PDA

View Full Version : OT: Dissection and partial resurrection of an LED lamp



PStechPaul
05-07-2015, 02:14 AM
Around December I purchased four LED lamps from www.banggood.com (http://www.banggood.com), for just a few dollars each, and I put one in a bedroom ceiling fixture which is on most of the time. Occasionally I noticed a flicker which would usually stabilize after awhile, but last evening it got worse and then just barely glowed. I replaced it with a spare, which seems fine, and proceeded to take it apart to determine the cause of its demise and possible repair.

This is the lamp (http://www.banggood.com/E27-LED-Bulb-9W-SMD-5730-AC-85-265V-Warm-WhiteWhite-Globe-Light-p-923266.html):
http://enginuitysystems.com/pix/LED_Lamp_1929.jpg

Inside:
http://enginuitysystems.com/pix/LED_Lamp_1930.jpg

The driver PCB assembly:
http://enginuitysystems.com/pix/LED_Lamp_1932.jpg

Closer view:
http://enginuitysystems.com/pix/LED_Lamp_1934.jpg

I applied a variable voltage to the LEDs through a 240 ohm resistor and half the LEDs lit brightly at about 27 VDC, and as I raised the voltage they flickered and then the other half lit up. I was able to use an ohmmeter to light up each LED individually and found one that was bad. There are two series strings of 9 in parallel. I made a short video of the results at various voltages:

http://enginuitysystems.com/pix/LED_Lamp_1938.AVI

More pictures follow...

PStechPaul
05-07-2015, 02:26 AM
This shows half of the LEDs lit dimly at 22.3 VDC:

http://enginuitysystems.com/pix/LED_Lamp_1936.jpg

I jumpered out the bad LED and one in the other string, and there is still some imbalance of brightness, but all are lit:

http://enginuitysystems.com/pix/LED_Lamp_1939.jpg

At higher voltage, where I could get about 10 mA current, all lit brightly so I had to put a paper towel over them, but still very visible difference in brightness:

http://enginuitysystems.com/pix/LED_Lamp_1940.jpg

This was at about 40 VDC and 20 mA:

http://enginuitysystems.com/pix/LED_Lamp_1942.jpg

This was with the DC supply connected to the input, which is normally connected to the 120 VAC mains. I found that the circuit would turn on and maintain a fairly consistent current (brightness) at 38 VDC, and it would stay lit until I reduced the input to 16 VDC.

bob_s
05-07-2015, 02:57 PM
So I assume you are going to measure Vf for each device, and then swap whichever between strings to achieve an average balance between the strings.

PStechPaul
05-07-2015, 05:56 PM
No, it is very difficult to unsolder and resolder the LEDs on the metal PCB which also functions as a heat sink. I might try to cut tracks and rewire with all (16 or 17) in series which would be about 50 volts, although I don't know the rating on the IC. Or I might add resistors (or diodes) to each string as needed to balance current. It's really not cost effective to spend hours fixing a $5 bulb, but it's a learning experience.

PStechPaul
09-18-2015, 09:52 PM
Well, I replaced that lamp with another of the same type, and it started to flicker and then blinked out. I replaced it with a CFL and took the cover off the recently deceased lamp. I found some dead bugs:

http://enginuitysystems.com/pix/electronics/LED_Lamp_9W_Repair_2346.jpg

I measured the LEDs with a DMM and found five of the 18 showed 50-800 mV drop on the diode test, while the others actually lit. I found that I could remove the DEDs (Darkness Emitting Diodes) by snapping them in half with a diagonal cutter and then pulling the metal pieces off the board while using a soldering iron. Then I made some fresh solder blobs on the pads and installed some smaller 1206 size white LEDs. These are probably 30 mA 100 mW maximum, whereas the lamp is 9W for 18 devices of 500 mW (180mA) each, and they are size 5630 (metric SMD size).

http://enginuitysystems.com/pix/electronics/LED_Lamp_9W_Repair_2345.jpg

The lamp works now, and is very bright. I didn't let it stay on very long. I might just try letting it stay on and see how long it lasts. The LEDs are fairly easy to replace, and the rest of the lamp seems well made, so I might get some real 500 mW LEDs and fix it right, along with the other one. New LEDs are about $0.20 each from Mouser, so about $3.60 (plus about an hour time) to fix:
http://www.mouser.com/ProductDetail/OSRAM-Opto-Semiconductors/GW-JDSRS1EC-FTGP-5O8Q-L1N2-120-R18/?qs=sGAEpiMZZMu4Prknbu83y6gBWitgFwb%2f4jmJSRGVqG%2 f4dSgrYQJ4EA%3d%3d

There are some US sources on eBay for about half that price:
http://www.ebay.com/itm/100x-White-5630-5730-Single-Chip-Smd-Smt-Leds-0-5W-UltraBright-Light-USA-/161241663864 (http://www.ebay.com/itm/100x-White-5630-5730-Single-Chip-Smd-Smt-Leds-0-5W-UltraBright-Light-USA-/161241663864?hash=item258ac08978)

And a lot less (about $0.02 ea) from China, but they are likely the same as those that failed in the bulbs in the first place:
http://www.ebay.com/itm/100-pcs-SMD-SMT-5630-5730-Big-chip-0-5W-High-power-Warm-White-LED-/261422340819

Doozer
09-18-2015, 10:33 PM
Paul,
I gotta ask you, as you might know.
Do manufactures overdrive these LEDs on purpose to give them
a finite life span? I mean, LEDs last an incredibly long time,
so to actually sell people replacement bulbs, do they intentionally
try and make them burn out in a certain number of such hours
to have a market for replacement bulbs?
I always wondered about this.

-Doozer

chipmaker4130
09-18-2015, 10:51 PM
Seems to me its always the support/driver components that fail. They like to use parts 'rated' for actual loads with no buffer. Then too, everything is designed for adequate cooling in free air, but typically used in some sort of fixture, either enclosed or at least causing stagnation. Same problem with most CFL's.

PStechPaul
09-18-2015, 11:22 PM
I think the answer lies in the Chinese manufacturing philosophy that is based on an extreme minimum bottom line cost and factoring in the ramifications of failure. In this case, I think the LEDs themselves were overrated by their manufacturer and the company that made the bulbs just selected the lowest bidder. They probably accepted the specifications of 50,000 hour life and had no way to verify that, other than run some of the lamps for a few weeks (maybe 500 hours). I usually have the lamp on about 16 hours a day and each lasted about 20 weeks or 2200 hours. That's actually about the same as an equivalent 75 watt incandescent lamp, but less than 10% of the expected life.

The driver circuitry is probably good. I used the driver PCB from a broken CFL to repair a fluorescent desk lamp a couple years ago, and it is now well into the second replacement bulb.

Banggood.com seems to promote quality and they have a fairly good warranty policy, but I think in most cases they require the defective merchandise be returned to them. Their shipping to the US is free, but I would imagine it would cost me at least the $6 cost of the bulb to send it back to China.

They do have a customer feedback mechanism and I have posted several, including on this bulb. I had thought the first was a "fluke" but with the second failure it seems to be a clear case of improper specifications and substandard components. I have bought quite a few things from them and I think I will write a new review and request at least a partial refund. They must approve all reviews, but I think they will accept an honest assessment of the problem, and they should publish it. I would also accept replacement lamps of a different model, and I actually purchased four of them, with two given as gifts.

boslab
09-19-2015, 02:57 AM
I wonder if it's also LED lamps getting in on the act?
http://topdocumentaryfilms.com/light-bulb-conspiracy/
Mark

andywander
09-19-2015, 10:43 AM
Hi Paul:

Are they wired as:

two series strings, of 9 in parallel

or

two series strings of 9, in parallel

?

I assume the latter....

Rosco-P
09-19-2015, 10:52 AM
I wonder if it's also LED lamps getting in on the act?
http://topdocumentaryfilms.com/light-bulb-conspiracy/
Mark

Are you sh!ttin me? Anyone reporting a similar problem with Cree bulbs? What good is a guarantee on a cheap crap off-shore item where it cost more to ship back the defective item for replacement than it's worth to begin with?

boslab
09-19-2015, 11:02 AM
But perhaps they ship more of them if they fail, Paul's got the right idea fixing them, they never counted on that!, perhaps we can all benifit in the long run as I for one wouldn't have a clue how to repair something made to fail,
I'm reading with interest how this works
Mark

Rosco-P
09-19-2015, 11:10 AM
But perhaps they ship more of them if they fail, Paul's got the right idea fixing them, they never counted on that!, perhaps we can all benifit in the long run as I for one wouldn't have a clue how to repair something made to fail,
I'm reading with interest how this works
Mark

Yes, we should all encourage them to flood us with more cheap crap, designed with a short service life and then waste our time fixing something that was poorly designed to start with. Makes perfect sense.

lakeside53
09-19-2015, 11:31 AM
I have zero issues with my Cree, Philips or Osram, and I have some of the early generation.

Not all semiconductors/leds meet "spec". Do the "out of spec" get dumped? na... not if someone wants to buy them. They are graded and sold off to the lower tier manfs.

PStechPaul
09-19-2015, 05:32 PM
I decided to let the lamp run until it failed. It had already run about 2 minutes. About 1 minute into the second test it blinked briefly. I felt the fixture and it was still cool, but when I touched the plastic globe, it was slightly warm. Then at about 5 minutes it flickered a lot and finally went out. After cooling a bit, one bank of LEDs flickered while the other was dead. I found that three of the five replacement LEDs had failed to a low resistance state, and one of the original LEDs had also failed.

It may be possible to run accelerated life tests by overdriving the LEDs, but I don't know what the function might look like. I would suspect that each doubling of current might give ten times less life, so for the 30 mA 100 mW LEDs, rated 50,000 hours, 2x might be 5,000, 3x for 500, 4x for 50, and 5x for 5 hours. The current is about 180mA, or 5x, and the life was only about 6 minutes, or 0.1 hours. Of course, the replacement LEDs were also cheap Chinese devices, and may have started out as 1000 hour MTBF, in which case 5x overload would result in 1000 * 10e-4 or 0.1 hour as experienced.

I do plan to order some OSRAM 500mW LEDs in various color temperatures and use them in the two lamps I have. With proper SMT equipment it probably takes maybe 20 minutes to R&R the LEDs, and I should then have two good lamps that may last 20,000-50,000 hours, which is 3-8 years at 67% duty cycle. I also have some smaller LED lamps that work marginally and perhaps can be rejuvenated in a similar manner. I also got them on eBay and they seemed to have rather low light output and poor (too blue) color temperature, but I don't think any of them actually failed.

Does anyone have experience with more costly name brand LED lamps? HD has the Feit brand of similar size lamps for $50/6, and the reviews are good. They are rated 25,000 hours, or 23 years based on 3 hours a day. And they are also dimmable, which mine are not:
http://www.homedepot.com/p/Feit-Electric-65W-Equivalent-Soft-White-BR30-Dimmable-Enhance-LED-Light-Bulb-6-Pack-BR30DM65-LED-MP-6-24/206036849

Those are actually 3.75" diameter while mine are 3.00", so they would not fit in my desk lamp fixture. But I found 60 watt equivalent Cree lamps at HD in a "normal" bulb size for only $5 each. But it had several poor reviews:
http://www.homedepot.com/p/Cree-60W-Equivalent-Soft-White-2700K-A19-Dimmable-LED-Light-Bulb-with-4Flow-Filament-Design-BA19-08027OMF-12DE26-3U100/205597078

boslab
09-19-2015, 06:37 PM
Yes, we should all encourage them to flood us with more cheap crap, designed with a short service life and then waste our time fixing something that was poorly designed to start with. Makes perfect sense.
I'm glad we agree, it's the perfect plan, they'll never see what hit em
Mark

MrFluffy
09-19-2015, 06:46 PM
I buy a brand called Tolum, which have approximate 6-8 times the life of the chinese ones off ebay and come with a 2 year no quibble return to base guarantee. The chinese ones are a fire hazard, I've had 3 or 4 fire up to the point they stank the house out from the resulting melted plastic mess.

What you post about the strings not balanced reminded me of this :-
https://www.youtube.com/watch?v=_wR_mFWeI6Y

Edit I meant to post this link which is the walkthrough of testing some led clusters but the above is fairly relevant
https://www.youtube.com/watch?v=NjKgPLeJ79Q

PStechPaul
09-19-2015, 08:20 PM
The LEDs in these lamps seem to be reasonably well balanced, as can be seen in this simple unedited video:
http://enginuitysystems.com/pix/LED_Lamp_1938.AVI

I think the failure is totally caused by intentional overrating of the specifications, probably by 2x, and these are actually maybe 300mW 100 mA LEDs. It is interesting that the 500mW 180mA Osram LEDs sold by Mouser are far cheaper than the next higher power devices of 200-250mA, still less than 1W, at which point the packaging changes and the devices are likely multiple chips in parallel and/or series.

I enjoyed the videos of the cheap LED modules that graphically illustrate the bad chips in the series strings.

lakeside53
09-19-2015, 08:27 PM
$50 for 6? lol that's not cheap at all, and Feit is low bidder Chinese to big box stores... HD here has a new low end brand -"TCP". $19.19 for 6 x60w equv..

I bought my last 25 Cree 4-flow "60W" for $1.97 each with the local power co instant subsidy. They are $5.99 without. I can get Philips locally for $4.99 (60W), and GE 60W for $10 for 3, and those aren't even subsidized.

Careful with "reviews".. Who bothers reviewing a light bulb except the whiner... Did they put them in sealed fixtures? That a no-no with led even at 9.5watts...

mattthemuppet
09-20-2015, 11:17 AM
Buy brand name led bulbs, like cree and Phillips, it's as simple as that. If you're going to spend a bunch of time playing around with led lights just make your own, it's no harder than what you're doing and you'll end up with a far better more reliable light.

I'm all for fixing stuff that's broken but starting with cheap crap is just a waste of time

PStechPaul
09-20-2015, 02:55 PM
That argument is similar to the ones that "horriblize" the HF and Grizzly lathes and mills in favor of higher quality Bridgeport and Clausing and other such machines. But, since it may not be possible to purchase high quality new machine tools, except high end industrial models, it often comes down to rebuilding, modifying, or repairing old "American Iron" or new "Chinese Crap". I like to accept a challenge, such as troubleshooting and repairing something, even a light bulb, and take some pride in accomplishment if I succeed. I can point to it and say, "I did that".

In the case of LED lamps, there is probably no avoiding the fact that some (or all) of the components are likely made in China or Malaysia or elsewhere, and the quality may or may not be better. I think the basic quality of my failed lamps is pretty good, and it appears only that the LEDs were highly overrated. This may have been deliberate on the part of the lamp manufacturer, but more likely because of the LED manufacturer cutting costs and misrepresenting the component specs.

One does need to look carefully at customer reviews. I generally pay more attention to the negative reviews and make sure they seem genuine. And some of the positive reviews may be posted by company "shills". Lack of negative reviews on a company website can indicate censorship, so that is also something to keep in mind. I will see if Banggood posts my latest review, and if they give me a refund or replacement as I requested.

Lew Hartswick
09-20-2015, 04:58 PM
It seems to me the thing to do is match the forward drop at the current you're going to run at for all in a string. Use a curve tracer. That should at least keep the dissipation equal.
There is just no way to anticipate the "catastrophic" random failures.
...lew...

Rosco-P
09-20-2015, 07:19 PM
Buy brand name led bulbs, like cree and Phillips, it's as simple as that. If you're going to spend a bunch of time playing around with led lights just make your own, it's no harder than what you're doing and you'll end up with a far better more reliable light.

I'm all for fixing stuff that's broken but starting with cheap crap is just a waste of time

+1 to the above.

Clearly demonstrated in this thread that some have more time than common sense.

mattthemuppet
09-21-2015, 02:20 AM
That argument is similar to the ones that "horriblize" the HF and Grizzly lathes and mills in favor of higher quality Bridgeport and Clausing and other such machines.

not really.

Cheap nasty ass poor tint terrible CRI LED bulb from random Chinese store = $3-4

Decent Cree or Phillips good tint, great CRI (yo mamma, I don't look like a zombie!) LED bulb from big box store = $6-9

Even a cheap as chips CFL is better than a bad LED bulb and I really don't like CFLs at all.

PStechPaul
09-21-2015, 05:06 PM
I just received a reply from Banggood:


Dear customer,

Thanks for your email.

We are so sorry for the inconvenience caused.

For the SKU093809*4 warm white in your order 3928984 led bubls problem.
Because the deal time on 2014-12-10 , so our system has empty. We can not
replacement the bulbs and cannot refund. But we can send the BG point. Or we can
send the coupon for some product of led category .

Hoping for your kindly understanding.

Best regards,

Natalie



I don't know why they would not consider the fact that the bulbs are supposed to last 25,000 hours, and two of them failed in succession in 8 months or less than 3000 hours each (based on continuous use). No idea how much the "coupon" is good for, or how many BG points they would offer, but they basically give one point per dollar sale and 100 points is worth $1.00, so it's just a 1% rebate.

Since similar bulbs are now available at comparable prices from HD and other local stores, I certainly won't bother buying any more from Banggood. I might still purchase other items, but I would first search for similar items on eBay or Amazon, where I think there is a better chance of getting some compensation for a defective product that does not meet specifications, although there may be a "statute of limitations".

As I said, there were quite a few negative reviews on the HD site regarding short lifetime failures of the Cree bulbs. I searched for others, and usually the early reviews were "glowing", but after a year or so there were a number of reported failures. The Cree bulbs may be assembled in the US, but the LEDs are made in China, and probably come from the same suppliers that sold the overrated LEDs that burned out in the Banggood bulbs.

http://www.gizmag.com/cree-led-light-bulb-review/27285/

http://www.designingwithleds.com/cree-60w-led-replacement-bulb-review-and-tear-down/

http://www.lunaraccents.com/educational-white-LED-life.html (General info on white LED lifetime determination)

http://www.dailymail.co.uk/news/article-2546363/The-great-LED-lightbulb-rip-One-four-expensive-long-life-bulbs-doesnt-like-long-makers-claim.html

http://www.cool.conservation-us.org/byorg/us-doe/lifetime_white_leds_aug16_r1.pdf

http://www.colorkinetics.com/support/whitepapers/LEDLifetime.pdf

http://www.theledlight.com/technical1.html

https://en.wikipedia.org/wiki/Light-emitting_diode (Extensive general information)

The second link is interesting as it shows a teardown and revelation that the LEDs are driven directly from the AC line and not isolated as are the Banggood lamps. This is probably not a significant problem, as the tough plastic material of the LED lamps is much less likely to break and expose dangerous electrical potentials, which IS a problem with incandescents and CFLs.

I have not found any material that plots expected life versus the amount of overcurrent. Most references state that longevity and failures are heat-related, but only state that lifetime is "greatly reduced" at excess current. It's probably difficult to provide reliable data on a device with an expected lifetime of 50,000 hours, or almost 6 years of continuous use. The usual life expectancy is extrapolated by measuring decrease in light output over a much shorter time and assuming a linear curve, and it seems that this could be done for higher current as well. I think I will ask the experts on the sci.electronics.design usenet newsgroup.

Rosco-P
09-21-2015, 06:50 PM
I just received a reply from Banggood:



I don't know why they would not consider the fact that the bulbs are supposed to last 25,000 hours, and two of them failed in succession in 8 months or less than 3000 hours each (based on continuous use). No idea how much the "coupon" is good for, or how many BG points they would offer, but they basically give one point per dollar sale and 100 points is worth $1.00, so it's just a 1% rebate.

Since similar bulbs are now available at comparable prices from HD and other local stores, I certainly won't bother buying any more from Banggood. I might still purchase other items, but I would first search for similar items on eBay or Amazon, where I think there is a better chance of getting some compensation for a defective product that does not meet specifications, although there may be a "statute of limitations".

As I said, there were quite a few negative reviews on the HD site regarding short lifetime failures of the Cree bulbs. I searched for others, and usually the early reviews were "glowing", but after a year or so there were a number of reported failures. The Cree bulbs may be assembled in the US, but the LEDs are made in China, and probably come from the same suppliers that sold the overrated LEDs that burned out in the Banggood bulbs.

The second link is interesting as it shows a teardown and revelation that the LEDs are driven directly from the AC line and not isolated as are the Banggood lamps. This is probably not a significant problem, as the tough plastic material of the LED lamps is much less likely to break and expose dangerous electrical potentials, which IS a problem with incandescents and CFLs.

I have not found any material that plots expected life versus the amount of overcurrent. Most references state that longevity and failures are heat-related, but only state that lifetime is "greatly reduced" at excess current. It's probably difficult to provide reliable data on a device with an expected lifetime of 50,000 hours, or almost 6 years of continuous use. The usual life expectancy is extrapolated by measuring decrease in light output over a much shorter time and assuming a linear curve, and it seems that this could be done for higher current as well. I think I will ask the experts on the sci.electronics.design usenet newsgroup.

You got exactly what you paid for from your supplier. You paid for cheap crap and got exactly that. Followed up by a chinglish, so sorry email.

How did you friend fare? The one you gave the two LED bulbs to?