Why can't we use heat to measure led efficiency and more ac cob experiments.

So, why can't we use heat to measure led efficiency?

What I did, is got a 250mL plastic container, cut a hole for light to go through, put one of those ac driverless cobs in there, and covered it with canola oil (159.6g)

By hooking the cob up to a power supply I was able to see how much energy is being input into the chip, and by monitoring the temperature of the oil determine how much heat was output. Here's some very preliminary results:

At 103V ac and 0.35A the cob heated the oil 5.6*C in 60s (21% efficient)
At 120V ac and 0.45A, 9.1*C (16% efficient)
however at 108V dc and 0.25A the cob heated the oil by 2.3*C (56% efficient)

Some major assumptions:
1) The heat is contained by the plastic
2) The light doesn't heat the plastic
3) The current reading is accurate (0.05A resolution)

What's coming next?

A better setup made from styrofoam, a resistor as heater to verify that the specific heat capacity of this oil is actually 1.9 and an inline ammeter that reads in 0.01A. Just optimistic about these very early results and wanted to share/know your thoughts and if I'm missing anything really basic.

I hope to do a write up on the architecture of these specific cobs and the theory of why they are so much more efficient when driven with dc power tomorrow.

sethimus said:

Click to expand...

Not sure what you mean. I want to know if the method is sound.. I mean energy is only leaving as light and heat right? I plan to toss in some chinese 'cree' and epistar too to see if they have the output as claimed. I don't have an integrating sphere or an expensive reference lamp, so a plastic container filled with oil and a thermometer is the best I can do right now.

How can we know for sure what we are getting if we can't test it ourselves? I want to try just bolting a led to an aluminum block and see it it's possible to determine the heat output from the air temperature and the block temperature, but the oil method seemed more likely to work.

Interesting

WTF lay off the drugs man

WTF putting canola oil on cobs in a plastic container just sounds like a stupid thing

im pretty sure going by spectrum and par is a much better way to tell efficiency

TrippleDip said:

Not sure what you mean. I want to know if the method is sound.. I mean energy is only leaving as light and heat right? I plan to toss in some chinese 'cree' and epistar too to see if they have the output as claimed. I don't have an integrating sphere or an expensive reference lamp, so a plastic container filled with oil and a thermometer is the best I can do right now.

How can we know for sure what we are getting if we can't test it ourselves? I want to try just bolting a led to an aluminum block and see it it's possible to determine the heat output from the air temperature and the block temperature, but the oil method seemed more likely to work.

Click to expand...

my aluminium u frames i glued my stripes on are usually as warm as ambient, efficiency and how much heat my leds emit are something i cared about the last time when i needed to chose the right heatsink for my cobs, about 2 years ago...

just use more of them and drive them low, with prices as low as they are now and heatsinks are no longer something expensive, don’t waste your time on something like that, especially with ac cobs...

Octastich said:

WTF lay off the drugs man

WTF putting canola oil on cobs in a plastic container just sounds like a stupid thing

im pretty sure going by spectrum and par is a much better way to tell efficiency

Click to expand...

Please tell me how it is stupid? I have no way of testing leds shipped to me for efficiency, etc. Decided to build my own lights and have samples of cree and epistar leds as well, I want to know what is the most efficient power to run them at but don't know how to figure that out.

Now to the idea itself. The oil is to capture the heat, the plastic is to keep the heat contained so it can be measured. Ideally all the light should leave through the hole but it doesn't. I think all the heat produced is captured though, afaik heat must transfer through one object to the next. Do you think this method of measuring efficiency is accurate? why or why not? How can it be improved?

I know ac cobs are a stupid idea but I had some lying around and didn't mind ruining one with oil. It also turns out that the oil can leak some high freq. current which is a little scary.

To clarify what exactly I am after, the cree cobs I got shipped claim 210 lumen/watt and they give me the spectrum, so I am hoping that i can compare the manufactures claimed efficiency against some actual standard but I quickly found out that reference lamps are 80-100 +shipping.

Ok, so the cob internals are a rectifier and capacitor followed by four parallel strings of 36 leds (144 in total) followed by five current regulators in parallel. Pretty basic but different than the ones in the youtube teardown so I thought it was worth mentioning. The leds are probably 1/2W and turn on at 2.5V.

Now on to why they are so inefficient. As you can see in my bad drawing that the current through the leds is constant, but the voltage swings from 91.5v (leds turn on) to 169v (peak voltage at 120v ac).

When the regulator is switching, the difference between the led on voltage and the input voltage is dropped across the sense resistors of the regulators, thus the power dissipated by the regulators increases with the square of the difference between the led on voltage and the input voltage, and we would expect a minimum of wasted power as close to the led on voltage as possible (in reality it will be 10 to 20v above the led on voltage because the regulators won't switch properly at low voltages)

Something interesting is that in all the tests above the leds output approximately 14W of power, that is to say despite the different efficiencies each test output the same amount of light during the part of the cycle when the leds are on, and the only difference was the power dissipated by the regulators. From this we can deduce that each regulator is regulating to 30mA, and the current through each led is about 40mA.

One takeaway from this is that the efficiency of these driverless cobs can be tripled just by running them on dc instead of ac current. More importantly it also doubles the light output because the leds spend 36% of their time off at 120v ac and 43% of their time off at 103v ac.

Another takeaway, and I guess one of the reasons that no one will touch this thread, is that it shows these $1.60 chips can be as efficient as hps with a similar spectrum. (In all fairness I expect that 56% efficient figure to come down to around 50% with further, more accurate testing).

Will report back on the efficiencies of my cree and epistar leds as well.

Thermodynamics is a very hard subject to work on in a practical setting. Ok, so you can theoretically measure how much heat builds up in your oil but how do you calculate how much the oil loses to the general environment while its being heated up? And how much heat is lost to the front of the cob? Especially if your chips have different form factors.

Rocket Soul said:

Thermodynamics is a very hard subject to work on in a practical setting. Ok, so you can theoretically measure how much heat builds up in your oil but how do you calculate how much the oil loses to the general environment while its being heated up? And how much heat is lost to the front of the cob? Especially if your chips have different form factors.

Click to expand...

Thanks, constructive criticism is super helpful. These are important questions. Yes I am trying to minimize heat loss to the environment, my next test will be in a styrofoam container for that reason. The entire cob is under oil, but there is still a 1.25"x1.25" hole through which there was air contact with the oil. I would note that while it heated in 60s, it took 20 min to cool back down. I have set up a logging system and hopefully I can figure out and correct for how much heat was lost to the environment by the rate of cooling.

As for the different chips I was hoping just to use different lids with the smallest possible hole in each. I have thought about using glass or polycarbonate/polyacrylate but thought the additional heating due to the reflection of the glass would be worse than convection carrying the heat away from a small opening. In the initial test I was hoping that the light shining through the oil would help offset the losses through the hole.

Any ideas to improve this setup are more than welcome, thanks.

You can also make a diy "sphere tester": get a 4" wide 12-16" white tube, set your light source in one end and measure the light in the other and compare.

Just realized that if I match the refractive index of my oil and my window then there should be minimal reflective losses (at the sodium d line, how true is this across the spectrum). Maybe I should have a second chamber to move the reflective surface further away from the led under test. I have some optical glass, which is nearly transparent, but conducts heat better than polycarbonate.

Rocket Soul said:

You can also make a diy "sphere tester": get a 4" wide 12-16" white tube, set your light source in one end and measure the light in the other and compare.

Click to expand...

Thanks, I have no way to calibrate unfortunately (calibration lamps are hella expensive) Even then I am not guaranteed the response of my meter will be accurate across the spectrum, especially with a cheap meter. And even then I still only have weak metrics for efficiency which don't translate across different chips with different spectrums.

The whole point of this was to figure out the optimal power for me to run my leds at, which I can do even without accurate efficiency numbers, just got interested in seeing if I could get accurate numbers out of this method.

Ok, so... DATA!






A brief summary:

• The ac cob leds really do seem to run a little above 50% efficiency when driven at about 110vdc
• The calculated heat capacity from the resistors is 1.9. This means that 1) the earlier results are also valid 2) the measurement method is accurate because 1.9 is the literature value for canola oil, 3) very little heat is lost through the container/hole
• The ac cobs then do run at around 15% efficiency when used as intended
• The cobs really do put out a relatively constant 14W of light
• These results are not really surprising as we were hitting 60% efficiency in consumer (blue) leds more than 5 years ago. But this is really good for phosphor coated wide spectrum leds. HPS is about 40% efficient, fluoros 16%, and incandescent 9% for reference.
• The regulators really do regulate to approximately 40mA through each led at 2.5V (144 1/2W leds in four parallel strings of 36, there is a circuit diagram a few posts up)

A more technical summary:

• Why are the curves flat on top? Because the cob or resistor internals are hotter than the liquid, during this period the chip is cooling down while the temperature of the liquid is constant - similar to an ice cube melting but in reverse. The width of the plateau is directly proportional to the amount of extra heat stored by the object. This explains why the led run at 44w appears to have a higher efficiency than at 27.5w (note the width of the plateau of the red line vs the green line). This is also apparent in the decreasing values of dT/J for the three resistors (note the increasing width of the plateau as power is increased), the resistors heat up fast and store more energy that is not seen by the sensors hence a lower temperature change per unit of energy input.
• How were the values calculated? Everything was calculated using the measured temperature rise and weight of oil (135g) and the constant of 1.9 (verified by experiment using the resistors) J/g/*C, where 1 J is 1 w*s. Just google specific heat capacity if you want to know more.
• How can I replicate this? This was performed with two 250mL plastic cartons (like from sour cream), one inside the other with a hole cut in the lid. The top container was filled with canola oil and the cob as well as 3 temperature sensors were submerged. Current was measured inline to 0.01A, temperature measurement had a resolution of 0.02 degrees. No other improvements were made over the original design, KISS.

Note: the first cob didn't survive. I left it submerged in oil for about a week and when I tried to test it again about half the leds were out. It drew 3w of power at 110v meaning the regulators were still doing a good job keeping them at about 40ma