630nm vs 670nm: what's the difference for sleep?

If you have scrolled a night light product page at half past eleven at night, you will have seen the numbers. 620nm. 630nm. 660nm. 670nm. True red. Deep red. Sometimes all of them on the same listing.

And you will have had the thought any sensible person has: is this a real difference, or am I being charged extra for a decimal place?

The honest answer is more interesting than either side of the marketing admits. The number does mean something. It just does not always mean what it is being used to sell.

Red light is asked to do two different jobs

Nearly all the confusion here comes down to one thing. The phrase "red light" describes two separate bodies of research that have little to do with each other, and a lot of product copy quietly blends them into one.

Job one: staying out of your melatonin's way

This is the job you care about at 3am. You want to see the cot, the nappy, the stairs, without your body deciding it is morning.

That decision is made largely by a photoreceptor system in the retina that has nothing to do with seeing. Two landmark studies from 2001, one from George Brainard's group in Philadelphia^[1] and one from Kavita Thapan and colleagues at Surrey^[2], built the first human action spectra for light-induced melatonin suppression. Both found the same surprising thing: the most potent region sat around 446 to 477nm^[1], with the best single fit at roughly 459nm^[2]. That is blue, nowhere near the sensitivity of ordinary vision, and it pointed to a photopigment nobody had accounted for. We now call it melanopsin, and the standardised melanopic sensitivity curve used in lighting peaks at around 480nm^[3].

What matters here is the shape of that curve, not just its peak. Melanopic sensitivity does not taper off politely towards the red end. It falls off a cliff.

That cliff is the whole reason a red light works at 3am. At 630nm, melanopic sensitivity is already down to roughly one part in 2,400 of its peak. You get light you can see by while delivering almost nothing to the receptor that would otherwise announce the morning.

Which raises the obvious question, and the one you came here for. If both reds are already flat on the floor of that chart, can the difference between them possibly matter?

It matters more than the chart above suggests

The trouble with Figure 1 is that a linear scale flattens everything below about one per cent into a single line. Plot the same curve logarithmically across just the red region and the picture changes completely.

The international standard for this, CIE S 026, publishes a melanopic weighting for every wavelength. At 630nm that weighting is 4.2 x 10^-4. At 670nm it is 2.0 x 10^-5. Deep red at 670nm carries roughly twenty-one times less melanopic weight than 630nm does.

Twenty-one times is not a rounding error and it is not a marketing number. It is arithmetic on the same standard that lighting engineers, sleep researchers and building certification schemes all work from^[3]. Watt for watt, 670nm hands the melatonin system about a twentieth of what 630nm hands it.

Two caveats we would rather give you ourselves. Both figures sit far out on the tail of the standard, in a region derived from a fitted model rather than measured directly in people, so treat twenty-one as an order of magnitude and not a decimal place. And this is weighting per unit of light, which means a bright 670nm lamp can still out-stimulate a dim 630nm one. The wavelength buys you headroom. It does not buy you permission to turn it up.

But the direction of the effect is not in dispute, and it does not wait on anyone running a trial. Further down the curve is quieter. 670nm is further down the curve.

You can see this in the experiments that went looking for a red effect. John Hanifin's team pushed monochromatic light at 460, 630 and 700nm at equal and very high photon densities, and found significant melatonin suppression at 460nm but only small reductions at the two red wavelengths^[4]. Helen Wright's group compared 595nm and 660nm against blue, blue-green and green: the long wavelengths produced no significant phase advance, while the short ones shifted the melatonin rhythm by roughly 40 to 65 minutes^[5]. And in a small study of 17 people, 630nm red light at night improved reaction time and reduced EEG markers of drowsiness, with only the white light condition significantly suppressing melatonin^[6].

Every one of those experiments points the same way: the longer the wavelength, the less the melatonin system notices it. 670nm is simply the furthest useful point along that line while still being a colour you can comfortably see a nappy by. That is the case for it, and it is a mechanism case rather than a marketing one.

Job two: photobiomodulation

This is a different field entirely, and it is where 670nm earned the rest of its reputation. Photobiomodulation studies what long-wavelength light does to mitochondria, the parts of your cells that make energy. The wavelengths that keep appearing in that literature sit roughly between 650 and 900nm, and the deep-red band is where almost all of the work has been done.

That concentration is worth noticing for a practical reason rather than a therapeutic one. When laboratories chose a red wavelength to build instruments and studies around, they overwhelmingly chose the deep end. Components, binning and supply followed. A manufacturer specifying 670nm today is reaching for a genuine deep-red part from that established band, not a general-purpose red.

Glen Jeffery's group at UCL has produced much of the human work. In one study, people aged 28 to 72 were tested on colour contrast and rod thresholds before and after 670nm exposure. Those under about 40 showed no benefit; those over 40 improved significantly on the blue colour axis and in rod thresholds^[7]. A follow-up found a single three-minute exposure improved cone-mediated colour contrast detection for about a week in people aged 37 to 70, though the timing of delivery mattered^[8]. A later study from the same group measured brain ATP synthase flux in older adults and reported an increase after 670nm exposure, while stating plainly that wider studies are needed and that cognitive benefits have not been shown^[9].

These are interesting findings. They are also small, early stage, and about ageing eyes and cellular energy, not sleep. The field has clear negatives too: a randomised trial of 135 adults using a 670nm eye patch twice daily for four months found no benefit over placebo for the eye condition it tested^[10].

Worth separating the two arguments, because most red light marketing runs them together. The mitochondrial research at 670nm is about cells, in specific populations, at specific doses. It is not a sleep claim and we will not use it as one. The sleep case for 670nm is simpler and stronger, and it lives in Figure 2.

630nm against 670nm, side by side

The thing that matters far more than the number

Two factors decide whether a night light costs you your sleep, and neither is the headline wavelength.

The first is whether there is any blue in there at all. This is the real consumer problem in this category. Some products sold as "red night lights" are not red LEDs at all, but ordinary white LEDs behind a red-tinted cover, so it is worth checking which you are buying. A white LED is a blue LED with a yellow phosphor on top, so its spectrum has a hard blue spike around 450nm. A red filter reduces that spike. It does not remove it. You end up with something that looks red while still delivering the wavelengths melanopsin is most sensitive to^[1][2][3].

A true red emitter is different in kind, not in shade. It makes light at the wavelength you want and produces essentially nothing in the blue, because there is no blue source to filter.

The second is how bright it is. This one is badly underrated. Joshua Gooley's team found ordinary room light under 200 lux in the hours before bed delayed melatonin onset in 99% of participants and shortened melatonin duration by around 90 minutes^[11]. Andrew Phillips and colleagues mapped individual sensitivity: the group needed only about 25 lux to halve melatonin output, and individuals ranged from 6 lux to 350 lux, a spread of more than fiftyfold^[12]. Children may be more sensitive still. A small study of ten preschoolers found around 88% suppression after an hour of bright evening light, with levels still depressed nearly an hour afterwards^[13].

The international expert consensus on indoor lighting recommends keeping the sleep environment below the equivalent of 1 melanopic lux, and under 10 in the three hours before bed^[14]. A blazing red lamp can still breach that. A dim one is very unlikely to.

What to actually check before you buy

1. Is it a real red emitter, or a filter? If the listing says "red light" but never names a wavelength, assume it is a tinted white LED until proven otherwise.
2. How low does it go? The lowest setting matters more than the highest. You want to be able to get it genuinely dim.
3. Does it flicker? Cheap dimming works by switching the LED rapidly on and off. Some people find it uncomfortable, particularly when tired.
4. Where does the light land? A light at floor level pointed away from your eyes beats a brighter one shining straight at you.
5. Does the wavelength claim come with a health claim attached? A brand quoting 670nm because it is further down the melanopic curve is telling you something useful. A brand quoting 670nm and then citing mitochondrial research at you is borrowing credibility from a field that was studying something else.

Where we land

So: is the number on the box a real difference, or a decimal place you are being charged for?

It is a real difference, and Figure 2 is the reason. Twenty-one times less melanopic weight for the same light is not a marketing flourish, it is what the international standard says when you read the two rows off the table. If you are choosing a light whose entire purpose is to stay out of your melatonin's way, choosing the wavelength that is an order of magnitude further out of its way is the obvious call.

Where we will not go is the next step, the one a lot of brands take: pointing at mitochondrial research and implying it says something about your 3am. It does not, and we would rather tell you that than borrow it.

This is why the Driftback Red Light Sleep Lamp is built around a true 670nm emitter and a warm 1700K amber, rather than a red tint over a white LED. Not because 670nm is magic, but because it is the quiet end of the curve, and because a genuine deep-red part is the one thing in this category you cannot fake.

Get that right, then get it dim and point it away from your face. That combination is as far from a lamp at 3am as this gets.