Posted by Marcel van der Steen in Explanation 16 Comments»

lli_bv-high-bay-industrial-led-fixt-48-pcs-cree_s_and_p_spectra_at_1m_distanceThis article explains about the S/P ratio. At very low ambient light levels this ratio indicates the maximum gain in efficiency one can obtain just by using different receptors in the human eye.
The teaser image shows the spectrum of the light of a streetlight, and also the two human eye sensitivity curves; the daytime (red) curve lower in sensitivity than the nighttime (black).

Rods and cones in the eye

The human eye makes use of a.o. rod and cones. Rods are different than cones, and also more sensitive. Some characteristics of both:

parameter Rods Cones
Location (roughly) Situated in the eye’s retina Situated in the eye’s retina
Location (more precisely) Around the fovea, the remainder of the retina. Mainly, and in high numbers, inside the fovea. It is there where most details are perceived.
Number 120 million 6-7 million
Optimal light level (qualitative) Low (night, moonlight, twilight) High (daytime)
Optimal light level (quantitative) < 0.001 cd/m2 > 0.6 cd/m2
Sensitivity (qualitative) High, bad for color perception, no red, less details visible, good for diffuse light Low (lots of light needed), perfect for colors, details well visible, good for directly incident light
Sensitivity (quantitative) 1700 lm/W 683 lm/W
Speed (qualitative) fast slower (20 ms slower with comparable light levels)
Overview eye rods_cones_eyeClick on the image for a zoom. Coming from this site. The image well indicates were the retina is and the position of the rods and cones.
Rods and cones rods_conesTwo cones amid a number of rods.

The rods are used primarily with low ambient light levels; when there is a lot of light they do not function. So when the light application is about generating light in low light ambient conditions (night time, twilight, moonlight) then one can make optimal usage of rods.

The eye sensitivity curves

As indicated there is a difference between the rods and the cones. The eye (spectral) sensitivity curves indicate how sensitive both are for the different colors (or, wavelengths) of light.


(Spectral) sensitivity curves of the human eye

It is clear that rods (under photopic light conditions) are sensitive for blue-green light and not for red light. Their sensitivity is clearly higher, maximally 1700 lm/W and for cones this maximum value is 683 lm/W (photopic light conditions).

When developing light for really low ambient light conditions (parking garages, night illumination) then it is more efficient to generate light with the majority of its energy in the scoptopic sensitivity area of the human eye. It is to be compared with moonlight, which is very cold white (blueish white). The human eye is during its evolution completely adapted to it.

Visually useful light

Take a randomly selected lamp. This gives light with a certain spectrum (power spectrum or also called color spectrum). This spectrum can be weighted against the spectral eye sensitivity curve.
All color spectrum meters, giving their resulting light values in lux, are calibrated to the daylight eye sensitivity curve (the photopic light intensity curve). So the reported lux value is valid when cones are working well, and these work well with high or at least sufficient ambient light conditions.
For low levels of ambient light one needs to weigh the spectrum of the illuminant against the scotopic (low light intensity) eye sensitivity curve, or maybe a combination of both. In the latter case this is called mesopic light level, where both the rods as well as the cones are active.

A publication dealing with this issue is coming from ASSIST. This publication works with a certain plan consisting of steps to be executed one after the other to come to a comparison between lamps. An example of a comparison: the are sodium lamps used for night illumination, which emit light with a certain color spectrum, which results in yellowish light. When using an other lamp emitting ligh with much more blue content (which has a better match with the rod’s sensitivity), then less light is possibly needed (compared to photopic/daylight levels). See below a short description of the steps.

1. the publication starts with mentioning recommended illumination values (in lux) on different types of public roads. One chooses the right type of public road and reads the recommended (minimum) illumination value.
2. this illumination value in lux is computed into a luminance (in Cd/m2), indicating an intensity (and it are intensity-differences that are actually perceived or seen by the human eye). So an incident amount of light energy in lux is converted into luminance in Cd/m2 (one uses an estimate for the reflection factor of the pavement, and one assumes a perfect diffuse surface). Since the recommended minimum lux values are determined on the photopic eye sensitivity curve, the resulting luminance calculated is also based on the photopic sensitivity curve and is called photopic luminance.
3. Once the needed photopic luminance value is known, the illumination level is known: either scotopic (<0.001 Cd/m2) or mesopic (in between), or photopic (>0.6 Cd/m2). When in the scotopic or mesopic level, then one have to take into account the human rods and their higher sensitivity as well.

3a. This can be done by using a table presented in the publication. In the first row the requested photopic sensitivity is given. Then the S/P ratio of the used illuminant needs to be known. As an example a high pressure sodium lamp is taken, with a known (low) S/P ratio of 0.63. This means little green-blue so not so efficient for the rods. The table gives, with the S/P ratio known and the requested luminance value, the unified luminance. This is a luminance that is taking both effect of rods and cones into account. It will only take the effect of rods into account when in scotopic or mesopic light conditions.
3b. The unified luminance is known. If now a lamp is taken with a higher S/P ratio, then starting with the needed unified luminance, one can find the belonging photopic luminance (with the known S/P factor) and then it is seen that less photopic luminance is needed to generate the same unified luminance. This is the gain! A lamp with a high S/P ratio and when used in scotopic or mesopic light conditions, does need less photopic luminance to get to the requested unified luminance.
3c. This new, lower photopic luminance, can be calculated back to the needed illuminance value (using the same reflection value) and the result is that less lux are needed!

The relation of visual useful light is as follows: at low light level conditions a high S/P ratio can lead to better use of the human eye sensitivities and eventually a saving on the light illumination values (when expressed in photopic condition) can be possible. However the S/P value is needed as an input.

The S/P ratio and its determination

The lamps measured by OliNo have, since a certain date, the S/P ratio as parameter defined. The next graph is given as example.


The power spectrum, scotopic and photopic sensitivity curves and the resulting weighted night and daylight power spectra (last ones at 1 m distance).

First in blue the power spectrum of the light of the light bulb concerned. It is a relative measure (assume linear scale) and does not have an y-axis as reference.
A black dashed line shows the scotopic eye sensitivity curve. The dashed red line shows the photopic eye sensitivity curve, which are both connected to the left y-axis.
The black line is the scotopic power spectrum (weighted) what results when the power spectrum is weighted against the scotopic sensitivity curve.
The red line in the photopic power spectrum (weighted) results when the power spectrum is weighted against the photopic sensitivity curve.

The surface below the scotopic power spectrum, divided by the surface below the photopic spectrum, gives as a result the S/P ratio. For this example lamp it is 1.25.

The example lamp taken was a halogen lamp. Its light has a lot of red and very little blue. As the sensitivity of the rods is about 3 times higher than that of the cones, still with little blue and a lot of red the resulting S/P ratio was still more than 1.0, meaning still some more efficiency is expected when rods use this type of light.
It is clear that the more blue is in the illuminance’s light, the higher the S/P ratio is and the more benefit can be obtained when these light is used in low ambient light confitions.
Now the very first image can be looked at and understood.

The application of the S/P ratio

One can beneficially use the S/P ratio when a lamp with high S/P ratio is used in scotopic or mesopic light conditions. Possibly street lighting, especially where it concerns only pedestrian of bicycle traffic, as they do not carry headlights that might generate too much light by themselves to make the mesopic or photopic light condition invalid. In cases where photopic or mesopic light conditions are valid, then a lamp with high S/P ratio would make more use of the higher sensitivity of the rods of a human eye and therefore the originally requested (photopic) illumination values can be lower. The steps outlined by ASSIST offer a possibility to select higher S/P ratio lamps and compare these with already existing situations, as see what gain is possible.
In this article an other possible approach is given to use the tables in the ASSIST publication, which can be easier.

Determine the requested illumination level and luminance

What we need is a specification of what illuminance value is needed. This can be retrieved from norms. As an example use the ASSIST publication, containing a table with minimum illumination levels for streetlighting. Assuming a local street, with low pedestrian traffic, an average reflection of 10 % which is also assumed diffuse, then the illuminance value is computed into a luminance value as follows:

This computed luminance is the photopic luminance, not taking into account the possible higher sensitivity of the rods in the human eye. This photopic luminance is correct when in photopic light conditions, bit when in lower light conditions one can make a beneficial effect of the higher sensitivity of the rods.

Evaluate the luminance and determine the unified-luminance

When the desired luminance falls into the scotopic (< 0.001 Cd/m2) or in the mesopic light level (> 0.001 Cd/m2 en < 0.6 Cd/m2) then it is useful to compute the found illuminance into a unified luminance. The S/P ratio of the lamp is needed for that.
Note that at requested luminance values of at least 0.6 Cd/m2 we are in the so called photopic light level and then rods do not function well meaning no benefit can be obtained from using the higher sensitivity of these rods over the cones; the unified luminance is then the same as the photopic luminance.
We start with the photopic luminance that is requeted and the S/P ratio of the lamp, and determin the unified luminance. This is the real experiences luminance (intensity) by the human eye. The computation is as follows:


P is the photopic luminance and
S is the scotopic luminance,
this latter can be computed with the S/P ratio:

When the S/P ratio > 1, then the computation will give an Lunified that is bigger than the P. This is correct, as we are in a scotopic or mesopic light level confition and here the experienced intensity of luminance is higher than a sole computation on photopic levels does make us believe.

Take the unified luminance the same as the requested limunance

We calculated an Lunified, and in the event the S/P ratio > 1, the Lunified will be bigger than P. So we have more luminance than needed/requested, and can go for less. There is in this potential for a saving in driving and power consumption of the lamp. It is all about the perceived luminance, and therefore we need to compute with the unified luminance.
So by putting Lunified equal to the requested or recommended value, one can compute back what is needed for P (with the S/P ratio) and then compute back what needed illuminance values are needed.

Use the ASSIST table for determination of the P and Ev, via the S/P and Lunified

ASSIST table nr 3 can be used. As the equation to compute Lunified cannot easily be used to compute P from Lunified, it is easier to use the table made by ASSIST, that contains many computed values already.

Example: having a lamp with S/P ratio of 2.05, and a requested P of 0.1 Cd/m2. The table indicates with these two inputs an Lunified of 0.1568 Cd/m2 (which could also have been computed with the earlier given equation for Lunified). Since only 0.1 Cd/m2 is needed, one can follow the row of the chosen lamp with S/P ratio of 2.05 to the left until Lunified is equal to 0.1, and read at the top row what P is equivalent herewith.


A part of table 3 from the ASSIST publication, and the way to come to the really needed P.

The left column shows S/P ratios, and the first horizontal row shows the P values. It becomes clear that 40 % reduction can be obtained; only 0.06 Cd/m2 for P is needed to get an Lunified equal to 0.1 Cd/m2, when using a lamp with an S/P ratio of 2.05 and with the mentioned low light levels!
With the 0.06 Cd/m2 for P the illuminance value can be computed which also results in a lower value.

Amount of saving

The amount of saving depends on:
The S/P ratio: the higher this ratio the more possible saving, as the light generated is in a frequency band where the rods are sensitive.
The requested light level. The lower the requested level (so the more towards the scotopic light level, or otherwise the farther away from photopic light levels) the more the rods function and the more benefit can be made from their higher sensitivity.
Note: it is important that a certain low light level is not disturbed by sudden high intensity light spots. For instance on a low traffic road for cars, even though the normal light level is low, when a car passes then the headlights disturb the low light level considerably and for the human eye it needs to adapt to this higher light level and after the car has passed, it needs to adapt back to the low light level again (which, depending on the intensity of the head lights and the age of the person, can take many minutes).

Beware on the effect of dimming

When lamps are dimmed, their color temperature can change as well. This means the spectrum changes and herewith the S/P ratio. For LEDs, the method of dimming (amplitude dimming opposed to PWM dimming) can have a different impact on the color temperature. These effects can be measured so that the S/P ratio is known also when lamps are dimmed.

16 replies on “S/P-ratio”

Can you further discuss the advantages of higher SP for road lights.
Can we attain 2.0 SP using 4300K led?
In what lux level does the rod still works.
Do you think 10 lux is acceptable for main road way applications.

Road lights are generally in a mesopic lighting situation. In such lighting situations our general eye sensitivity does not correspond to the V_lambda (photocpic) sensitivity, but instead is somewhere in between the photopic and scotopic sensitivity.
Then a spectrum should be weighed against this mesopic sensitivity curve and then we will see that for leds the spectra belonging to higher colortemperatures will deliver a higher S/P ratio and therefore more possible saving.
2.0 with 4300 K seems a bit high to me. I would gueaa 1.5 for 4200 K.
The rods work at very low lux levels, for instance moonlight gives 0.25 lux and we could almost read a newspaper (one article, not more since it would make us tired).
For main road way applications the 10 lux you mention seem enough.
For these applications the term luminance L_v is used as well. There is a relation between E_v and L_v, wihch is: L_v = rho * E_v / PI, where rho is the reflectivity of the road surface (about 0.1-0.3 I guess).

Thanks for your reply. We are into bidding process., and our client is asking us to produce SP ratio of not less than 2.0 at 3500k – 4300K. My reseach shows that some manufacturers like Lemsis was able to meet this by mixing up green, red and warm color. Pls check http://lemnislighting.com/

We are using CREE LED by Betaled (Ruud Lighting), and we were informed that the IESNA (Illuminating Engineering Society of North America) has not endorsed the (S/P)n method in lighting. Do you think a high SP can be use as gauge in selecting a good LED today. What do you think is the effect of mixing up the green and red LED. Will this be better than the White LED of CREE.

Further, we are considering a wider road for city driving. Where there are lots of vehicles on the road. Do you think a high SP ratio will still be effective specially when the headlights are ON. As you have explained, high SP is only good at scotopic and mesopic light condition. Generally, when you need a well lighted road, and when the safety of the people is our prime concern against accident and crime., do you think it is better to employ white LEDs rather than going back to warm colors.

In response to id 3. Yes, Lemnis indeed was able to book good results with some tryouts that they did. They work with a lot of green and blue and therefore with an SP of about 3. They also tend to add red in order to stimulate the green and red cones such that we still get a color impression. The colors might not be natural and sometimes too saturated, but there IS a lot of color and the overall view is magnificent, compared to traditional lighting.
Then about the endorsement. Neither does the Netherlands endorse the use of S/P as factor to use to save on lumen when considering road lighting applications. I understood in a course on illumination that I follow, that is was because the team that investigated could not determine well whether the rods could take over cone functionality with regard to sharp vision. Also they said that with the headlights the eye afterwards needs to adapt to low light output constantly and therefore does not well get into the mesopic sensitivity.
It is hard to say, I suppose that tests shoudl be done where streetlights are installed and then it must be seen. The Lemnis experience is that many more details are seen and much more contrast, while the amount of light measured in traditional scotopic lumen is very low. So it might not be yet well explained, but there seems to be some gain in all this.

In response to ID4. Romeo, you really need low light levels to be below scotopic level. That 0.6 Cd is about to be equivalent to L = rho*E / pi ==> E = 10lx (rho = reflectivity assumed to be 0.2).
So when E> 10lx then we are already in scotopic region.
when safety is of concern the more light the better the overview is. However people tend to drive faster as well hence decreasing overall safety.
In many countries outside Netherlands there is as a general rule NO streetlights at all above roads outside city borders. So people need to work with their own lights and eventually reflectoes in the road.

Thank you very much for all the info. If Netherlands do not endorse the SP ratio, therefore it should not be use as a gauge in determining a good LED luminaire. In the same way, safety on the road requires more illumination, the low lux level will be very difficult specially in countries where there are more rainy days/nights. This condition will endanger the lives of many commuters.

@7, the recommendation of the group was to do more research on it (as I recall it). So it is not yet endorsed but need more research. I think it is strange to not take into account the SP effect since if you look at streets lit with for SP optimized lamps then the visibility is good, also compared to traditional lights.
Theory states that rods are more sensitive but serve for overall vision and cannot give details. Theory also states that cones are for details, in fovea area but do not work (well) at low intensities. So it seems hard to theoretically explain the nice results that can be obtained by SP optimized streetlighting.
I think I have to do this once more, to see whether I can actually read in moonlight. Since Ev < 1 lux it is then low in mesopic region. I thought I WAS able to read (I thouhgt I did this once in the past) but am not sure. If reading is possible then detail vision is also possible.

@8, I agree with you in general. Since I’ve been explaining mesopic / scotopic vision during the last 10 years while selling induction lamps. However in our case where the buying entity requires an SP ratio of 2.0+, and disqualifying CREE’s 4300K w/ SP of 1.65. Do you think it is theoretical and legal to do that. What is the difference of 1.65 from 2.0 when we are talking of the rod vision at low lux level, as against the busy roads where tens of thousand vehicles passes every hour. Do you believe it is relevant.

@9 Difficult to say whether it is relevant. I think S/P ration of 1.6 related to 2.0 might not be that relevant. However S/P can go much higher than that, for instance > 3.0 even. Then > 3.0 compared to 1.6 there is more difference. I guess that the customer had to draw a line somewhere.

And also the complete spectrum will have its effect. So light with spectrum A with SP=3 might give a different (color) perception than light with spectrum B with SP=3. Therefore, when all theory is not yet known, it is important to play with spectra as well to see what the results are.

Marcel, can you email me at romiebrite@yahoo.com, or call me at +12175964227. I just want to be enlightened what is the difference of SP ratio of 1.65 as against 2.0+. What is the total effect of this in terms of efficiency. Is this parameter very important in the proper selection of a good LED luminaire.

I want to know if there’s any oficial recognition of the sp ratio as a standard for government lighting normativity in any country?
In Mexico it´s all based on luxes and it has been very difficult to make the normativity people to understand and accept as a new way of measuring visibility, so any information would help.


Dear Alejandro,
In The Netherlands the recent update of the guideline for stretlights did acknowledge that there is a benefit but did not quantify it. SAo it is now up to the designer or architect to define what is acceptable. Very difficult, Fieldtests then need to be done.
In UK there is a reduction in class allowed when you work with led streetlighs, in essence anything other than SOX Sodium low pressurelamps. That helps.

The current recommendation, as now being implemented by cities like New York City, Los Angeles, Phoenix, and Tucson is to use LEDs with a CCT of 3000K or less at the lowest brightness needed. The American Medicial Association also recommends this, as they have found blue light exposure at night to be harmful to our health. It’s also harmful to our wildlife and creates much, much more light pollution than the old streetlights most cities have been using. The warmer streetlights are becomming much more affordable, and many find them more pleasant at night, as white light at night is often harsh, cold feeling and creates unsafe glare.

Leave a Reply

Your email address will not be published. Required fields are marked *