LED ceiling lights, part 2: learning from past mistakes

Almost five years ago, I published a review of some light fixtures, where I talked about the problems I ran into back then. Now I needed similar fixtures again, but since I already had a bad experience with this kind of product, I approached the selection a little differently this time.


First, I should mention that I needed a lot of fixtures for the bathroom, hallway, and kitchen. To keep things simple, I decided to order identical ceiling lights.
What I didn't like last time:
1. The power rating was overstated: instead of 5 W, I got 2.82
2. The reflector created bright spots on the walls that were irritating.

To be fair, though, three of the four fixtures are still working perfectly. The fourth failed six months ago. The problem turned out to be the COB LED module itself; I suspect it overheated because the fixture is above the gas stove.

Anyway, during the 11.11 sale I bought 24 fixtures. After all the coupons and so on, they cost me $180. Of course, you might ask: couldn't I have bought them locally? No, I honestly tried to find something decent, but the junk I was offered for that price was completely unacceptable. They didn't even have a proper heatsink...
The alternative was halogen. In fact, the wiring was already set up for it: all the transformers were in another room, with 3x2.5 mm² cable running to the fixtures. But halogens use more power, and a decent lamp holder costs around the same $7–8, plus the halogen bulb itself and the transformer.

The seller was a bit odd. Since I was ordering so many, I split the order into three equal parts, with three different recipients. But in the end, he gave me six tracking numbers, one of which didn't work. I ended up receiving five parcels, containing 6+6+4+4+4 fixtures instead of the 8+8+8 I had originally planned.
There really were a lot of boxes.


Their condition ranged from very neat to badly dented, and even wet (water was literally dripping out of one bag), but fortunately nothing was damaged in the end.


The fixtures themselves were unharmed because they were wrapped in plastic bags, and the drivers were simply lucky.


The seller offers:
three light color options
two housing color options
four power options
two options: dimmable or non-dimmable

I ordered mostly silver ones — link — and a few white ones — link.
The first mistake I corrected: I ordered extra fixtures. I needed 19–20; I thought about getting 22, but if I ordered 23, the extra one was practically free thanks to the discounts, so I bought 24.


The shape is slightly different: the silver ones have a bevel around the edge, while the white ones don't. But overall, it doesn't really matter. I ordered the white ones for the hallway, but after handling them, I thought I actually liked them even more. The only thing that held me back was the bathroom; I was worried the painted finish might cause problems. I was also concerned about damage, since one review mentioned it, but everything was fine—not a single scratch.


You can aim the light slightly by rotating the LED module relative to the base, though in this case it was more of a nice bonus.


There was no lesson learned here: the previous fixtures also had good heatsinks. I chose these based on the heatsink size too, since the LEDs need to be cooled properly.


The fixture dimensions depend on the power rating. The seller offers 5, 7, 10, and 12 W versions. I bought all 10 W models, so their dimensions are as follows:
1. Base without springs
2. Base with springs
3. Outer diameter
4. Height.



The driver connects via a plug, and the wires from the fixture are additionally insulated with a fabric sleeve.


Let's disconnect the driver for now and take a look inside. At first I tried to unscrew the heatsink, but it wouldn't budge, so I had to remove a couple of screws holding the base and LED module together.


Then simply unscrew the heatsink together with the reflector and remove the diffuser.


The diffuser is where the second improvement comes in. Last time, the glare from the reflector was clearly visible on the walls, making them look either dirty or like they had some strange wallpaper.


A slapdash design: the COB is simply glued to the heatsink, and the reflector is glued to the COB LED array. I liked the previous ones better.


The LED module consists of 20 small LEDs mounted on a single substrate and potted in one solid mass. The connection (as I later found out) is 2P10S, i.e. 10 in series.



The drivers come in slightly comical enclosures; they even have fake slots for plastic screws on the plastic "terminal blocks" :)
The rated current is 300 mA, which at 33 V should give the rated 10 W.


They come apart very easily; there are a couple of holes in the bottom cover for mounting the driver.


The boards are tiny, considerably smaller than the enclosure itself.


Still, for a rated 10 W, they have a decent transformer and overall design.


1. The input has a capacitor with a stated capacitance of 10 µF, which is fine for our mains supply and the rated power.
2. A resistor is used instead of a fuse.
3. The driver is built around an FT8350 PWM controller.
4. The output has a 47 µF, 50 V capacitor, with a resistor in parallel. They even remembered to add an interwinding capacitor to reduce radio interference.


I found both the datasheet and an application example online.



The datasheet also included a typical application circuit.


But that's not all: it turned out that the board in my drivers is identical to the one shown in the description.


It also lists operating temperatures and efficiency at different supply voltages, with an output for 12 LEDs connected in series at 250 mA. It's all in Chinese, but it's quite possible to figure out what is what.


Before confirming receipt, I checked all 24 lights. In one of them, the output capacitor was loose inside the driver enclosure—it had either come unsoldered or was never soldered in. Other than that, not a single problem.


The light is even, so no complaints there. The trade-off is lower light output, though—the diffuser is quite dense.




The seller offered three color temperatures: cool, warm, and neutral. I chose warm because I know cool white is usually too cool, and warm suited the finish, which was also in warm tones.
Basically, I got more or less what I was looking for, but on closer inspection you can see that the color has a greenish tint. It's a little less noticeable in person, but it's still there.
On the left is a photo under ordinary incandescent lighting, with automatic white balance calibrated against a sheet of paper. On the right is the same sheet lit by this fixture, with no further white balance correction.


Third time's the charm: learning from my mistakes. Since I knew perfectly well that the actual power wouldn't match the rating anyway, I bought a 10 W light instead of a 7 W one.
The reasoning was a little different, actually. I needed about 5 W, but for reliability I was thinking of going from 7 W down to 5 W. Since they usually don't actually deliver 7 W, though, I went with a 10 W one :)

In reality, it turned out that the power was close to the rated value.
1, 2. Power draw from the mains is 9 W right at startup and 9.3–9.4 W after a minute of warm-up; it doesn't change after that.
3, 4. The voltage across the module drops slightly as it warms up, settling at about 30.8 V.
5, 6. The current barely changes as it warms up and stays around 257 mA.

So the LED array gets about 7.91 W, almost 80% of the rated power. I don't think the difference is that big, and I didn't open a dispute because I was expecting an even worse result.


And of course, heat tests. First I just grabbed a couple of lights, one silver and one white, and turned them on.


I ran them like that for about 15 minutes, but then I remembered that the lights would be installed in the ceiling void, so I covered one of them with a plastic jar :)


The results:
1. 15 minutes after switching it on, the heatsink was at 42 degrees. Then I covered one of the lights.
2. The uncovered light after another 20 minutes: 55 degrees
3. The covered light after the same 20 minutes: 60.5 degrees.

I think those are pretty good results.


1. The driver heated up to 77 degrees; the hottest component was the transformer.
2. Temperature under the reflector: 53 degrees.
3. The LED array itself was around 65 degrees.

In reality, you can safely add 10–15 degrees to the last two readings. Although I didn't switch off the light, it took some time to take it apart, and the assembly cooled down a little in the process, so the temperature dropped. In any case, I didn't see any overheating, even under normal operating conditions.


Next, I tried to measure the illuminance. I hung the light 1.5 m above the floor and placed a smartphone underneath it to use as a lux meter. The second measurement was taken half a meter to the right of the optical axis.


Everything was going fine until I decided to compare the readings from two smartphones as an experiment.
The results from the Lenovo P2 are on the left, and those from the UMI Plus E are on the right. Quite a difference, huh?


I mentioned earlier that I wanted to reduce the lights' power at the expense of brightness, in favor of a longer service life. To do that, you need to change the values of the driver's current-sense resistors.
In this case, there are two resistors, 1.74 and 1.65 Ω, connected in parallel.


In the schematic, this is the resistor connected to the CS (Current Sense) pin.


If you're designing the resistor value from scratch, it depends on the required output current. But I wanted to reduce the power by about 1.5 times, so I took the simpler route.
1. I calculated the total resistance of the resistors. You can do this with a calculator, but I was so lazy that I just went online and entered the values into a special calculator.


The value is 0.847 Ω.
2. To reduce the current by a factor of 1.5, I needed to increase this value by the same factor, giving 1.27 Ω.
3. Since it's better to use two resistors, I multiplied 1.27 by 2 and got about 2.5 Ω. In simple terms, two 2.5 Ω resistors in parallel give a resistance of 1.25 Ω, which is close to what I needed.
4. I didn't have any 2.5 Ω resistors, so I went to the store and bought a hundred 2.4 Ω resistors, and just in case, another hundred 3 Ω ones (2.4 + 3 gives 1.33 Ω in parallel).


I soldered in a pair of 2.4 Ω resistors to try it out.


In the end, I got roughly what I wanted: input power of 6.4 W, current of 181 mA, and voltage of about 29.9 V. The voltage drops a little as it warms up, but not as much as it did before the modification.
The LED array now gets 5.4 W instead of 7.9 W, which I'm completely happy with. That works out to 0.27 W for each of the array's 20 LEDs.


The brightness has decreased, but not by all that much. The illuminance compared with the unmodified version: before the modification is at the top, after at the bottom. The Lenovo phone is on the left, and the UMI on the right.


The temperature has also dropped noticeably. These readings were taken after about 40–50 minutes of warm-up:
1. Open
2. Closed
3. Driver.

You could say the assembly barely heats up at all, so I can leave it like this.


When I was separating the drivers from the lights to modify them, I happened to notice that some of them had ordinary wires inside a fabric sleeve, rather than ShVVP cable. I opened it up and saw that the driver was different.


These drivers are slightly larger on the outside, and the transformer is clearly a little more powerful, but the output current is the same.


It is built around a BP3125 PWM controller, and the current-setting resistor values are slightly different.
The driver itself is also different: it has a separate rectifier to power the controller, and the board layout is different overall, with some components located under the transformer.


Basically, it's just another controller. I'm not too fond of the little flame in their logo :)


Although the current-setting resistor values are different, their total resistance is the same, so the same 2.4 Ω resistors I used above will work for the modification.


In the end, my only complaint is the greenish tint of the LED arrays. It's barely noticeable in real life, but you can see it under certain conditions. Otherwise, there's no overheating—which is unusual for Chinese devices—the design is decent, and the price is reasonable. Even the difference between the actual power and the stated rating isn't that large, which also surprised me :)

That's all for now, but only for now, since I'll be modifying the drivers further to extend their service life. To be continued.