Light Compare

Light Compare


A Comparison Between Light Sources Used in Planted Aquaria

By Ivo Busko

How to compare bulbs

Criteria for comparing bulbs to be used in planted freshwater aquaria always include some measure of personal taste. We do not only want plants to grow well, we also want the aquarium to look good. Of course, we should try to provide in the first place "strong" or "bright" lighting (whatever those terms mean), in most cases guided by rules of thumb of the type "Watts per gallon". This is not the subject of this article however. We concern ourselves here with efficiency instead. That is, given a fixed amount of electrical energy (Watts), how can we get the maximum "light" possible, simultaneously keeping the good looks of our tank ?

If one has a lot of room on top of the tank, there are no concerns with electrical energy economy, and one has a large pocket, the primary selection criteria for light bulbs should be based on their visually meaningful parameters such as lumen output, color temperature and color rendering index. If plants don't like it, just add extra bulbs until they do. If however there are space constraints (due for instance to exotic tank geometry, which is my case !) and/or budget concerns both for initial setup as well as for long term maintenance (my case as well !), one should try to design an efficient lighting setup. To this end every efficiency and throughput characteristic of each potential light bulb must be taken into account in greater detail. In particular, more appropriate parameters than just the lumen output should be considered. But what are these parameters, and what values should we seek for ?

My aim with this article is to provide this additional information on light bulbs used (or potentially usable) in planted aquaria, with numerical results expressed in standard physical units. It is not my objective to recommend particular types or brands of bulbs; the information is presented in here with the sole objective of enabling the reader to take more informed decisions when designing his/hers lighting system.

The following section introduces some of the terminology used in the work; you can skip it if you are familiar with the subject. The next sections briefly discuss the technique and data used, and present the results. There is an appendix with pointers to the primary data sources, a detailed description of the computational steps and errors, and an example of a practical use of the data presented here.

Lumens, lux, PAR..., what are they ?

Artificial light sources are usually evaluated based on their lumen output. Lumen is a measure of flux, or how much light energy a light source emits (per unit time). The lumen measure does not include all the energy the source emits, but just the energy with wavelengths capable of affecting the human eye. Thus the lumen measure is defined in such a way as to be weighted by the (bright-adapted) human eye spectral sensitivity. If we plot this sensitivity as a function of the wavelength of the light (building the so called photopic curve), we see that it has an approximately bell shape, peaking up at a wavelength of around 550 nanometers (nm), the "green" region of the light spectrum, and decreasing at both longer (red) and shorter (blue) wavelengths. See the plot here. The consequence is that two light sources that emit the same total amount of energy can have vastly different lumen ratings, depending on how much of that energy is concentrated around the 550 nm region.

Another quantity often quoted when talking about light output is lux. Lux is a measure of illumination, not flux. Flux refers to the light energy that leaves the source. Illumination refers to the light energy that reaches the receiving surface. Lux is equivalent to lumens/m2. Lux cannot be computed only from the know data of a light source. Additional information regarding the illumination geometry, reflectors, distances, intervening media (glass covers, water) must be taken into account.

Other quantities used to describe light quality associated with its visual characteristics are color temperature and color rendering index (CRI). Color temperature is defined as the temperature that a perfect electromagnetic radiator ("black body") would have to have to emit light with the same "color" as the light source in question. Higher color temperature means bluer color, lower temperature, redder color. Color temperature is expressed in degrees Kelvin (from Lord Kelvin, the 19th century physicist, and which means degrees Celsius above absolute zero). CRI measures how close to their "true color" a light source can render objects illuminated by it. A "perfect" light source would have a CRI of 100, lower values mean that the colors are shifted from their "true" hue and saturation. Many people are familiar with the color shifting that takes place when one buys clothes in a store with artificial illumination and then realizes that under natural (sun) light the colors are not quite the same. Had the store used high-CRI light bulbs that color shift would be much smaller or not noticeable at all.

It is easy to guess from the wording in the above paragraph that these two parameters are also strongly related to the human eye response characteristics. In fact, the technical definition of the term "color" used above is directly based on psychophysics experiments performed with human subjects and standardized by the CIE (Commission Internationale d'Eclairage) about 60 years ago. In other words, color temperature and CRI are parameters entirely based on the human visual system characteristics and may carry absolutely no meaning when applied in other contexts.

Laboratory experiments showed that the photosynthesis process that takes place in plants when submitted to intense light has a very different spectral response than the human eye. In fact, photosynthesis is the least efficient in the region around 550 nm. Most of the light capable of inducing the photosynthesis reaction is either red or blue. In other words, plant leaves mostly reflect green light, while they absorb red and blue with higher efficiency. An experimental fact that confirms this statement, independent of any laboratory measurement, is the fact that many plants look green ! Portable field instruments used to quantify photosynthesis in growing plants often exploit this fact by using as light source a pair of red and blue LEDs (Light Emitting Diode) instead of a white light source.

The curve that results from plotting photosynthesis efficiency as a function of wavelength is named "Photosynthesis Action Spectrum". It is the equivalent of the photopic curve for photosynthesis. The curve is typically double-peaked, with maxima around 420 (blue) and 670 (red) nm and a "valley" around 550 nm. The curve drops sharply below 400 nm and above 700 nm. The peaks are broad and not as pronounced as the central peak in the photopic curve. There is still significant response in the green region around 550 nm. See a typical curve here. Many plant species can show specific action spectra that differ markedly from that "average" curve. In some extreme cases there is no response at all in one of either red or blue regions. The important point is that photosynthesis has a much broader wavelength response than the human eye, with less dependency on specific, narrow wavelength regions. Thus, light sources that look very different to us may "look" similar to a plant. Conversely, light sources that look similar to us may "look" very different to plants, all depending on their specific spectral distributions.

In some instances we see references to "plant growth spectrum" as well. This is not to be taken as equivalent to the action spectrum though. The action spectrum has a precise meaning in terms of quantity (in moles/sec/leaf surface area) of CO2 consumed by the plant subject to measurement. "Growth", on the other hand, can be defined in many different ways (height ? weight ? weight of dry plant mass ?) that can be even very species-dependent, so it hardly makes a good standard for comparison purposes.

Based on the Photosynthesis Action Spectrum, light bulb manufacturers came up with fluorescent "plant bulbs". They basically emit most of their light in the wavelengths that are more efficient for photosynthesis, namely the red and blue ends of the visible spectrum. As expected, these light sources look dim to the human eye and consequently have poor lumen ratings. Also, their color temperature and CRI ratings have little, if any, meaning. After all, these bulbs were not designed to be "seen" by humans...

The standard measure that quantifies the energy available for photosynthesis is "Photosynthetic Active Radiation" (aka "Photosynthetic Available Radiation") or PAR. Contrary to the lumen measure that takes into account the human eye response, PAR is an unweighted measure. It accounts with equal weight for all the output a light source emits in the wavelength range between 400 and 700 nm. PAR also differs from the lumen in the fact that it is not a direct measure of energy. It is expressed in "number of photons per second", whose relationship with "energy per second" (power) is intermediated by the spectral curve of the light source. One cannot be directly converted into the other without the spectral curve.

The reason for expressing PAR in number of photons instead of energy units is that the photosynthesis reaction takes place when a photon is absorbed by the plant, no matter what the photon's wavelength (or energy) is (provided it lies in the range between 400 and 700 nm). That is, if a given number of blue photons is absorbed by a plant, the amount of photosynthesis that takes place is exactly the same as when the same number of red photons is absorbed. For convenience, number of photons is usually reported in the literature in micromole units, or microEinsteins. One microEinstein is equivalent to 6.02 1017 photons. Another important difference is that usually PAR is quoted as an illumination measure akin to lux, thus related to the receiving surface. PAR is typically reported in microEinstein/second/m2.

Thus we see from the above that, to evaluate light sources for use in plant applications, we cannot in principle rely entirely on an human-based criterion, the lumen rating. Unfortunately, manufacturers provide little information in that regard. Power consumption in Watts and lumen ratings are easy to get, and for many bulbs spectral plots do exist. Many of these are not depicted in physically meaningful units though (such as Watt/nanometer), making it difficult to compare different products. PAR figures are never quoted because they depend on the detailed illumination geometry, which varies from setup to setup.


However, having access to the spectral plot in relative units, and the lumen and Watt rating for a bulb, it is possible to derive several useful bulb parameters. For instance, an overall efficiency factor can be computed comparing the theoretical lumen output the bulb should have, with its actual lumen output. This efficiency factor is independent both from the bulb's spectrum and from the human eye photopic response, contrary to the often used lumen/Watt efficiency factor. The efficiency factor thus computed can be used then to normalize the spectrum from relative to absolute physical units such as Watt/nm. It then becomes a simple matter to compute other quantities in standard physical units.

For instance, one can compute the total amount of photons generated per second in the interval 400 to 700 nm, which is related directly to the bulb's PAR characteristics. If all other parameters that affect the light input into the aquarium (reflectors, glass cover, water depth and transparency, physical dimensions) are kept constant, this flux-like PAR measure can be used directly to compare different bulbs, without the hassle of converting the measure to illumination units. Another possibility is to weight the PAR measure with an average Photosynthesis Action Spectrum, thus generating a figure of merit akin to the lumen rating, but targeted towards plant use, not human use. The figure of merit thus created is usually named in the literature "Photosynthetic Usable Radiation" or PUR. All these figures taken together should make the selection of a particular bulb a more objective process.


For carrying out the computations, I used only spectral curves and bulb data I was able to get from the web, as well as some web-published photosynthesis action spectra. I also had to write a short computer program. Details and pointers to original data can be seen in the appendix.

Initially I got data for mostly normal-output fluorescents, since this is the type of bulb which I was primarily interested in. But the methodology is general and applicable to any light source. The most recent results include data for metal halides, power compacts, HO and VHO fluorescents, an incandescent halogen, two mercury vapor bulbs, a high pressure sodium, as well as a solar spectrum.

The main problem when comparing normalized light spectra is related to spectral resolution. Roughly speaking, this is the amount of detail a spectral plot has regarding light intensities at neighboring wavelengths. The smaller the resolution, the more detail and information the spectral plot conveys. Published bulb spectra span a relatively large range of spectral resolutions, and accurate comparisons can only be made in between spectra of the same, or about the same, resolution. Fortunately the majority of these published spectra have resolutions in a narrower range, in between 5 to 10 nm. This enables relatively fair comparisons among most of the bulbs, generating errors in the computed parameters of a few percent only. The few spectra in the sample that have smaller (better) resolutions were numerically degraded to a nominal 5 nm resolution in order to be directly compared with the main body of data. A few spectra with very poor resolutions are presented separately and cannot be reliably compared with the others.


The main results are presented in the tables that follow. The Hagen bulbs have published spectra which clearly show systematic distortion in the emission profiles not caused by the spectrophotometer characteristics, but more likely by doctoring at the marketing department. Such data cannot be compared with other bulbs unless very approximately, so their data is reported in a separate table. Relative comparisons among the Hagen bulbs should still be possible though (with a grain of salt).

By the same reason pointed in the previous section, it is very difficult to visually compare spectral plots with very different resolutions. In particular, the narrow, strong emission features characteristic of modern fluorescent phosphors may look very different in plots with even a small difference in resolution, turning objective visual comparisons difficult. This fact must be kept in mind when examining the library of normalized spectral plots. Integral quantities such as the ones reported in the tables are much less prone to the effects of varying spectral resolution and make a much better objective criterion. Spectral plots can be useful though, always keeping in mind the resolution effect.

Columns in the first two tables list the following quantities:


  • Power: the bulb's rated power.
  • Maximum lumen output: this theoretical value depends only on the bulb's spectrum and rated power. It is the lumen output that the bulb would have if all electrical energy input to the bulb were transformed into electromagnetic energy.
  • Rated lumens: (initial) taken from bulb's specs.
  • Efficiency: the ratio between rated lumens and maximum lumens, or overall efficiency.
  • PAR: the bulb's output in PAR. The units are just uE/sec. Elsewhere in the literature PAR is usually defined as a measure of illumination (like lux). Thus it should be computed at the receiving surface in units of uE/sec/m2 or equivalent. The PAR figures in this article are a measure of flux (like lumens). To convert them to uE/s/m2 one should enter with complicated geometric and transparency factors that are specific to each individual setup. The tabulated values are appropriate to use in relative comparisons among different bulbs, and also as a starting point if one wants to compute the illumination created by a specific setup.
  • PAReff: the PAR/Watt efficiency ratio. With plant applications in mind, this parameter should be the primary criterion for quantifying bulb efficiency. Thus, in the following two tables, bulbs are ranked in order of decreasing PAR efficiency.
  • PUR: obtained by weighing the photon spectrum with an "average" photosynthesis action curve. The Total column lists the sum of all photons in the range 400-700nm weighted by the action spectrum. Since there is no clue in these figures about the relative amount of red and blue photons, I also computed PUR in the 400-500 nm range only (blue) and 600-700 nm range only (red).
  • R/B: The ratio between the red and blue PURs. This measure is in some way analogous to the color temperature of the bulb. Color temperature, however, is defined in terms of the eye photopic response. R/B is defined in terms of the photosynthesis action spectrum. It is a measure of the "color temperature" that the plants, not the human eye, "see".


Bulb Power
Effic. PAR
ADV850 32 9700 3100 0.32 46.2 1.44 22.8 11.5 7.2 0.63
HPS Dlx 100 22650 7300 0.32 140 1.40 72.8 10.1 53.8 5.33
MHN 150 34500 11250 0.33 207 1.38 116 45.0 47.7 1.06
Iwasaki65 150 37700 12000 0.32 199 1.33 107 46.6 35.2 0.75
Dulux54 55 16400 4800 0.29 72.2 1.31 36.3 19.6 10.3 0.52
Pentron41 HO 54 17800 5000 0.28 69.5 1.29 32.0 13.7 12.0 0.88
Panasonic67 96 28600 8100 0.28 123 1.28 60.7 30.6 17.3 0.57
Aquarelle 38 8100 2380 0.29 48.2 1.27 29.1 18.8 6.9 0.37
MH 250 82500 23000 0.28 310 1.25 152 67.6 32.2 0.48
TLD950 36 9100 2350 0.26 42.8 1.19 23.4 8.5 10.0 1.18
GE SPX65 40 11600 3050 0.26 46.2 1.15 24.0 13.8 5.1 0.37
PLL950 55 14800 3800 0.26 62.8 1.14 32.8 15.5 10.7 0.69
Triton 40 9000 2200 0.24 43.2 1.08 25.1 14.9 7.2 0.48
GE Fresh & Salt 40 10000 2350 0.24 42.6 1.06 23.2 12.4 7.6 0.62
Cool White 40 12600 3050 0.24 42.4 1.06 20.5 9.3 5.7 0.61
Daylight Dlx 40 10400 2550 0.25 42.3 1.06 23.2 11.9 6.2 0.52
Ott CF 23 5000 1200 0.24 24.0 1.04 14.4 7.5 4.6 0.61
Gro-Lux 40 5100 1200 0.23 41.2 1.03 27.4 9.7 15.5 1.6
VitaLite 40 8200 2340 0.29 41.0 1.02 22.2 8.6 8.7 1.01
Warm White 40 14000 3100 0.22 40.7 1.02 18.1 6.1 6.6 1.08
Cool White Dlx 40 9500 2250 0.24 40.5 1.01 22.4 8.6 9.4 1.09
Warm White Dlx 40 9400 2200 0.23 39.8 1.00 21.4 5.9 11.6 1.97
Perfecto 40 6800 1500 0.22 39.5 0.989 25.3 9.8 12.6 1.28
C50 40 10100 2250 0.22 39.2 0.980 21.2 8.1 8.8 1.09
Osram Biolux 40 10200 2400 0.24 38.1 0.953 20.4 10.2 4.5 0.44
P&A 40 8900 1900 0.21 37.7 0.943 20.9 5.4 12.0 2.22
VHO Cool White 115 36000 7500 0.21 105. 0.916 51.4 23.0 14.2 0.62
AgroLite 40 7800 1600 0.21 33.6 0.841 19.2 5.4 11.1 2.05
GE Freshwater 40 8500 1425 0.17 30.8 0.771 18.4 6.9 9.1 1.33
TL950 32 13000 2000 0.15 22.7 0.709 8.9 2.0 2.4 1.19
MV 100 42300 4300 0.1 46.2 0.46 17.8 7.1 4.3 0.60
Wonderlite 160 31500 3125 0.1 56.9 0.356 32.1 14.3 10.3 0.72
Halogen 60 1100 730 0.65 15.8 0.263 8.7 1.2 6.1 4.69
Sun light 40 9300 2000 0.22 39.9 0.999 22.8 8.6 10.0 1.17


Bulb Power
Effic. PAR
PowerGlo 40 8900 2200 0.25 43.2 1.08 25.7 14.9 5.8 0.39
SunGlo 40 13100 3100 0.24 42.4 1.06 20.6 9.7 4.8 0.49
AquaGlo 40 4600 960 0.21 38.5 0.964 27.9 11.5 14.6 1.27
FloraGlo 40 12100 2180 0.18 34.3 0.857 16.7 3.4 9.2 2.69

The bulb names mean:


ADV850 Philips Advantage fluorescent, 5000K F32T8/ADV850
HPS Dlx High Pressure Sodium deluxe Philips Ceramalux Comfort C100S54/C/M
MHN Philips dense-line emitter metal halide 4100K CRI 80 (MHN150/TD/840)
Iwasaki65 Iwasaki 6500K metal halide
Dulux54 Osram Dulux L 5400K 82 CRI compact fluorescent
Pentron41 HO Osram/Sylvania T5 HO fluorescent 4100 K (FP54/841/HO)
Panasonic67 6700K Panasonic compact fluorescent PC96W67K
Aquarelle Philips Aquarelle 10,000 K fluorescent for freshwater aquaria
MH generic, non-coated metal halide 4000K CRI65 (from Philips catalog)
TLD950 Philips full spectrum fluorescent 'TL'D/90 de Luxe 5300 K CRI > 95
GE SPX65 GE SPX65 6500 K fluorescent
PLL950 Philips PL-L/950 5300K high-CRI (91) compact fluorescent
Triton Interpet Triton
GE Fresh & Salt GE AquaRays Fresh & Saltwater fluorescent (F40T12/AR/FS)
Cool White generic 4100 K cool white fluorescent F40T12CW (average of two spectra)
Daylight Dlx GE Daylight Deluxe fluorescent
Ott CF Screw-in full-spectrum compact fluorescent
Gro-Lux "Original" Sylvania Gro-Lux (not the wide spectrum variety)
VitaLite Vita-Lite fluorescent
Warm White generic 3000 K warm white fluorescent F40T12WW (average of two spectra)
Cool White Dlx. generic 4200 K cool white fluorescent deluxe F40T12CWX (average of two spectra)
Warm White Dlx. generic 3000 K warm white deluxe fluorescent F40T12WWX (average of two spectra)
Perfecto Perfecto-A-Lamp (a wide-spectrum grolux)
C50 Full spectrum T12 5000K fluorescent: GE Sunshine (or Chroma 50), Philips Colortone, Sylvania Designer (average of three spectra).
Osram Biolux Osram Biolux fluorescent
P&A GE Plant & Aquarium fluorescent
VHO Cool White Very High Output version of the 4100 K cool white fluorescent (Osram F48T12CW/VHO/LT)
AgroLite Philips Agro-Lite fluorescent F40T12AGRO
GE Freshwater GE AquaRays Freshwater fluorescent (F40T12/AR/FR) (a modified, lower efficiency grolux)
TL950 Philips TL950 5000K fluorescent very high CRI (98) F32T8/TL950
MV Deluxe Mercury Vapor Philips H38MP-100/DX 3700 K, CRI 45
Wonderlite Self-ballasted screw-in mercury vapor R40 flood ligth with special "plant" spectrum made by Westron
Halogen Spot halogen Philips Masterline Par 16 (60PAR16/H/NSP) 2950K
Sun light Theoretical bulb that perfectly reproduces sun light (at 5000 K) with the average efficiency of a full spectrum NO fluorescent.
PowerGlo Hagen PowerGlo fluorescent
SunGlo Hagen SunGlo fluorescent
AquaGlo Hagen AquaGlo fluorescent (a wide-spectrum grolux)
FloraGlo Hagen FloraGlo fluorescent

We can draw several conclusions from these data:

The most important conclusion in my opinion is that the efficiency of converting electrical energy into PAR light energy is not that different for the several bulbs and technologies included in the sample. The majority of bulbs in the sample deliver approximately (within a 20% range) the same amount of uE/s/Watt in the 400-700 nm range, about 1 uE/s/Watt. High intensity discharge lamps and high-end fluorescents tend to be more efficient, but not by a large factor. On the other hand, the efficiency of converting electrical energy into visible light energy can be very different among the several types. In other words, the lumen/Watt efficiency can encompass a very wide range, about 200% in this sample. This effect can be quantified by the correlation coefficient between these two quantities, which is 0.56 in this sample. This tells us that no correlation exists between the lumen and PAR output, or, in other words, the lumen efficiency is a very poor criterion for selecting bulbs. We should strive instead for raw power (Watts), since PAR/Watt is on a first approximation the same for all bulbs. The reason for this lack of correlation is of course the broad band response of plants to light, as opposed to the narrow band response of the human eye. Some bulbs are somewhat better than others in converting Watts into photons though, so when efficiency is a major design factor, one should stick with the highest PAReff bulbs.


  • The ordering in decreasing PAR efficiency is almost identical with the ordering in decreasing overall efficiency. This effect is expected since both parameters ultimately measure the efficiency of conversion of electrical energy into electromagnetic energy, over a wide spectral band and without regard to the spectral shape. If we ignore the halogen bulb, both parameters span a relatively narrow range of a factor 2. This is also a consequence of the fact that both parameters in fact depend only on the underlying physical processes used to convert electricity into light. We may conclude that the most popular existing lighting technologies are not capable of conversion efficiencies larger than about 30%.
  • A few bulbs do not follow the trend described above. These are the broad spectrum ones: Vita-Lite, Biolux, Wonderlite, the halophosphor "Deluxe" bulbs, and most markedly, the halogen. The reason is that these bulbs deliver a significant fraction of their total electromagnetic output outside the 400-700 nm range. For the broad spectrum fluorescents and MHs this fraction is about 8-15%, for the halogen it is more than 90% ! The result is that they have a significantly poorer PAR emission given their overall efficiency. In other words, they spend a fair amount of electricity to create light which both the human eye and the photosynthesis process cannot see. Of course, there might be other processes that benefit from that ultraviolet and infrared energy, but strictly from the perspective of optimizing raw PAR emission and lumen output these bulbs aren't the best choice. It also must be noted that the halogen bulb is not strictly comparable with fluorescent and discharge lamps in this regard, since the underlying physical processes responsible for light and heat emission are different in each case. Thus its apparently high overall efficiency factor is in fact an artifact caused by the very definition of this efficiency factor in the first place.
  • Some of the bulbs with the worst efficiency are:


    • the notoriously inefficient halogen incandescent, about four times less efficient than the average fluorescent in producing PAR photons. The 60 Watt halogen produces about 0.01 Watt/s/nm in the green-yellow spectral region, as compared to a, say, 40 Watt Cool White fluorescent which produces about 0.04-0.05 Watt/s/nm in the same region. Note however that for this particular type of spot light bulb, which has an enclosed parabolic reflector, the very narrow bean partially compensates for this low efficiency. The ultimate comparison would be in between the illumination at the illuminated spot, and the illumination created by the other bulbs when under a very efficient reflector. The spot created by the halogen spot is very small though, and of limited use (perhaps as accent light).
    • the mercury vapor bulbs, about halfway between the halogen and the worst fluorescents. The Deluxe MV probably should be avoided, since besides its low efficiency, it provides a way too unpleasant light (too low CRI, too yellow). A regular (non-Deluxe) MV is even worse since it lacks entirely any red emission. The Wonderlite though seems to have overcome the color problems. The light is reported to be white and with good color rendering. Should be an option to consider in non-standard applications, since its enclosed reflector should partially balance out the lower efficiency.
    • the very high CRI fluorescent TL950, about 30% less efficient (in PAR units) than the average fluorescent. Note that its lumen/Watt efficiency is not that bad though, even outperforming in this respect some of the older high-CRI bulbs. This bulb is a tri-phosphor that seems to use special phosphors that emit light at different wavelengths than the "normal" tri-phosphors found in other bulbs. The overall emission is packed tightly around the 550 nm region, with minimal emission at the blue and red ends of the spectrum. Probably these phosphors were specifically tailored to achieve the extremely high CRI, at the expense of other performance factors. High CRI ratings are usually associated with low efficiency, but the newer TLD/950 and PL-L/950 bulbs (both European...) seem to break this trend. They provide both relatively high efficiency and high CRI.
    • the Philips Agro-Lite and the GE Aqua Rays Freshwater bulbs.
    • the VHO version of the Cool White tube. Low efficiency seems to be a normal characteristic of VHO tubes.
  • The most efficient bulbs in the sample are the metal halides, the high pressure sodium, and the tri-phosphors ADV850, Dulux, Panasonic, Pentron HO, and Aquarelle, about 20-30% more efficient than the average fluorescent. Interesting enough, a generic, traditional metal halide does not perform so well when put side by side with more evolved types such as the dense-line emitter MHs (MHN and Iwasaki65). The HPS deluxe was included in the sample only for completeness, since it has a too low color temperature (2200 K) to be of use as the main light source in planted aquaria. It might be useful as a replacement for halogen/incandescent bulbs used as accent lights though. Normal HPS bulbs were not analyzed due to their poor CRI (around 20).
  • High performance fluorescent tubes are capable of generating the same, or even slightly more, light output per Watt than a MH bulb. In particular the Philips ADV850 operated under standard conditions even outperforms the MHs. Considering that these inexpensive T8 bulbs can be overdriven by electronic ballasts with high ballast factors (> 1), thus delivering even more light, they are possibly the best option to light a planted aquarium in terms of performance/cost factor.
  • As an interesting exercise we could rank the bulbs according to their PUR efficiency ratio instead. The following table lists the bulbs so ranked.


Bulb PUReff
MHN 0.77
Aquarelle 0.76
HPS Dlx 0.73
Iwasaki65 0.71
ADV850 0.70
Gro-Lux 0.69
Dulux54 0.66
TLD950 0.65
Panasonic67 0.63
Perfecto 0.63
Triton 0.63
Ott CF 0.62
MH 0.61
GE SPX65 0.60
PLL950 0.60
Pentron41 HO 0.59
GE Fresh & Salt 0.58
Daylight Dlx 0.57
Cool White Dlx 0.56
VitaLite 0.55
Warm White Dlx 0.54
C50 0.53
P&A 0.52
Cool White 0.51
Osram Biolux 0.50
AgroLite 0.48
GE Freshwater 0.46
Warm White 0.45
VHO Cool White 0.44
TL950 0.28
Wonderlite 0.20
MV 0.18
Halogen 0.15

The PUR ranking roughly repeats the trend observed with PAR efficiency ranking: the same high-efficiency bulbs in terms of PAR/Watt rank high in the PUR/Watt list. In other words, the two parameters correlate well. The outstanding exception here is the generic MH, which has a spectrum well matched to the photopic curve. The main effect of using PUR to rank a bulb is the expected better performance of "plant" bulbs that were specifically designed with the Photosynthesis Action Spectrum in mind. Once again, the ADV850 bulb shows extraordinary performance, with a PUR efficiency similar to the "best" plant bulb, the original Gro-Lux.

Another interesting exercise is to rank the bulbs by their lumen/PAR ratio:


Bulb lumen/PAR
MV 93.1
TL950 88.2
Warm White 76.5
MH 73.6
Pentron41 HO 71.9
Cool White 71.8
ADV850 67.1
Dulux54 66.5
GE SPX65 66.1
Panasonic67 65.9
Osram Biolux 62.9
PLL950 60.5
Daylight Dlx 60.3
Iwasaki65 60.3
C50 57.4
VitaLite 57.1
Cool White Dlx 55.6
Warm White Dlx 55.3
GE Fresh & Salt 55.2
Wonderlite 54.9
TLD950 54.9
MHN 54.4
HPS Dlx 52.1
Triton 50.9
P&A. 50.4
Sun light 50.0
Ott CF 50.0
Aquarelle 49.4
AgroLite 47.6
GE Freshwater 46.3
Halogen 46.2
Perfecto 37.9
Gro-Lux 29.1

This parameter does depend exclusively on the shape of the bulb's spectral curve. It measures how well matched is this curve to the photopic curve, and how far it is from a perfectly flat (in photon units) spectrum. Again, the large range in lumen/PAR values in the above table is a direct consequence of the lack of correlation between the two parameters.

The lumen/PAR measure would be useful, for instance, in helping picking out bulbs that both look bright to our eyes, and deliver a fair amount of light at the red and blue ends of the spectrum. The highest ranking bulbs in the table have spectra extremely well matched to the photopic curve, thus lacking red and blue emission. The lowest ranking bulbs, on the contrary, look dim to our eyes but deliver a larger fraction of their output into the red and blue ends of the spectrum. Flat spectrum bulbs, as expected, are the ones showing the best balance, ranking close to the center of the list.

Another measure of spectral balance can be conveied by how close the R/B ratio is to true sun light. As expected, the high-CRI TL950 and TLD950 rate very close to sun light. The inexpensive C50 and some of the halophosphors rate next, together with the more expensive Vita-Lite. Plant bulbs tend to put out an excess of red light, and higher efficiency bulbs tend to be bluer. The extreme case among the fluorescents is the Aquarelle bulb, which ranks as the most efficient fluorescent in PUR/Watt and is the bluest as well. The exception to this "rule" is the HPS Deluxe bulb, which creates a lot of red light despite its high efficiency. If one believes, as many people do, that the red and blue regions of the spectrum govern plant growth in different ways, one should take the R/B parameter into account when selecting a light bulb. Note that there is no straightforward relationship with color temperature, which is a human eye-based criterion and meaningless in this context (since it is defined over a too narrow wavelength span).

It is easy, from the tabulated data, to compute figures for multi/mixed bulb configurations, by just adding the individual bulb's measures, weighted by the number of bulbs of each type in the mix and scaling the contributions by the actual bulb power. So it should be easy to come up with optimum mixes given the constraints of ones' configuration. Of course, these constraints may play a significant role in the final result. For instance, a T5 bulb such as the Pentron HO may benefit from its better optical coupling to specular reflectors, minimized restrike, and higher operating temperature than PC, T8 and T12 bulbs, thus resulting in higher overall efficiency of the ligth fixture. See the example. Remember that a small additional error may appear when scaling by the bulbs rated power. Both fluorescent and metal halide bulbs of different sizes/powers may have slightly different efficiencies of conversion of electrical energy into light.

And bear in mind: these results are only as good as the manufacturer's published spectral curves, lumen and Watt data allow them to be. I estimate that errors in the computed parameters should be a few percent at most for the best data, up to 10-15 percent for the worst ones. See the appendix.


Data sources:

Bulb spectra:


Bulb lumens:

Photosynthesis action spectra:

Photopic curve:

The first posted version was made with a curve digitized from The current table uses CIE's 1988 table (thanks to Roger Miller)

Lumen definition:

Steps to compute the tabulated quantities:

  • digitize the published spectrum at suitable values of wavelength and rebin to a small wavelength step (1 nm).
  • (optional) numerically degrade the resolution to 5 nm.
  • numerically integrate the digitized spectrum over its entire wavelength range, getting the total power S in relative units.
  • knowing the bulb's total power consumption P in Watts, normalize the digitized spectrum ordinate scale by multiplying the relative values by P/S. This gives the bulb's spectrum in Watts/nm, assuming that the conversion efficiency from electrical to electromagnetic energy is 100%.
  • multiply the Watts/nm spectrum by the eye photopic curve (normalized to unity peak). Integrate the result over the same wavelength range used above, and multiply the resulting integral by 683 lumens/Watt. This gives the theoretical maximum lumen output of the bulb.
  • divide the bulb's rated lumen output by the theoretical one. This gives the overall efficiency factor.
  • multiply the spectrum in Watts/nm obtained above by the efficiency factor. This gives the spectrum in true Watts/nm.
  • convert the spectrum from Watts/nm to uE/sec/nm using the appropriate physical relation and constants (uE/sec = power in Watts X wavelength in nm X 8.36 10-3).
  • integrate the uE/sec/nm spectrum in the 400-700 nm range, the result is the bulb's PAR measure in units of uE/sec.
  • multiply the uE/sec/nm spectrum by a suitably digitized version of the photosynthesis action spectrum (normalized to unity peak).
  • integrate this spectrum in the 400-700 nm range to get the PUR measure.
  • integrate this spectrum in wavelength ranges 400-500, and 600-700 nm to get the blue and red PUR measures, respectively.

The above procedure computes PAR in uE/s units and not in uE/s/m2. This is so since the computation starts with a flux measure (lumen) as opposed to an illumination measure (such as lux = lumen/m2). So there is an unknown area factor involved in the translation. In other words, the computation is at the source of light, while PAR is usually defined at the destination of light.

Sources of error:

The main source of error is the published spectrum itself. It appears that manufacturers either use very low resolution spectrophotometers, or else smooth the curves on purpose (to hide proprietary information ?). A few spectra have very good spectral resolution though, but most show broader than expected emission features. The tri-phosphor bulbs in particular emit most of the energy in a few narrow wavelength regions, and if these regions are not well sampled, the lumen and PAR output can show significant errors. Spectra for the "Glo" bulbs (Hagen) are very smoothed out and show suspect shifts in the wavelength of the main emission features, so expect larger errors in these. I did some experimentation changing the emission lines a bit, both in width and intensity. I also numerically degraded the spectral resolution of the best spectra in the sample by known values to check how much this effect affects the measured quantities. Based on these experiments I can estimate that the theoretical maximum lumen output can be wrong by as much as 5-8% in the worst cases, and likely 1-2% in the best cases.

A second source of error is the published lumen output. For some bulbs the manufacturer provides both initial and design lumen specs. Most bulbs are listed with just a single lumen rating, which in most cases is assumed to be initial lumens. But one can never be certain, and comparisons are only accurate if a consistent lumen parameter is used. The results are also dependent on the assumption that the bulbs' published wattages are the actual wattages drawn by the bulbs, and not nominal values.

A less important source of error affects only the PUR measures, and is due to the fact there is no single photosynthesis action spectrum that can be applied to all plant species. Some species show a very intense response in the blue and almost none in the red, others show a more balanced response. The curve I chose has the red and blue peaks with similar values.


Lets try to use the data presented here in a real situation. I will get the lighting configuration currently on my 46 gallon bow front aquarium and try to estimate what should be the light level at the substrate. The lighting arrangement consists of two strip fixtures. One carries two 55 Watt power compact fluorescents, the other a single 3' 30 Watt T-12 fluorescent tube.

The total PAR emission of the two PC bulbs (one Dulux54 and one PLL950), according to the table, is 72.2 + 62.8 = 135 uE/s. For the 30 Watt tube (an AquaGlo) we scale by the power factor: 35.8 / 40 X 30 = 26.9 uE/s.

We must account for the fact that all figures in the table were computed under the assumption of initial lumen output. We might want to apply a correction factor to account for the mid-life loss, say 10%. So the two mid-age PC bulbs will deliver 122 uE/s and the mid-age 30 Watt tube delivers 24.2 uE/s.

Each fixture has a different type of reflector, so their contributions to the actual amount of light that make its way into the water should be accounted for separately.

The raw PAR values presented in the main table quantify the total emission of each light bulb in all spatial directions. A bare light bulb placed above the water surface will have most of its emission lost outside the water, due to the lack of a reflector, to the critical angle at the air/water interface, and to partial reflection at that interface. For a fluorescent tube placed parallel and near the water surface, we can estimate the loss due to these factors to be of the order of 75%. That is, only about 25% of the light that leaves the bulb makes its way into the water. Adding a reflector increases that efficiency considerably. Lets assume that the PC fixture reflector can divert about 50% of the escaping light back into the water. Lets presume that the the single fluorescent strip reflector can divert back 30%. Bear in mind that all these factors are very crude approximations of the real situation.

So the total amount of light emitted by the PC fixture into the water would be 122 X 0.25 + (122 X 0.75) X 0.50 = 76.3 uE/s. For the single fluorescent fixture, 24.2 X 0.25 + (24.2 X 0.75) X 0.30 = 11.5 uE/s.

Lets not forget the glass cover. Assume (according to measurements archived at the Krib site) that 10% is lost due to dirt glass. Now we can add the total amount of PAR light that makes its way into the water: (76.3 + 11.5) X 0.90 = 79.0 uE/s.

As for the light that actually reaches the substrate, we should only take into account the absorption and scattering of light by the water. Due to total reflection at the internal glass/water interface, an aquarium acts in a very similar way as a wave guide (an optical fiber is another example). So there is no attenuation by a geometrical inverse-square law effect. According to data posted on the APD list, a 16" deep freshwater layer should absorb about 50% of the light. Thus the total amount of PAR light that reaches our substrate is 79.0 X 0.50 = 39.5 uE/s.

Now lets recall that our PAR figures are for total flux, not illumination (flux divided by area of receiving surface). Lets assume as a first approximation that the light flux inside the water illuminates the entire substrate area evenly. This is not a too bad approximation in the case of long fluorescent tubes that span the entire aquarium length. Some areas will get somewhat less light, some areas more, but the average figure will not be too far. For my 36" long X 14" wide (average) aquarium, the surface is 0.32 square meters. So our final figure for the PAR illumination is 39.5 / 0.32 = 123 uE/s/m2.

Now we can finally compare this level of illumination with what plants actually require. Data posted on the APD suggest that the compensation point for photosynthesis in aquatic plants is in between 15 and 80 uE/s/m2, and that anything below 100 uE/s/m2 should be considered low light. If that is the case, we see that my setup should be able to drive photosynthesis above the compensation point for most plants, but not much. I shouldn't expect for instance to see pearling at the low levels near the substrate. And that is effectively what I observe.

One can also work backwards and, starting from a pre-defined illumination requirement, work upwards on the water column to find how much PAR flux should be provided by the light fixtures, and thus what combination of bulbs/reflectors and the like should be used.

Printed on DPH with permission of Ivo Busko

Copyright 1996/2010 Discus Page Holland.
All rights reserved