All articles
Lighting Specification

CRI vs TM-30: What the Colour Rendering Number Leaves Out

Archlior
CRI vs TM-30: What the Colour Rendering Number Leaves Out

Almost every LED datasheet gives a CRI figure, often 80, or 90 for products aimed at better projects. For many clients it is the only colour rendering number they ever see. It is also an average of eight test colours, none of them a strong red. A light source can score well on those eight and still make a red velvet chair, a terracotta tile or a walnut veneer look dull.

That is less a flaw in the products than a limit of the number. TM-30, a newer method from the Illuminating Engineering Society, shows much of what the single number hides. Knowing what each measures, and what to ask for, is one of the simplest ways to protect the materials in a scheme.

CRI vs TM-30: The Short Answer

CRI, strictly its general index Ra, is a single average of how eight test colours look under a light source compared with a reference light. TM-30 uses 99 test colours and reports fidelity (Rf), saturation (Rg) and a graphic showing which colours shift. Regulations and datasheets still use CRI; TM-30 shows the detail behind it. For most projects, specify both: a minimum Ra, a minimum R9 for strong red, and a TM-30 report.

What CRI Actually Measures

The colour rendering index is defined by the CIE in CIE 13.3. It compares how a set of test colours look under a light source with how they look under a reference light of the same colour temperature. Below 5000K the reference is a Planckian radiator, the idealised source that incandescent light approximates; above it, a model of daylight. A score of 100 means the test colours look the same under both.

The method specifies 14 test colours. The general index, Ra, which is what "CRI" on a datasheet means, is the average score for the first eight, which cover the hue circle. Six more represent strong red, yellow, green and blue, a complexion colour and foliage. They produce special indices, R9 to R14, which are not part of Ra.

The one most often singled out is R9, strong red. Because it is not in the average, a light source can have a respectable Ra and a poor R9. It can even be negative: one manufacturer's published example of a standard LED shows Ra 79 with an R9 of minus 1.4.

Why One Number Is Not Enough

An average hides where the shortfall is. Two light sources with the same Ra can render the same red fabric quite differently, because the eight colours in the average are not where they differ. Lighting manufacturer XAL makes a further point: the eight CIE test colours are not colours often seen in daily life. And because Ra measures only the size of colour differences, it cannot tell you whether a light makes colours richer or duller.

Ra 80 is the usual floor. EU ecodesign rules for light sources, Regulation (EU) 2019/2020, set a minimum of CRI 80, with exceptions, including outdoor and industrial applications where lighting standards allow less. EN 12464-1, the European standard for lighting indoor workplaces, sets a minimum of Ra 80 for typical office tasks such as writing, reading and data processing; our guide to office lighting design covers offices in more depth.

Homes, hospitality and valuable finishes usually justify more. The 2022 edition of California's building energy code set a CRI of at least 90 and an R9 of at least 50 for most light sources qualifying under its Joint Appendix 8. Our own guides recommend CRI 90 as a minimum and 95 where finishes matter, with R9 above 50, and our guide to lighting artwork at home goes further for paintings.

What TM-30 Adds

TM-30 is the Illuminating Engineering Society's method for evaluating colour rendition. It was first published in 2015, is now an American National Standard, and its current edition is ANSI/IES TM-30-24. Instead of eight test colours it uses 99, chosen from colours that appear in everyday life and spread evenly through colour space. It reports two main numbers and a picture.

Rf, fidelity. How closely, on average, the 99 colours match their appearance under the reference light, on a scale of 0 to 100. A value of 100 is an exact match.

Rg, gamut. Whether colours are, on average, more or less saturated than under the reference. Above 100 means an increase in saturation; below 100, a decrease.

The colour vector graphic. A diagram showing which hues are pushed towards more saturation and which are muted, so you can see at a glance whether reds, for instance, are rendered weakly. TM-30 also reports local values for each hue, including Rcs,h1, the change in saturation of reds, and Rf,h1, the fidelity of reds.

TM-30 also handles the reference light more smoothly. CRI switches abruptly from a Planckian reference to daylight at 5000K; TM-30 blends the two between 4001K and 4999K.

The CIE has published its own colour fidelity index, Rf, in CIE 224:2017, which shares its core calculation with TM-30. But the CIE is explicit that its Rf is not a replacement for Ra, and that it should not be used for rating and specifying products or for regulatory minimum requirements. So, for now, Ra remains the number in regulations and on datasheets, and TM-30 is what tells you what that number is hiding.

How to read a TM-30 report

A TM-30 report looks dense, but for most architectural work a tutorial from Pacific Northwest National Laboratory suggests four values are enough: Rf, Rg, Rcs,h1 and Rf,h1. Read them like this:

Rf. Closer to 100 means more accurate colour overall.

Rg. Around 100 means saturation close to the reference. Above 100, colours look richer on average; below 100, duller.

The colour vector graphic. The black circle is the reference. Where the red line runs outside the circle, the light is increasing the saturation of those hues; where it runs inside, it is reducing it. Check the red end of the circle first.

Rcs,h1 and Rf,h1. A positive Rcs,h1 means reds look more saturated than under the reference; a negative value means duller reds. Rf,h1 shows how accurately reds are rendered, which is not the same thing: two lights can have the same Rf,h1 while one makes reds richer and the other makes them duller.

CRI and TM-30 at a Glance

CRI (Ra)TM-30
Published byCIE (CIE 13.3)Illuminating Engineering Society (ANSI/IES TM-30-24)
Test colours8 in the average, 14 in total99
Main resultsRa, plus special indices R9 to R14Rf, Rg, the colour vector graphic and local values
Strong redR9, reported separatelyIncluded, and reported as Rcs,h1 and Rf,h1
Richer or duller coloursNot shownShown by Rg and the graphic
Reference lightSwitches to daylight at 5000KBlends between 4001K and 4999K
In the regulations cited hereYes, as a minimum CRINo

Accurate Is Not Always Preferred

Fidelity measures accuracy, not preference. Annex E of TM-30, added with the 2018 edition, sets out three design intents: colour preference, colour vividness and colour fidelity, each with priority levels. For preference, the top level asks for an Rf of at least 78, an Rg of at least 95, and a red saturation (Rcs,h1) between 1 per cent below and 15 per cent above the reference.

In other words, the light people prefer is not necessarily the most accurate one. A slight boost to reds falls within the preferred range; a clear loss of red saturation does not.

That has limits. Push saturation far enough and light flatters what it should show accurately. A source with high Rg and low Rf makes colours look more vivid than they really are, and our supermarket lighting guide explains why that matters when the product is food.

How to Specify Colour Rendering

What follows is design practice, built on the metrics above.

Keep Ra in the specification, but not on its own. It is the number regulations and standards use. Add R9 as a separate requirement.

Ask for the TM-30 report. Request Rf, Rg, Rcs,h1, Rf,h1 and the colour vector graphic. If a product has never been measured, that is worth knowing too.

Match the metric to the intent. Fidelity for art, colour-critical work and retail where customers judge colour, from galleries and boutiques to the clothes and hair colour covered in our guides to walk-in closet lighting and salon lighting. Preference for homes and hospitality, such as restaurants, where people want materials, food and skin to look good.

Look at the materials under the light. Numbers narrow the choice; a mock-up with the real fabrics, timbers and stone decides it. Watch for metamerism, too: two samples can match under one light and visibly differ under another, because they reflect light differently across the spectrum. Match paint, fabric and stone under the light they will live under. Our guide to dental clinic lighting shows how seriously colour-critical work takes this.

Treat colour temperature as a separate choice. CRI is measured against a reference of the same colour temperature, so high colour rendering is possible at 2700K or 4000K. Our guide to colour temperature explains the options.

Do not confuse rendering with consistency. Two fittings with identical CRI can still differ in colour from batch to batch. That is a question of binning, specified in SDCM (MacAdam steps), and our LED strip lighting guide covers it.

Expect a trade-off with efficacy. Industry guidance is that, with LEDs, the lower the CRI, typically the higher the efficacy in lumens per watt, and that enhanced phosphors raise CRI at the cost of efficacy; the US Department of Energy notes that TM-30 can help in weighing that trade-off. Spend high colour rendering where it is seen. Our guide to lumens and watts explains how light output is measured.

An example specification line, to adapt to each project: light sources Ra 90 or above with R9 of 50 or above; TM-30 report to be provided, including Rf, Rg, Rcs,h1, Rf,h1 and the colour vector graphic; colour consistency SDCM 3 or less; one colour temperature and one product family within each space.

Checklist

Specification

  • Ra stated, with a minimum suited to the project: 80 as the regulatory floor, 90 or above for homes, hospitality and valued finishes
  • R9 specified separately
  • TM-30 report requested: Rf, Rg, Rcs,h1, Rf,h1 and the colour vector graphic
  • Design intent agreed: fidelity, preference or vividness
  • Colour consistency (SDCM) specified separately from colour rendering

On site and in the studio

  • Mock-up with the real materials under the specified light
  • Finishes matched under the light they will live under, to catch metamerism
  • Colour temperature chosen separately from colour rendering
  • High colour rendering placed where it is seen, given the efficacy trade-off

Five Mistakes

  1. Specifying CRI alone. It averages eight test colours and leaves out strong red.
  2. Assuming CRI 90 means good reds. Ask for R9.
  3. Dropping Ra from the specification. Regulations and standards still use it; TM-30 adds to it rather than replacing it.
  4. Chasing the most saturated light. High Rg with low Rf makes colours look more vivid than they really are.
  5. Matching finishes under the wrong light. Samples that match in the studio can differ on site; check them under the light they will live under.

Frequently Asked Questions

What is the difference between CRI and TM-30? CRI (Ra) averages how eight test colours compare with a reference light. TM-30 uses 99 test colours and reports fidelity (Rf), saturation (Rg) and a colour vector graphic showing which hues shift. CRI is what regulations use; TM-30 shows the detail behind it.

What is a good CRI for a home? Ra 80 is the regulatory floor for most light sources in the EU. For homes, Ra 90 or above is a sensible minimum and 95 where finishes matter, with R9 above 50, especially where food, timber and skin are seen close up, as in kitchens and at bathroom mirrors. The 2022 edition of California's energy code set CRI 90 and R9 50 for most light sources qualifying under its Joint Appendix 8, which shows where better practice sits.

What is R9 and why does it matter? R9 is the CRI score for a strong red test colour. It is not part of the headline CRI figure, so a light source can have a good Ra and a poor R9, and R9 can even be negative. Reds, warm timbers and skin tones are where a weak R9 tends to show.

Is TM-30 replacing CRI? Not yet. Regulations and standards still use Ra, and the CIE says its own fidelity index, Rf, is not a replacement for Ra and should not be used for rating products or for regulatory minimums. TM-30 is best used alongside CRI, to show what the single number hides.

What do Rf and Rg mean? Rf is fidelity: how accurately the 99 TM-30 test colours are rendered on average, on a scale of 0 to 100, where 100 is an exact match. Rg is gamut: whether colours are on average more saturated (above 100) or less saturated (below 100) than under the reference.

Does a higher CRI use more energy? Typically, yes, for the same light output. Industry guidance is that, with LEDs, the lower the CRI, typically the higher the efficacy, so high colour rendering costs some efficiency. Use it where materials and people are seen.

Does colour temperature affect CRI? Not directly. CRI compares a light source with a reference of the same colour temperature, so a 2700K source and a 4000K source can both score highly. Colour temperature and colour rendering are separate choices.

What CRI do I need for artwork? Higher than for general lighting. Our guide to lighting artwork at home recommends CRI 90 as a minimum and 95 or above for work you care about, with R9 above 50, and above 80 where possible.

More on Colour and Light

Notice

This article is published by Archlior for general information and educational purposes. It is written for architects, interior designers and specifiers as an introduction to colour rendering metrics.

It is not professional lighting design advice, a product endorsement, or a statement of compliance for any project.

Regulatory figures are summarised from the sources listed and may change; check the current version of any regulation or standard that applies to your project. Manufacturer and industry statements are labelled as such. The sections on specification describe design practice, not rules.

Archlior accepts no liability for any loss, cost or damage arising from reliance on the information in this article. Readers act on it at their own discretion.

Last reviewed: October 2026.

Sources and further reading