EMS-Trade
Cooling
Loop consumables

Document ECT-01 · REV C · ISSUED 2026-09 Six lines · 48 sheets

Sheet 40 · HPC & research · the long-lived loop

The loop outlives the equipment it was built for

One loop carries equipment from two generations for 5–10 years, and the sample that would define its normal state was never kept.

Request the specification file One business day, with the file attached

Life band 5–10 yearsFiltration 1 µm absoluteAt charge ISO 4406 17/15/12

A research machine room: an older row of cabinets with dense cabling on the left and a current row on the right, both fed from one coolant pipe run clamped under the tray above them
FIG 1 ·

One manifold, two generations, one fluid. The loop is older than every machine on it.

5–10 yearsService band of one loopLonger than the equipment it was sized for
1 µmAbsolute rating in circuitGrades from 0.05 to 100 µm on one housing
0.1 mLeak location resolutionOn 500 m of rope per channel, alarm inside 30 s
17/15/12ISO 4406 at charge and at top-upSampled at the head of the run, not at the tank
01

Four facts that decide how this loop is specified

Pain statement · the conditions this industry actually runs under

Service band
One loop is specified once and then runs for 5–10 years. Equipment is replaced inside it several times, and the fluid and the filtration stay as they were specified on day one.
Mixed generations
A legacy row and a current row are fed from the same manifold. The joints either side of the tee were made ten years apart and are wearing on different clocks.
The third year
Corrosion inhibitor is consumed over that life, and what is left of it is not visible from the aisle. Depletion is reported from the third year onward, usually as a conductivity reading nobody expected.
No baseline
The reading that would settle it is the one taken on the day the loop was filled. It is usually not held, so the first sample cannot be called normal or abnormal.

A baseline is not a document the supplier keeps for you. It is a sample drawn on the day the loop is filled, and it belongs in the same file as the loop drawing.

02

What a ten-year loop looks like from the aisle

Pain plates · mixed generations, the element at the end of its life, and the probe nobody has a number for

A row manifold with an older cabinet branch on one side and a current rack branch on the other, hoses of different lengths and two different clamp styles on the same pipe
FIG 2 ·

The tee between the two generations is a joint, and it was made ten years after the first one.

A pleated filter cartridge cut open on a bench with dark deposits packed into the pleats, an unused cartridge beside it and a drip cloth under both
FIG 3 ·

What the element holds is what has been circulating: the filter is the loop's only account of it.

The probe is where the baseline argument lands. It is pulled, wiped and put back at every service, and its reading stays an opinion until there is a first sample to hold it against. Holding that sample costs one bottle and one line in the file.

A conductivity probe drawn half way out of a pipe section on a bench, with a thin film of deposit on the electrode and its terminal block beside it
FIG 4 ·

Conductivity is the reading that reports inhibitor depletion, and it is only meaningful against a baseline.

02

What we will put in writing for a ten-year loop

Schedule R · acceptance conditions for the loop, not for the CDU

Schedule R · the conditions this industry is quoted on
No Condition Value Unit Basis
01 Service band of one loopHose, seals and filtration 5–10 years Specified once, then carried through several equipment generations
02 Inhibitor service lifeIn reported loops Third year Depletion is reported from the third year onward, as a conductivity shift
03 Baseline sampleOn the day the loop is filled 1 sample Conductivity, pH and cleanliness, kept with the loop drawing
04 Cleanliness at charge and top-upLoop fluid 17/15/12 ISO 4406 Top-up fluid is gated to the same grade as the first fill
05 Cartridge absolute ratingFilter grade in circuit 1 µm Grades from 0.05 to 100 µm on one housing
06 Filter change differentialAcross the housing 0.5–0.7 bar Read continuously, not at the annual service
07 Leak location resolutionLocating sensing rope 0.1 m On 500 m of rope per channel, alarm inside 30 s
08 Shell design pressure304 and 316L 1.0 MPa Hydrostatic proof per batch
09 Hose working pressureDASH −4 to −10 1.38 MPa About 200 psi, burst above 5.5 MPa
10 Hose temperature bandWater-based coolant −40 to +99 °C The band claimed for this application

The value column is the figure we will be held to, and every figure in it is one this loop is bought against. Where a tolerance is not published, the cell stays empty rather than carrying an estimate.

03

Where the two readings on a ten-year loop come from

Evidence plates · the rope that runs with the pipe, and the differential across the housing

A braided sensing cable run along a coolant pipe under a tray, tied to the pipe at intervals and turning a gentle bend past a support bracket
FIG 5 ·

The rope is laid at the pipe, not at the room: it moves when the loop moves.

A differential pressure transmitter clamped on two impulse tubes taken across a filter housing in a pipe run
FIG 6 ·

Filter loading is read continuously, because an annual reading cannot show when it moved.

05

The loop as a mixed-generation row is actually built

Section through the loop · consumables called out where they sit

1 2 3 4 5 6 legacy row on the left, current row on the right, one manifold

Consumables on this section

  1. 1 ·Row manifolds, one per generation, tied into one loop
  2. 2 ·Legacy branch, made ten years before the tee beside it
  3. 3 ·Current row branch, short and direct off the same manifold
  4. 4 ·Sampling valve at the head of the run, not at the tank
  5. 5 ·Filter with a differential pressure transmitter on the return
  6. 6 ·Locating sensing rope along the pipe, 0.1 m resolution
06

The three files this industry is held as

Archive form · no client name and no project name is held on any of them

Three configurations recur where a loop outlives its equipment. Each is held as a file: the loop drawing, the consumable list, the baseline sample and every sample after it.

One manifold with two generations of branch on it: a long routed hose on the legacy side and a short direct hose on the current side, each with its own isolation valve
FIG 7 ·

File A. Two generations, one manifold, isolation per branch.

A ring main on the rear of a cabinet row with supply and return running side by side, a sampling valve and a filter at the head of the run
FIG 8 ·

File B. A ring per row, and the sample drawn at the head of it.

A bench loop in a laboratory: a short run taken off a wall main to a bench instrument, with a quick disconnect and a sight glass in the run
FIG 9 ·

File C. The bench loop: the smallest file in the set, and the same schedule.

07

The standards a research loop is written against

Conformity · the document set held for this sheet

OCP
Cold Plate Coolant Base Specification v1.3, with the Reservoir and Pumping Unit Spec v1.0 for the facility side
ASHRAE
Water Cooled Servers, supply-water classes W17–W45.
T/CAEE 085—2026
EPDM hose and hose assemblies for cold-plate liquid cooling, water-based coolant
ISO 4406
17/15/12 at charge and at every top-up, verified on a sample
ISO 9001
Held by the partner carrying the work, not claimed here as ours
EN 10204 3.1
Material certificates for every metal part in the loop, kept with the baseline
UL 94 V-0
Flame rating on the hose cover compound
RoHS · REACH
Elastomer and seal material compliance
CE · UKCA
CE marking and UKCA, for European delivery
04

How those standards are demonstrated

Evidence plates · the batch bench, the replaceable interface, and how a run ships

A hydrostatic and burst test bench with a coiled braided hose under test behind a steel guard, a pressure gauge and a control cabinet at the side
FIG 10 ·

1.38 MPa working and burst above 5.5 MPa, proven per batch rather than per drawing.

Four light grey thermal pads fanned out on a neutral ground, each square cut with a matte surface
FIG 11 ·

The second thing that gets replaced inside a long-lived loop: the interface, at 1.2–7.5 W/m·K.

A run for a loop that will be topped up and re-branched comes terminated: ends fitted and proof tested before dispatch, so a twelve-year-old line is never cut into on a live row.

A terminated hose length on a bench, coiled and held with wrapping tape, with a bag and a folded drawing beside it
FIG 12 ·

Terminated and tested before dispatch, so the row is never the place the length is decided.

05

What every one of those files consumes

Specimen strip · real parts, one cell each · contain fit, the part is never cropped

Four light grey thermal pads fanned out on a white ground, each pad square cut with a matte surface
SP 1 · Thermal pad 1.2–7.5 W/m·K through-plane
A cream braided sensing cable coiled on a white ground with waterproof connectors at both ends
SP 2 · Sensing cable 500 m per channel, 0.1 m resolution
Section drawing of a coupling with a seal groove and hatched body, dimension letters only, no brand and no model number
SP 3 · Coupling section Seal groove and mating face called out
A grey thermal pad sheet on a white ground with two red pads laid across one corner
SP 4 · Pad sheet Cut to the die footprint on site

SP 1–4 ·Supply-side samples: a thermal pad at 1.2–7.5 W/m·K, a sensing cable rated 500 m per channel at 0.1 m resolution, a coupling section with its seal groove called out, and the pad sheet a bench loop is cut from.

00

What stops at the boundary of this sheet

Scope · the part of the loop that is not supplied here

We supply the loop: hose, couplings, seals, filtration, sensing rope and the instruments that read them. The cooling distribution unit and the cold plate are not ours to quote, and neither is the installation, the design or the commissioning of either.

On a long-lived loop that matters for a second reason: the unit sized for the first generation is usually still in place for the third, and it is not what wears out.

Supplied the loop itselfNot supplied CDU · cold plate · installation
A facility side pumping unit in a plant room: a pump, a tank and a filter housing on one frame with pipework leaving the top, a service aisle in front of it
FIG 13 ·

The pumping unit and the tool side stay outside this sheet. What we quote starts at the take-off.

00

The other sheets in this set

The same schedule, five different industries · one row per sheet

01 Colocation A hall that cannot be shut down, on a return shared with other tenants. SLA 99.99% · 52.6 min
02 Hyperscale & AI training Rack power from 10–15 kW to 100 kW+, and a coupling pair added with every rack. 10–15 → 100 kW+
03 Edge & telecom No round, no one on call, and a leak that can run for weeks. −20 to +55 °C
04 Semiconductor Particles read against yield, on wetted parts declared as 316L and PFA. 0.05–100 µm
05 EV & power electronics Bench loops broken and remade weekly, and glycol errors that surface weeks later. 1.38 MPa · −40 to +99 °C

Every sheet in this set is written in the same order: the pain, the schedule it is quoted on, the loop as it is built, the standards, and the configuration files held for it.

DOCUMENT REQUEST

Four documents are held for this sheet. Tick the ones you need.

Tick a document, name the row or the bench loop you are planning against, and the file comes back attached.

The only promise on this page

One business day, with the file attached.

Nothing else on this sheet is a commitment. Price, stock and lead time are not promised here.

The four documents held for this sheet laid out on a bench: a folded specification sheet, a certificate sheet, a test report and a rolled outline drawing
FIG 14 · Specification sheet · EN 10204 3.1 certificates · burst and hydrostatic report · outline drawings
If the form cannot send, sales@ems-cool.com takes the request instead. A request is never accepted and then dropped.