Introduction: A closed CDU loop holds a fixed water volume, so every temperature swing turns into a pressure swing that an expansion tank has to absorb and smooth out.
Walk past a liquid-cooled row in the morning and again in the afternoon and the pressure gauge often sits on a different number, even though no valve was touched and no pipe changed shape. The piping is still sealed; the water inside it simply warmed up a little. In an open system that small volume change would raise a tank level and disappear. In a closed coolant distribution unit (CDU) loop it has nowhere to go, so it becomes pressure. Understanding that chain — heat, expansion, compression, pressure — makes the expansion tank look less like an accessory bolted into a cabinet and more like the part of the loop that keeps everything else predictable.
Why Temperature Change Creates a Pressure Problem in a Closed CDU Loop
A sealed loop has no free surface, and that single fact explains most of the pressure behaviour inside a CDU. In an open tank system, warmed water simply raises the level and surplus air leaves through a vent. In a closed loop the same mass of deionized water is trapped between the pump, the brazed plate exchanger, the manifolds and the server cold plates, and the steel around it is stiff. Water itself behaves like a very poor spring: it barely compresses under pressure. So when heat arrives from the servers, the small increase in volume needs somewhere to go. It can flex the pipework, lift a relief valve, or compress something genuinely compressible — a trapped gas cushion. The scale of that volume change is easy to underestimate. Water in the 10-45 °C band that packaged CDUs typically work in expands by roughly 0.2 to 0.3 percent for every 10 °C of warming, and that relationship is well documented for volumetric thermal expansion. Put the numbers into a 1,000-litre secondary loop and a 20 °C swing adds something close to five litres of water that did not exist a few hours earlier. Five litres is a lot of displacement in a rigid, fully flooded network, because the only remaining destinations are compression of the fluid itself or elastic strain in the piping and fittings. Facility engineers see the consequence long before they calculate it: the gauge on a closed loop drifts upward through the working day as supply temperature rises, then settles back overnight as load drops. That drift is why pressure margins in closed piping systems are treated as a design subject rather than a nuisance. Pressure piping standards such as ASME B31.1 exist because closed systems are engineered around expected operating pressure plus a margin, and thermal expansion is one of the inputs that margin has to cover. A loop designed with no room for expansion does not fail politely; it waits for the warmest day, the heaviest load, or the moment a valve isolates a section.
What Happens Inside a Closed Loop as Water Temperature Changes
The sequence is worth walking through in order, because each step explains why the gauge moves at all and why the fix is a buffer rather than a stronger pump.
- Heat enters the water at the cold plates. Server load rises, the return line warms first, and the mixed temperature across the loop climbs before the supply sensor settles at a new steady value.
- The water expands by a small fraction of its volume. Density falls, so the same mass occupies slightly more space, and a few tenths of a percent is normal for the temperature window a data hall actually uses.
- Because nothing can leave the loop, the extra volume compresses whatever is compressible. With a gas cushion present, the cushion absorbs it; without one, the pipe walls, gaskets and fittings take the strain instead.
- Pressure settles at a new balance point. The loop holds that higher value while the water stays warm, then contraction pulls pressure back down as the load falls and the loop cools again.
Notice that the circulating pump is not the source of this behaviour. A pump adds head and moves the same water around; it does not create net volume. The pressure movement comes from volume change combined with the stiffness of the containment, which is why two loops of identical length can behave very differently depending on how much compressible volume each one carries. A loop with a generous gas cushion turns a five-litre expansion into a modest pressure rise. The same five litres in a fully flooded, cushion-free loop has to be absorbed by metal. This is also why supply temperature stability and pressure stability are linked but not identical. A CDU holding a steady 5 °C approach temperature still lets loop pressure rise as the secondary water warms, because the water itself is doing the expanding. Operators who watch only the temperature readout can be surprised by a gauge that climbs on a warm afternoon while every temperature value looks normal.
How the Expansion Tank Supports Stable Pump Inlet Conditions
An expansion tank is a small vessel with a trapped gas volume, connected to the loop so that water can move in and out of it. When the loop warms, water pushes into the tank and compresses the gas; when the loop cools, the gas expands and pushes water back into the network. The gas is the spring the loop otherwise lacks. Because gas compresses easily, several litres of extra water produce a gentle pressure rise rather than a sharp one, and loop pressure tracks a narrow band across the day instead of a sawtooth. That band matters most at the pump. A circulating pump wants its suction pressure to stay comfortably above the point where the water begins to boil locally at the inlet, and cavitation damage starts well before anyone hears anything dramatic. A loop whose pressure swings widely with temperature forces the pump to operate across a moving inlet condition, which shows up as unstable flow and extra noise at part load. The buffer keeps the inlet condition close to what the pump curve assumed. In hydronic practice the tank connection is usually taken on the suction side for exactly that reason, so the pump feels the cushion rather than the raw expansion of the whole network. Zejiu builds an expansion tank into the GLU series cabinet alongside the pumps, brazed plate exchanger, isolation and regulating valves, and internal metal piping, so the pressure buffer ships as part of one assembly rather than a field-installed accessory. The standard secondary fluid in that loop is deionized water. Tank volume is not a single universal figure: loop water volume, expected temperature swing and static height all decide how much gas cushion is needed, and on the GLU series the vessel is matched to the project during configuration rather than fixed as a catalogue constant. A useful way to think about sizing is that the tank only has to cover the difference between cold and warm states, not the whole loop volume. A loop that swings 15 °C needs less cushion than one that swings 35 °C, and a tall building puts more static head on the pump than a single-storey hall. That is why the same packaged CDU arrives with different tank arrangements for different buildings even when the heat load is identical.
Conclusion
The expansion tank is the quiet reason a closed CDU loop stays predictable. Water in a sealed circuit expands and contracts with temperature, and because nothing can leave, that volume change becomes pressure unless a compressible gas cushion is waiting to receive it. The tank absorbs the expansion, returns the volume on cooldown, and keeps pump inlet conditions inside a steady band across the working day. Anyone reading a specification can treat the internal tank as evidence that the package was designed around thermal volume change rather than around a perfectly rigid loop. The GLU series product information shows how that vessel sits among the other integrated components for readers who want to see the arrangement.
FAQ
Q: Why does a closed CDU loop need an expansion tank if the piping is already sealed?
A: Sealing is precisely the reason it needs one. In a sealed loop the water cannot leave or vent, so when it warms and gains volume, something inside the loop has to give. Water barely compresses, which leaves the pipe walls, gaskets or relief path to absorb the displacement. A tank provides a compressible gas cushion so a few litres of expansion produce a modest pressure rise instead of stressing the whole network.
Q: How does water temperature change affect pressure in a closed cooling loop?
A: Water expands by roughly 0.2 to 0.3 percent per 10 °C in the range these loops normally run, so a 1,000-litre circuit gains several litres across a 20 °C swing. In a fixed volume that surplus has to compress the gas cushion or strain the pipework, and the resulting pressure rises with temperature and falls again on cooldown. That is why gauges drift through a working day without any valve being touched.
Q: Does an expansion tank replace the need for pressure relief or air removal?
A: No, those are three separate jobs. The tank buffers normal thermal volume change and keeps daily pressure swings gentle, a relief device protects the loop against abnormal overpressure, and air removal handles entrained gas that would otherwise cause noise, poor heat transfer or false pressure readings. A packaged CDU normally includes the buffer and valve hardware together, while air removal stays a function of vents and separators.
Sources / References
Volumetric (Cubic) Thermal Expansion
No comments:
Post a Comment