What Is CRI in Lighting? Ra, R9, and TM-30 Explained

Every desk lamp and monitor light bar listing carries a CRI figure, usually 90 or better, and almost none of them say what the number measures. So here is what is CRI in lighting explained from the standard that defines it rather than from a product page. CRI is a score, topping out at 100, for how closely a lamp reproduces the colours of eight pastel test samples compared with a reference source of the same colour temperature. It leaves saturated red out entirely, and it is easy to claim and hard to check. This guide covers how the number is built, what 80 and 90 mean, and what to ask for instead.

What CRI measures, and how the number is built

The method is CIE 13.3-1995, which the International Commission on Illumination calls the Test-colour Method: light a set of test-colour samples with the lamp under test, light the same samples with a reference illuminant, and measure how far each colour shifts. Eight samples feed the General Colour Rendering Index, written Ra. The CIE specifies that they “cover the hue circle, are moderate in saturation, and are approximately the same in lightness.” A DOE fact sheet on CRI and LEDs calls them pastels.

The reference changes with colour temperature. Per the same fact sheet, a lamp below 5000 K is scored against a black-body radiator, roughly an incandescent bulb, and a lamp above 5000 K against a defined daylight spectrum. The average shift across the eight samples is subtracted from 100, so a perfect match scores 100. Incandescent scores at or near 100 by definition.

Two consequences follow. CRI is relative: a 2700 K lamp at Ra 95 and a 6500 K lamp at Ra 95 are faithful to different references, so the number says nothing about how warm or cool the light is. And CRI averages the easy colours. The CIE supplies six further samples, “strong red, yellow, green, blue, complexion and foliage colours,” numbered R9 to R14, for Special Colour Rendering Indices. They never enter Ra.

Why R9 matters more than the headline number

R9 is the strong red sample, and it separates a genuinely good LED from one tuned to pass the test. The Department of Energy’s LED Basics page is blunt: “CRI is far from a perfect metric and is especially poor at predicting the fidelity of saturated reds, for which the supplemental value R9 is often used.” A phosphor-converted white LED can render the eight pastels well enough for Ra in the 80s or even 90 and still be thin in deep red. Because R9 sits outside the average, the headline figure never shows the gap.

Regulators noticed. Section 9.7 of the EPA’s ENERGY STAR Lamps specification, version 2.1 required a solid-state lamp to have “a color rendering index (Ra) ≥ 80” and “an R9 > 0,” calculated per CIE 13.3-1995 from IES LM-79-08 measurements, with the spectral power distribution reported from 380 nm to 780 nm at 5 nm steps or finer. That is the check a buyer wants: the red sample and the spectrum behind it.

Past tense is deliberate. The EPA’s sunset memo ended those specifications effective December 31, 2024, so nothing external now forces a vendor to pair Ra with R9. The pairing survives as a rule you apply yourself.

At a desk the red sample is not academic. Skin tones on a video call, a printed proof, red-coded wires and resistor bands all sit in the part of the gamut Ra ignores.

What 80 and 90 mean, and why two CRI 80 lamps can look nothing alike

LED Basics states that “a minimum CRI of 80 is recommended for interior lighting” and that “CRI of 90 or higher indicates excellent color fidelity,” with cost and efficiency trade-offs as fidelity climbs. Ra 80 is fine for screen-and-paper work; Ra 90 with a published R9 is the floor for colour-sensitive work.

The problem is what a threshold hides. In a 2017 LEUKOS paper, Comparing Measures of Average Color Fidelity, Michael Royer of Pacific Northwest National Laboratory compared CIE Ra with the newer IES Rf across a large set of real, experimental and theoretical spectra. The abstract reports “a range of IES Rf values of at least 50 to 86 for SPDs having a CIE Ra value of 80.” Two lamps can both print “CRI 80” on the box while one renders colour far more faithfully than the other. DOE’s summary of the study, covering nearly 5,000 light sources, adds that CRI “more strongly penalizes increases in the vibrancy of reds, due to the outdated models of vision and color that underlie the calculation.”

Raising the bar does not fix the metric. Royer’s TM-30 guidance in LD+A notes that a CRI 90 floor “may filter out some unacceptable light sources, but also filters out many highly desirable light sources.” A lamp that renders reds more vividly than incandescent is marked down for the deviation.

TM-30 and CIE Rf: what replaced the eight samples

IES TM-30, first published in 2015, scores a lamp against 99 colour samples and reports two headline values. Rf is a fidelity index on the familiar 0 to 100 scale. Rg is a gamut index centred on 100: above it the lamp increases average saturation relative to the reference, below it colours flatten. LED Basics says the pair “can provide a more comprehensive evaluation of color rendering.”

The CIE’s own 99-sample index, CIE 224:2017, carries a caveat worth reading before assuming CRI is dead: CIE Rf “is therefore not a replacement of the general colour rendering index, Ra, neither for the purpose of rating and specification of products nor for regulatory or other minimum performance requirements.” CRI remains the regulatory number; TM-30 is the fuller description. A manufacturer who publishes Rf, Rg and R9 for a desk lamp is volunteering the harder figures, which is the strongest positive signal a spec sheet can send.

What CRI cannot tell you

CRI says nothing about how much light reaches the desk, what colour temperature it is, whether the driver flickers at low dim, or whether the fixture throws glare into the screen. Those decide comfort over a working day and are covered in desk lighting for screen work. Where the lamp sits relative to the monitor matters as much as what it emits; Home Desk Guide’s lighting and monitor placement for long sessions covers that side.

How to read a desk light listing

• Ra 80 or better for general work. Ra 90 or better, with R9 published, for colour work.
• “CRI 97,” “sunlike” and “full spectrum” with no R9 and no spectral plot are claims, not specifications.
• A published spectral power distribution, an LM-79 report, or TM-30 Rf and Rg values each count for more than a higher Ra without them.
• Compare CRI only between lamps of similar colour temperature; the reference changes at 5000 K.
• Do not pay a premium for Ra 95 over Ra 90 for ordinary work. The visible difference is in red rendering, which Ra does not measure.

The monitor light bar buying guide applies these rules to the bars that publish real numbers.