Burned out 3 boost converters trying to power a 100W LED... at my wit's end... please help.

Thread Starter

-Ty-

Joined Feb 5, 2017
83
Hey all,

I'm trying to build a photography light based on a video on Youtube: VID, and I've been running into an issue with my DC-DC boost converters burning out for a reason I can't identify.

The circuit is pretty simple: A 120W AC-DC power adapter is outputting 12V to the boost converter (LINK), which is bumping it up to 32V, with a constant-current limit of 3A. The boost converter is then outputting to a 100W LED chip rated for a forward voltage of 30-34V and a max current of 3.5A. The power adapter is also feeding into a dc-dc step down module which is regulating the current to under-volt a 12V computer case fan, for cooling.

Now, despite burning out the first board for a reason I think i understand, I got the second board to work steadily and without failure. That was, until I turned it off one time.

I had been running the LED for about a half hour (and this was the fifth time ive done so), and then when I was done, i flicked the switch in the circuit to cut the power between the laptop adapter and the rest of the equipment, and this is when I noticed that the powder adapter's LED light was flickering, as it had done when the last board burned out and was shorting. Realizing what was about to happen, i went to pull the plug, but I was too slow and the second board burnt out. The third and most recent time I rebuilt the system from scratch, I actually made improvements to the design in the video, choosing to connect all the components on a through-hole PCB, instead of having them just floating in the air, to ensure secure and reliable solder connections. I also made sure to use the best, highest-quality switch and potentiometer i could find. The system worked fine, running for 15 minute intervals every time. I started final assembly, and started gluing the components into their case, and so i decided to test the system one more time before it was all glued in for good. I hooked up my multimeter in series to read the current passing through the LED, and turned it on. Everything worked perfectly, but the multimeter was giving a negative current reading. Just for the sake of my OCD, I turned everything off, disconnected the multimeter, and swapped the leads so that the value would read as positive. When i turned the light back on, without having changed anything else, it turned on for a second, and then immediately burned out the MOSFET on the boost converter. I'm at a loss...

The component that is burning out is the NCE6075K N-Channel Enhancement MOSFET (DATASHEET), and I don't know enough about electronics to figure out why exactly.

I'm so frustrated and upset with this darn project. It was supposed to be easy and simple and cheap but ive sunk tons of money into these boards, and they take weeks to get here each time. I can't keep doing this. I've even been adding heatsinks to the boost converters from the start, AND they've been actively cooled by the fan that's cooling the LED.

I've checked, double-checked, and even quintuple-checked the wiring, and its EXACTLY the same as his is in the video. His system worked fine, mine keeps blowing.

Please, any help is greatly, greatly appreciated. I just don't know whats wrong.

I've attached some pictures of the circuit below. I hope it helps.

PS. Please note that I am aware that this circuit is not wonderfuly designed.. I KNOW that the power switch should be switching the power supply, not the LED (though to be fair, I actually added another switch in front of the power supply, and thats the one i was using when testing, and when it died), and I KNOW that fans shouldnt be slowed by under-volting, but instead by PWM. The point of this post is simply that the design in the video --be it good or bad--, when built by the guy in the video, WORKS, and mine, when following the exact same steps, using virtually identical parts, DOESNT.

20180202_231902.jpg 20180202_231912(1).jpg 20180202_231942.jpg 20180202_232020.jpg
 

Attachments

Ylli

Joined Nov 13, 2015
1,086
My guess would be that you are getting an adverse reaction between the two SMPS you have in series. I would ditch the computer power unit and build up a simple transformer/diode bridge/capacitor.
 

Alec_t

Joined Sep 17, 2013
14,280
The long wires from the board to the LED may have enough inductance to create voltage spikes which kill the FET when things are switched off.
 

Ylli

Joined Nov 13, 2015
1,086
I KNOW that the power switch should be switching the power supply, not the LED (though to be fair, I actually added another switch in front of the power supply, and thats the one i was using when testing, and when it died)
Actually, I think you will have more luck by leaving the switch between the computer power block and the LED power supply. Always have the switch between the supplies off whenever you power the computer supply on or off.

[The issue here is that a switching power supply draws more current at low input voltages than it does at higher input voltages. If the input of the second supply is connected to the output of the first, when the driving power supply tries to come up to voltage, it sees a very high current demand from the second supply and may go into current limit. Now you have the first supply sitting in current limit, and the second supply trying to run at its minimum operable input voltage and drawing as much current as it can. Perfect prescription for failure. If the first power supply is already on and up to full output voltage when the second supply is turned on, then there is enough output capacitance in the driving supply to get the second supply 'over the hump' and into normal operation.]
 

Thread Starter

-Ty-

Joined Feb 5, 2017
83
Actually, I think you will have more luck by leaving the switch between the computer power block and the LED power supply. Always have the switch between the supplies off whenever you power the computer supply on or off.

[The issue here is that a switching power supply draws more current at low input voltages than it does at higher input voltages. If the input of the second supply is connected to the output of the first, when the driving power supply tries to come up to voltage, it sees a very high current demand from the second supply and may go into current limit. Now you have the first supply sitting in current limit, and the second supply trying to run at its minimum operable input voltage and drawing as much current as it can. Perfect prescription for failure. If the first power supply is already on and up to full output voltage when the second supply is turned on, then there is enough output capacitance in the driving supply to get the second supply 'over the hump' and into normal operation.]
Ylli, that explanation makes good sense, i suppose. After the MOSFET on the DC boost converter dies, the whole system shuts down, obviously, because of the short circuit. If i leave the dead boost converter on, then the AC-DC power supply will make a buzzing noise when its on, and won't deliver any power to the fan voltage regulator or the fan. If i remove the dead board, it works properly again.

Despite the initial high current demand you said the power supply would be experiencing, it seems to be undamaged after all these burnt boards, and continues to function...

So.. what can I do to make this light work? In the video that im basing this off of, he switches the system from before the power supply, so why doesn't it fry his board the way it might be doing to me? I mean, i get that i can just always be sure to supply power to the AC-DC supply BEFORE turning on the LED, but I don't like the idea of having a light that will completely break if i forget to do this even once...

What workaround is there?
 

Thread Starter

-Ty-

Joined Feb 5, 2017
83
An LED power supply that is better protected. DickCappels pointed one out.
The power supply isn't what's dying though, it's the DC-DC Boost converter.

An unfortunately, that LED power supply costs more than every other component in this build combined. Its not a realistic solution for the problem, when it somehow works for the guy in the video just fine :l
 

Ylli

Joined Nov 13, 2015
1,086
Other than trying to add some protection to the DC-DC converter, putting the switch between the computer power supply (and bring sure that one is on first) is the only thing I have to offer. Have you tried contacting the author of the original article?
 

Thread Starter

-Ty-

Joined Feb 5, 2017
83
Other than trying to add some protection to the DC-DC converter, putting the switch between the computer power supply (and bring sure that one is on first) is the only thing I have to offer. Have you tried contacting the author of the original article?
I guess that's what I'll do for now, then.... even though that's how my other ones were when they died too.... and I guess I'll just cross my fingers and hope for the best? I tried contacting the author but as a Youtube creator he doesnt really get back to his fans...especially considering how hard it is to diagnose something like this over the internet.

By the way, what kind of "protection" could I add? A diode? I haven't worked with diodes yet so i wouldn't know how to select the right one.
 

Ylli

Joined Nov 13, 2015
1,086
By the way, what kind of "protection" could I add? A diode? I haven't worked with diodes yet so i wouldn't know how to select the right one.
Without a schematic of the converter, I'd have no idea. I did look at the FET, and it already includes a built in protection diode.
 

Thread Starter

-Ty-

Joined Feb 5, 2017
83
Fair enough. What makes sense to me is trying to protect the DC-DC converter, since that's the part that's dying.. specifically its MOSFET. The ONLY percievable difference between my build and the one in the video is that the boost converter I use seems to be SLIGHTLY different in terms of chip layout... after all, this board is one of those nameless Chinese products, where 25 companies produce the same thing, and i've noticed slight discrepancies between my board and some of the others, even though in terms of specifications they're the exact same. I could understand if using sub-par components would reduce the load the board could carry.. but reduce it by 150 watts? Seems doubtful to me..

How can i protect THIS board? By the way, i find it extremely weird that this board, rated for 250W of dissipation, has no heatsinks.. while boost converters with half the wattage rating are given pretty beefy ones... What's up with that? I will admit that this 250W board isn't built on a fiberglass PCB, but is instead seemingly laminated onto aluminum.

I've considered using THIS board, despite it's lower power rating.. the fact that it follows a completely different circuit layout makes me think it might not be prone to the same failure mode..

I've also considered THIS one, except it is probably too big to fit in my enclosure.
 

Ylli

Joined Nov 13, 2015
1,086
Feeding one switching power supply from another switching supply can be problematic. Without them sitting on the bench in front of me, and the use of my meters and scope, I really don't have any more ideas.
 

ebp

Joined Feb 8, 2018
2,332
[edit] in reply to "The long wires from the board to the LED ..."

If that is the case, the wires should be twisted together (should be in anyway). When they are in close proximity the equal and opposite magnetic fields nearly cancel, dramatically reducing the inductance. All power connections should be done this way.
 

ebp

Joined Feb 8, 2018
2,332
I HATE switchers. I've designed many, from milliwatts to a kilowatt (that could be made to deliver well-regulated microwatts). I've sent many a FET to a firey doom. I've blown up big electrolytic capacitors. I've spend days tracking down problems. I HATE switchers.

I'm going to toss out a bunch of points here, but apart from one issue, I can't see anything fundamentally wrong with what you are doing.

The lack of a heatsink is strange, but if the FET is running at less than 100°C or so it should be OK, though knocking 20°C off of that would improve reliability. The electrolytic capacitors are far more likely to die prematurely from overtemp than the FET. To add a heatsink you'll probably need to use some of the soft, thick (a millimetre or two) compliant thermal interface material between the back of the PCB and the sink.

Can you identify the control IC on the boost converter? Is there evidence of a current sense resistor between the source of the FET and ground? Just looked at pic at vendor site - definitely current sense R's

A good controller will do cycle-by-cycle current limiting which should protect the FET from any sort of overcurrent situation. This is true even if the controller is a voltage-mode type (i.e. not an inner current-mode loop with a voltage-mode loop wrapped around it). Cycle-by-cycle current limiting is almost magical in its ability to protect the FET.

Can't tell for sure, but it looks like the toroid is on a core painted green with one blue side ?? Probably a Chinese counterfeit of Micrometals type 52 material (yes, they really do make counterfeit Micrometals cores). Improperly chosen core materials or poor inductor design can lead to saturation of the core, but powdered iron saturates very softly and a cycle-by-cycle current limit should deal with impending saturation.

What are the bits on the perf board and the pot for? If they are replacing a trimmer or fixed resistor on the boost board [again, determined that from pic on vendor site] and that trimmer is in a feedback path, using an off-board pot with long wires that aren't twisted together is asking for trouble. Noise pickup in those wires can seriously screw up the control circuit. If one of the wires connects to ground of the boost board, try using co-ax with the shield as the ground conductor. If neither connects to ground, use a shielded twisted pair, again with the shield to ground. A pot used as a rheostat (as opposed to voltage divider) should always have the unused terminal connected to the wiper so it can't go completely open-circuit if the wiper loses contact with the element. An open-circuit pot could either make the regulator shut down or go to max, depending on where it is in the circuit.

The FET has plenty of margin for both current and voltage. Generally, unless a FET is already really hot it is quite hard to kill it with moderate overcurrent. The thing is rated for 75 A continuous. Overvoltage can cause sudden death if the available energy is large, but it's a 60 V part so there is good margin when running at 35 V (the FET in a boost converter must deal with the maximum input current and the maximum output voltage). However, if something prevents the FET from turning off when it should (e.g. noise scrambling the control IC), the FET short-circuits the input supply, discharging all the capacitors in the path, ultimately little more than the resistance of the inductor to limit current.

Diode problems can kill the FET. As the FET begins to turn off each switching cycle, the inductor current commutates from the FET to the diode and the output capacitors. If the diode goes open, all the energy stored in the inductor will go through the FET, and it will raise the voltage across the FET to whatever level is required. This is very unlikely to be a problem unless the diode is getting so hot it is melting its solder joints. I've certainly seen semis that have melted their joints to the PCB.

Boost converters generally have quite poor dynamic response - a step change in load current will result in a sudden transient decrease or increase in output voltage for increase or decrease current, respectively. The time it takes to settle back to low error is usually considerably longer than for a buck ("right half plane zero" problem). There is nothing that can be done about this, as long as the design isn't worse than it needs to be. Disconnecting the load could cause some voltage overshoot at the output, but it's unlikely to be enough to kill it. Do you keep the voltage set to just a few volts above that required to drive the LED? Disconnecting the input supply or switching off the other supply shouldn't be damaging. The input current will rise as the voltage at the boost input decays, but once more a decent design should take care of that and prevent damage. EVERY boost converter should be expected to have to cope with that.

Startup is usually the most stressful time for a switcher. Many controllers have a soft-start feature that slows ramp-up to the steady-state operating point. If soft-start is implemented off-chip and not done very carefully, it may fail to reset properly during a brief loss of input power, and thus allow hard-start when power is reapplied.

At this point, I return to the external pot. Get rid of it for testing. If the circuit fails with the original trimpot installed on the PCB, then something else is at fault, but you must eliminate that issue first.
Get all of the other connecting wires twisted together in their respective pairs.

Did I mention that I hate switchers?

===
It is usually perfectly acceptable to run DC fans at reduced voltage. I have two or three that have been running continuously for several years that way.
 

be80be

Joined Jul 5, 2008
2,072
This may not add up but I have played around with a bunch of SMPS out of printers cell phone chargers and the likes.

I've seen this happen on power up some would put out really high voltage till they loaded up.
I've not lost nothing. But I have a 5 volt that I drop to 3.3 volts It hits 18 volts till the baby turns on it happens fast.
The 3.3 handles the 18 volts ok and as soon as the esp is up and running the SMPS hits a rock hard 5 volts but with not much load
it starts up high.
 

Thread Starter

-Ty-

Joined Feb 5, 2017
83
EBP:

First, I'd like to take a moment to thank you for your very detailed, and comprehensive post. I appreciate you taking the time to cover so much for me. I'll admit, a lot of it is over my head, as I'm very much a n00b at electronics, but nonetheless, thanks.

Now, instead of trying to reply to each point one by one, i've simply quoted you below, and have added my thoughts in red.

~~~~~~~~~~~~~~~~~~~~~~~~
I HATE switchers. I've designed many, from milliwatts to a kilowatt (that could be made to deliver well-regulated microwatts). I've sent many a FET to a firey doom. I've blown up big electrolytic capacitors. I've spend days tracking down problems. I HATE switchers.

I'm going to toss out a bunch of points here, but apart from one issue, I can't see anything fundamentally wrong with what you are doing.

The lack of a heatsink is strange, but if the FET is running at less than 100°C or so it should be OK, though knocking 20°C off of that would improve reliability. The electrolytic capacitors are far more likely to die prematurely from overtemp than the FET. To add a heatsink you'll probably need to use some of the soft, thick (a millimetre or two) compliant thermal interface material between the back of the PCB and the sink. I actually have added heatsinks to the back of the "aluminized PCB", if that's the correct term. I used silicone heat transfer adheasive, and the biggest heatsinks I could find that were the same size as the board. They do get pretty toasty with use... to the point where its uncomfortable to touch them for extended periods of time, which i believe lies around the 60-70 degree mark for most people.

Can you identify the control IC on the boost converter? Is there evidence of a current sense resistor between the source of the FET and ground? Just looked at pic at vendor site - definitely current sense R's

A good controller will do cycle-by-cycle current limiting which should protect the FET from any sort of overcurrent situation. This is true even if the controller is a voltage-mode type (i.e. not an inner current-mode loop with a voltage-mode loop wrapped around it). Cycle-by-cycle current limiting is almost magical in its ability to protect the FET.

Can't tell for sure, but it looks like the toroid is on a core painted green with one blue side ?? Probably a Chinese counterfeit of Micrometals type 52 material (yes, they really do make counterfeit Micrometals cores). Improperly chosen core materials or poor inductor design can lead to saturation of the core, but powdered iron saturates very softly and a cycle-by-cycle current limit should deal with impending saturation. With eyesight like that you should enter competitive marksmanship...Yes, the donut Toroid is blue underneath, on the side that faces the board and which you're not supposed to see, and is just painted green on top. I've also noticed that this board manufacturer doesn't apply the silicone adhesive that some of the other manufacturers do on the trimpots and other components... I'm starting to wonder if my purchasing of a sub-quality board is what's leading to my issues. Something very strange that I noticed is that some of these boards (the ones used in most videos) seem to have a SLIGHTLY different layout and chip placement than the one i ended up getting. You can see what i mean by comparing MY BOARD with THIS BOARD. Note the slightly narrower PCB, and the lack of a second SS56 chip below the two trimpots. If it really is a different chip layout/circuit, then maybe that's the reason things are going wrong?

What are the bits on the perf board and the pot for? If they are replacing a trimmer or fixed resistor on the boost board [again, determined that from pic on vendor site] and that trimmer is in a feedback path, using an off-board pot with long wires that aren't twisted together is asking for trouble. Noise pickup in those wires can seriously screw up the control circuit. If one of the wires connects to ground of the boost board, try using co-ax with the shield as the ground conductor. If neither connects to ground, use a shielded twisted pair, again with the shield to ground. A pot used as a rheostat (as opposed to voltage divider) should always have the unused terminal connected to the wiper so it can't go completely open-circuit if the wiper loses contact with the element. An open-circuit pot could either make the regulator shut down or go to max, depending on where it is in the circuit. You know, i never knew that when dealing with such small currents and voltages, inductance through wires was something you had to even think about. I would NEVER have thought about this, so i thank you for making me a better builder. I also really thank you for the tip about unused pot terminals. Very smart. I suppose thats why the LED's brightness flickers pretty wildly as im moving the Pot wiper, but then settles as soon as I stop.

As for the bits on the perf board, I'm not going to ask you to watch the 15 minute video i posted in my original post, but basically, the guy removes the Constant-voltage trimpot on the boost converter, and replaces it with this circuit, so that a one-turn potentiometer will replace the 20-turn trimpots, but can be set to operate between a specific minimum and maximum brightness (read: voltage). The two trimpots in the circuit set the lowest-resistance and highest-resistance paths through the circuit, so as long as they're set, you could short the full-sized pot and the circuit would still be safe. The first few times I tried building this, i just soldered the components in a free-floating manner, and i found that the connections were weak and that even moving the wires around would cause all sorts of weirdness. I rebuilt it this time with the components on a perfboard to increase the strength and reliability of the connections, and as far as I can tell, that seems to have worked.


The FET has plenty of margin for both current and voltage. Generally, unless a FET is already really hot it is quite hard to kill it with moderate overcurrent. The thing is rated for 75 A continuous. Overvoltage can cause sudden death if the available energy is large, but it's a 60 V part so there is good margin when running at 35 V (the FET in a boost converter must deal with the maximum input current and the maximum output voltage). However, if something prevents the FET from turning off when it should (e.g. noise scrambling the control IC), the FET short-circuits the input supply, discharging all the capacitors in the path, ultimately little more than the resistance of the inductor to limit current. This is partly what i find so infuriating about this whole deal. If the MOSFET is rated for 75A continuous and 60V, how the HELL am i blowing it off a 120w, 12V 10A power supply and a 100W LED? And I have confirmed that it IS the MOSFET that's dying, as the terminals short on a multimeter continuity test, and... you know.. magic smoke and poof and whatnot.

Diode problems can kill the FET. As the FET begins to turn off each switching cycle, the inductor current commutates from the FET to the diode and the output capacitors. If the diode goes open, all the energy stored in the inductor will go through the FET, and it will raise the voltage across the FET to whatever level is required. This is very unlikely to be a problem unless the diode is getting so hot it is melting its solder joints. I've certainly seen semis that have melted their joints to the PCB.

Boost converters generally have quite poor dynamic response - a step change in load current will result in a sudden transient decrease or increase in output voltage for increase or decrease current, respectively. This is really interesting to know, thank you. So then, what would be a better system for creating a dimmable 100W LED? I get using an LED driver, but i haven't found any that are freely dimmable off a pot. The time it takes to settle back to low error is usually considerably longer than for a buck ("right half plane zero" problem). There is nothing that can be done about this, as long as the design isn't worse than it needs to be. Disconnecting the load could cause some voltage overshoot at the output, but it's unlikely to be enough to kill it. Do you keep the voltage set to just a few volts above that required to drive the LED? Yes, i think its set to around 32V? The chip maxes at 35, i believe. Disconnecting the input supply or switching off the other supply shouldn't be damaging. The input current will rise as the voltage at the boost input decays, but once more a decent design should take care of that and prevent damage. EVERY boost converter should be expected to have to cope with that. Yeah, tell that to the Chinese Ebay market ;)

Startup is usually the most stressful time for a switcher. Many controllers have a soft-start feature that slows ramp-up to the steady-state operating point. If soft-start is implemented off-chip and not done very carefully, it may fail to reset properly during a brief loss of input power, and thus allow hard-start when power is reapplied.

At this point, I return to the external pot. Get rid of it for testing. If the circuit fails with the original trimpot installed on the PCB, then something else is at fault, but you must eliminate that issue first. A good idea in principle... in practice, having to wait another month and a half for a new booster... oi. I suppose I have to.
Get all of the other connecting wires twisted together in their respective pairs. Will do.

Did I mention that I hate switchers? ;)

===
It is usually perfectly acceptable to run DC fans at reduced voltage. I have two or three that have been running continuously for several years that way.


~~~~~~~~~~~~~~~~~~~~~~~~
 
Last edited:
Top