How much capacity does a heat pump lose during defrost?

How much heating capacity is lost while the heat pump defrosts?

At 30°F, the tested Carrier system loses about 4% of integrated capacity at minimum heating operation and 13% at maximum operation, calculated from NREL's February 2022 regressions. Table A shows how that loss changes with outdoor temperature. These are humid-test cycle losses for one equipment pair; use whole-cycle measurements at the installed unit's conditions for sizing.

NREL's February 2022 report includes defrost electricity and indoor cooling. Thermal kW is heat delivery rate; electrical kW is input power. ODU means outdoor unit; IDU means indoor unit. Integrate each over hours for its respective kWh. [NREL/TP-5500-80802 (February 2022)](https://docs.nlr.gov/docs/fy22osti/80802.pdf)

Method: Anvilfield estimates subtract Figure 18's capacity multiplier from one and express the result as a percentage. For outdoor dry-bulb x in °F above the study's 25°F control breakpoint, the published regressions are f_min = -0.0001371x² + 0.01273x + 0.7045 and f_max = -0.0003917x² + 0.03786x + 0.08490. Calculations retain those coefficient digits before rounding. Tests used comfort mode, Auto defrost and sea-level corrections against matched low-frost references. Keep these equations within the tested model and above-25°F control branch. At 30°F, a 10 kW reference at maximum operation becomes 8.68 kW under this regression. [NREL/TP-5500-80802 (February 2022)](https://docs.nlr.gov/docs/fy22osti/80802.pdf)

Why warm supply air does not prove the heat pump is heating

A resistance heater downstream of the indoor coil can warm supply air while the coil extracts indoor heat. Locate the probes and meter the strips separately to distinguish coil output, air-handler output and occupied-space delivery. With the indoor fan stopped, a cold coil probe alone does not establish sustained forced-air negative capacity: airflow and stored heat must enter the complete-cycle balance. [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996) [Daikin IOD-1065B (January 2025)](https://daikincomfort.com/docs/default-source/dh6te-r-32/im-iod-1065b.pdf) [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

The [heat-pump fundamentals guide](/field-guides/hvac/heat-pump-fundamentals-cop-defrost/) explains the refrigerant cycle behind these measurements.

Which published capacity and efficiency number are you comparing?

Match the equipment pair, speed, temperatures, airflow and electrical boundary before comparing performance numbers.

For covered consumer central air conditioners and heat pumps, Appendix M1, checked November 5, 2026, governs efficiency representations from July 7, 2025 until the applicable SCORE/SHORE standards compliance date, incorporates AHRI 210/240-2024, and uses Region IV heating certification bins. The CFR prevails where its provisions conflict with incorporated standards. Match the rating program to the equipment class; M1/HSPF2 and the central-heat-pump controls verification procedure do not cover every rooftop, VRF or hydronic system. [DOE final rule (January 7, 2025)](https://www.energy.gov/sites/default/files/2024-12/cac-tp-fr.pdf) [Current Appendix M1](https://www.ecfr.gov/current/title-10/chapter-II/subchapter-D/part-430/subpart-B/appendix-Appendix%20M1%20to%20Subpart%20B%20of%20Part%20430)

What NEEP data and ENERGY STAR verification tell you

NEEP's Cold Climate Air Source Heat Pump Specification Version 4.0, effective January 1, 2023, excludes defrost-cycling and drain-pan-heater power from its expanded performance table. Keep those exclusions with operating-point comparisons. [NEEP specification v4.0](https://neep.org/sites/default/files/media-files/cold_climate_air_source_heat_pump_specification_-_version_4.0_final.pdf)

ENERGY STAR's March 2025 Controls Verification Procedure tests native controls at 5°F; its official one-hour test starts no earlier than ten minutes after pretest defrost. When steady state is not reached during the four-hour pretest, successive 30-minute capacity and power averages must be within 2% for dynamic equilibrium. This verifies low-temperature control behavior at a defined test condition. [ENERGY STAR CVP (March 2025)](https://www.energystar.gov/sites/default/files/asset/document/ENERGY%20STAR%20Cold%20Climate%20Heat%20Pump%20Controls%20Verification%20Procedure_1.pdf)

Use the rating's stated integration boundary; adding a blanket defrost correction to an already integrated value counts the loss twice.

How do you calculate capacity over a complete defrost cycle?

Sum every interval's signed heat, including indoor cooling, then divide by elapsed hours for average capacity. Divide the same heat total by the electrical energy consumed during exactly that period for [COP](/field-guides/glossary/cop/). NREL's 2022 frost characterization supplies the signed integration principle. [NREL/TP-5500-80802 (February 2022)](https://docs.nlr.gov/docs/fy22osti/80802.pdf)

Calculation box: H = Σ(q × Δt); E = Σ(P × Δt); cycle capacity = H/T; cycle COP = H/E; capacity loss = 1 - H/H_reference. Here q is thermal kW, P is electrical kW, Δt and T are hours, and H and E are thermal and electrical kWh. H_reference is matched clean-coil output integrated over the same elapsed time.

Define the clean-coil reference from low-frost tests or a fitted performance curve at matched outdoor dry-bulb, return conditions, airflow and compressor operation. Record moisture separately because it drives frost accumulation; compare before/after control changes within matched moisture regimes. Comparing post-defrost high-speed output with pre-defrost low-speed output confounds the control response with coil clearance. Retain gradual capacity degradation as frost accumulates, as well as reversal and recovery.

Consecutive defrost-termination timestamps define a repeatable complete cycle, with precedent in the historical 2025 CFR frost test. That window contains recovery after its first termination and ends at the next termination; adding another recovery window counts part of the following cycle. A consistently applied reversal-start-to-next-reversal-start boundary also encloses a full cycle. [Annual CFR (January 1, 2025), historical M1](https://www.govinfo.gov/content/pkg/CFR-2025-title10-vol3/pdf/CFR-2025-title10-vol3-part430-subpartB-appM1.pdf)

The worked balance separates interrupted heating, indoor extraction and recovery shortfall. Under changing loads, use longer continuous records and label reference-normalized losses as estimates.

Define recovery before collecting data

Record reversal, return to heating, fan restart and auxiliary shutdown as separate timestamps. Anvilfield's analysis convention ends recovery when output reaches at least 90% of the matched clean-coil reference for five continuous minutes, while comparable heating demand and operation persist. This is a reporting convention, not a manufacturer acceptance threshold.

When the thermostat satisfies, compressor speed changes materially or another event begins first, mark recovery interrupted or unresolved. Retain chronological totals without inventing an endpoint. Integrate each recovery once using nonoverlapping seasonal samples.

What instruments and logging intervals capture a defrost event?

Synchronize signed heat, active power and operating mode. RH means relative humidity. PNNL-35724, March 2024, specifies one-second power/mode records, one-minute indoor temperature/RH and five-minute outdoor temperature/RH in its field-validation plan. The faster Anvilfield specification resolves transitions. [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996)

Daikin IOD-1065B, January 2025, requires disconnecting all supplies before installation or service and warns that multiple supplies may be present. Install electrical instrumentation under that isolation requirement before restoring power for logging. [Daikin IOD-1065B (January 2025)](https://daikincomfort.com/docs/default-source/dh6te-r-32/im-iod-1065b.pdf)

True-power meters capture active energy and power factor. Verify current-transformer polarity, range, phase association and whether a parent meter already includes each branch. Nominal voltage multiplied by measured amperes does not establish inverter input energy.

Calculate total air-side heat from dry-air mass flow multiplied by supply-minus-return [enthalpy](/field-guides/glossary/enthalpy/). When moisture change is negligible, density-corrected mass flow multiplied by heat capacity and temperature rise gives sensible heat. Preserve the sign; a standard-air 1.08 factor does not correct for actual density. Use the [psychrometric chart guide](/field-guides/hvac/psychrometric-chart-reading-field-guide/) to resolve humidity ratio and enthalpy consistently. [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996) [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

For water-side heat, multiply mass flow by actual fluid heat capacity and signed leaving-minus-entering temperature. Record fluid temperature, glycol concentration and the pump/buffer boundary, as required to interpret the air-side and water-side monitoring methods reviewed by Annex 88. [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

Where to put the probes

Draw the measurement boundary around the fan, strips and outdoor-air intake. Use representative, well-mixed inlet and outlet measurements; keep temperature probes out of direct radiant view of electric elements. NREL's 2023 field study could not use strip-on thermal readings because its thermocouples were too close to the elements. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf) Return air measured before an economizer and supply air after it do not isolate coil output without accounting for ventilation. Validate airflow in every fan regime, including defrost and recovery, using the [air-balancing procedure](/field-guides/hvac/air-balancing-report-procedure/). [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996) [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

Shield outdoor sensors from sun, meltwater and discharge recirculation. Preserve ductless head airflow during capture tests. Cross-check paired sensors together and record their response times; a five-second logging interval cannot resolve a transition that the sensor takes a minute to follow.

How long should the monitoring run?

NREL's 2022 humid laboratory characterization used 8 to 12 hours per condition. Capture complete native cycles in matched regimes for event comparisons, and the declared season for seasonal results. Report valid hours, weather coverage and uncertainty. [NREL/TP-5500-80802 (February 2022)](https://docs.nlr.gov/docs/fy22osti/80802.pdf)

A dry monitoring period with zero events establishes operation in that observed weather. Record its hours and moisture conditions, then extend into frosting weather to estimate defrost losses. Keep forced service defrost separate: it tests sequence operation rather than native event frequency.

How long and how often should defrost run?

NREL's May 2023 report separates typical site-average durations from longer individual events. Judge installed operation against the applicable control sequence and coil clearance. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf)

Follow the installed model's sequence. [Daikin IOD-1065B (January 2025)](https://daikincomfort.com/docs/default-source/dh6te-r-32/im-iod-1065b.pdf) [Trane RT-SVD03N-EN (August 2022)](https://support.trane.com/hc/en-us/article_attachments/28249817280525)

A duration distribution is more useful than one average

For each 5°F dry-bulb bin, retain complete-event count, valid outdoor-unit operating hours including defrost, wall-clock hours, moisture, P25, median, P75, P90, maximum and time-limit termination count. Label wet-bulb and dry-bulb explicitly: NREL's duration plots use wet-bulb, while the 2026 PNNL paper uses dry-bulb and RH. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf) [PNNL defrost study (March 23, 2026)](https://link.springer.com/article/10.1007/s12053-026-10424-7)

PNNL's March 2026 Figure 4 summarizes distributions of site-level mean durations within weather bins. Its plotting filter requires at least ten cycles from at least five sites per bin; its boxplots do not supply pooled individual-event P90 values. [PNNL defrost study (March 23, 2026)](https://link.springer.com/article/10.1007/s12053-026-10424-7)

Anvilfield's event-reporting convention shows individual durations below ten complete events per bin. At ten or more, report linearly interpolated quantiles using sorted-data position 1 + (n - 1)p, with p equal to 0.25, 0.50, 0.75 or 0.90. This display threshold does not establish statistical sufficiency or stable tail estimates.

Count recovery separately from reverse-cycle duration

Report reversal-to-heating duration separately from recovery-to-reference duration. Following PNNL's operating-hour approach, frequency is complete events per valid outdoor-unit operating hour; prevalence is defrost hours divided by those operating hours, including reverse-cycle operation. Label wall-clock percentages separately. [PNNL defrost study (March 23, 2026)](https://link.springer.com/article/10.1007/s12053-026-10424-7)

Retain partial events in chronological energy and prevalence totals; label complete-event counts separately. Add RH or dew-point strata because cold dry air and damp air at the same temperature can produce different frosting behavior. Check coil clearance and sequence after time-limit termination. [NRCan heat-pump guide](https://natural-resources.canada.ca/energy-efficiency/energy-star/heating-cooling-heat-pump)

How do you allocate auxiliary and parasitic energy?

Count every circuit once, then allocate energy among modes. PNNL's measurement plan and NREL's field accounting distinguish component meters from mode allocations. [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996) [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf)

Add resistance output only when the measured heat excludes it; downstream supply measurements already contain strip heat. If a whole-air-handler meter includes strips, subtract the separately metered strip input to calculate heat-pump-only electricity. State whether the result is heat-pump capacity, complete air-handler delivery or occupied-space delivery; [duct leakage](/field-guides/hvac/duct-leakage-testing-sealing/) changes the last boundary.

Meter heater stages rather than deriving kWh from a [thermostat call](/field-guides/hvac/thermostat-types-installation-smart-controls/). Include pan/crankcase heaters, controls and standby on their actual circuits and timestamps within the declared seasonal boundary. [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996) [NEEP specification v4.0](https://neep.org/sites/default/files/media-files/cold_climate_air_source_heat_pump_specification_-_version_4.0_final.pdf)

What changes for dual fuel, ductless, and rooftop systems?

Dual-fuel total-energy performance requires measured fuel and a declared higher- or lower-heating-value conversion basis. NREL's Site 3 lacks measured gas, so its report does not calculate complete-system COP for that site. An electricity-only ratio omits part of the input. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf)

Trane's cited sequence enables gas heat during defrost for its covered dual-fuel packaged heat pumps. Preserve the installed controller's sequence when allocating auxiliary energy. [Trane RT-SVD03N-EN (August 2022)](https://support.trane.com/hc/en-us/article_attachments/28249817280525)

What changes for hydronic and stored-heat defrost?

A heat pump can extract water-side heat while a buffer continues delivering heat to emitters. Report heat-pump-to-buffer and buffer-to-building boundaries separately when both are instrumented. Include relevant pumps and account for stored-energy changes over the integration period. [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

The [water-source and geothermal guide](/field-guides/hvac/water-source-geothermal-heat-pump/) covers systems without the exposed outdoor air coil driving this frost test. Keep residential air-to-air, rooftop and hydronic results attached to their source populations. [NRCan heat-pump guide](https://natural-resources.canada.ca/energy-efficiency/energy-star/heating-cooling-heat-pump)

How much uncertainty can hide the defrost penalty?

Anvilfield estimates a 5.83% standard uncertainty in COP, or 11.66% expanded uncertainty at k=2, for the declared steady 10 K heating-rise budget below. NIST TN 1297, 1994 edition, sections 5 and 6 supplies the propagation method, including covariance for correlated inputs. [NIST TN 1297 (1994)](https://www.nist.gov/system/files/documents/2017/05/09/tn1297s.pdf)

Method: combine independent relative standard uncertainties by the square root of their squared sum, retaining full precision until the final result. Convert catalog error limits or calibration expanded uncertainties to standard uncertainties before using this method. Density and heat capacity are held fixed in this estimate; an installed budget also includes their uncertainty, mixing, lag, drift and covariance. Under NIST's distribution and uncertainty conditions, k=2 gives approximately 95% coverage.

This budget describes one steady heating point. Propagate signed cycle energy from the actual samples, including sensor correlation, timing and baseline uncertainty. Near zero temperature difference, report absolute heat uncertainty because relative error grows rapidly as net heat approaches zero.

Sampling cannot remove calibration bias or clock offset. Record [duct static pressure](/field-guides/hvac/duct-static-pressure/) and fan regime so airflow changes do not masquerade as heat-pump degradation.

What supports a defensible improvement claim?

Calculate uncertainty in the difference between tests, including reference uncertainty and shared-sensor covariance. Two unrelated tests with roughly 12% expanded uncertainty do not establish a measured 3% improvement merely because their displayed COPs differ.

Anvilfield's event-test rules synchronize clocks within one second, exclude partial cycles from event comparisons and retain gaps in a coverage ledger. Never fill missing defrost samples with clean-coil averages.

Worked examples: published energy and a complete cycle balance

These NREL winter 2021 to 2022 observations retain Table 11's monitoring periods and component boundaries. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf)

Using NREL's winter 2021 to 2022 data, Anvilfield estimates that Site 7's 2,266 kWh of reverse-cycle ODU plus total IDU electricity, including operating strips and excluding recovery, equaled 41.67% of its 5,438 kWh of compressor-based heating electricity, which excludes auxiliary heater energy. This electrical mode ratio is not a thermal derate. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf)

NREL's compressor-only COP uses that heating mode's samples. Its system heating-season COP estimates auxiliary-assisted heat through Equations 25 and 26 because strip-on thermal readings were unsuitable. Those heating-mode equations do not supply a measured all-mode signed seasonal ratio including defrost and standby. [NREL/TP-5500-84745 (May 2023)](https://docs.nlr.gov/docs/fy23osti/84745.pdf)

Method: Anvilfield estimates the electrical ratios from rounded published values. Representative recovery energy multiplies Table 13's statistical-mode runtime by corresponding mean power: Site 1's 6.5 minutes at 11.9 kW yields 1.29 kWh. Integrate recovery records for seasonal recovery energy.

A one-hour signed cycle

Anvilfield estimates that a 60-minute cycle with 48 minutes at 10 kW, five minutes at negative 3 kW, and seven minutes delivering 0.67 kWh averages 8.42 kW, a 15.83% capacity reduction before auxiliary heat. The thermal boundary includes blower heat and excludes resistance heat; electricity includes outdoor and indoor units but excludes separately metered strips. Separately metered resistance heat is added once in the complete-system results.

Method: the phase durations and powers are declared Anvilfield estimate inputs using NREL's signed integration principle. Recovery uses unrounded net heat of 40/7 kW and strips of 15/7 kW. The clean-coil reference stays at 10 kW for one hour; displayed cells round after integration. [NREL/TP-5500-80802 (February 2022)](https://docs.nlr.gov/docs/fy22osti/80802.pdf)

Extraction and recovery explain the capacity reduction beyond the 8.33% reversal-time fraction. Auxiliary heat raises complete delivered capacity to 9.08 kW and changes COP from 2.69 to 2.40. Net heat already includes blower heat, so adding fan watts again would overstate delivery.

How do you average COP over a heating season?

Seasonal COP equals total signed thermal kWh divided by total electrical kWh over the same time and boundary. Use elapsed sample durations or synchronized energy registers; ordinary or time-weighted COP means fail when input power varies. [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

Method: Anvilfield estimates integrate each row's thermal and electrical kW using exact 20/60-hour and 40/60-hour durations. Thermal energy totals 14/3 kWh and electricity totals 7/3 kWh, giving COP 2.00. The arithmetic COP mean is 2.30; the time-weighted mean is 2.07. Weighting COP by electricity gives 2/7 × 3.00 + 5/7 × 1.60 = 2.00.

Preserve negative heat and off-cycle input. Resolve duplicate timestamps, clock changes and meter resets; separate measured totals from gap-filled estimates.

Anvilfield's seasonal worksheet sums 18,000 kWh of delivered heat and 7,870 kWh of electrical input, including defrost, recovery, auxiliary heat and standby within its declared boundary, for an estimated seasonal COP of 2.29. Its three active-bin COPs have an arithmetic mean of 2.83, which does not represent the season.

Method: hours, loads and integrated COPs are declared estimate inputs. Each active bin's delivered heat is divided by its integrated COP; 0.06 kW of standby for 2,000 hours adds 120 kWh once. The defined heating season spans 5,100 hours. [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

Build the weighting table around weather and load

Carry local hours, delivered load, thermal/electrical energy, event count, moisture and valid-data coverage with every weather bin. Assign standby to its actual timestamps and keep each timestamp in one active or off-cycle category. Preserve low-load cycling and distinguish auxiliary-assisted performance from compressor-only performance.

Label an actual monitored-season ratio, a valid-data-only ratio and a weather-normalized estimate separately. An empty cold or damp bin remains a coverage gap: bound its contribution or extend monitoring. Temperature-only reweighting does not recover absent moisture or load regimes.

Why HSPF2 is not the COP at your building

Dividing HSPF2 in Btu/Wh by approximately 3.412 expresses its rating ratio in W/W while retaining rating weather, controls and auxiliary assumptions. Current M1 uses Region IV certification bins; a local-hour weighting worksheet describes a different climate and operating boundary. [DOE final rule (January 7, 2025)](https://www.energy.gov/sites/default/files/2024-12/cac-tp-fr.pdf) [Current Appendix M1](https://www.ecfr.gov/current/title-10/chapter-II/subchapter-D/part-430/subpart-B/appendix-Appendix%20M1%20to%20Subpart%20B%20of%20Part%20430)

What does the loss cost, and who needs the test?

Anvilfield estimates that delivering 1,000 thermal kWh at COP 2.40 instead of 3.00 adds 83.33 electrical kWh, a 25% increase. For the job's energy cost, multiply that additional electricity by the customer's marginal billed rate.

Method: divide the fixed heat delivery by each declared COP and subtract unrounded electrical totals. Demand charges require coincident metered kW and the applicable tariff. Attribute the COP change to defrost only through a matched reference.

Scope testing around access, circuits, instrument installation/removal, monitoring duration and analysis. A service visit checks the sequence; a capacity comparison needs complete matched cycles; seasonal performance needs coverage of the declared season.

Commercial winter programs compare equipment type, access and operating schedules across a territory, then use the [commercial heat-pump guide](/field-guides/hvac/heat-pumps-for-commercial-buildings/) to connect findings to retrofit planning. Winter testing routes start with the buildings: [PorchRocket Ground Control](/lists/rooftop-hvac) returns lists selected for two or more visible rooftop HVAC units, and the equipment register identifies the heat pumps to instrument.

Is the loss caused by defrost, a fault, or the test setup?

Reconcile the measurement boundary, then compare control behavior with the installed manual and actual coil clearance. [PNNL-35724 (March 2024)](https://www.osti.gov/servlets/purl/2475996) [IEA EBC Annex 88 (October 2024)](https://annex88.iea-ebc.org/Data/publications/EBC_Annex_88_Subtask_A_Report_2024.pdf)

Use the [reversing-valve and defrost guide](/field-guides/hvac/heat-pump-reversing-valve-defrost/) for persistent icing and valve/sensor diagnosis. Follow the [refrigerant charging guide](/field-guides/hvac/refrigerant-charging-subcool-superheat/) when verifying charge rather than assigning every capacity shortfall to defrost control.

Which changes actually reduce the loss?

Correct verified obstruction, drainage, airflow, sensor attachment or control-setting faults using the installed model's instructions. Fewer reversals alone do not establish improvement: a longer interval can increase frost degradation and extend the next reversal. Compare integrated heat, electricity and coil clearance at matched moisture, temperatures, airflow, demand and compressor regime. [NREL/TP-5500-80802 (February 2022)](https://docs.nlr.gov/docs/fy22osti/80802.pdf) [Daikin IOD-1065B (January 2025)](https://daikincomfort.com/docs/default-source/dh6te-r-32/im-iod-1065b.pdf) [Trane RT-SVD03N-EN (August 2022)](https://support.trane.com/hc/en-us/article_attachments/28249817280525)

After an OEM-supported change, recheck room temperature, minimum supply temperature while the fan runs, recovery and auxiliary energy. PNNL's March 2026 study of 22 prototype homes found median room-temperature drops below 1°F across analyzed strategies and temperature bins, using each event's largest drop among the home's room sensors through defrost and 15 minutes afterward. Several fan-on strategies delivered cold supply air despite those small room-temperature changes; track draft complaints separately. [PNNL defrost study (March 23, 2026)](https://link.springer.com/article/10.1007/s12053-026-10424-7)

For a sizing decision, compare matched cycle-average capacity with the building load, using the [Manual J load guide](/field-guides/hvac/cooling-heating-load-calculation-manual-j/) for residential calculation. The [balance point](/field-guides/glossary/balance-point/) is the capacity/load crossing; a low outdoor operating limit answers a separate question. Check the short-event deficit against room-temperature drift or storage as well: meeting an hourly load does not guarantee comfortable supply air throughout reversal.

Field checklist and the report to retain

Identify models and boundaries, verify acquisition, then capture complete events before calculating performance.

Report whole-cycle heat-pump capacity and complete-system COP with their respective boundaries, monitoring dates and valid coverage. Identify observed-season results and normalized estimates separately.

Review record and citation

Sources checked and review date: November 5, 2026. Byline: Anvilfield. Update log: final review clarifies rating applicability, probe placement, meter boundaries and comfort-study methods; verifies numerical examples; and removes repeated summaries.

How to cite this page: Anvilfield, How much capacity does a heat pump lose during defrost?, 2026, https://anvilfield.com/field-guides/hvac/heat-pump-defrost-seasonal-performance-test/.

Related tools

FAQ

People also ask

• How does a heat pump heat in winter?
• What does a reversing valve do on a heat pump?
• What is a Manual J load calculation?
• What is a psychrometric chart used for?
• What is air balancing?

Sources checked

• [S1] PNNL-35724 (March 2024) DOE CCHP Field Validation Data Collection and Analysis Plan, Table 3: logging intervals and measurement channels. Sources checked November 5, 2026.
• [S2] NREL/TP-5500-84745 (May 2023) Field Validation of Air-Source Heat Pumps for Cold Climates, section 5.1, Table 11, section 5.6 and Table 13: winter 2021 to 2022 energy, COP boundaries, durations and recovery. Sources checked November 5, 2026.
• [S3] NREL/TP-5500-80802 (February 2022) Performance Assessment of High Efficiency Variable Speed Air-Source Heat Pump in Cold Climate Applications, sections 3.2.2 and 3.5.3; Figure 18, printed p.33: humid tests and integrated capacity regressions. Sources checked November 5, 2026.
• [S4] NEEP specification v4.0 Cold Climate ASHP Specification Version 4.0, effective January 1, 2023: expanded-table power exclusions. Sources checked November 5, 2026.
• [S5] Daikin IOD-1065B (January 2025) DH6TE090/120 Installation and Service Reference, p.12: timing, sensor condition, long-line setting and override. Sources checked November 5, 2026.
• [S6] Trane RT-SVD03N-EN (August 2022) ReliaTel Microprocessor Controls Service and Diagnostic Support Manual, printed pp.141 to 142: initiation, termination and dual-fuel sequence. Sources checked November 5, 2026.
• [S7] NRCan heat-pump guide Heating and Cooling With a Heat Pump: moisture, control types, supplementary heat and ground-source distinction. Sources checked November 5, 2026.
• [S8] IEA EBC Annex 88 (October 2024) State of the Art, Subtask A: performance monitoring, air-side/water-side measurement and dynamic/seasonal boundaries. Sources checked November 5, 2026.
• [S9] ENERGY STAR CVP (March 2025) Cold Climate Heat Pump Controls Verification Procedure, revised March 2025: native controls, pretest and equilibrium. Sources checked November 5, 2026.
• [S10] DOE final rule (January 7, 2025) 90 FR 1224: revised M1 incorporating AHRI 210/240-2024 and separate M2 structure. Sources checked November 5, 2026.
• [S11] Current Appendix M1 10 CFR part 430, subpart B, Appendix M1: July 7, 2025 applicability, CFR precedence and Region IV certification bins. Read on eCFR; curl returned an access interstitial. Sources checked November 5, 2026.
• [S12] Annual CFR (January 1, 2025), historical M1 January 1, 2025 annual edition, historical section 3.9: termination-to-termination frost-test boundary. Historical numbering. Sources checked November 5, 2026.
• [S13] PNNL defrost study (March 23, 2026) Evaluating the impacts of defrost mode in cold climate residential heat pumps, Energy Efficiency 19, article 25: 22 prototype sites, Figure 4 site-average duration distributions and comfort findings. Sources checked November 5, 2026.
• [S14] NIST TN 1297 (1994) Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results, sections 5 and 6: propagation, covariance and coverage. Sources checked November 5, 2026.