DC HVAC cheat sheet

Sensible heat ratio (SHR) in a data hall typically runs 0.90-1.00 because IT loads release almost no latent heat; latent removal comes mainly from infiltration and people.

Heat load balance: Q_total = Q_IT + Q_lighting + Q_people + Q_infiltration + Q_solar + Q_fan. IT equipment usually dominates at 80-95% of total in a well-designed hall.

Room sensible heat: Q_s = 1.08 x CFM x delta-T (imperial, Btu/h) or Q_s = 1.21 x L/s x delta-T (metric, W); latent: Q_l = 4,840 x CFM x delta-W (grains/lb).

Air density correction: standard air is 1.2 kg/m3 at 20 C sea level; at 1,500 m altitude density falls roughly 15%, reducing mass-flow cooling capacity for the same volumetric flow.

ASHRAE TC 9.9 recommended inlet envelope: Class A1 allows 18-27 C dry-bulb and 5.5-15 C dew point; excursions outside the envelope raise failure rates and void many warranties.

Dew point, not relative humidity, governs condensation risk on cold surfaces; keep supply air dew point below the coldest IT inlet surface to avoid liquid water.

Heat transfer modes in a hall: convection to room air dominates; radiation between racks and walls is minor; conduction is significant only through raised-floor slabs and walls.

Psychrometric rule: at constant moisture content, cooling dry-bulb raises relative humidity; the coil must remove moisture to lower dew point.

Free cooling viability depends on the difference between outdoor wet-bulb (evaporative) or dry-bulb (air-cooled) and the required supply air temperature, not on outdoor dry-bulb alone for evaporative systems.

CRAC = compressor-based direct-expansion unit inside the hall; CRAH = chilled-water coil unit. CRAH units need a chiller plant or district cooling source; CRAC units are self-contained.

DX capacity control: compressors cycle or modulate (digital scroll, inverter); chilled-water control modulates a 2-way or 3-way valve on the coil. 3-way valves keep constant plant flow but waste pump energy.

Air-cooled chiller efficiency (EER) typically 2.8-3.5 at design; water-cooled centrifugal chillers reach COP 5.5-7.0 with condenser water at 29-32 C.

Cooling tower approach = leaving condenser water temperature minus entering wet-bulb; good design approach is 3-5 K. Towers are the plant's wet-bulb-dependent efficiency lever.

Economizer modes: air-side (outside air damper), water-side (strainer cycle or plate heat exchanger), and refrigerant-side. Water-side economizers use a plate HX between tower water and chilled water.

Redundancy nomenclature: N, N+1, 2N, 2(N+1). Concurrent maintainability means any component can be serviced without dropping load; fault tolerance means a second failure after the first does not drop load.

Fan types: EC plug fans modulate smoothly and are often placed in a plenum; belt-driven centrifugal fans need belt and bearing maintenance. Fan power scales with the cube of speed (affinity law).

Humidification options: steam canister (electric), infrared, ultrasonic, and evaporative. Over-humidification is a common energy waste in halls with tight envelopes.

Pump and fan affinity laws: flow proportional to speed, head proportional to speed squared, power proportional to speed cubed. A 50% speed cut reduces power to about 12.5%.

Hot-aisle/cold-aisle layout: racks face each other front-to-front in cold aisles and back-to-back in hot aisles; servers draw from the cold aisle and exhaust to the hot aisle.

Containment types: cold-aisle containment (CAC) encloses the cold aisle with roof panels and end doors; hot-aisle containment (HAC) encloses the hot aisle and returns air to the CRAC/CRAH.

Bypass air is cold supply air that never reaches IT intake; recirculation is hot exhaust air that reaches IT intake. Both degrade cooling and are measured by RCI (rack cooling index).

Rack Cooling Index: RCI_HI = 1 - (total degrees of intake temperature above the upper limit)/(maximum allowable exceedance). 100% means all racks within the envelope.

Return Air Temperature Index and Supply Air Temperature Index compare actual to ideal temperatures; a return index near 100% indicates little bypass and good mixing control.

Perforated tile free area controls cold-aisle pressure: 25% open tiles are common; oversizing open area in one zone starves distant racks. Balance tiles by measured airflow, not count.

Raised-floor plenum velocity should stay under about 2.5 m/s (500 ft/min) to limit friction loss and turbulence; deeper plenums reduce static pressure drop.

Blank panels in unused U-spaces and brush strips at cable openings are the lowest-cost containment measures; an open 1U slot can leak a large fraction of a rack's cold air.

Cable management under the floor blocks plenum flow; route cable trays overhead or along the sides of the plenum to preserve the pressure field.

Cooling capacity conversions: 1 ton refrigeration = 3.517 kW = 12,000 Btu/h; 1 kW IT load requires roughly 0.28 tons of cooling plus overheads.

PUE = total facility energy / IT equipment energy. Design PUE is modeled; annualized PUE is measured over 12 months and includes all overheads (cooling, losses, lighting).

pPUE (partial PUE) scopes a boundary such as one hall or one module; useful when the facility shares plant across tenants.

DCiE = 1/PUE expressed as a percentage; a PUE of 1.5 equals 66.7% DCiE.

WUE = total water use (liters) / IT energy (kWh). Evaporative cooling trades water for electricity; WUE and PUE must be read together.

CUE = total carbon emissions / IT energy; depends on grid carbon intensity and on-site fuel use.

Cooling load components: IT, UPS and PDU losses (typically 2-6% of IT), lighting (1-2% of IT), people, infiltration, and fan/coil gains.

Design margin discipline: specifying 2x actual IT load wastes capital and often degrades part-load efficiency; variable-speed plant performs best at 40-70% load.

Chiller IPLV/NPLV weights part-load points and is a better comparison metric than full-load COP for data centre duty, which runs mostly at part load.

Set-point policy: raise supply air temperature toward the upper ASHRAE envelope (e.g., 22-24 C inlet) to extend economizer hours and improve chiller efficiency.

Dead band and staging: avoid simultaneous heating and cooling or humidification and dehumidification; interlock set-points with at least 1-2 K separation.

Filter maintenance: pre-filters and final filters load up and raise fan energy; monitor differential pressure and change on pressure, not on a fixed calendar alone.

Coil cleaning restores heat transfer; fouled coils raise approach temperature and can cut capacity by double-digit percentages.

Condenser water treatment controls scale, corrosion, and biological growth; poor treatment raises condenser approach and chiller kW per ton.

Commissioning and retro-commissioning: verify economizer damper travel, sensor calibration, valve and damper actuator stroke, and airflow balance; sensor drift is a top hidden energy waste.

Variable-speed drives on pumps and fans, plus EC fans, deliver the largest part-load savings; confirm minimum speed limits and control curves are set for the actual load profile.

Air-side economizer control must use differential enthalpy or dry-bulb with a high-limit lockout; a stuck-open damper can freeze coils or overload dehumidification.

Energy monitoring: trend kW/ton, PUE, supply and return temperatures, and valve/damper positions; persistent valve positions near 100% signal undersized coils or low delta-T syndrome.

Low delta-T syndrome: chilled-water flow rises while delta-T falls because of coil fouling, 3-way valves, or control valve bypass; plant kW/ton worsens even as capacity appears adequate.

Chiller sequencing: stage chillers on actual load and on condenser water temperature, not only on return water temperature; avoid short cycling which erodes efficiency and reliability.

Condenser water reset: lower tower water temperature where the chiller allows (often to 20-24 C) to cut compressor lift and kW/ton, subject to minimum condenser temperature limits.

Free-cooling changeover: switch to water-side economizer when outdoor wet-bulb is low enough to produce chilled water at the required supply temperature via the plate HX.

Preventive maintenance intervals: quarterly filter checks, semi-annual coil and belt inspection, annual chiller overhaul review, and continuous monitoring of critical alarms.

Capacity testing: perform integrated load bank or live-load tests to confirm N+1 or 2N performance at design conditions and to validate failover sequences.

Humidifier and dehumidification control: use dew-point control rather than relative-humidity control to prevent simultaneous humidification and dehumidification.

Alarm rationalization: prioritize high-temperature, high-humidity, water-leak, and power-path alarms; flood of low-priority alarms hides real events.

Energy efficiency benchmarking: compare annualized PUE, WUE, and kW/ton against peers with similar climate and redundancy; normalize for climate and IT load factor before drawing conclusions.

Sensible cooling dominates: IT equipment converts nearly all electrical input to sensible heat, so coil sensible capacity, not total capacity, sets the real limit in a data hall.

Supply air temperature rise across the hall equals IT load divided by (mass flow x specific heat); a 100 kW rack at 1.0 kg/s and 1.005 kJ/kg-K rises about 100 K, so flow must scale with density.

Altitude correction: at 2,000 m, air density is about 1.0 kg/m3 versus 1.2 at sea level, so the same CFM delivers roughly 17% less cooling.

Dew point control: supply air dew point should stay below the coldest surface in the hall; typical target is 5.5-15 C per ASHRAE A1.

Latent load sources in a tight hall are small: infiltration through doors, people, and any open humidifier; most halls need humidification only in dry climates.

Enthalpy difference across the coil equals total cooling; dry-bulb difference alone gives only sensible cooling.

Chilled-water supply/return design delta-T is commonly 6-8 K; low delta-T forces more flow and raises pump and chiller energy.

Primary-secondary pumping decouples plant flow from distribution flow; primary-only variable flow is simpler but needs careful chiller minimum-flow control.

Air-cooled vs water-cooled: air-cooled avoids water and towers but has lower efficiency and is limited by ambient dry-bulb; water-cooled needs make-up water and treatment.

DX CRAC units commonly provide 20-60 kW each; CRAH units can be larger and are fed by the central plant.

Redundancy tiers: N, N+1, 2N, and 2(N+1); 2N duplicates the entire path, while N+1 adds one unit to the minimum required.

Fan coil and in-row units place cooling close to the load; in-row units handle high-density racks (10-30 kW+) that room-level CRACs cannot reach.

Airflow per rack: estimate from server fan curves or nameplate CFM; a 10 kW rack typically needs about 2,000-2,500 m3/h of cold air at a 10-12 K rise.

Cold-aisle pressure is maintained slightly above hot-aisle pressure; containment doors should have a small pressure relief or the aisle can over-pressurize.

Hot-aisle containment returns hot air directly to the CRAC/CRAH return; it also enables higher supply temperatures and more economizer hours.

Chimney and ducted exhaust racks capture hot air at the source and are an alternative to aisle containment in mixed-density halls.

Measure containment performance with temperature sensors at rack intakes top, middle, and bottom; top-of-rack intake is usually the hottest point.

Airflow balancing: adjust perforated tile open area and damper positions until each rack intake is within the ASHRAE envelope, then recheck after any rack change.

IT load estimate: nameplate watts overstate real draw; use measured power or vendor typical power (often 40-70% of nameplate) for load calculations.

Cooling capacity in kW = mass flow (kg/s) x specific heat (kJ/kg-K) x delta-T (K); for air use 1.005 kJ/kg-K.

PUE = (IT + cooling + losses + lighting) / IT; a PUE of 1.2 means 20% overhead, and 1.0 is the theoretical floor.

Annualized PUE uses total kWh over a year; instantaneous PUE varies with weather and load and should not be compared directly.

WUE in L/kWh: evaporative cooling can push WUE to 2-5 L/kWh while air-cooled designs approach 0.

Part-load efficiency: chillers are most efficient at 40-70% load; oversizing moves operation away from this band.

Supply air temperature reset: raise supply temperature when all rack intakes remain inside the envelope; each 1 K rise can extend economizer hours meaningfully.

Humidification control: use dew-point or absolute-humidity control; relative-humidity control causes simultaneous humidification and dehumidification.

Economizer lockout: use a high-limit switch or enthalpy comparison; a failed damper actuator is a common cause of frozen coils and energy waste.

Filter differential pressure trending reveals loading; change filters on pressure rise, and verify fan curves after filter changes.

Valve and damper stroke testing during commissioning catches actuators stuck at 0% or 100%, a frequent hidden energy loss.

Trend logging: kW/ton, PUE, supply and return temperatures, and valve positions; persistent 100% valve position signals undersized coils or low delta-T.

Chiller staging on load and condenser temperature avoids short cycling and keeps units in their efficient band.

Low delta-T syndrome: chilled-water delta-T falls below design while flow rises; causes include fouled coils, 3-way valves, and control valve bypass.

Condenser water reset lowers compressor lift and kW/ton, subject to the chiller's minimum condenser temperature limit.

Free cooling: engage the water-side economizer when outdoor wet-bulb allows the tower to produce chilled water at the required supply temperature.

Maintenance intervals: quarterly filter checks, semi-annual coil and belt inspection, annual chiller review, and continuous critical alarm monitoring.

Capacity testing: integrated load tests confirm N+1 or 2N performance and validate failover sequences at design conditions.

Alarm rationalization: prioritize high temperature, high humidity, water leak, and power-path alarms so real events are visible.

Benchmarking: compare annualized PUE, WUE, and kW/ton against peers in similar climates and redundancy classes, normalized for load factor.

Convection from server surfaces to room air dominates heat removal; radiation and conduction are secondary in a well-mixed hall.

Cold surfaces below the local dew point condense water; keep supply air dew point below the coldest surface temperature.

At constant moisture content, cooling dry-bulb raises relative humidity; the coil must remove moisture to lower dew point.

Free cooling viability depends on outdoor wet-bulb for evaporative systems and dry-bulb for air-cooled systems, not on dry-bulb alone for towers.

ASHRAE A1 envelope: 18-27 C dry-bulb and 5.5-15 C dew point; excursions outside the envelope raise failure rates and can void warranties.

Sensible heat ratio near 1.0 means coil capacity is mostly sensible; oversizing latent capacity wastes energy on unnecessary dehumidification.

CRAC units are self-contained DX; CRAH units use chilled water from a central plant and have no compressor.

Redundancy: N is minimum, N+1 adds one unit, 2N duplicates the path, and 2(N+1) duplicates plus one spare.

Concurrent maintainability allows servicing any component without dropping load; fault tolerance survives a second failure.

Fan affinity laws: flow scales with speed, head with speed squared, and power with speed cubed; a 50% speed cut leaves about 12.5% power.

EC plug fans modulate smoothly and are often plenum-mounted; belt-driven centrifugal fans need belt and bearing maintenance.

3-way valves keep constant plant flow but waste pump energy; 2-way valves with variable flow save pump energy but need minimum-flow control.

Bypass air is cold supply air that never reaches IT intake; recirculation is hot exhaust that reaches IT intake. Both raise intake temperature or waste cooling.

Rack Cooling Index (RCI) measures how well rack intakes stay within the envelope; 100% means all racks are within limits.

Return Air Temperature Index near 100% indicates little bypass and good mixing control.

Blank panels and brush strips are the lowest-cost containment measures; an open 1U slot can leak a large fraction of a rack's cold air.

Raised-floor plenum velocity should stay under about 2.5 m/s to limit friction and turbulence.

Perforated tile open area controls cold-aisle pressure; balancing by measured airflow, not tile count, prevents starving distant racks.

1 ton refrigeration = 3.517 kW = 12,000 Btu/h; 1 kW IT load needs roughly 0.28 tons plus overheads.

PUE = total facility energy / IT energy; annualized PUE covers 12 months and includes all overheads.

pPUE scopes a boundary such as one hall or module and is useful when plant is shared across tenants.

DCiE = 1/PUE as a percentage; a PUE of 1.5 equals 66.7% DCiE.

WUE = liters of water / IT kWh; evaporative cooling trades water for electricity, so WUE and PUE must be read together.

CUE = carbon emissions / IT kWh and depends on grid carbon intensity and on-site fuel use.

Set-point policy: raise supply air temperature toward the upper ASHRAE envelope to extend economizer hours and improve chiller efficiency.

Dead band and staging: interlock set-points with at least 1-2 K separation to avoid simultaneous heating and cooling.

Coil cleaning restores heat transfer; fouled coils raise approach temperature and can cut capacity by double-digit percentages.

Condenser water treatment controls scale, corrosion, and biological growth; poor treatment raises condenser approach and chiller kW/ton.

Retro-commissioning verifies economizer damper travel, sensor calibration, and actuator stroke; sensor drift is a top hidden energy waste.

Variable-speed drives and EC fans deliver the largest part-load savings; confirm minimum speed limits match the actual load profile.

Chiller sequencing stages units on actual load and condenser water temperature, not only on return water temperature, to avoid short cycling.

Free-cooling changeover: switch to the water-side economizer when outdoor wet-bulb can produce chilled water at the required supply temperature.

Capacity testing: integrated load bank or live-load tests confirm N+1 or 2N performance and validate failover sequences at design conditions.

Alarm rationalization: prioritize high temperature, high humidity, water leak, and power-path alarms; a flood of low-priority alarms hides real events.

Benchmarking: compare annualized PUE, WUE, and kW/ton against peers with similar climate and redundancy, normalized for load factor.

Preventive maintenance intervals: quarterly filter checks, semi-annual coil and belt inspection, annual chiller overhaul review, and continuous critical alarm monitoring.