LED light colour, CRI and experiments

The experiments on this page are in particular meant to see what is the best light colour of LEDs for bicycle headlamps. But there are some other experiments to do with how far a headlamp should be from your eyes in case of fog, and more.


1 LED colour comparison in the garden: Neutral white vs. cool white

The colour usually chosen for LED bicycle lamps is cool white because that is the type of colour LED makers can produce most light output with, i.e. this is a 'bigger is better' choice, but it's more complicated than that. Neutral or warm white give better colour rendition for those colours that one sees at the edge of the paved road, and for mountain biking this means obviously a better colour rendition on the entire 'trail'.

Here is a picture of a bit of green in the garden:
colour, daytime

About the same position, but at night, illuminated using a torch with cool white LED (F23, with Cree XP-E, LED colour WC):
colour, at night, illuminated by a cool white LED light

The same position as before, again at night, illuminated using a torch with neutral white LED (Liteflux LF2XT, with Cree XP-E, LED colour 4C):
colour, at night, illuminated by a neutral white LED light

Did you notice the purple flowers in the first picture? Did you notice them in the last picture? A few weeks ago these flowers were more blue than purple as they are now, and at that time the flowers suddenly jumped out at night when I used the cool white LED light. During the day they were inconspicuous. This means blue is accentuated much more than other colours giving a view that's quite different from what you experience in daylight. Even more so because blue doesn't appear much in nature! When you look at the above pictures, these flowers don't jump out as much as they did before (of which I unfortunately didn't make photographs), but they do appear more bluish and they stand out more. Also, you can see the picture from the cool white LED is bland; all the greens sort of look alike and the whole scene looks flat, whithout depth. Now look at the picture with the neutral white LED. You now see the greens are all different, with various bits of yellow/brown which hardly show up in cool white light. For mountain biking in particular, neutral white is much better than a cool white light. But for road riding too, neutral white gives better colour rendition of the colours you will encounter in grass on the side of the road, trees far ahead etc.

Final example of the same small flowers (but in another spot in the garden) 2 weeks later. Their colour now is near violet, and with a cool white torch, they look to be a pale blue. Using the neutral white torch, the real colour shows...
colour, at night, illuminated by a cool white LED lightcolour, at night, illuminated by a neutral white LED light

2 LED colour comparison on various road surfaces: cool, neutral and warm white LEDs

After seeing the difference that neutral and cool white gave in my garden, I began experimenting more to see what changing the LED colour brings to a bike lamp by using a few torches as bike lamps and doing other experiments with them.

Here's a ANSI white colour graph (for Cree LEDs, as those are the ones I test with):
ANSI-white colour graph
And here are the torches I experiment with, from left to right:

2.1 Colour comparison test on dry asphalt using torches

General impression: Neutral white is probably the best as it gives a more agreeable colour for opposing traffic, more depth in the objects you see on the side of the road (i.e. not all the greens becoming one green ocean, and yellow/brown colours are visible instead of almost gone as they are with cool white). For the roads themselves the colour doesn't matter that much, although warm white gives much better results under certain circumstances (road types), depending of course on colour/composition of the road surface.

2.2 Colour comparison test on wet asphalt using torches

A brief first test I did wasn't conclusive although I think the more warm the light, the better (and less annoying to oncoming traffic) it is under all circumstances (this is only true for the asphalt road itself, not the green beside it, as you can see from other experiments I did and pictures I made)

Another test done on wet asphalt, including sections where a shallow layer of water was laying on it, with the following road types: rough asphalt, new smooth dark asphalt, and the reddish variety that's often used for bicycle paths. From all the above torches I prefer the Liteflux's colour (i.e. 4C). It gives better rendition of the colours on the road than the XP-G 5B1, much better than the cool white WC one. Warm white is perhaps slightly better but not enough to offset the large drop in efficiency. So in wet conditions 4C seems an excellent choice. Pictures to be made at a later date.

The following pictures were made a lot later on 28 October 2010, the same day I made beamshots of the Frankenlamp with XP-G modification, i.e. on a wet road 2, with torches using:
(1) XP-E WC cool white, (3) XP-G R4 5B1 neutral white, (4) XP-G Q4 8B1 warm white:
Wet road 2: XP-E WC Wet road 2: XP-G 5B1 Wet road 2: XP-G 8B1
(14 M pixel images)
Note that these pictures were made as an afterthought, so I want to make a systematic set of pictures on various road surfaces at a later date.

2.3 LED light colour tests on wet asphalt on a bicycle: Neutral white (XP-G R4 5B1 in modified Lumotec a.k.a. the Frankenlamp) vs. cool white (Edelux)

Lamps used: 1. Frankenlamp, LED XP-G R4, colour 5B1 (neutral white), 2. Edelux, cool white, slightly greenish light.

2.3.1 Night ride after lots of rain (2010-9-15)

This was ca. 1 hour after it stopped raining, some bits of road in various places were already getting dry due to strong wind.

All in all having tested quite a bit the difference is clear: Neutral white is much better.

2.3.2 Night ride during rain (around midnight 2010-9-15 - 2010-9-16)

Fully wet road, cycling in the rain, lots of puddles, the superiority of neutral white over cool white as I experienced the previous night was again confirmed under these circumstances on the various road surfaces I tested on (various types/colours of asphalt)...

2.3.3 Beamshots 28 October 2010, wet asphalt

These were made on a wet road 2 (See this separate page about camera settings, lamp settings and roads). I didn't keep very strict to those settings, esp. not putting the camera 40cm behind the lamp but a bit more, also the camera aim is to a point ca. 20 m on the road. I wasn't interested in getting a perfect beamshot as the road was wet...

In this first picture you can just about make out the marker on the left at about 25 m. The reach of the beam in the centre of the road is also about 25 m (a bit more if I point it higher). It's of course not as good as an Edelux but pretty useful and much better than the original beam you get with the halogen bulb.

Beamshots on road 2, wet: Camera height: ?, camera aim: on the road ca. 20 m from the lamp.

(original 14 Mpixel)

I decided to compare a bit on the side of the road with the Edelux to see how nicely the cool white and neutral white show objects beside the road (the 2nd Edelux shot is a bit overexposed as I pointed it too low):

Beamshots on road 2, wet: Camera height: ca. 1.80 m, camera aim: ?, position: ? m behind the lamp.

(original 14 Mpixel)

(original 14 Mpixel)
Beamshots on road 2, wet: Camera height: ca. 1.80 m, camera aim: ?, position: ? m behind the lamp.

(original 14 Mpixel)

(original 14 Mpixel)

The above pictures are really not that suitable to properly compare the light colour as it would be better to have the same light beam with different LED colours but along the with the various other pictures of LED colours and what you see with them that I made with some torches/flashlights, it should give a reasonable impression of how the cool and neutral white beams differ in how you see the road and what's beside it.

3 XM-L LED colour comparison and beam shape with various lenses

LEDs that I'm experimenting with:


3.1 Pictures of the test setup

LiPo battery, h6flex, and XM-L LEDs:

3.2 Wallshots (for the light colour)

Wall shots on a very white wall. Camera: white balance = 4600K.
Light sources: LF2XT (torch), 1A, 1D, 2T, E5, E7, easywhite E7, E8 (XM-Ls are bare, no lenses):

3.3 Garden pictures with green/flowers and shots of a tile path with leaves on it

In the garden, green with flowers. Camera: white balance = 4600K.
Light sources: LF2XT (torch), 1A, 1D, 2T, E5, E7, easywhite E7, E8 (XM-Ls are bare, no lenses):

In the garden, path + stuff beside it + leaves on the path. Take note of the leaves and the brownish colours, especially the difference between the tiles and the leaves.
Light sources: LF2XT (torch), 1A, 1D, 2T, E5, E7, easywhite E7, E8 (XM-Ls are bare, no lenses):

3.4 Beamshots of the lenses with various LED colours on a road

Pictures to come soon.

4 Colour comparison tests in fog using torches

The first test was done when I didn't have my new camera (at that moment this was the ST70) with me and will perhaps be redone with pictures next time there is dense fog. The second test gave enough information to confirm the results and by studying the pictures closely, the results of the first test in dense fog shouldn't be a surprise...

4.1 Colour comparison test in dense fog using torches (2010-9-21, night/very early morning)

Warm white is superior, and cool white is very poor under these circumstances. Warm white clearly showed the objects on or beside the road, at good distance. Cool white gave a hazy view, as the reflected light is similar in colour to the fog itself (which you can see at night as well from even the little light there is at that time) which means it looked to be even foggier than it was, and even more contrast was lost than cool white does under normal circumstances (so it's hard to distinguish objects in various colours from each other or from the fog).

4.2 Colour comparison test in less dense fog using torches (2010-11-1, night)

The following pictures were made a lot later on 1 November 2010, in less dense fog than on 21 Sept., on road 1, with torches using:
(1) XP-E WC cool white, (2) XP-E Q4 4C neutral white, (3) XP-G R4 5B1 neutral white (with a wide beam), (4) XP-G Q4 8B1 warm white:

1st set: torches were held underneath the camera, close to the camera lens:
Fog on road 2: XP-E WC Fog on road 2: XP-E 4C Fog on road 2: XP-G 5B1 Fog on road 2: XP-G 8B1
(14 M pixel images)

2nd set: almost the same conditions, torches as close as possible to the camera lens, underneath the camera:
Fog on road 2: XP-E WC Fog on road 2: XP-E 4C Fog on road 2: XP-G 5B1 Fog on road 2: XP-G 8B1

3rd set: torches held at arms length from the camera (ca. 60 cm):
Fog on road 2: XP-E WC Fog on road 2: XP-E 4C Fog on road 2: XP-G 5B1 Fog on road 2: XP-G 8B1

The effects aren't as dramatic as in really dense fog, but some things are clear:

On 17 Nov. 2010 I experimented with the distance from lamp (torch) to my eyes/camera and for a torch with wide beam: At about 35 cm away there was no foggyness ('wall of light') any more. I've made some pictures with the torch at a few cm, 10cm, 20 cm and 30cm from the camera lens but they are blurry so I will make new ones soon. On a previous fog-ride with the Philips SLD dynamo lamp I put my head fairly low, as close to the lamp as I dared when riding (about 45 cm from the lamp to my eyes), and there was no problem with any wall of light phenomenon.

(pictures to be added at a later date when it mists again, the ones I made on 17 Nov. were poor as I didn't have a tripod with me)

I will do the distance test with a more powerful torch or bike lamp in the future to see if that changes anything about the required distance from lamp to eyes before there's no hinder from the lamp's beam reflecting back into your eyes. Update: I've done this with a Lupine Betty in the mist. The result is that even with much more powerful lamp, from 60cm away there is no wall of light phenomenon, but at 30cm it wasn't all that bad. This shows that the earlier results for far lower power lamps holds up. If you keep the lamp about 60cm away from your eyes you will not notice any 'wall of light'.

The following were made on road 1, the fog effects didn't show much except when aimed further, so I decided to make these more to show what the side of the road looks like in all these LED light colours:
Fog on road 1: XP-E WC Fog on road 1: XP-E 4C Fog on road 1: XP-G 5B1 Fog on road 1: XP-G 8B1
(14 M pixel images)

5 Colour rendition index (CRI)

An email about this subject made me put my thoughts about this in writing. I had already thought it out but hadn't written anything down and as I have now done it in an email, I might as well put those thoughts here too:

CRI = colour rendition index, and tells you how accurate (with a given method of adding deviations) the light is to that produced by black body radiation, where 100 = perfect, i.e. distribution of light is the same as that of a black body radiating heat/light. The CRI is, or rather can be important for accurate colour reproduction, but:

  1. I only deal with LED lamps as others don't produce enough light for the power (incandescent bulbs such as halogen), or produce horrible light (HID/Xenon).
  2. Those 2 above are also too fragile (short lifespan, HID can't do a lot of on/off cycles).
  3. CRI is similar for most LED types, about 75-80. There are some higher ones, but none of the big manufacturers seems to concentrate on this, large output comes first :)
  4. The CRI only says how accurate the colours are to a black body radiation at a given temperature (with a given way of adding/averaging deviations from the output to that of the blackbody at that given temperature) and not how accurate they are compared to sunlight (which is approximately black body radation of 6500K depending on the precise conditions)... Incandescent bulbs are of course very close to blackbody radiation as they are hot metal that radiates away light. However, incandescent bulbs glow at far lower temperatures than the equivalent of the sun which is why they make everything appear yellowish. They do give a broad continuous spectrum of light which is why you don't get the effects of in particular cool white LEDs, which show all greens as almost the same colour. So, colour reproduction is bad compared to daylight, but you can very well pick out the differences in colours (in contrast to cool white LEDs) which is why some people still use them for MTB riding.
  5. LEDs usually have a big spike in the blue right before the light output drops off and then it slowly increases in the direction of red, see this picture of the XM-L series as an example which shows more emphasis on higher wavelengths in neutral and warm white compared to cool white:

Now think about what this mean: If you use a LED with 6500K (daylight, overcast which should show all the colours as you are used to them), then it will be the awful blue light as in Xenon lamps and as in cool white LEDs. These do not work well to show the colours you see beside the road! And they do not work well because of their low CRI, too much light is just in a spike in the low wavelength area, i.e. blue.

So, you need to lower the colour temperature (neutral white LEDs, ca. 4500 K for example) to get good (or better) colour rendition of those colours you will usually encounter (green, yellow, brown, gray of the road etc.).

I just don't think taking CRI into account is useful at this point. If and when LEDs produce so much power that manufacturers will pay more attention to this it could be, or perhaps also not because you don't need high CRI if you produce the right colours for the objects/surroundings you normally see on/next to the road. So if the green/brown/yellow colours are fairly accurately illuminated as they are in daylight, then you really have what you need. And that was what I wanted to show on this page with LED experiments.

So, unless a LED has a horribly low CRI, I stand by my advice of a neutral white LED being the best choice (almost irrespective of CRI) to illuminate the road/trail such that the colours you will see are closer to how they look in daylight, and to give more distinction between the greens and browns which all just disappear in a single green colour using a cool white LED.

6 Fog: Beam with cutoff vs. circular symmetric beam

In some places mention is made that a beam with cutoff such as car headlamps have, and that all approved lamps for dynamos in Europe have, is better in fog as you will get blinded by a wall of light otherwise. I tested this in autumn 2009 in fairly heavy fog at night, by using a Schmidt Edelux and while cycling I tilted the lamp up and down many times to see what difference it makes. When aiming high a very large amount was going up into the air. The results were clear: There was almost no difference in how well I could see the road (the light scattering when the lamp was pointed up, gave more light on the road and all in all the difference between all the positions, in what one can see of the road, was very small), and I did not get blinded by light scattered by the fog. A single 200 lumen lamp is therefore in no way enough to cause this wall of light phenomenon.

Update 2010-2-6: Again in fairly heavy fog at night, I tested with a torch (flashlight) that gives the same amount of light as the Edelux (both using Cree Q5 LEDs) but of course has a symmetric (circular) beam. I then observed what might be called a wall of light. It wasn't a wall of light though, but a lot of light being reflected into your eyes from the beam just after it leaves the lamp (this gives the same problem as when a car approaches you from the opposite direction: you don't see much at all except the light from the car's headlamps). I then proceeded to try angling the lamp up and down, and when angling down the problem got naturally less as the road close to me reflected more light back to me, so even though my eyes were adjusted to the large amount of scattered light from the fog, I could see the road clearly. But of course, the road was only lit up to close distances...

The next test was to put the torch (flashlight) lower. I had kept it at chest height at first, and moving it down to handlebar height it was already much better. There was not a lot of fog-scattered light that made my eyes adjust to a higher brightness, and I did not need to angle the lamp down to be able to see the road better. Lowering the torch (flashlight) still further to fork crown mounting height (where my Edelux is mounted), it was even better, and there wasn't any influence of the scattered light on the visibility of the road. This confirms the results of my earlier test with the Edelux: It just doesn't matter what beam shape a lamp has, the 'wall of light' phenomenon is almost entirely related to the light beam's distance to your eyes (at least for a 200 lumen headlamp). This means, the best mounting position, in case of fog, is as far away from you (i.e. as low and as far forward) as possible.

In section 4.2 I have described some more experiences which confirm all this.

And finally: The tests at the end of 2010 with the Lupine Betty 2011 at full power in the fog made it clear that the amount of light of the lamp is not really an issue in the 'wall of light' phenomenon, the results of the earlier tests with the Edelux and a torch were confirmed: the distance to your eyes of a reasonably bright light source is by far the biggest component of what causes a 'wall of light'.

7 Foglamps for bicycles?

It was misty again in the night of 17 Nov. 2010 and I was up early, so I did do tests with mist lamps: 2 torches mounted near the axle of the front wheel.

Result: Doesn't add anything over a Philips SLD. The broader beam of the one on the right hand side was completely useless. The light colour (the one on the right was neutral white, the one on the left warm white) also didn't help. So, the concentrated beam of bicycle lamps with cutoff, really make it so that you can see farther ahead with them than with a standard symmetric beam, also in fog, and using a symmetric beam as foglamp low to the ground certainly doesn't help.

Perhaps an asymmetric beam would help and I will try this next time there's fog with an Edelux or Philips SLD mounted very low, near the axle of the front wheel.

8 The Frankenlamp: Experiments with a halogen bicycle headlamp modified with a neutral white LED

I've modified a Lumotec oval senso plus, which originally uses a halogen bulb with which you can hardly see anything, by putting a rectifier, smoothing capacitor of 4700µF and Cree XP-E R2 cool white LED into it. Actually, some of it is not inside it, but hanging outside for now which is why it's a Frankenlamp :)

Experiment 1: I used a standard bridge rectifier, no Schottky diodes (which have a very low voltage drop, but are probably overkill when using 1 LED), as I just wanted to test if/how well the LED modification works. At first I didn't use a smoothing capacitor which gave very flickery light at low speeds. Also for some reason the Lumotec's standlight didn't work in that case. I was unhappy with the light output, it was too unfocused, not a good illumination of the road surface. Another point that's interesting is that the vibrations in the bike from the dynamo hub started earlier (ca. 17 km/h) and I could feel them better, than when I added a smoothing capacitor. This means the electronics in a LED lamp can have an influence on vibration in the bike. The influence of the electronics on vibrations is probably also why I didn't get more vibrations with the Ktronik triple-XP-G on the dynamo hub, than with the Edelux. In particular it wouldn't surprise me if the tuning capacitors (with which one can select the power curve from the dynamo hub), have an influence on the vibrations...

Experiment 2: I then experimented with the LED connected to a power supply, and manoeuvred the LED such that it was near the focal point for the reflector, and directed such that most of the light went downwards. Projected on the wall, the lamp gives stray light and a smallish blob that will be the main beam on the road.

After adding a smoothing capacitor (1000 µF), I made another test ride and the result was much better. The flicker is a bit stronger than what one usually gets with lamps attached to a hub dynamo, so not too bad but it could be better, there's a lot of light where I want it (on the road instead of scattered all over the place) and the standlight now works...

In the first picture you see a bit of the aluminium plate onto which I glued the LED, sticking out (the plate is bent into a U-shape, and one side of that goes into the reflector housing).

The result is reasonable: I get a lot more light on the road than with the original halogen, I also get light at lower speed, but there is still too much spill light going upwards so when I ride on a street with trees beside it, they are noticeably lit up... I will try to improve on this by reverse mounting the LED or just increasing the angle that it's currently at.

Experiment 3: Instead of aluminium plate I used thin aluminium as used in some baked products you can get in the supermarket. This gives good enough cooling and allows me to deform and thus position the LED where I want it. Reverse mounting the LED (so no light from the LED leaves the front plastic directly) doesn't work well. I can't get the light focused enough.

Experiment 4: Same setup, but slight changes to the position which is now similar to that used in experiment 2, but adjusted a bit to give an as narrow slit of light on the wall as I could when using the power supply on the LED. This gives me a very good beam, about as wide as the Edelux, a bit less bright and not as much throw. Too much spill light gets into the air and I doubt it would be StVZO approved with this lighting pattern, but it's all an experiment anyway. I wanted do some tests as to how much oncoming traffic is hindered by the direct light from the LED, but I didn't get round to it.

Experiment 5: Same setup with the thin aluminium foil but here I use an XP-G R4 5B1 (neutral white). It was fiddly, couldn't get it focused as tightly as the XP-E R2 cool white (yes, I've encountered the XP-G focusing problem!) but it puts out a very nice beam which, if I didn't have the Edelux, I would be really happy with. The Edelux is brighter and has a bit better beam but I do have a much nicer neutral white illumination of the road and what's beside it, plus more near-field light. So I'd say experiment succeeded, also because I now pointed the LED downwards instead of up with which I could also get a hotspot on the wall (hotspot on the wall = light beam on the road) but all the spill light goes downwards. This also results in very good near-field illumination. And I'll say it again: I really like neutral white instead of cool white!

After this I did tests on wet roads with this lamp which confirmed the results of my torch light colour test. See earlier on this page. Btw. Cooling of the LED didn't appear to be a problem even though it's just on a small piece of thin aluminium... Perhaps because the lamp is partly open and plenty of air flows round it to cool the aluminium.

To email me go to the email page