet range of 18 to 27°C so equipment running on the benches is within its envelope.
The test room is 20 by 30 feet with the same ceiling, 7,200 cubic feet, holding six racks at 30 kW each, 180 kW in total, with servers assumed to run a 25°F temperature rise. It is also ISO 8, because the servers inside are the product and are opened there between test runs. A makeup air unit supplies dried, filtered outdoor air to both rooms for pressurization and ventilation.
Assembly room: particles and air changes govern
The sensible load is 25 kW of benches, 85,300 BTU per hour; lighting at 0.8 watts per square foot, 6,550; twelve people at 275 BTU per hour sensible, 3,300; and fan heat from 16 fan filter units at an assumed 250 watts each, 13,650. The total is 108,800 BTU per hour. At a 15°F supply difference that load needs 108,800 ÷ (1.08 × 15) = 6,720 cfm, about 14 air changes per hour.
The basis of design sets 20 air changes per hour, 9,600 cfm, for particle control and recovery with people and packaging in the room, so particle control governs and the supply runs 10.5°F below the room. Packaging is a particle source worth removing at the boundary: unpack cardboard in the gowning vestibule or a material airlock, never inside the assembly room. The air changes by ISO class page shows how to check 20 against the room's own particle release, and the choice between fan filter units or a ducted air handler follows the same criteria as any cleanroom.
Test room: rack heat governs
At 180 kW and a 25°F rise, the racks need 3.16 × 180,000 ÷ 25 = 22,750 cfm, about 190 air changes per hour. The data center cooling calculation derives the formula and adds bypass air and cooler fan heat; at an assumed 0.3 watts per cfm the coolers' fans add about 6.8 kW, bringing the load to about 187 kW.
Moving that much air through a room's ceiling and walls would turn every door into a pressure event. The case study keeps the cooling air inside the test room instead: the racks face a cold aisle, their hot aisle is contained, and close-coupled coolers take hot air from the contained aisle, cool it on chilled water, and return it to the cold aisle within the room. Only makeup air and door leakage cross the room's boundary.
With coolers rated at an assumed 40 kW each, N is five units for 187 kW and N+1 is six, so one cooler can fail or be serviced without the racks losing cooling. The redundancy design page explains why a room of servers needs full airflow on the remaining units rather than a reduced-airflow compromise.
Pressure between the data hall, assembly room, and test room
The assembly room holds the open chassis, so it sits highest: 0.03 inches of water column above the data hall. The test room sits 0.01 inches below the assembly room and 0.02 inches above the data hall, so air moves from assembly toward test and from both toward the data hall. Each room's offset is sized from its door and wall leakage at those steps, as the pressure cascade calculation does for a pharmacy suite.
Keeping cooling air inside the test room is what makes that cascade holdable. If the test room were cooled by pulling data hall air through transfer grilles and exhausting it back, 22,750 cfm would cross the boundary through openings sized for that airflow, and those openings would set the room's pressure instead of the offset. A vestibule between the assembly and test rooms gives door use a buffer, so opening the test room door does not connect the two rooms directly.
Filtration that works with server fans
The test room's recirculating air passes through the coolers, not through ceiling filters, so its particle control comes from filters at the coolers and from HEPA or MERV 13 filtration on the makeup air. Every filter added at a cooler adds pressure drop against the cooler's fans and reduces airflow unless the cooler is selected with the filter in place. Ask for cooler capacity and airflow with the chosen filter at its replacement pressure drop.
Servers pull air through themselves with their own fans, and they do not tolerate a starved cold aisle. A filter at a cooler that loads until airflow falls below the racks' demand does not show up as a filter alarm; it shows up as hot air recirculating over the containment to server inlets. Monitor cooler filter pressure drop and cold aisle temperature together.
Acoustic treatment in a clean space
Servers at full test load and close-coupled coolers in one small room produce more noise than an office or an assembly bench area. OSHA's occupational noise rule sets an action level of 85 dBA as an eight-hour time-weighted average, which triggers a hearing conservation program, and a permissible exposure limit of 90 dBA, above which feasible engineering or administrative controls are required. Measure the test room at full load during commissioning and plan exposure controls from that measurement.
Acoustic treatment has to be cleanroom-compatible. Absorbent panels need sealed, non-shedding, cleanable facings, and fibrous absorbers cannot be exposed to the room air. A vestibule and a wall between the test room and the assembly room keep test room noise out of the space where people work all shift. The article on sound power versus sound pressure explains how equipment sound ratings translate into what a person in the room hears.
Humidity, static, and the thermal envelope
Both rooms share one moisture target, set by the stricter of the two uses. ASHRAE's thermal guidelines recommend a server inlet dew point between minus 9 and 15°C, and its contamination guideline keeps relative humidity below 60 percent. The dew point floor addresses electrostatic discharge risk for people handling boards, and the ceiling protects against conductive dust.
Control the makeup air unit to a dew point inside that band, as the cleanroom humidity control guide describes, and keep the assembly room's recirculation coil and the test room's coolers dry. A cooler condensing moisture in a room controlled by its makeup air unit competes with that control and puts water where servers are.
Fire protection and failure response
NFPA 75 sets minimum fire protection for IT equipment areas, covering detection, suppression including gaseous and water mist systems, HVAC, and emergency response. In a room recirculating 22,750 cfm, smoke is diluted and moved quickly, so detection has to be chosen and located for that airflow, and the fire protection engineer and the authority having jurisdiction decide the system.
The failure responses follow the rooms' priorities. Loss of a cooler is covered by N+1. Loss of power stops coolers while servers on uninterruptible power keep producing heat through the generator transfer, so the coolers' fans and controls belong on protected power too. The assembly room follows the positive cleanroom restart order in the failure sequences guide, supply first, and is held for recovery before chassis are opened again.
| Requirement | ISO 8 assembly room | High-density test room |
|---|---|---|
| Cleanliness target | ISO 8 in operation, with HEPA-filtered supply | ISO 8, with recirculating air filtered as ASHRAE recommends for data halls |
| What sets the airflow | 20 air changes per hour for particle control and recovery | Rack heat, about 22,750 cfm at a 25°F server rise |
| Cooling equipment | Ducted air handler or fan filter units with a dry coil | Close-coupled coolers serving a contained hot aisle |
| Pressure | 0.03 in. w.c. above the data hall | 0.01 in. w.c. below the assembly room, still above the data hall |
| Humidity | Dew point inside the ASHRAE recommended range, below 60 percent RH | Dew point inside the ASHRAE recommended range for server inlets |
| Noise | Bench work with people present through the shift | OSHA 85 dBA action level for anyone working in the room |
| Fire protection | NFPA 75 and local code for an IT equipment area | NFPA 75 and local code, with detection chosen for 22,750 cfm of recirculation |
Spaces a person and the air pass through, from the data hall into the assembly room and on to the test room, with the pressure and airflow role of each.
- Data hallKept to ISO 8 by its own filtration, and the pressure reference
- Gowning vestibuleGowning and unpacking boundary, positive to the data hall
- Assembly roomISO 8 at 20 air changes per hour, 0.03 in. w.c. above the data hall
- Test room vestibuleDoor buffer between the assembly room and the noisy test room
- Test roomContained hot aisle with close-coupled coolers recirculating 22,750 cfm internally
- Makeup air unitDried, filtered outdoor air for pressurization of both rooms
Airflow needed to carry rack heat at server temperature rises of 25°F and 35°F, from CFM = 3.16 × watts ÷ ΔT, for total rack loads from 60 to 240 kW in the 7,200 cubic foot test room.
- 25°F server rise
- 35°F server rise
- At 180 kW and a 25°F rise the test room needs about 22,750 cfm, roughly 190 air changes per hour of its 7,200 cubic foot volume.
- ISO 8 particle control at 20 air changes per hour would need only 2,400 cfm, so cooling airflow is nearly ten times the particle airflow and decides the design.
- Servers running a 35°F rise cut that airflow by 29 percent, to about 16,250 cfm, but the rise belongs to the server fans and cannot be chosen by the HVAC design.
Questions people ask about this
What ISO class should a data center be kept at?
ASHRAE's data center contamination guideline recommends ISO 14644-1 Class 8, reached by continuously filtering room air with MERV 8 filters and filtering air entering the data center with MERV 11 or MERV 13 filters. Areas where chassis are opened or assembled can justify tighter control in operation.
Can a data hall's cooling serve a cleanroom built inside it?
It can share chilled water, but the cleanroom needs its own HEPA-filtered supply, its own pressure control, and a makeup air path so its class and pressure do not depend on data hall airflow. Pulling the cleanroom's air straight from the data hall hands its cleanliness and pressure to a space controlled for different goals.
How loud can a high-density test room be for workers?
OSHA sets an action level of 85 dBA as an eight-hour time-weighted average, which triggers a hearing conservation program, and a permissible exposure limit of 90 dBA, above which feasible engineering or administrative controls are required. Measure the room at full test load during commissioning and plan controls from the result.
Does an ISO 8 room need HEPA filters?
ISO 14644-1 states particle limits, not filter types. ASHRAE reports that data halls can reach ISO 8 with MERV 11 or 13 filtration of incoming air, but a room with people handling open equipment benefits from HEPA-filtered supply so it holds the class in operation and not only at rest.
Evidence recordSource verification pending
standards body publication, government guidance · editorial review
This page is awaiting source verification against the documentation in its evidence record: ASHRAE Technical Committee 9.9, ASHRAE, International Organization for Standardization, United States Occupational Safety and Health Administration, via eCFR and National Fire Protection Association, via ANSI technical literature. Its documentation class and intended scope are shown here while that check is pending.
- Documentation class
- standards body publication, government guidance
- Scope of the definition
- Confirm against the exact model manual