How to select OCP liquid-cooling quick disconnects

What must an OCP quick disconnect be qualified to do?

Qualify the coolant and temperature range first, then the pressure states, hydraulic loss, mating geometry and acceptance evidence. Matching dimensions settle only the interface question.

OCP UQD v2.0.0, effective September 22, 2026, covers non-combustible, single-phase water/glycol service and specifies an operating reference envelope of 5°C to 65°C and 100 psig. Use separate component qualification for refrigerants, two-phase fluids and immersion fluids. A higher coupling catalog rating does not increase the rating of its hose, manifold or cold plate.

UQD identifies handmate connectors; UQDB identifies blindmate connectors; a hybrid is a UQD plug entering a UQDB socket. One coupled set, or pair, contains one plug and one socket. A supply-and-return branch normally contains two coupled sets.

Locate the connection within the technology cooling system (TCS) and identify its relationship to the coolant distribution unit (CDU). The direct-to-chip loop architecture guide explains those boundaries. OCP passive LHP guidelines, Rev 1.0 (November 21, 2024) address a separate loop-heat-pipe architecture.

OCP UQD v2.0.0 §13 assigns acceptance of component data for server-level safety compliance to the end user or certifying party. Include the equipment safety assessment and applicable piping-test basis in the submittal alongside UQD conformity evidence.

Write the selection basis before requesting a quotation

State the coolant trade name, concentration and concentration basis, additives, minimum cold-start temperature and maximum return or stagnation temperature. Obtain density and viscosity at both the coldest flowing condition and normal operation, using the formulation-specific approach in OCP PG25 guidelines, Rev 2.0 / v1.0.0.

List maximum static pressure, pump shutoff and transient exposure, permitted service pressure, branch flow, QD loss allowance, available space and end connections. Identify these as project requirements. Request both part numbers and revisions, wetted bill of materials, conformity declaration, property tests and applicable mixed-vendor reports; OCP UQD v2.0.0 §9 requires at least five samples per verification test.

Which UQD parts can mate?

Match configuration, nominal size and version before comparing brands. The matrix applies OCP UQD v2.0.0 §§4, 6, 7 and 10.

OCP UQD v2.0.0 requires suppliers to cross-validate with at least three other suppliers using at least three samples per cross-verification test, but explicitly excludes cross-version combinations from that testing scope. Keep supplier A's plug with B's socket separate from B's plug with A's socket.

Mating matrix and mixed-vendor evidence

OCP UQD v2.0.0 §10 requires cross-verification of fluid loss, connection force, pressure drop and seal washout in both supplier orientations. This subset does not establish mixed-pair burst performance, full endurance or the customer's overdrive-device qualification. Identify additional project tests separately.

Check the required size/version markings on both halves and obtain a conformity declaration naming the specification revision. OCP UQD Rev 1.0 (2020) identifies the legacy interface; v2 intends backward compatibility, while its non-normative appendix identifies tip, chamfer, stroke, envelope and engagement issues for mixed-version deployments.

Anvilfield's spare-control rule is to list both part numbers, revisions, coolant approval and mating-report ID. Review a supplier, seal-compound or interface change before substitution.

Configuration decisions apply OCP UQD v2.0.0, effective September 22, 2026. Release requires evidence for the exact installed combination.

Which pressure rating controls the coupling?

Use the lowest documented allowable pressure among the exposed components in the state being used. Gauge pressure describes containment; differential pressure describes flow loss. A coupling losing 2 psi across its valves can still contain a much higher static pressure, so those two figures answer different selection questions.

Danfoss's September 2026 UQD fact sheet rates UQD08 at 8.9 bar when connected but only 6.9 bar for each disconnected half, so its connected rating cannot be applied to the separated valves. The table compares Danfoss UQD, September 2026, Danfoss UQDB, July 2026 and Stäubli RD7110301B (2024); pressures remain in their published bar units, with Stäubli's PS designation retained.

Read pressure and temperature as an approved combination for the exact seal, state and termination. Stäubli's family brochure does not separately tabulate disconnected-half ratings; obtain those instructions for the selected product.

Keep elastomer properties separate from assembly ratings. Danfoss's September 2026 UQD sheet lists EPDM-P material capability from -40°C to +150°C, while the assembly operating range is 5°C to 65°C. Neither the rubber range nor a minimum burst result authorizes extending the assembled product's service or hydrotest exposure.

Sources: Danfoss UQD AI423953312825en-000106, September 2026; Danfoss UQDB AM480054893377en-000106, July 2026; Stäubli RD7110301B, ©2024. PS and TS mean maximum allowable pressure and allowable temperature. These catalog envelopes are not normalized hydraulic results.

How do you compare QD pressure-drop curves fairly?

Compare pressure-drop curves at identical fluid formulation, concentration, temperature, flow direction and engagement. Match terminations and hose arrangements; identify whether the measurement covers the coupled set or the complete hose assembly.

OCP's UQD v2.0.0 specification, effective September 22, 2026, requires pressure-drop testing with water at 25 ±5°C, at least five samples and at least six flow points spanning 25% to 150% of rated flow, with fixture losses subtracted. Its §9.1.4 requires both directions and reporting the minimum Cv at rated flow. OCP UQD v2.0.0 supplies the method.

The Stäubli RD7110301B (2024) water curve states density 998 kg/m³, viscosity 1.08 cSt and socket-to-plug flow; the Danfoss UQD, September 2026 curve states water at 60°F. Use the same actual fluid conditions when ranking products.

The curve report that makes two quotations comparable

Request raw fixture loss, net coupling loss, temperature, formulation, density, viscosity, direction and sample population. OCP UQD v2.0.0 §9.1.4 asks for a fit consisting of a quadratic flow term plus a linear flow term: ΔP = A Q² + B Q, with flow in L/min and loss in psi. Retain coefficients and losses at rated and 150% rated flow.

For size 04, Anvilfield estimates six test points of 1.61, 3.22, 4.83, 6.44, 8.05 and 9.66 L/min by multiplying OCP's 6.44 L/min reference by 25%, 50%, 75%, 100%, 125% and 150%, respectively. This is one sequence meeting OCP's flow-point rule. For glycol, obtain corrected curves or test the actual formulation and temperature: specific gravity corrects density in the Cv equation, not viscosity or incomplete valve opening.

Measure installed loss and flow at steady fluid temperature, using calibrated instruments and approved taps bracketing the stated assembly. Record accuracy, zero and tap locations. The differential instrument must resolve QD loss and withstand loop static pressure; a high-range system gauge can obscure coupling loss.

Is rated flow the maximum flow?

Rated flow is a characterization point. OCP UQD v2.0.0 §8 directs users to supplier limits and gives a general velocity reference of 5 m/s through the minimum internal area. Danfoss's September 2026 UQD04 sheet distinguishes 6.4 L/min rated flow from its 9.4 L/min reference at 5 m/s.

Select matching geometry, calculate required Cv, check each directional curve against the pressure allowance, then verify the supplier's service limit. Carry the accepted comparison into liquid cooling loop commissioning.

Worked example: size supply and return QD sets

At 1.7 US gpm and water specific gravity 1.00, Anvilfield estimates that two coupled QD sets in series lose 4.78 psi at Cv 1.1 each, compared with 2.26 psi at Cv 1.6 each, using the UQD04 reference values in OCP UQD v2.0.0. The lower-Cv selection consumes 2.12 times as much pressure allowance at the same flow.

To estimate one set's loss, divide US-gpm flow by its flow coefficient (Cv), square the result and multiply by fluid specific gravity. Add supply and return losses using their actual flow directions. The table uses OCP UQD v2.0.0 §8's size-04 endpoints and an Anvilfield project allowance of 3 psi across both sets.

Method, calculated November 5, 2026: ΔP_set = SG × (Q/Cv)²; ΔP_total = ΔP_supply + ΔP_return. For two identical sets, Cv_required = Q × √(2 SG/ΔP_allowance). Use US gpm and psi; SG is 1.00. Convert unrounded results before display rounding.

Each equal-loss set needs at least Cv 1.39 to meet the 3 psi allocation. Danfoss UQD, September 2026 publishes UQD04 Cv 1.37 at 1.7 US gpm; substituting that input gives Anvilfield estimates of 1.54 psi per set and 3.08 psi total. That exceeds the allocation by 0.08 psi. Obtain matched-fluid directional data and measurement uncertainty before deciding whether the assembly meets the allowance; revise the selection or allowance if it fails.

Both sets carry the full flow of this series branch. Parallel server branches share manifold differential pressure and their flows add at the manifold; multiplying one branch's QD loss by the rack's branch count overstates the required pressure. Add hose, manifold, filter, cold-plate and control-device losses separately, then verify available pressure through CDU pressure and flow acceptance.

A constant-Cv sensitivity calculation gives 44% more coupling loss when flow rises 20%, because loss scales with flow squared. Check the actual curve over that range before spending the remaining pump allowance.

Anvilfield estimates calculated November 5, 2026 from OCP UQD v2.0.0 reference inputs. The 3 psi allocation is a project input.

How do you qualify UQD seal compatibility?

Approve the complete wetted assembly for the exact coolant, not the word EPDM on a quotation. Record seal compound and cure, metals, plastic valve parts, lubricant, adhesive and hose. OCP UQD v2.0.0 §12 requires compatible wetted materials; OCP PG25 guidelines, Rev 2.0 / v1.0.0 provide formulation-specific screening guidance.

Danfoss's September 2026 UQD and July 2026 UQDB sheets identify peroxide-cured EPDM-P; Stäubli RD7110301B (2024) identifies EPDM JE and advises against mineral-origin oils and fats. Installation lubricant and cleaning residues therefore belong in the compatibility record alongside the circulating coolant. Include approved cleaning and assembly products in the work instruction.

OCP water-based fluid guidelines, undated final §4.1 recommend stainless-steel quick connects and caution against plated parts until their suitability is demonstrated, because repeated use can expose base metal. Obtain the wetted bill of materials and wear evidence.

What exposure evidence is enough?

OCP's September 15, 2026 PG25 guidelines, Revision 2.0 / Version 1.0.0, describe typical material exposure of six weeks at 65°C to 85°C. The screening ranges below support material review; the component manufacturer sets acceptance for the actual seal and application. Record specimen type, coolant concentration, exposure time, temperature and measurement method.

A static coupon screen measures material change; require assembly evidence for retained sealing, valve function, wear, particles, extractables and lubricant washout after exposure.

OCP PG25 §8.12 makes retained function part of its tensile and elongation screening guidance; §9.4 warns that short-term screening does not predict long-term reliability. Transfer the approved formulation and sampling requirements into the coolant chemistry and sampling program.

Screening guidance is from OCP PG25 Revision 2.0 / Version 1.0.0, September 15, 2026. Apply component-specific functional acceptance.

Will the connector fit and connect at service pressure?

Budget insertion force at service pressure and verify the entire engagement path. Include hose reactions, hand access, panel spacing, valve stroke and latch visibility. Match end-connection standard and thread as well as interface size: UQD04 does not mean a 1/4-inch pipe thread.

Using OCP UQD v2.0.0 §8's size-04 blindmate SI allowance of 65 N plus 9.62 N/bar, Anvilfield estimates approximately 131 N per pair at 6.89 bar, or 263 N for two pairs before chassis friction. Calculated November 5, 2026: two-pair force = 2 × (65 + 9.62 × 6.89) N, rounded after calculation. Use the selected supplier's measured curve for chassis design. OCP's SI and customary force-scaling rows are not exact conversions; this estimate uses its SI row throughout.

OCP UQD v2.0.0 §9.1.3 tests connection force at 0, 40 and 100 psig and an insertion speed of 40 ±10 mm/s; UQDB testing includes maximum radial offset. The specified pressure is applied to both halves. Require that condition and force over travel in the report; a one-side-pressurized result is a different test.

Stäubli RD7110301B (2024) lists 0.8 mm maximum UQDB04 coupling misalignment; Danfoss UQDB, July 2026 lists 1 mm radial compensation. Stäubli's separate 3 mm guiding-arrangement claim concerns the guiding system and does not increase its coupling's inter-axis limit.

Retention and hard stops need their own evidence

Handmate acceptance requires latch engagement; blindmate and hybrid acceptance requires retention and compliance that maintain the specified hard-stop engagement. OCP/Intel blindmate methodology explains the historical rack-tolerance problem; current acceptance follows OCP UQD v2.0.0. Require evidence at the worst allowed offset and insertion path, including the customer's retention mechanism: a closed drawer does not establish full valve opening.

What spill and air-inclusion limits should you specify?

Specify fluid released during cycling, air entering during reconnection and ongoing external leakage as separate criteria. OCP UQD v2.0.0 §8 limits size-04 fluid loss to 0.03 mL per cycle, but its performance table supplies no universal numerical air-inclusion limit. Agree that limit separately with the OEM and system owner.

The table combines OCP UQD v2.0.0's water fluid-loss method, which references ISO 18869 §12, with the catalog maxima in Danfoss UQD, September 2026 and Danfoss UQDB, July 2026. One cubic centimetre equals one millilitre. The nonzero OEM air values show why dry-break construction cannot be treated as proof of zero gas entry.

Specify collection method, initial wetting, fluid, pressure, temperature, cycle count and instrument resolution. Retain gas-volume reference absolute pressure and temperature, and define whether a limit applies per cycle, to an average or to a worst result. Obtain laboratory qualification reports before the installed visit; rack acceptance checks identity, engagement, tightness and flow.

Anvilfield's procurement rule is: release the coupling when the fluid-loss and air-inclusion reports each meet their own specified limit under the agreed fluid, pressure, temperature and cycle method; hold reports with missing conditions, and reject a confirmed exceedance of either limit. Apply the applicable OCP fluid-loss ceiling and any tighter contractual OEM guarantee; use the separately agreed numerical air criterion for gas inclusion.

Contain collected cycling fluid away from electronics. Continued stationary weeping requires its own leak disposition, and the liquid-cooling leak-detection checks still need to prove containment and alarms.

Danfoss UQD September 2026 and UQDB July 2026 catalog maxima require their agreed test conditions. OCP values refer to UQD v2.0.0, effective September 22, 2026.

Can you disconnect under pressure or use the QD for filling?

Connect, disconnect, fill and drain only within the selected product's and system's service procedure. Operating pressure establishes containment capability under stated conditions; it does not by itself authorize every coupling action. Check pressures on both halves and provide the specified collection and isolation arrangement.

OCP UQD v2.0.0 §9.1.5 qualifies closed-loop seal washout using water or PG25 at 60°C, 100 psig and twice rated flow for 20 cycles in each direction, with no observed leakage or seal washout. The test limits the transient to 30 psi above pre-disconnect pressure. That controlled test condition does not authorize routine operation of a 100 psi installation at 130 psi.

Open-loop filling and draining impose different differential pressures and valve velocities. OCP UQD v2.0.0 calls for progressive introduction of flow and pressure and supplier guidance for this use. Specify the approved fill rig and service method, and coordinate the branch's isolation and return to service with concurrent maintenance of liquid cooling.

Where must the hydrotest stop?

Stop the construction hydrotest at the documented boundary of approved equipment exposure. The actual adopted piping code, owner specification and equipment instructions establish that boundary. A supplier qualification test and a construction piping test answer different questions, as reflected in OCP UQD v2.0.0 §9.2.4 and OCP water-based fluid guidelines, undated final §4.3.

OCP UQD v2.0.0 specifies a post-cycle hydrostatic leak test at 100 psig for at least three minutes with no visible leakage; its separate 300 psig minimum burst requirement does not authorize a field hydrotest at 300 psig. OCP §9.2.4 requires purging air, achieving the specified internal pressure in a mated pair and pressurizing a disconnected half in the direction that closes its valve. Any construction-test allowance must cover the exact state, pressure, medium, temperature and duration.

For the boundary table's 150 psig project test, separate a Danfoss UQD06 unless exact OEM authorization covers that exposure. A higher-rated substitute still requires checking the hose, terminations, manifold and every other exposed component.

Anvilfield's field rule is to mark the piping and instrumentation diagram with rated temporary blanks, removed equipment, vents, drains, monitored boundaries and reinstatement steps. Record gauge elevation and the pressure at the lowest exposed component so static head is included in its test exposure. A dust cap is not a pressure-rated blank, and a disconnected QD valve is not automatically an approved test boundary. Identify the pressure-facing half and paths by which pressure could reach excluded equipment.

Use the chilled-water hydro-test package for the piping procedure. After testing, restore the final arrangement and verify disturbed joints; a passed test of temporary closures does not establish the tightness of connections assembled afterward.

The 150 psig piping test is an Anvilfield project input. Exact equipment test instructions and the approved project boundary govern construction exposure.

How long should qualification and service evidence cover?

Allow for six weeks of typical material exposure under OCP PG25 guidelines, Rev 2.0 / v1.0.0, plus the separate durability sequence when new qualification is required. OCP UQD v2.0.0 §9.2.2 specifies operational soaks of 72 hours at 5°C and 72 hours at 65°C, at 100 psig, followed by hydrostatic testing.

OCP UQD v2.0.0 §9.2.1 separately specifies shipping soaks of 72 hours at -40°C and 72 hours at +70°C using air at 28 to 30 psig. Anvilfield estimates 288 soak hours, or 12 days, by summing all four 72-hour holds. Add transitions, mechanical cycling, leak checks and laboratory queue time.

Obtain existing compliant reports before commissioning new laboratory work. OCP UQD v2.0.0 §8 specifies at least 5,000 coupling cycles, five years of storage life from manufacture and ten years of service life following storage. These are specified life requirements; replacement follows OEM instructions and recorded condition.

OCP UQD v2.0.0 §9.2.3 performs mechanical cycling at 40 psig with water or PG25, followed by hydrostatic leakage testing. Retain sample identities through the sequence: §9 requires thermal and mechanical cycling on the same sample set, while time-zero property tests can use parallel sample sets.

Anvilfield's service checks are inspection at receipt and before each mating, with immediate investigation of new leakage, damaged surfaces or abnormal force. Record cycles, fluid condition and exposure; set overhaul from the OEM maintenance plan.

Keep chemistry monitoring separate from coupling replacement

OCP's September 2026 PG25 guidelines §11.2 recommend qualified-laboratory analysis of each individual TCS loop at least annually, alongside startup and onsite checks. Keep the chemistry program connected to the approved materials record, and obtain quotation-specific durations for sample fabrication, laboratory queue and retesting.

Field acceptance checklist and coupling qualification record

Use this Anvilfield matrix and checklist to release a documented pass, hold missing evidence and reject a demonstrated mismatch. Every gate must pass; a score cannot average away a failed pressure or compatibility requirement.

The purchaser or owner's designated engineer approves selection and substitutions; the OEM supplies ratings and qualification evidence; the installer records as-built parts and assembly; the commissioning witness records installed results. Assign witness and hold points in the inspection and test plan.

File the branch ID, both part numbers and revisions, lot/manufacturing dates, coolant and lot, report links, instrument IDs, witness/date and each gate's result. Attach repairs and retests to the original deficiency in the turnover proof packet, with the approved spare list.

Anvilfield coupling qualification matrix applying OCP UQD v2.0.0 and the named OEM editions. Each result belongs to the exact installed plug/socket combination.

Why does an installed QD fail its flow or leak check?

Measure pressure loss across the QD before changing pump settings or condemning the cold plate. Compare the result with the accepted baseline at the same fluid, temperature, flow, direction and engagement. A branch-level loss measurement includes hoses and other devices and cannot by itself locate a restriction inside the coupling.

Compare event timing with the branch's last accepted condition. The matrix applies OCP UQD v2.0.0 tests and the cited coolant guidance.

Isolate the branch and verify pressure is relieved under the approved equipment procedure before opening a suspect joint. Keep cycling wetting separate from stationary leakage and confirm the wet location before replacing a seal. After correction, repeat the accepted installed test point and retain the before-and-after values at matching conditions.

Anvilfield diagnostic sequence, based on OCP UQD v2.0.0 and the cited coolant materials guidance.

How much work should the qualification package include?

Anvilfield estimates 22 labor hours for one coupling-family document review, one installed-verification visit and one evidence report. Hardware, supplier laboratory qualification, travel and remedial work are separate quotation items. Obtain compliant supplier reports before scheduling field verification.

Define the quotation unit: a half, a coupled plug/socket set or a complete supply-and-return branch. Itemize terminations, compatibility evidence, dimensional checks, mixed-pair tests, temporary fixtures, instruments, witness attendance and rework, with ownership of each report stated.

Public mechanical packages often split installation, testing and commissioning into separate bid items. Estimators pursuing public cooling installation and commissioning work use PorchRocket's Bid Control to find HVAC and refrigeration bids and early project signals from budgets and capital plans.

Method note: Anvilfield estimates use task-hour allowances set November 5, 2026. Sum the four tasks below; an additional eight-hour witness day brings the allowance to 30 hours. Build the bid from actual loaded labor rates and supplier quotations, and schedule laboratory exposure, queue, shipping and retests separately.

Anvilfield estimating inputs, November 5, 2026: one family, one visit and one report. Supplier laboratory qualification is a separate scope.

Sources checked and citation

Sources checked November 5, 2026: the ten primary documents listed below, including OCP UQD v2.0.0 and the downloaded Danfoss and Stäubli editions.

Update log, 2026-11-05: Clarified compatibility guidance, force-test pressures, hydrostatic test configurations and qualification schedules; added measurement uncertainty and a coupling qualification matrix; revised scope allowances.

How to cite this page: Anvilfield, How to select OCP liquid-cooling quick disconnects, 2026, https://anvilfield.com/field-guides/datacenter/liquid-cooling-quick-disconnect-selection/.

Related tools

FAQ

People also ask

• What pressure do you test chilled-water piping to?
• What is direct-to-chip liquid cooling?
• What is direct-to-chip liquid cooling?
• What is a CDU in a data center?
• What coolant is used in data center liquid cooling?

Sources checked

• OCP, Universal Quick Disconnect (UQD), v2.0.0, effective September 22, 2026. Opened with curl and checked November 5, 2026.
• OCP/Intel, Universal Quick Disconnect Specification, Revision 1.0, September 4, 2020. Opened with curl and checked November 5, 2026.
• Danfoss Hansen UQD fact sheet, internally AI423953312825en-000106, September 2026. Opened with curl and checked November 5, 2026.
• Danfoss Hansen UQDB fact sheet, internally AM480054893377en-000106, July 2026. Opened with curl and checked November 5, 2026.
• OCP, Guidelines for Using Propylene Glycol-Based Heat Transfer Fluids in Liquid Cooled Racks, Revision 2.0, Version 1.0.0, September 15, 2026. Opened with curl and checked November 5, 2026.
• OCP, Guidelines for Using Water-Based Transfer Fluids in Single-Phase Cold Plate-Based Liquid-Cooled Racks, final PDF, no explicit revision/date on cover. Opened with curl and checked November 5, 2026.
• Stäubli, Universal Quick Disconnect UQD & UQDB, RD7110301B, ©2024. Opened with curl and checked November 5, 2026.
• OCP, Guidelines for Implementing Passive Technology Cooling Systems Using Loop Heat Pipe Technology, Revision 1.0, November 21, 2024. Opened with curl and checked November 5, 2026.
• OCP/Intel, Methodology: Development and Testing of Blind-Mate Universal Quick Disconnect, rev-1-0 publication URL; no revision/date printed on cover. Opened with curl and checked November 5, 2026.
• OCP, Cold Plate Cooling Loop Requirements, Rev 2, no publication date printed on cover. Opened with curl and checked November 5, 2026.