Voltage Regulation for LED Lighting

[geoffschultz]

A couple of years ago I installed DoctorLED (http://www.doctorled.com/p24.htm) Mars Dome interior lights throughout BlueJacket. This was the first shipment of these to the US and I literally had to wait for them to get off of the container and into my hands before I could depart for the NW Caribbean. They may have changed the design since then, but this is what I've got.

Anyhow, about a year later I realized that they were getting dimmer and dimmer and recently took a failed unit over to a friend who's an excellent electronics engineer to determine what had failed. He found 2 problems with the units:

1) There's a layer of thermal paste which is supposed to conduct the heat between the circuit board that the LEDs are mounted on and the aluminum backing plate which also functions as a heat sink. For all intents and purposes, there was too little paste to provide contact between the two. However, after running a working unit for about 10 minutes, there was no appreciable heat buildup, so we don't suspect that my problem was due to thermal overheating.

2) The basic design of the circuit is a transistor which limits the voltage to 9V to 2 parallel stacks of 3 LEDs in series. This provides 3 V to each LED, which is just fine. However, the transistor is too slow to handle fast transient voltage spikes, and there's no voltage dump circuitry on the boards to handle this. A spike probably occurred somewhere along the line and fried some of the LEDs.

DoctorLED wants to sell me replacement circuit boards at about 1/2 of the cost of the whole unit, which is still expensive for a design flaw especially when they are supposed to last some 50,000 hours. I may do that, but I want to protect against this happening in the future.

-- Geoff

 

[Maine Sail]

If they

If they are three years old you might consider new bulbs with better circuitry, more output and a warmer color.

The Sensibulbs are simply amazing, bright and warm and also have the widest beam angle I've tested. I have one of the original Sensibulbs going on four or five years now and it is still as bright as the day I bought it

It is very frustrating to hear of your problems. I hope it works out..

 

[geoffschultz]

The DoctorLED bulbs are warm white, not daylight. There's nothing in the SensiBulb documentation to indicate that they have handle voltage spikes any better than DoctorLED. This could have been a 1 time glitch that fried them, but I want to keep it from happening again.

 

[Blitz]

Try new - you'll like it

If they are three years old you might consider new bulbs with better circuitry, more output and a warmer color.

The Sensibulbs are simply AMAZING, bright and warm and also have the widest beam angle I've tested. I have one of the original Sensibulbs going on four or five years now and it is still as bright as the day I bought it..

I have to agree, Sensibulbs are amazing in how bright they are (assume they are brighter than what you have). Not sure if it is worth to keep the older technology in this case.

 

[tsmwebb]

Since they are designed to take up to 30v DC they ought to be able to take most spikes on a 12 volt system... I suspect if you did spike them it was a one time deal-- if you were getting those kinds of voltages on your bus very often you'd probably notice other electronics letting out their smoke.

I'm not sure if this is relevant but: "white" leds use phosphors in a way similar to florescent lights. They all get dimmer with time. Some early versions of bright white leds had phosphors that got used up pretty quickly particularly if run hot. If the leds are still functional but dim it is possible that they have just been used up and that there isn't a voltage problem at all...

--Tom.

 

[Witzend]

Tom has it correct, E spikes in this system that would damage an LED would probably never happen unless you get struck by lighting and then they just would not work at all.
Now that they are running dimmer they won't be produce the same heat so drawing the conclusion that they were not sunk properly is incorrect.

 

[tsmwebb]

...I was thinking about a 12 V DC to DC power supply. Any suggestions?

Could you just add a transient voltage suppressor (eg:http://www.onsemi.com/pub_link/Collateral/1N6267A-D.PDF)? I ask because I'm hoping somebody with real EE creds will set me right. I'm in way over my head here, but a DC/DC converter seems kind of tangential to the problem (assuming the problem is spikes) and redundant and expensive both in terms of $$'s and watts...

--Tom.

 

[Witzend]

Again, barring a lighting strike, the only E spikes you would ever see would be a result of your alternator and as long as you don't open and close the circuit from the alt to the bat. while running (and if you did you would have bigger problems) your not going to need such a transient E suppressor. This type of device is used in applications where sensitive CMOS devices need to be protected. Your LEDS are not in that category.

 

[walt]

I would agree that the problem is not likely voltage transients.

The LED light fixture in the link from the initial post is fairly high current (for a dome light that is) - the spec says .25 amps at 12 volts and .13 amps at 24 volts. That is 3 watts and in a very small area, this can produce some high localized temps. If there was not good thermal paste between the LED's (surface mount?) and a heat sink, I would think this is more likely the problem. Im not sure what test you did for heat but if you could have touched the die for the LED's which had poor heat sinking, they would have felt hot.

Note also that the power the fixture uses is fairly constant between 12 and 24 volts - at about 3 watts. This means the "regulator" is a switching type - ie, it has a low power switched DC to DC converter function. A simple linear voltage regulator would have had twice the power at 24 volts vs 12 volts. Typically LED driver IC's actually hold the current into the LED's constant rather than holding the voltage constant. If you just try and hold the voltage constant to a LED, the light output will vary a lot from hot to cold and from part to part. The LED would get brighter as temp goes up. Based on the literature also saying the light output is constant, I would guess this design uses constant current drive to the LED' rather than a 9 volt regulator. Note that with the constant current, if you measured the voltage on the diodes, the design current may have produced 9 volts on the diodes so it "looks" like a 9 volt regulator.

I dont think the problem was voltage tranients but the simplest way to solve this would be to put something like a 100 uf 25 volt electrolytic cap on the DC voltage just before it goes into the LED fixture.

 

[Witzend]

When semiconductor devices get hot their resistance goes down, as a result they start drawing more current. Walt is correct in that the regulator is supplying more current to keep the voltage at 9v, after all that is what a regulator is suppose to do. This keeps the LED at a constant luminosity despite the temperature. My guess is that the LED has gotten too hot and as a result this damaged the junction, when it cooled down it now has a greater resistance across the junction and can not draw enough current to light up fully. If you want to test this theory, bypass the regulator and connect them directly to a variable bench DC power supply. I bet you can get them more excited with 10v, but in the end they are probably toast.

 

[rardiH36]

At the risk of escalating this thread to a higher level of complexity, a suggestion that you post your situation on www.allaboutcircuits.com

Specifically, this url (http://forum.allaboutcircuits.com/showthread.php?t=18185) contains the history of my questions on the forum several months ago when I wanted to convert a great post Christmas deal on very bright warm white led's ($2.50 for each string of 35 leds) that I thought I could use to DIY into my existing incadescent dome fixtures (7 of them) for my 1980 Hunter 36. The suggestions provided by the circuit forum regular SgtWookie have worked splendidly. (My take is that he's like the equivalent on the circuit forum as MaineSail is to Sailboatowners.com - very knowledgeable individuals in their respective fields).

Although it looks like there are several grounded (no pun intended) solid state electronic people on the Sailboatowners.com website, a post on the circuits forum might provide some alternative insight. Just a thought.

 

[Maine Sail]

There's nothing in the SensiBulb documentation to indicate that they have handle voltage spikes any better than DoctorLED.

Other than this:

Sensibulb™ Specification:
Size: 1.00" x 1.00" x 5/8"
Weight: 10g
Typical Current Usage: 0.16 Amps(warm) - 0.2 Amps(cold)
Voltage Range: 10-32 Volts DC
Typical Operating Temperature: 140° F
Ultra Violet Emissions: 0.00

The operating voltage of the Sensibulb can be as high as 32 volts. This means even a short spike to 18 volts or 20 or 24, on a 12 volt system is well within it's range of normal operation. I am going to guess you have a 12V system not 24? If you had a voltage spike anywhere near 32 volts on a 12 volt system a lot more would be toasted than just a couple of bulbs..

Even the Dr. LED bulbs operate on a 10-30 volt range for normal operation so I question whether the problem is a voltage spike. In order for this to be caused by a spike wouldn't it need to be over 30 volts? This clearly would have certainly caused other issues if it happened..

 

[geoffschultz]

First off, I want to thank everyone for their suggestions and feedback. I've been doing a 150 mile/round-trip trek to the boat every day to get it ready to launch on Friday, so my only chance to respond is first thing in the morning before I head back to the boat.

After looking at the SensiBulbs, they do indeed look like a great alternative. Maine Sail published a great comparison of various LED bulb replacements on the SSCA web site (http://ssca.org/phpBB3/viewtopic.php?f=4&t=7969&start=45). I don't know if it's reproduced here, but it's well worth looking at.

The correction to my original post stating that the circuit design provided current limiting instead of voltage control was indeed correct. This is what happens when one starts writing 05:30 after just rolling out of bed and not being fully awake (and not completely conversant in the subject.)

I've asked my friend, John, who looked at the LED lights with me, to step into this thread and provide a much more professional analysis than I can provide. Hopefully he'll do so today.

Based upon John's analysis, I still believe that the problem was probably caused by a voltage spike. According to John, this is a significant issue for the auto industry as they move to LED based systems, and is not easily (cheaply) solvable. Based upon my understanding of the problem, the issue is that transient spikes can occur in a very short duration, which is faster than the circuitry can respond to the spike. While the light may be designed to run from 10V to 32V steady state, if that 32V spike only lasts 1 ms, a sizable portion of that spike may make it to the LEDs, burning them out.

The reason that we don't believe that the issue is heat related, is that after 10 minutes of operation (using a fully functional unit), the circuit board where the 36 LEDs were mounted was only warm to the touch. An IR meter showed a maximum temp of 120F. Charts showing LED longevity vs temp showed a lifespan of 100K hours at this temp. While the thermal paste clearly wasn't doing its job, we just didn't feel that was the issue.

-- Geoff

 

[MCEngineer]

Hello-

The details of the circuit are as follows:

* Total of 36 LEDs
* Each (good) LED has a measured Vf of 3.0~3.1V at 25C
* Total of 6 transistors and 12 resistors. One resistor establishes the target current for the high setting and the other establishes it for the low setting.
* The LEDs are wired (3) in series for each series-string for a nominal Vf of 9~9.3V
* The transistor is wired as a constant current source and hits saturation when the input voltage is below ~10V, and operates in linear mode above 10V to provide constant current to the string. It dissipates the voltage and power whenever Vin exceeds 10V.
* Since there are (6) transistors and (12) LED series strings, there are two string pairs per transistor.
* We measured ~180 mA on high / 110 mA on low(on a good fixture) = 15 mA (high) / 9.1 mA (low) per LED which is reasonable.
* The failed unit shows LEDs with Vf of 4~6 volts. Interestingly, all of the failed LEDs emitted a tiny amount of light, but again, their Vf was way out of spec, indicating damage to the junction.
* The LEDs are standard T-5 through hole packages which are difficult to heat sink. At 15 mA, they are drawing 46.5 mW. The total of 36 LEDs consume a power of 1.67 watts regardless of the supply voltage. The remainder of the power is dissipated in the (6) transistors.
* We ran the good fixture for ~15 minutes at 14V and measured the LEDs / PCB temperature with an IR thermometer and saw 110~119 degrees F.

My first reaction was that the LEDs were over heating since the heat sink compound was not in contact with the PCB and barely contacted a handful of LED pins on the back side. It is impossible for me to replicate the exact environment (ambient temperature, airflow around the fixture with the boat, etc), it seems unlikely that the LED temperature was much more than 150~160F in Geoff's boat. If we have anyone on this forum who is a thermodynamics engineer (not my specialty), please speak up

The failed LEDs looked like (page 2) of this article:

http://www.us.schott.com/lightingimaging/english/download/visileds_extraordinary_illuminations_brighten_up_the_industry_1003_us.pdf

I work with a number of automotive component suppliers and they are all typically required to survive multiple 40V load dump and other high voltage transient conditions. I typically see regulation circuits with input voltage clamps and transient suppression to prevent (as some posters have already mentioned) damage to the CMOS sensors and microprocessors contained within these modules.

Without a failure anlisys of the LEDs themselves, I can't conclude the exact conditions of failure (temperature or current spike). However, since so many LEDs failed on the "Dim" fixture (around 50%) in such a short period of time (1000~2000 hours) and our measured bench temperature was _relatively_ low, we suspected voltage spikes may have played a significant role in the demise of these fixtures. I suggested Geoff add a large amount of thermo paste to ensure complete PCB contact on the replacement PCBs, and seek out transient suppressors (if they are affordable and relatively easy to obtain) for the circuit supplying these fixtures.

Hope this helps!

John

 

[Witzend]

As your article indicates these devices are subject to electrostatic discharge, typically in the 3-10Kv range.. Static electricity discharge is an unlikely event in a humid marine environment, and you would need to have been handling these devices prior to failure. So if this is your theory, where would these transient spikes be coming from? Not.... poor quality LEDs

 

[geoffschultz]

As your article indicates these devices are subject to electrostatic discharge, typically in the 3-10Kv range.. Static electricity discharge is an unlikely event in a humid marine environment, and you would need to have been handling these devices prior to failure. So if this is your theory, where would these transient spikes be coming from? Not.... poor quality LEDs

I would guess that relays opening and closing would be a prime suspect. I have a large relay on my refrigeration system and there's also the starter for the engine and generator. I've also had close lightening strikes. Unfortunately I don't know if there was just 1 event that caused this or a series of events which gradually caused the issue.

-- Geoff

 

[GuyT]

From your post I would conclude that the problem is current sharing since two strings are paralleled without any ballast resistor. If there is no provision for current sharing, the possibility exists for one string to take more current - this expedites the end of life for the LED.
I seriously doubt the problem is Transient related - the post clearly says that output diminished with time - Transient is a one time deal.

 

[geoffschultz]

From your post I would conclude that the problem is current sharing since two strings are paralleled without any ballast resistor. If there is no provision for current sharing, the possibility exists for one string to take more current - this expedites the end of life for the LED.
I seriously doubt the problem is Transient related - the post clearly says that output diminished with time - Transient is a one time deal.

There are 36 LEDs in 6 banks. A transient could take out separate banks. If it were heat related, I would have expected all of the bulbs to dim at approximately the same rate. Instead I see entire banks out. Also note that I never said that the "output diminished with time."

-- Geoff

 

[GuyT]

Geoffschulz, I assumed from your statement that quote "Anyhow, about a year later I realized that they were getting dimmer and dimmer" that these lamps were failing.

The more probable cause is not a transient that mysteriously killed entire strings of LEDs but more from a long term problem of current sharing causing more current to go through one string (which heated up that string) which caused premature failure.

A quality LED fixture house will match Vf of the LED's since they have a negative temperature Coefficient. They think that if they increase the number of LEDs in a string to three, it diminishes the current sharing problem - and it does but.... it is still not a good solution for poor quality LED's which have terrible Vf distribution.

My money is still with poor quality LEDs that did not share current correctly - not a transient theory. Just my 2 cents.

 

[tsmwebb]

...
My money is still with poor quality LEDs that did not share current correctly - not a transient theory. Just my 2 cents.

I've got a buddy who runs a marine LED lighting company and I've had a beer with him while discussing the practical engineering issues of LEDs... So, I'm an expert Ha, ha... But, I do know that he's very particular about the source and brand or his LEDs and he does a good deal of continuing testing because he found that many white LEDs on the market have problems both in terms of their initial specs and performance over time. Since you're working with a very early, fresh off the boat and untested lot of Dr LED bulbs, I, with my vast knowledge of the field, also suspect that you're looking at a bad batch of LEDs.

I also am not impressed that Dr LED by way of backing their bulbs is just offering you 1/2 price innards on a product that has obvious manufacturing defects.

--Tom.