An inline conductivity sensor on a direct liquid cooling loop reads 15 µS/cm. In a deionised water loop, that number is a warning: the water is picking up ions faster than the polisher is removing them, and copper surfaces in the cold plates are being exposed to an increasingly aggressive fluid. In a PG25 loop filled with a standard inhibited glycol, the same 15 µS/cm would mean something far worse: the coolant has effectively been replaced by water, because the fluid as delivered should read in the thousands.
Conductivity is one of the most useful measurements on a DLC loop, and one of the most frequently misread. The sensor works the same way in both fluids, but the numbers, the direction that signals trouble, and the sensor needed to measure them are completely different.
The short answer is this. In a DI water loop, conductivity is a purity measurement: the target is a very low value, and any rise means contamination or corrosion. In a PG25 loop, conductivity is a stability measurement: the target is the baseline of the specific fluid installed, and a change in either direction matters. A rise points to contamination or glycol degradation, while a fall points to dilution or inhibitor depletion. Monitoring both fluids well depends on knowing the right baseline, choosing a sensor ranged for it, and alarming on change rather than on a generic limit.
What conductivity tells you in a cooling loop
Conductivity measures how easily a fluid carries electrical current, which depends on the concentration and type of dissolved ions. In a closed cooling loop, those ions come from a small number of sources: the fluid’s own inhibitor package, metals dissolving from wetted surfaces, contaminants introduced during fill or maintenance, carbon dioxide absorbed from air, and anything that leaks in from the facility water side of the CDU.
Industry guidance recognises that conductivity in the technology cooling loop can rise over time through ionic contamination, CO2 permeation, materials leaching, or the addition of inhibitors, biocides or buffers. Because each of these changes conductivity, the reading on its own does not identify the cause. It tells you that the chemistry has moved, and how fast. Diagnosis comes from interpreting that movement against the right baseline and cross-checking it with other parameters.
DI water loops: conductivity as a purity measurement
In a DI water loop, the goal is to keep dissolved ions to an absolute minimum. ASHRAE’s water-cooled server guidance lists a technology cooling system conductivity range of 0.2 to 20 µS/cm, and some OEMs set tighter limits; IBM, for example, specifies conductivity at or below 10 µS/cm for the system-side loop and recommends a deionising step to remove ions that leach from loop surfaces.
The reason for keeping conductivity this low is that pure water is chemically hungry. It readily dissolves ions from the metals it touches, and once ions are present, the water becomes more effective at carrying corrosion currents between dissimilar metals such as copper cold plates and stainless manifolds. Rising conductivity is therefore both a symptom of corrosion and an accelerator of it.
What makes conductivity rise in a DI loop. The common causes are carbon dioxide absorbed during fill or through permeable hoses, which forms carbonic acid and raises conductivity even with no contamination present; ions leaching from new pipework, solder flux and fittings in the first weeks after commissioning; an exhausted mixed-bed polishing cartridge that is no longer removing ions; top-up water that was not deionised; and a leak through the CDU plate heat exchanger from the facility side.
How to read the trend. A slow, steady rise usually points to leaching or polisher exhaustion and calls for a check of the polishing resin. A rise that plateaus at around 1 µS/cm after a fill often reflects CO2 absorption rather than harmful contamination. A sudden step change is the signature of a heat exchanger leak or a contaminated top-up, and it calls for immediate investigation, because facility water can bring chlorides, hardness and bacteria with it.
Conductivity as a control signal. In a well-designed DI loop, the conductivity reading also tells you when to act. It can trigger an alarm to replace the polishing resin, or open a side-stream through the polisher when conductivity crosses a set point and close it once purity is restored.
PG25 loops: conductivity as a stability measurement
PG25 behaves very differently. The corrosion inhibitor package is made of dissolved, often ionic, compounds, so an inhibited glycol has a much higher conductivity than DI water by design. The key point is that there is no single correct value, because the number depends entirely on the formulation.
The spread between products is wide. Dow lists its DOWFROST LC 25 data centre fluid at an electrical conductivity above 2,000 µS/cm. Crystal Clean’s PGDC, a 25% propylene glycol direct-to-chip coolant with an organic acid inhibitor package, lists a typical conductivity of 3,260 µS/cm at 20°C. At the other end, Clariant markets Antifrogen LC25 PG as a low-conductivity fluid, with electrical conductivity below 40 µS/cm. All three are PG25 coolants, and their correct readings differ by nearly two orders of magnitude.
This has two practical consequences. First, any alarm limit must come from the fluid supplier’s datasheet and the commissioning baseline, never from a generic figure. Second, a conductivity sensor ranged for one PG25 product may be completely unsuitable for another.
What a rise means in a PG25 loop. Conductivity climbing above baseline usually indicates contamination, either from facility water entering through a heat exchanger leak or from top-up with untreated water containing dissolved minerals. It can also reflect glycol degradation: when propylene glycol oxidises under heat and dissolved oxygen, it forms organic acids, which ionise and push conductivity up while pulling pH down.
What a fall means in a PG25 loop. A falling reading is just as important and is more often ignored. It typically means dilution, where water top-ups have lowered the concentration of both glycol and inhibitor, or inhibitor depletion, where the protective compounds are being consumed. Either way, the loop is losing corrosion protection, and in the case of dilution it may also be falling below the 25% glycol concentration at which the fluid resists microbial growth on its own.
Separating the causes: conductivity is strongest when paired
Because several causes move conductivity in the same direction, the reading becomes far more useful when it is read alongside glycol concentration and pH. The table below shows how the combination narrows down the cause.
| Fluid | Conductivity | Paired reading | Most likely cause |
|---|---|---|---|
| DI water | Slow rise | pH stable | Material leaching or polisher resin exhaustion |
| DI water | Rise to around 1 µS/cm after fill | pH falls slightly | CO2 absorption from air |
| DI water | Sudden step up | Turbidity may rise | Heat exchanger leak or contaminated top-up |
| PG25 | Falls | Glycol concentration falls | Dilution from water top-ups |
| PG25 | Falls | Glycol concentration stable | Inhibitor depletion |
| PG25 | Rises gradually | pH falls | Glycol oxidation forming organic acids |
| PG25 | Sudden step up | Glycol concentration falls slightly | Facility water entering through heat exchanger leak |
A conductivity sensor alone would flag every row of this table as “something changed”. With glycol concentration and pH alongside it, most of these events can be identified before anyone takes a sample.
Choosing and installing the sensor
Match the sensor range to the fluid. DI water needs a contacting sensor with a low cell constant designed for ultra-pure measurement, because a general-purpose probe cannot resolve the difference between 1 and 5 µS/cm with any confidence. An inhibited PG25 at several thousand µS/cm needs a sensor ranged for that level. Pyxis covers both ends of this range. Its ST-725 ultra-low conductivity sensor is built for DI water and low-conductivity glycols such as the Clariant product, while its standard-range conductivity sensors measure up to 300 mS/cm, comfortably covering high-conductivity PG25 formulations in the 2,000 to 3,500 µS/cm region. The point is to specify the sensor for the fluid actually in the loop, and to confirm the range again if the coolant is ever changed to a different product.
Get the temperature compensation right. Conductivity changes strongly with temperature, and DLC loops run warm. Pure water has a non-linear temperature response that needs a dedicated compensation curve, while glycol mixtures respond differently again because viscosity, and with it ion mobility, changes sharply with temperature. Using a default water compensation coefficient on a glycol loop produces readings that move with return temperature rather than chemistry. Set the compensation for the actual fluid, and record the temperature alongside every conductivity reading.
Install for a representative, bubble-free sample. A side-stream flow assembly on the CDU return, with isolation valves, gives a sample that reflects what the IT equipment has just seen and allows the sensor to be removed for verification without draining the loop. Air bubbles on the electrodes cause erratic low readings, so avoid high points where air collects.
Measure the makeup water too. Many contamination events start with the water used for top-up. A conductivity check on the fill and top-up supply catches untreated water before it enters the loop, which is far easier than diagnosing it afterwards.
Setting alarms that mean something
The most effective conductivity alarm scheme uses three reference points: the conductivity of the fill water, the baseline recorded after commissioning once the loop has stabilised, and an allowable band around that baseline.
For DI water, the main alarm is an upper limit set from the OEM specification, supported by a rate-of-change alarm to catch step events early. For PG25, alarms are needed in both directions around the commissioning baseline, again supported by a rate-of-change alarm. A heat exchanger leak may produce a step that stays inside a wide absolute band, and only a rate-of-change alarm will catch it on the day it happens.
Why this matters for Malaysian facilities
Direct-to-chip loops are being commissioned across Malaysia now, many of them in Johor and many on facility water that increasingly comes from recycled or reclaimed sources. Reclaimed facility water typically carries far higher conductivity than any technology loop, which makes a heat exchanger leak easier to detect if the monitoring is in place, and more damaging if it is not. Getting conductivity monitoring right at commissioning, with the correct baseline, range and alarm logic, costs little compared with discovering a leak through corrosion or fouling at the cold plates.
For the wider picture on coolant selection and the other parameters in a monitoring package, see our articles on choosing between PG25, DI water and immersion and what each coolant monitoring approach catches.
Autoflo Technology is the Malaysian distributor for Pyxis Lab inline sensors and coolant monitoring panels. For help selecting a conductivity sensor and range for a specific DI or PG25 loop, contact us at info@autoflotechnology.com.