Why It Looks Like That
Article 19

How to Choose and Adjust TV Color Temperature for D65

D65 is defined by chromaticity coordinates, not a preset name or Kelvin label. Use a documented preset as a starting point, then measure the grayscale before adjusting white balance.

By · Published · Updated

Direct answer

In brief

For BT.709 SDR and BT.2100 HDR, target D65 at CIE 1931 x = 0.3127 and y = 0.3290. A preset labeled Warm, Expert, or 6500 K is only a starting point, not proof of D65. Without measurement, use the best documented reference-oriented preset, inspect a grayscale ramp for obvious tint, and leave fine controls at reset. With measurement, correct broad errors with 2-point controls, use multipoint controls only when needed, and recheck EOTF, clipping, color, and each SDR or HDR path separately.

A television’s whites can look:

The visible tint affects every color built from the same red, green, and blue channels.

That is why white balance matters.

The reference target for BT.709 SDR and BT.2100 HDR is D65:

D65 has a correlated color temperature of approximately 6500 K. [4]

The word approximately matters.

The coordinates define the television target more precisely than the Kelvin number.

The practical answer

Without measurement:

  1. Select the reference-oriented picture mode.
  2. Reset it.
  3. Turn off variable room and energy processing for the baseline.
  4. Choose the preset documented as D65-oriented for the exact model.
  5. Inspect a grayscale ramp for obvious blue, yellow, green, magenta, or changing tint.
  6. Leave 2-point and multipoint controls at reset.
  7. Record the preset and signal mode.
  8. Use the result as a practical baseline, not a certified calibration.

With measurement:

  1. Measure several grayscale levels.
  2. Compare all candidate presets.
  3. Choose the preset with the best broad tracking and usable controls.
  4. Correct large broad errors with 2-point controls.
  5. Correct remaining localized errors with multipoint controls when necessary.
  6. Recheck EOTF, luminance, clipping, color, and uniformity.
  7. Repeat for SDR and each HDR path that stores or behaves differently.

D65 is a white-point chromaticity

Recommendation ITU-R BT.709-6 defines the assumed reference white for equal primary signals as D65:

Recommendation ITU-R BT.2100-3 uses the same D65 coordinates for HDR system colorimetry. [3]

Those x and y coordinates describe the chromaticity of reference white under the CIE 1931 system.

They do not describe luminance.

The same D65 chromaticity can be measured at:

The white point stays D65 while the brightness changes.

A standards-based card identifies D65 television reference white as CIE 1931 x equals 0.3127 and y equals 0.3290, while a separate card labels approximately 6500 kelvin as correlated color temperature shorthand. An arrow states that the coordinates define the target more completely than the Kelvin label.
D65 is defined by chromaticity coordinates; approximately 6500 K is only a correlated-temperature shorthand.

D65 is a standard illuminant

ISO/CIE 11664-2:2022 defines D65 as a standard illuminant intended to represent average daylight.

It provides the relative spectral power distribution from 300 to 830 nm at 1 nm intervals. [4][5]

A television does not have to emit the same spectrum as daylight.

Its red, green, and blue primaries combine to produce the target chromaticity.

This distinction helps explain why:

The left panel plots the official CIE standard illuminant D65 relative spectral distribution from 380 to 780 nanometres using the CIE data set. The right panel shows three conceptual red, green, and blue display-emission peaks combining to a D65 chromaticity target. A note says that a television is calibrated to the target chromaticity and is not required to reproduce the daylight spectrum.
A display targets D65 chromaticity without reproducing the daylight spectrum itself.

D65 is not exactly 6500 K

The CIE standard says D65 has a correlated color temperature of approximately 6500 K. [4]

A Kelvin value is useful shorthand.

It is not the full target.

What correlated color temperature means

CIE S 017:2020 defines correlated color temperature as the temperature of the Planckian radiator whose chromaticity is nearest to the test chromaticity under the stated colorimetric method. [6]

The definition contains the word nearest.

A test white does not have to lie exactly on the Planckian locus.

The missing green-magenta axis

Two chromaticities can have similar CCT while differing in the direction perpendicular to the Planckian locus.

One can look greener.

The other can look more magenta.

A display can report approximately 6500 K and still miss D65.

This is why a calibration report should include chromaticity, not only Kelvin.

A schematic chromaticity plot shows a Planckian locus and one iso-temperature line. Two white points lie on opposite sides of that line with similar correlated color temperature but different green or magenta displacement. D65 is marked separately at x 0.3127 and y 0.3290. The figure states that the geometry is conceptual rather than a coordinate-accurate standards chart.
Two whites can share a similar CCT while differing in chromaticity and tint.

Color temperature versus white balance

A television menu can label a control:

These are related but not identical concepts.

Color-temperature preset

A coarse preset that selects a factory white-balance table.

White point

The measured chromaticity of a selected white or gray level.

Grayscale tracking

The measured red, green, and blue balance across multiple signal levels.

White-balance controls

User controls that change the RGB balance at broad or localized parts of the grayscale.

The calibration goal is not only one white point.

It is neutral grayscale tracking across the usable range.

Choosing the preset

Start with the intended picture mode

Use the reference-oriented mode selected in Article 15.

That can be:

The correct name depends on the model.

Filmmaker Mode

The UHD Alliance public technical specification lists:

That makes Filmmaker Mode a strong candidate.

It does not establish the measured result.

Measure the actual television.

A four-stage flow starts with the Filmmaker Mode program target of D65, passes through manufacturer implementation and individual panel variation, reaches a measured grayscale result, and then allows optional measurement-based calibration. The diagram states that a program target does not certify the measured white point of a particular television.
Filmmaker Mode’s D65 requirement is a target, while actual grayscale performance remains sample and condition dependent.

Preset names cannot be universalized

A television can label its presets:

The same label can produce different chromaticity on another product.

The warmest preset can be too warm.

A middle preset can be closer to D65.

One preset can have good high-level white and poor dark-gray tracking.

Use model-specific evidence.

Three conceptual television preset cards labelled Warm 2, Expert, and 6500 K point to three different measured chromaticity locations around D65. The figure states that preset names describe manufacturer intent, not a standardized or guaranteed measured result. No brand is assigned to any label.
A preset name is a starting label, not proof of a measured D65 result.

A numerical preset

A preset labeled 6500 K states an intended CCT.

It does not guarantee:

Measure it.

Compare all plausible candidates

Do not assume the factory-selected preset is best.

Measure candidate presets at several levels.

A useful comparison includes:

Include lower levels when the meter and display are stable.

Choose the preset that:

Why D65 may look warm

Adaptation

Human vision adapts to the surrounding and prevailing white.

A viewer who has watched a cool preset can adopt that bluish white as normal.

When D65 is selected, it can initially appear:

That first impression is not a measurement.

Viewing conditions matter

CIECAM16 is explicitly viewing-condition-specific and includes a chromatic adaptation transform. [7]

Color appearance can change with:

A colored room light can make a correct display appear tinted. [10][11]

No universal adaptation time

There is no factual basis for saying every viewer adapts in:

Adaptation can occur over several time scales and can remain incomplete.

A person who moves between cool and D65 displays may never maintain one stable adopted white.

Use measurement rather than a calendar.

A practical visual adaptation method

For visual comparison:

  1. Set the intended neutral room lighting.
  2. Remove colored decorative light.
  3. Select the candidate preset.
  4. Watch varied content for several minutes.
  5. Display a grayscale ramp.
  6. Avoid rapid switching for the first judgment.
  7. Compare another preset only after documenting the first result.
  8. Return to the measured target when impressions conflict.

This improves consistency.

It does not replace a meter.

A conceptual three-stage perceptual diagram shows a viewer first adapted to a cool display, then viewing a D65 display that may initially appear warm, while colored ambient light also shifts appearance. The final panel states that measured screen chromaticity does not change merely because the observer adapts. A badge and footer identify the visual examples as conceptual simulations rather than measurements.
D65 can look warm after a cool display, and ambient white can change appearance, but the measured target remains D65.

What can be checked by eye

Obvious tint

A neutral grayscale can reveal:

A full grayscale ramp is more useful than one patch.

Nonlinear coloration

Look for:

This indicates tracking error rather than one global color-temperature offset.

Real content

Useful checks include:

Real objects are not measurement standards.

A white shirt can be:

Use them to find obvious inconsistencies, not to define D65.

What the eye cannot prove

A visual check cannot certify:

A display can look neutral after adaptation and still measure wrong.

The left panel shows one conceptual 50 percent gray patch that appears neutral. The right panel shows a ten-step grayscale ramp whose dark steps are magenta, middle steps are neutral, and bright steps shift blue. A conceptual simulation badge and footer state that the drawing is not a photograph or measurement. The takeaway says one neutral point does not prove grayscale tracking.
A neutral-looking mid-gray does not establish neutral tracking through shadows, midtones, and highlights.

Measuring grayscale

Measure several levels

Recommendation ITU-R BT.2035-0 says the 10-step grayscale should be measured to confirm that nonlinear processing is not distorting it and that there should be no visual coloration. [2]

The exact number of points in a home workflow can vary.

One point is not enough.

Record each point

For every grayscale point, record:

A Kelvin-only chart is incomplete.

Use the correct reference

The reference should state:

Without those details, a numerical error has limited meaning.

Low-level measurement

Near black can be difficult because:

Use appropriate integration and pattern methods.

Do not make large corrections from unstable readings.

High-level measurement

Near white can be affected by:

Check the result after white-level calibration.

Pattern selection

Grayscale windows

Spears & Munsil documents equal-energy windows for meter-based grayscale and gamma calibration.

The patterns keep total emitted light more constant as the center level changes. [9]

That can reduce certain display power and iris interactions.

It does not make every television immune to temporal or local processing.

Three conceptual equal-energy grayscale patterns show a dark center, middle-gray center, and bright center. Concentric surrounding bands change in the opposite direction so total pattern energy remains approximately constant across the sequence. The figure states that this is a vendor pattern design for meter-based work and does not eliminate every display interaction.
Equal-energy patterns vary the measured center while compensating the surround to hold total emitted energy more constant.

Window size

Choose a documented pattern size.

A very small window can interact with:

A full field can interact with:

Use the selected calibration workflow consistently.

Signal path

Confirm:

A correct file can be changed before it reaches the panel.

Pattern sequence

Record:

A display can respond differently after bright or static patterns.

Measurement instruments

Tristimulus colorimeter

A colorimeter measures light through filtered channels designed to approximate color-matching functions.

Advantages can include:

Limitations can include:

Spectroradiometer

A spectroradiometer measures the spectrum over wavelength intervals.

Advantages can include:

Limitations can include:

No instrument is perfect.

Profiling

A common professional workflow profiles a colorimeter to a spectroradiometer on the actual display.

The resulting correction is:

Do not copy another person’s meter profile.

Instrument disagreement

BT.2035 warns that commercial measurement instruments may produce different results because they use different sensing technologies. [2]

It also discusses matching differences across display technologies.

A report should not hide meter uncertainty.

Observer metamerism

Two displays can measure to the same standard-observer coordinates and look different to an individual observer because their spectra differ.

This is especially relevant when comparing:

The standard target remains D65.

A perceptual match between unlike displays is a separate workflow and must be labeled.

A measurement workflow runs from a verified pattern generator to the display, then to a spectroradiometer used as a spectral reference, a colorimeter profiled on that display, calibration software, and a verification report. Each stage lists required records such as signal format, pattern size, meter identity, profile, mode, firmware, room, and date. The diagram states that profiling is meter, display, and condition specific.
A defensible calibration records the generator, display state, spectral reference, colorimeter profile, software target, and final verification.

Two-point white balance

What it does

A 2-point system usually offers broad RGB adjustment in:

Common labels include:

The exact affected range is model-specific.

Start with measurement

Measure the reset mode.

Identify:

Do not move all controls by eye.

Protect luminance

Changing one channel changes luminance.

A common practical method is to reduce excess channels rather than raising every deficient channel, especially near white where clipping is possible.

That is not a universal rule.

The correct direction depends on:

Measure after every change.

High-region adjustment

High controls can affect:

After adjustment, recheck:

Low-region adjustment

Low controls can affect:

After adjustment, recheck:

Iterate

High and low controls often interact.

A safe sequence is:

  1. adjust broad high error
  2. adjust broad low error
  3. remeasure the full range
  4. repeat with small changes
  5. stop when improvements become unstable or cause collateral error
A conceptual graph shows low-region and high-region two-point white-balance controls with broad overlapping influence across the grayscale rather than two isolated code values. A full-range RGB response curve runs through both regions. Notes state that the exact control ranges are model specific and that adjusting either region can change luminance, clipping, EOTF, and neighboring levels.
Two-point controls should be treated as broad, interacting correction regions that require full-range remeasurement.

Multipoint white balance

Use it only when needed

Multipoint controls can correct remaining localized errors after 2-point adjustment.

They are not automatically superior.

A smooth small error can be less objectionable than a jagged corrected curve.

Point labels can be misleading

A control labeled 10 percent may not affect only a 10 percent signal.

Its effective position can shift with:

Map the controls through measurement.

Use small changes

Large corrections can create:

Use the smallest change that provides a repeatable improvement.

Recheck neighboring levels

After adjusting one point, measure:

Do not assume the control is local.

Check ramps

A numerical chart can improve while a smooth ramp develops visible bands.

Use:

Pattern and real-content checks serve different purposes.

A conceptual multipoint graph compares a smooth grayscale-error curve with a jagged correction created by moving one menu point too far. The selected point also shifts its neighboring points. A banding warning card states that a better numerical chart can still create visible contouring, so adjacent levels, the full ramp, EOTF, and real content must be checked.
Use small multipoint corrections and verify neighboring levels, the full grayscale, EOTF, and visible ramps.

Adjustment order

A practical order is:

  1. Room and viewing condition
  2. Picture mode
  3. Optional processing
  4. Signal range
  5. Black level
  6. White level and SDR luminance or HDR mode
  7. Best color-temperature preset
  8. 2-point white balance
  9. Full grayscale and EOTF
  10. Multipoint white balance if required
  11. Full grayscale and EOTF again
  12. Color-management calibration
  13. Verification

The controls can interact.

Return to earlier checks when necessary.

A conceptual flow shows an RGB white-balance correction improving chromaticity at a grayscale point while also changing measured luminance. A before-and-after EOTF plot then shows the luminance curve moving. The final step calls for remeasurement of grayscale, EOTF, black and white level, channel clipping, and color. The plot is explicitly conceptual rather than a measured television result.
White-balance corrections must be verified against luminance tracking, clipping, black and white level, and color.

SDR calibration

Target

For BT.709 SDR:

Use the selected SDR reference-white luminance and EOTF.

Check the whole range

Measure from near black through reference white.

Avoid unstable levels.

Include enough points to reveal:

Day and night modes

A brighter SDR mode can retain D65.

Changing luminance does not require changing white chromaticity.

The room can still change perceived white.

Measure both modes when they store separate tables.

PQ HDR calibration

Target chromaticity

BT.2100 uses D65:

Pattern behavior

PQ measurements can be affected by:

Use the HDR workflow established in Articles 13 and 18.

White balance and PQ tracking

Correcting RGB balance can alter luminance.

Measure both:

Do not improve one while ignoring the other.

High-level controls

High white-balance adjustment can change:

Verify all of them.

HLG calibration

Target chromaticity

HLG also uses D65 under BT.2100.

Relative system

HLG output depends on:

Measure grayscale and luminance within the selected HLG system configuration.

Do not use PQ code-to-luminance expectations.

Internal apps and inputs

A television can store white balance:

Test storage behavior before doing extensive work.

A calibration can be lost or copied unintentionally when:

Record the exact state.

A television base-settings card branches to SDR, HDR10, HDR10 Plus, HLG, Game HDR, and internal-application cards. Each branch is labelled potentially separate or shared, depending on the model. A warning says that one measured white-balance table must not be copied across signal states without verification.
Verify SDR, each HDR format, Game Mode, and internal applications separately before assuming one table applies.

Uniformity and viewing angle

Spatial uniformity

Display a neutral gray at several levels.

Inspect:

A center white-balance calibration cannot remove spatial variation.

Viewing angle

Color can change off axis.

Measure and view from the primary seat.

If several seats are important, document the compromise.

Meter position

Record:

A different position can produce a different result.

A conceptual television uniformity map shows a center measurement at D65, a cooler upper-right corner, a warmer left edge, and a slightly green lower region. The primary center measurement position and viewing axis are marked. A note states that global white-balance controls can correct the measured region but cannot remove spatial tint variation across the panel.
Global controls can improve the measured region but cannot remove spatial tint nonuniformity across the screen.

A no-meter workflow

A reader without measurement equipment can make a safe improvement.

Step 1: Use the reference-oriented mode

Select the best documented film or reference mode.

Step 2: Reset it

Remove unknown prior changes.

Step 3: Stabilize the room

Use:

Step 4: Identify the intended preset

Use the exact manual or a reliable measurement review for the exact model and size.

Look for wording such as:

Do not infer from the word alone.

Step 5: Inspect a grayscale ramp

Look for:

Step 6: Compare only plausible presets

Avoid Vivid or cool presets when the goal is D65.

Do not assume the warmest is best.

Choose the candidate with the most neutral and consistent visual grayscale after stable viewing.

Step 7: Stop

Leave:

at reset.

Call the result:

Do not call it calibrated D65.

A measured workflow

Step 1: Document the system

Record:

Step 2: Verify the pattern path

Confirm that the generator produces the intended:

Step 3: Warm and stabilize

Follow the display, meter, and workflow documentation.

Do not publish one universal time.

Step 4: Measure all candidate presets

Measure a representative grayscale in each.

Choose the best starting preset.

Step 5: Set broad white balance

Use 2-point controls with small changes.

Protect:

Step 6: Remeasure

Do not begin multipoint correction until broad tracking is stable.

Step 7: Use multipoint only when justified

Correct localized repeatable errors.

Check neighboring points and ramps.

Step 8: Verify

Measure:

Step 9: Repeat for other signal paths

Verify SDR, HDR10, HDR10+, HLG, Game Mode, and apps as applicable.

Two parallel workflows compare a no-meter setup with a measured calibration. The no-meter path uses a reference-oriented mode, a documented D65-oriented preset, a grayscale-ramp check for obvious tint, and leaves fine controls at reset. The measured path verifies the generator, measures multiple grayscale levels, uses broad two-point correction before optional multipoint correction, and verifies EOTF, clipping, color, and uniformity. A footer reserves the word calibrated for a measured and verified result.
A documented preset and visual ramp check can form a safe baseline, while calibration requires measurement, adjustment, and verification.

Troubleshooting

The image looks yellow after selecting the reference preset

Possible causes include:

Use a stable room and measurement.

Do not change the preset from one first impression.

The image looks blue

Possible causes include:

Measure x and y.

Do not diagnose only by Kelvin.

Gray looks green

Possible causes include:

A warmer or cooler preset may not solve it.

Gray looks magenta

Possible causes include:

Measure several levels.

Highlights are neutral but shadows are tinted

Check:

Do not change high Gain first.

Shadows are neutral but highlights are tinted

Check:

One gray field is correct but the ramp changes color

That is grayscale-tracking error.

Use several measured points.

A single white-point result is insufficient.

Measurements change between runs

Check:

Do not calibrate unstable data.

Meter and vision disagree

Check:

Keep the standards-based result and label any perceptual matching separately.

SDR is correct but HDR is tinted

Check:

Do not copy SDR controls automatically.

A firmware update changed the result

Remeasure.

Do not assume the old calibration remains valid.

What correct setup looks like

By measurement

A well-set mode should show:

By eye

A practical unmeasured baseline should show:

Visual success is not numerical certification.

What calibration can fix

Calibration can reduce:

within the control resolution and panel capability.

What calibration cannot fix

Calibration cannot:

A white-balance report should include:

Where this leaves us

The practical target is simple:

The calibration process is not simple enough to reduce to:

Preset names are not standards.

Kelvin alone is not chromaticity.

Visual adaptation is not measurement.

For a no-meter setup:

For calibration:

The next article covers gamma and EOTF.

White balance determines whether the grayscale is neutral.

Gamma or EOTF determines how luminance is distributed through that grayscale.

Sources and standards

Primary ITU and CIE colorimetry standards, Filmmaker Mode specifications, and calibration-pattern guidance used in this article. Last verified July 31, 2026.

  1. International Telecommunication Union Radiocommunication Sector. Parameter Values for the HDTV Standards for Production and International Programme Exchange. Recommendation ITU-R BT.709-6, June 2015.
    https://www.itu.int/dms_pubrec/itu-r/rec/bt/r-rec-bt.709-6-201506-i%21%21pdf-e.pdf

    ↩↩↩

  2. International Telecommunication Union Radiocommunication Sector. A Reference Viewing Environment for Evaluation of HDTV Program Material or Completed Programmes. Recommendation ITU-R BT.2035-0, July 2013.
    https://www.itu.int/dms_pubrec/itu-r/rec/bt/R-REC-BT.2035-0-201307-I%21%21PDF-E.pdf

    ↩↩

  3. International Telecommunication Union Radiocommunication Sector. Image Parameter Values for High Dynamic Range Television for Use in Production and International Programme Exchange. Recommendation ITU-R BT.2100-3, February 2025.
    https://www.itu.int/dms_pubrec/itu-r/rec/bt/R-REC-BT.2100-3-202502-I%21%21PDF-E.pdf

    ↩↩↩

  4. International Commission on Illumination. Colorimetry - Part 2: CIE Standard Illuminants. ISO/CIE 11664-2:2022(E), 2022.
    https://cie.co.at/publications/colorimetry-part-2-cie-standard-illuminants-0

    ↩↩↩

  5. International Commission on Illumination. CIE Standard Illuminant D65 Data Set. Based on ISO/CIE 11664-2:2022.
    https://cie.co.at/datatable/cie-standard-illuminant-d65

    ↩

  6. International Commission on Illumination. International Lighting Vocabulary: Correlated Colour Temperature. CIE S 017:2020, entry 258.
    https://cie.co.at/eilv/258

    ↩

  7. International Commission on Illumination. The CIE 2016 Colour Appearance Model for Colour Management Systems: CIECAM16. CIE 248:2022.
    https://cie.co.at/publications/cie-2016-colour-appearance-model-colour-management-systems-ciecam16

    ↩

  8. Ultra High Definition Alliance. Filmmaker Mode Technical Specifications. Official About page, reviewed July 31, 2026.
    https://filmmakermode.com/about/

    ↩

  9. Spears & Munsil. Pattern Help Text, 2nd Edition. Official calibration documentation, reviewed July 31, 2026.
    https://spearsandmunsil.com/portfolio-item/pattern-help-text/

    ↩

  10. International Telecommunication Union Radiocommunication Sector. Television Colorimetry Elements. Report ITU-R BT.2380-2, 2018.
    https://www.itu.int/dms_pub/itu-r/opb/rep/R-REP-BT.2380-2-2018-PDF-E.pdf

    ↩

  11. International Commission on Illumination. Ambient White Point, of a Display. CIE S 017:2020, entry 17-32-034.
    https://cie.co.at/eilvterm/17-32-034

    ↩
Related reading: Why Accurate D65 White Balance Often Looks Yellow Understand why a D65-aligned white balance can initially look yellow or warm. Next step: Gamma and BT.1886 Continue from D65 and neutral grayscale into BT.1886, SDR EOTF measurement, room adaptations, gamma patterns, and separate PQ and HLG verification.