Field note · Liquid cooling · September 29, 2026
Separate the loop before the alarm.
A water quality alert has little value until the team knows which cooling circuit it describes, who owns that circuit, and what action the evidence supports.
This is an operating framework based on published guidance, not a report of AquaMesh performance at a data center.
First boundary
Facility water and server coolant are different systems.
An ASHRAE TC 9.9 white paper distinguishes facility water systems (FWS) from technology cooling systems (TCS), which a coolant distribution unit commonly separates. The paper warns that applying the quality specification for one loop to the other can either impose needless cost on the facility side or leave the IT side at risk. It also notes that ownership may differ between facilities staff, the IT owner, equipment provider and service provider.
Before setting a threshold, draw that boundary on the plant diagram. Identify the sample point, fluid, treatment program, materials, control owner and equipment manufacturer's requirements. A signal on a cooling tower does not establish the condition inside a separated server loop. Nor does a server-side result alone identify a tower fault.
Existing coverage
Use each signal for the question it answers.
- 1
Thermal and hydraulic tags
Trend supply and return temperature, flow, pressure differential, pump and valve state, filter differential pressure, and load by operating mode. These records can localize a heat transfer or flow problem. They do not, by themselves, identify its chemical cause. The Lawrence Berkeley National Laboratory liquid cooling controls guide describes the data inputs and ownership choices involved.
- 2
Conductivity, pH and turbidity
ASHRAE describes these as an intermediate online monitoring option, with cost and risk determining whether it is warranted. Conductivity responds to dissolved ionic content; it is not a specific contaminant assay. The USGS conductance reference explains that the reading depends on the types and amounts of dissolved substances and temperature. Interpret a change alongside makeup water, dosing, temperature compensation and the correct loop's specification.
- 3
Samples and equipment guidance
Maintain the sampling and analysis required by the fluid and equipment providers. ASHRAE's paper calls for a monitoring frequency and an agreed response to out-of-range values; it explicitly retains periodic bacteria samples. A conductivity, pH or turbidity trend does not replace those tests or prove biological safety.
- 4
A defined missing observation
If existing tags flag a change but cannot distinguish the physical state that would change an operator decision, write down that missing variable and the point of measurement. Targeted spectral sensing could be evaluated for a suitable signature with paired reference samples, interference and fouling checks, and false-alarm review. No spectral result should be treated as a validated concentration, bacteria test or corrosion-rate measurement without application-specific validation.
Operating test
Reconstruct the decision at the time it mattered.
Choose one historical water quality or cooling event. Align the existing BMS or historian trends with lab sample collection time, dosing and maintenance records, operating mode, and CDU status. Ask which loop changed first, whether the evidence identified a likely cause, and who had authority to isolate, sample, inspect or hold equipment under the site's procedures. Compare with a normal run before declaring an anomaly.
Then define a prospective test: a condition that triggers an operator-reviewed action, an acceptable false-alarm rate, and the downstream signal or sample that would verify the response. ASHRAE emphasizes an agreed plan for out-of-range results; this is where monitoring becomes a decision rather than another chart.
Start with the loop, the owner and the response window. Add an instrument only when the evidence required for that decision is missing from the installed controls and sampling program.
Sources and scope
Primary references.
- ASHRAE TC 9.9, Water-Cooled Servers: Common Designs, Components, and Processes, especially water quality and monitoring sections. Published guidance, not an AquaMesh validation.
- U.S. Geological Survey, Specific Electrical Conductance, measurement basis for conductivity.
- Lawrence Berkeley National Laboratory, Liquid Cooling Controls Team white paper, control system data inputs.
The historical review and prospective test are AquaMesh's proposed method. They are not a claim of a particular site's results or a tested AquaSpectra performance level.