Heat of LED vs HPS

Slippertalk Orchid Forum

Help Support Slippertalk Orchid Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.


Well-Known Member
Apr 7, 2012
Reaction score
Fairbanks, Alaska, USA
Instead of hi-jacking Dot's excellent GH thread, a new thread about they myth of LED efficiency (continuation from):


After Bjorn mentioned that the heat generation of LED isn't so different from HPS, I was curious about this topic, too. Probably, only a few people care about this, but I thought I'll write it down since I spent a bit of time looking around the info (and learning about it).

So there are several different ways to state "efficiency of light bulbs".

luminous efficacy: expressed as lumen per watt

how efficiently electricity energy get converted to HUMAN visible light. This is the easiest statistics you can see for consumer products. But it isn't the best for plants because human eyes are more sensitive to green, and the light with lots of green seems to be efficient to human, but not to plants.

photon efficacy (or photon efficiency) of photosynthetically active radiation (PAR): expressed as micro mol/J (=micromol/s/W)

micromol/s is the NUMBER of photons emitted per unit time. The number of photons between 400-800nm of wavelengths is related to amount of photosynthesis. So this is the most relevant efficiency for plants; it tells how effectively electric energy is converted to the light relevant to plant.

radiant efficiency (= Wall-plug efficiency): % (= Watt/Watt)
Each photon has energy, and the amount of energy is determined by it's wavelength (a blue photon with shorter wavelength carries more energy than a red photon). By looking at the emission spectra, you can calculate how much total energy the emitted light carries. So this efficiency measurement tells us what proportion of electric power is converted to light. Let's say this efficiency is 30%. Then 70% of energy is wasted, and released as heat.

By looking around, I found some number of radiant efficiency of LED and HPS. Both of them are around 30%. So if you put 400W of LED or 400W of HPS in a same size enclosure, the temperature of the enclosure will be similar. So Bjorn is right about it. Here is the link (in 3rd and 5th paragraphs; search for "wall-plug efficiency"):

Table 1 of
has radiant efficiency (which they call electrical efficiency) of red, blue, and white LEDs.

But as the 1st article mentioned, radiant efficiency of LED is dramatically influenced by the design. For example, with modern COB (chip-on-board) LEDs (e.g. Bridgelux Vero, Cree CXB etc), you can achieve 40% radiant efficiency easily. If you design LED to massively under drive (e.g. 100W capable LEDs with 10W driver), you can even achieve >50%. So in this case, LEDs can release less heat than HPS for a given input watt.

So LED could release less heat than HPS, but it seems that in general, both of them release similar amount of heat.
At 350 mA a Cree Cool white should give about 36% efficiency. This drops to 23% efficiency at 1500 mA. You get more micromol/s/W of light at lower currents.

I have four 7w COBs running over my aquaria. the operating temperature of -10 to 40 oC so I have had a thermometer on the heat sink monitoring things. I'm running them about 30% of designed current and still get plenty of light (enough for the aquarium plants to pearl). By over-installing more LEDs than needed and under-driving the LEDs you can boost efficiency with little loss of actual light output.

Thanks for the data.
Thanks guys, interesting stuff. So, since I build my own set-ups, if I add anoter 100w chip to the existing ( two 100wchips), (paralell or in series? ) I should be able to squeeze more light out of the driver (e.g32V/3000ma)?
I use Cree XT-E chips of 100 or 50w for my setup, color mix of warm white and white 3000and 6500K 50/50. This gives a light that gets close to natural, and the plants seem to grow nicely as well. Might not be the most efficient chips, but they suit me and are easily available. However if it was possible to squeeze some extra PAR out of them, I think that would be worth the investment.
Yes, you can. If you add parallel circuits, you'll lower the current for each diode. Then the efficiency goes up. So you'll get a bit more output with the same amount of energy consumption. But there is a diminishing return. What's the current configuration (# of diodes, driving current, driver model)?

If you want additional light (with more energy consumption), making 1 or 2 COB light is probably easier than adding bunch of XT-E, though. The cost of 1x Bridgelux Vero 29 driven at around 80-100W is around $65. 2x Vero 29 driven at 50W each would cost about $110 (more initial cost but higher efficiency). With the 2x 50W configuration, radiant efficiency is above 40%, I believe.
This is a bit off-topic, but I came across this interesting experiment.


The radiant efficiency of LEDs (i.e. (power of light emitted)/(input electric power)) can be as high as 50-60% with DIY LEDs. The experiment showed that the efficiency of LED can be as high as 230%! 69 pico watts of light came out from input electricity of 30 pico watts. This sounds impossible, but actually, it seems to be sucking up heat to covert it to light energy (I don't understand the exact mechanism). Maybe all of us can grow high altitude orchids easier and cheaper!
Can I add a few cents to this topic?

Although it's true that 100W LED = 100W HPS based on power consumption, LEDs can be biased to emit a certain frequency of light useful
to plants where as HPS will emit a broad spectrum of light of which a big portion of frequencies are not usefull for plants and are just wasted as heat (like UV). Whereas, LEDs can have very specific frequency that are useful for plants, so that a 100% of the 100W output can all be RED light or BLUE. No wasted frequency/energy.
In other words you get more of the needed light from LED than from HPS of the same power consumption.

Thank you.
That is true,color is possible to tailor with LED's. But since many of us like to see natural colors of our plants, at least I tend to end up with white LED's whose eficciency is more or less equal to what you get with HPS. When that is said, HPS does not give nice light, white LED's give much better. But then there is the price difference and logivety. If I build the lamps from components aquired on e.g. e-bay the price of LED get aprox double the price of a cheap HPS setup. But the logivety of the LED is more variable.
So a direct comparison gets difficult. I do play around with LED but I have no plans of replacing HPS with LED at the moment. For new installations the question gets a bit different, then LED's are absolutely candidates. But not the colorful so-called grow-lights. I og for white LED's:D
LED's don't emit UV rays like HPS does. That's the kind of heat that just wasted unless you need heater for your greenhouse. If one is using HPS at the same time cooling your GH with a fan or an evaporator cooler to bring the temp down, then your wasting energy twice.

Be careful with LED just because they draw less power, 50W of LED light will burn your plants if you place them too close. Leaves will turn yellow pretty fast.
I agree, gego, HPS has other light emission. Some of the emission spectra of HPS which you can find on internet is a bit misleading because the longer wavelength is cut off. But this seems to be a better representation of HPS emission:

This is Fig. 14 of this paper (it contains emission spectra of MH and other light source, too):
Elvidge, C. D., Keith, D. M., Tuttle, B. T., & Baugh, K. E. 2010. Spectral identification of lighting type and character. Sensors, 10(4), 3961-3988.

The largest peak is at 819nm, which is infra-red, and not useful for plants (400-700nm is useful for photosynthesis). The Y-axis is in radiance, so it is power (watt) not the number of photons. I didn't know that so much energy is used for this IR range until you point it out (I think you meant IR not UV because UV emission seems to be minimal as you expect for light used near human). This is the reason why we feel HPS is warm (throws heat forward).

With LED, even white ones emit mostly visible range (and relevant for plants) (400-700nm). But this doesn't mean that LEDs give more plant relevant light (PAR) than HPS for a given input watt (at least not yet). Take a look at the "Photon efficiency" column of Table 3 of PLOS One paper I linked above. This paper is measuring only PAR (400-700nm), and the photon efficiency here doesn't consider the infrared peak of HPS. So even though HPS waste energy to IR emission, the top performing HPS is similar to the top performing LED (around 1.7 micromol/J) in terms of PAR. Most of inexpensive LEDs (excluding high quality DIY LEDs) have much lower efficiency than 1.7 micromol/J, though. The paper also has a long term cost analysis of HPS vs LED.
Hi Naoki,

This study is not geared towards plant application. HPS output is not variable. But LED can be varied to maximize the LED output to the type of light a plant needs. It this study, they are using blue LED with coating to transform some of the blue to green ang yellow. Transforming from blue to other colors will have losses and not the best way to do it. The best way to do it is to have individual LEDs for each color.
Also, all those light sources except LED run directly from a 110V or 220V outlet. LEDs run at very low voltages, (1-2.5VDC depending on how much it's driven) a big chunk of losses are coming from power supplies to step down and convert AC to DC voltage. Depending on the topology of the power supply and of course the cost, efficiency can be as low as 50% but could be designed to go up to 90%. Since LEDs are expensive, a manufacturer will usually skimp on the power supply. The good thing about LEDs are that they can be designed and tailored for plants and you really don't need a lot of power because you can bring these source very close to your plants without burning your plants. HPS needs to be placed way farther and that distance is a loss. Not to mention that if you have a big green house you may need to cool down the heat.
Gego, are you talking about the Sensors paper? I referred it just to get the HPS emission spectra, and I agree that other than the spectra, there is nothing relevant for plants.

You should take a look at the PLOS One paper I linked in the first message:

The LEDs which achieved 1.7 micromol/J are red+blue types or white+red.

Yes, monochromatic LEDs are a little bit more efficient than phosphor-based white LEDs. There is a loss during the phosphor phase. However, with the advance in the white (much bigger market, so companies put more effort there), the difference is getting smaller. With modern white COB LEDs (Cree CXB3590, CXB3070, Bridgelux Vero 29 etc), you can achieve radiant efficiency of 60% if you do DIY and underdrive. Yes, you can do the same underdriving with blue + red to increase the efficiency, but it becomes pretty tedious since you need to use lots of individual LEDs. Once you start to look into the cost comparison, underdriven white COB becomes a really attractive option to get high efficiency with low investment. This is a part of the reason that some companies of grow light are starting to migrate to this path over the traditional red+blue. If I can get decent quality monochromatic COBs (there are some sketchy ones in ebay), I would consider going back to R+B, but I'm pretty happy with Vero29 and CXA3070.

AC/DC conversion efficiency of the constant current driver is about 90%. Even the cheap ones ($12) I use from ebay achives this (it is easy to measure the loss at this stage). If you go with Meanwell HLG-185H-C, you can achieve 94% efficiency. So it is true that you should pay attention to the driver AC/DC conversion efficiency, but you can get pretty efficient ones for cheap.
I know these things because I work with LEDs and semiconductor and I design power supplies. That's why I said you can't compare LEDs with HPS, they are not orange to orange.

So you can measure the AC power input and the load to your power supply?
How do you measure the total power dissipated by the load which are the LEDs?

As your over all power from the total number of LEDs increases, higher efficiency can be achieved on the power supply. 94% is pretty good. Maximizing the number of LEDs you connect in series and loading the power supply up to 80-90% of the rated output of the power supply, will maximize your efficiency. You probably know this already.

I'm excited to hear from the real professional in LED industry! I'm self-taught about electronics, so I understand the biology side, but I would love to learn more about the electronics side!

I don't quite understand why you can't compare LED with HPS, though.
Bugbee's group has been doing horticultural research about light, and is there a methodological problem in their paper? To me, it sounds pretty fair comparison of LED vs HPS (and others). Minor issue is that PAR put same weight to each photon between 400-700nm instead of weighing them with the action spectrum of photosynthesis (i.e. yield photon flux, YPF). But this is rather minor, right?

It may not be accurate, but I measure the AC power input with Kill-a-watt. Then from the DC side of the driver I get the current and Vf with digital multimeter, so that' how I calculated the load by LEDs. Then I calculate the AC/DC conversion efficiency by DC LED watt / AC input watt. I have this kind of efficiency measurement in Message #16 of
Is this wrong?

Yes, I try to find the driver which will operate at the near maximum output for a given LED (and I know that the Meanwell's published efficiency is the maximum efficiency and realized efficiency can be lower). One thing I don't know is what happens to the overall efficiency if you use the dimmer (like HLG of Meanwell). As you decrease the current, LED radiant efficiency goes up. But I'm guessing the driver efficiency goes down, right? Do you know what happens at the system wide? It obviously depends on the specs of the LEDs and drivers, but is there a general statement (e.g. something like Meanwell HLG and Cree CXB, which are pretty common combinations these days)?
It may not be accurate, but I measure the AC power input with Kill-a-watt. Then from the DC side of the driver I get the current and Vf with digital multimeter, so that' how I calculated the load by LEDs. Then I calculate the AC/DC conversion efficiency by DC LED watt / AC input watt. I have this kind of efficiency measurement in Message #16 of
Is this wrong?

That's right. If your watt meter is good then the rest should be good. One thing, you must use the voltage of the output of the DC supply and the current output. To maximize efficiency of the power supply, when running on current source, maximize the number of LEDs in series so that the output of the power supply is at the rated output voltage or close to it.
LEDs are just like any diodes, depending on the quality, there is a diminishing return in terms of efficiency. The more you drive the diode, the higher the power loss, the diode junction heats up, then more losses. So you add more heatsink or use a bigger blower and more input needed. But you don't need to drive them hard. Instead of driving them hard, bring those LEDs closer to the plants, because YOU CAN. Get quality 120 deg lenses, space them out according to what level of light you want to provide to the plants. You can adjust the bands to your preference, have morning lights then mid day lights and afternoon lights and you can even alter the position of the source of the light accordingly. You just can't compare HPS with LEDs.
gego, what is best: to have the LED (here COB) in series or in parallell if you want to maximise efficiency? Assume that the driver is rated according to one COB. Just curious because I am playing around with COBs.
Bjorn, as you know, constant current drivers have a range of voltage where the constant current can be provided. For example,
HLG-185H-C1400 can provide 1400mA in the region 71-143V.

To get the maximum efficiency, you want have the forward voltage of LEDs close to (but slightly less than) the maximum of this region. Many good COBs requires around 36V (varies depending on the driving current, so you need to check the datasheet). For example, Cree CXB-3070 uses slightly less than 36V (see p.6 of this datasheet): http://www.cree.com/~/media/Files/C...Modules/XLamp/Data and Binning/ds CXB3070.pdf
So if you connect 4 CXB3070 in series (4 * 36=144), it will use slightly less than 144V. This is at the maximum range of HLG-C driver. This means that the CC driver would be operating at the maximum efficiency.

If you want to further increase the efficiency, you can make multiple strings of serials in parallel. So let's say you make 2x serial loops and connect them to a driver in parallel (4 cxb3070 per loop, 8 cxb3070 total). Then each serial loop get 700mA (1400mA divided by 2), and the voltage is same < 144V. So the driver efficiency is same as before (1 serial loop of 4 CXB3070), but the efficiency of each COB LED increase.

In general serial is safer. If you have 2 parallel circuit and one LED breaks, then the other loop get double the amount of current. If the current is within the range, it will be ok, but if not, it might damage the other LEDs. However in the above example of 2x (4 serial CXB3070), it is underdriven. So even one circuit dies, it will get only 1400mA (max current for CXB3070 is 2800mA). So this is completely safe.

In serial, if one LED breaks open, no current goes through, so no chance of damaging the rest of good LEDs. When the no current is running, it may stress the CC driver, but I think good drivers can handle the situation (of infinite resistance due to open circuit), but gego probably knows better about the protection circuit of CC driver.

The total watt is about 200W in either 4 serial or 4+4. If you have 8 COB connected in this way, the radiant efficiency is around 60% I believe. Alternatively, HLG-185H-C700 (700mA, 143-286V) can be used to drive 8 in a single serial loop.

Indeed, this HLG-185H-C 700 or 1400 and CXB3070 (or CXB3590 if you can afford) is a popular combination among MJ people. But the DC output voltage is pretty high, so you have to be careful.

Which COB are you thinking, Bjorn?
naoki, I have been playing around with Cree XT-E, particularly like this combo of white and warm white
What I have done recently is to use two of these 100W in parallell on a 100W driver (typ 24-36V, 3000mA). Visually, there is not much difference in light from the paralel coupled and the single, but I have not measured anything. My hope would be to improve life a bit, heat is always an issue dealing with these things:p
btw here is a Picture of my 200W set-up (1x100W + 2x50W) run with a powersupply of 2 X 100W in parallell (makes it possible to use one cable for the power supply)

upload pics
That looks like a nice heatsink! Is it pre-drilled or custom order? I haven't seen this type of LEDs; it's not technically COB (chip-on-board), but it is much easier to deal with the individually mounted diodes.

From the spec, it has 10 XT-E in series. "100W" one has 2 parallel circuit and "50W" has a single serial circuit. This is why your configuration of connecting these 2 kinds in parallel works.

Here is the data sheet of XT-E:
http://www.cree.com/~/media/Files/Cree/LED Components and Modules/XLamp/Data and Binning/XLampXTE

When you were running 100W unit with 100W CC driver, each diode was getting 1500mA!!! This is the maximum current allowed from the data sheet. Most of us generally use much much lower current. You can see the efficiency vs driving current in this page:

Select the following option:

Model: Cree XLamp XT-E {White} (this is near the bottom of the options)
Flux: R5 [139] (this is "bin")

Current Range: Coarse (0.15-2A) (this option is top right of the page)

Note that I'm not sure about the bin of yours, so I used the highest efficiency XT-E which are commonly available (R3, R4, R5, with R3 have lower efficiency).

If you are driving it at 1500mA, you are getting only 93 lumen/W (total of 491.4 lm per LED). This is slightly better than fluorescent light. So the single 100W board is outputting 9828.9lm.

Now after connecting them in parallel, each diode is getting 750mA. Now the efficacy goes up to 124lm/W (total output of 299.6lm per LED). So a single 100W board is emitting 5992.5lm. Now you have 100W+50W x 2, then the total output from the 3 boards is 5992.5 * 2 = 11985lm. So over all, you increased the total output about 21.9% by adding the additional parallel circuit. The difference is probably something you can't detect without measuring, though.

Note that in the Cree Product Characterization website, I left the junction temperature to be the default 25C, but if you are driving hard, this temperature would be much higher (maybe 50C or more?), so the actual output (and efficiency) is lower than the values I mentioned above.

When I make LED fixtures, I usually start from the nominal current (or testing current) in the datasheet, and goes lower than that (to get high efficiency). The nominal current is 350mA for XT-E, at 350mA, XT-E achieves 151lm/W. But you'll need lots more diodes, so the initial cost significantly goes up. I goes for efficiency because the electricity is expensive in alaska. But you mentioned the electricity is pretty cheap in Norway, so it is probably better to go with lowering the initial cost as you have done.

Just for a comparison, if you use Cree COB CXB3590, this is the number. 2x CXB3590 parallel in the same 100W driver: 155.7lm/W, total output (from 2x): 8178.5 * 2 = 16357lumen. From 1x CXB3590 to 100W driver: 121.4lm/W, total: 13677lumen. But CXB is much more expensive ($50-70 per each). So $20 for your XT-E with 20 diodes, is a pretty good deal.
In greenhouse, you might want to cover up the solder points with something. It's unlikely something happens with that much distance, but just for the safety. Kapton tape or some silicone caulk?

Latest posts