CRI, R9 and TM-30: what to specify for decorative and hospitality lighting

At a glance

Figures read on 6 November 2026. TM-30 Annex E figures are as printed in a PNNL tutorial for the TM-30-20 edition and apply at 200 to 700 lux. None is a UK requirement for decorative lighting except the floor of 80 for light sources.

What each number measures

Colour rendering is the effect of a light on the colours of objects, judged against a reference light (CIE International Lighting Vocabulary, entry 17-22-107). Three scoring systems are in use, and a specification can ask for any of them. They answer different questions. The drawing below shows each in one picture.

• CRI, written Ra. The CIE method of 1965, corrected in 1995 as CIE 13.3 with its technical recommendations unchanged from CIE 13.2 of 1974. The US Department of Energy describes it as eight standard colour samples of moderate lightness, spaced about equally in hue, and six special samples. Each sample's score is 100 minus an adjusted colour difference against a reference light of the same colour temperature. Ra is the mean of the eight (CIE e-ILV 17-22-111). A score of 100 means colours look identical to the reference, and negative scores are possible.
• R9. The special score for the strong red sample, TCS09. R9 to R14 belong to the six special samples and are not part of Ra. The DOE fact sheet singles R9 out because the rendition of saturated red matters for skin tones, among other materials, and says an R9 above 0 is generally considered acceptable because the colour space used often exaggerates shifts in red.
• TM-30, from the Illuminating Engineering Society. It scores 99 colour samples taken from real objects (paints, textiles, inks, skin, food, flowers, foliage and plastics) and reports a fidelity index Rf (0 to 100), a gamut index Rg (100 means the average saturation matches the reference, above 100 more saturated) and a colour vector graphic for 16 hue bins. The same method gives local values per bin, such as Rcs,h1 (the change in saturation of the reds) and Rf,h1 (the fidelity of the reds). The DOE names ANSI/IES TM-30-24 as the current edition. The CIE has a technical report, CIE 224:2017, that describes a colour fidelity index Rf based on the IES one, with 99 samples; it calls it a research report and says it does not address colour quality beyond fidelity.

Why Ra alone can mislead

Ra is easy to ask for and easy to meet, which is why it is everywhere: the UK rules for light sources set a floor of 80 in it. Its limits are well documented, and they matter most for the warm, saturated colours of hospitality interiors.

• One average hides the pattern. The DOE says two sources with the same CRI can render colours very differently, to the point where a scene looks appealing under one and unappealing under the other, because CRI gives the size of the difference and not its kind.
• It penalises saturated reds. The DOE says outdated colour science often leads CRI to penalise sources that saturate reds, which many people prefer. The PNNL tutorial adds that the CIE itself acknowledges Ra deviates from a scientifically accurate fidelity measure and that Ra values do not always correlate well with what people see.
• Ra and Rf drift apart. In a data set of 1,529 spectra of real lighting products, the PNNL author found that for sources with a CRI of 80 the difference between Rf and CRI can exceed 20 points. So a step of 5 points in Ra is smaller than the gap between two sources with the same Ra.
• R9 has its own scale. The tutorial notes that the scale for R9 differs from the other special scores, because of the colour space used in CIE 13.3. A small R9 is not directly comparable with a small Ra.
• Neither system is mandatory by itself. The tutorial says IES recommends users move from CRI to Rf, that the CIE stopped short of recommending Rf replace Ra, and that neither is a mandatory standard unless a regulation or a client points to it.

What the rules and programmes ask for

No rule we read sets a colour rendering figure for decorative or hospitality lighting. What exists is a legal floor for sources, regional rules that name higher figures, and a recommended set of criteria in TM-30. The drawing puts the figures on three scales and the table says where each comes from.

TM-30 gives specification criteria in its Annex E, in three design intents (preference, vividness, fidelity) and three levels. The figures below are as printed in the PNNL tutorial for the TM-30-20 edition; they apply to scenes with many colours lit to 200 to 700 lux, and each cell must be met in full. The tutorial says the intent called preference, a pleasing and natural-looking space, could be the dominant intent in retail, office, hospitality or residential lighting, and that specifiers writing for light levels below 200 lux should build their own criteria. Many evening scenes are dimmer than that.

Where this is used: the tutorial lists the DesignLights Consortium (preference level 3 for most indoor products), ANSI/ASHRAE/USGBC/IES 189.1-2017 (a combination of P2 and F3) and the WELL Building Standard (a modified P1). Each also offered a route using CRI when the tutorial was published.

What to specify for decorative and hospitality lighting

Practice is to ask for more than the floor. The advice below is our reasoning and not a rule: vendor sheets disagree on what R9 should be, and we found no UK source that settles it. It rests on the figures above.

1. Ask for Ra and R9 together. For light seen close to people, food, timber, leather and fabric, Ra of at least [90] and R9 of at least [50] is the pair California uses for residential high-efficacy sources. It clears the UK floor of 80 with margin, and it clears the bare R9 above 0 that DOE calls acceptable. Say whether the figure is declared or measured; a tested value can sit 2.0 below a declared one.
2. Ask for TM-30 data as well. The spectral power distribution is already needed to calculate it, and the DOE points to a free PNNL online spectral calculator as the reference implementation. Ask for Rf, Rg, Rcs,h1 and the colour vector graphic. Choose a design intent, for example [P2], or write your own limits, with the red bin tightened where reds matter. State the TM-30 edition.
3. Ask for the figures where the light will be used. Dimming and tuning change a spectrum. The UK Regulations assess a source at its factory default settings (Sch 3 para 11), which tells you little about a dimmed evening scene. The DOE gives a rule of thumb of CRI 80 or more at the lowest setting of a dim-to-warm product, and says a mock-up is the best judge. Ask for [100], [50] and [10] per cent, and at the ends of any tuning range. See dim-to-warm, tunable white and RGBW.
4. Judge on the real finishes. A score is an average over test samples, not your brass or your leather. Light the approved samples with one production-standard fitting, at the evening scene and the lowest level, before the run is made. See shop drawings and mock-ups for bespoke lighting.
5. Lock the source. A change of LED module, lamp or driver can change the spectrum. Write that no change is made after approval without a new test report that is no worse.

What to look at, material by material

The table says what to look at, not how a finish will behave. No score predicts how a particular brass, lacquer or hand-blown glass will look under a given lamp, and we do not give numbers for materials. Use it to decide which measure to ask for and which sample to put on the mock-up bench.

Two cautions. The colour of the light itself is a separate matter from rendering: Ra compares a source with a reference light of the same colour temperature, so it says nothing about whether the light looks warm or greenish. For that see colour temperature by space. And light level changes what people see: see lighting levels by space.

Light sources visible in [space] to have a colour rendering index Ra not less than [90] and R9 not less than [50], at full output and at [50] and [10] per cent, as measured on the supplied model by [accredited laboratory] and not only declared. Report to show the measured values at the lowest dimming level reached. These are the specifier's figures, not a legal requirement; the UK floor for a light source is Ra 80 [where SI 2021/1095 applies].

The maker is to supply, for each light source and fitting family, the spectral power distribution file and an ANSI/IES TM-30 report [edition TM-30-24] giving Rf, Rg, Rcs,h1, Rf,h1 and the colour vector graphic, at [2700] K and at [100, 50 and 10] per cent output. Design intent [P2 of TM-30 Annex E, or: Rf not less than [ ], Rg between [ ] and [ ], Rcs,h1 between [ ] and [ ] per cent].

Approved samples of [timber, leather, brass and fabric] for [space] to be lit by one production-standard fitting at the evening scene and at the lowest dimming level, and accepted in writing by [lighting designer] before production. The mock-up to be repeated if the light source, driver, optic or tuning curve changes.

No change of LED module, lamp, driver or tuning curve after approval without a new test report showing Ra, R9, Rf, Rg and Rcs,h1 not worse than the approved values and the written approval of [lighting designer]. The approved model numbers are to be listed on the schedule and in the handover pack.

Standards and sources

• SI 2021/1095, Lighting Products Regulations 2021 (Sch 1, Sch 2 para 3 and Table 1, Sch 3, Sch 9): revised text, no outstanding effects shown; read 6 Nov 2026.
• Approved Document L, Vol 1 (2021 and 2026 editions) and Vol 2 (2026), England: Vol 1 2026: para 5.57, Table 5.7; Vol 2: para 5.65.
• CIE 13.3-1995; CIE 224:2017; CIE S 017:2020 (e-ILV): publication pages and e-ILV read; CIE 13.3 itself not read.
• ANSI/IES TM-30-24 (IES); PNNL tutorial by Royer (Leukos, 2022): Annex E figures as printed for TM-30-20; confirm against TM-30-24.
• US DOE, LED Color Characteristics fact sheet (2016): CRI, R9; dated.
• California Energy Commission, 2019 Energy Code: Residential Lighting (JA8-2019): JA8-2022 not read.
• BS EN 12464-1:2021: listed current by BSI; not read; check for any colour rendering figure for your space type.

Questions we are asked

What CRI should I specify for a hotel or restaurant?

No rule we read sets one. The legal floor for an LED or other light source in Great Britain is 80. Practice for light seen close to people, food and finishes is higher: Ra of at least 90 with R9 of at least 50 is the pair California uses for residential high-efficacy sources, and is a sound starting point. Ask also for TM-30 data and judge the real finishes on a mock-up.

Is R9 important?

Yes, where reds matter: skin, timber, leather, copper, meat and red fabric. R9 scores one saturated red and is left out of Ra, so a source can have a good Ra and a poor R9. Published thresholds vary: the US DOE calls an R9 above 0 acceptable and California asks for above 50. Ask for the figure, and for the TM-30 red-bin values, Rcs,h1 and Rf,h1, which do the same job on a clearer scale.

What is the difference between CRI and TM-30?

CRI is one average over eight pastel samples. TM-30 uses 99 samples of real objects and reports fidelity (Rf), gamut (Rg) and the shift in each of 16 hue bins, so it shows which colours change and whether they are made more or less saturated. The IES recommends a move from CRI to Rf. Neither is mandatory unless a regulation or a client names it.

Does a higher CRI use more energy?

A little, and the UK rules allow for it. The efficiency limit for a light source includes a CRI factor of (CRI + 80) / 160, which we computed as 1.06 at CRI 90 and 1.09 at 95, against 1.00 at 80. So the rule gives a source with CRI 90 about 6 per cent more power allowance than one at 80. The 2026 Approved Document L for dwellings also lists a row for sources of CRI 95 or more in its table of specialist sources.

Is CRI 95 worth it over CRI 90?

We found no source that says how much difference people see between 90 and 95. Two sources with the same Ra can differ more than that: in the PNNL data set Rf and CRI differed by more than 20 points at a CRI of 80. Ask for the TM-30 values and look at your own samples under the lamp, at the dimmed level the room will use.

Do the numbers change when the light is dimmed?

They can. Dimming and tuning change what a source emits, and the UK Regulations assess a source at its factory default settings. A US DOE page gives a rule of thumb of CRI 80 or more at the lowest setting of a dim-to-warm product and says a mock-up is the best judge. Ask for the figures at the levels you will use.