I got some 100W LEDs to make a lamp, realized with a boost converter I could use an old ATX power supply and bingo I'd be fine.

It worked at first, until I didn't like the color so I decided to order a full spectrum LED, it turns out to be purpleish in color but looked great when mixed with the white light.

The specs on the purple LED said 30-36V, 3.5A, so I gathered if I went with 36V @ 2.7A, I'd probably be Okay(higher voltages provide better lumen per watt efficiency), but I went with 30V anyway, just to be safe because my meter can't exceed 2A. I

The boost converter failed on the purple light, I did notice that the converter was getting kind of hot on the MOSFET side so I replace the boost converter and add a fan, that worked for about 2 weeks and the MOSFET dies again on the purple light.

That got me angry and I started researching boost converters and how they're made, thinking maybe I could just replace the MOSFETs. After killing a dozen or so MOSFETs, I decided to make my own boost converter. I ended up having the same hot MOSFET and voltage drop issue( I forgot to mention there was always a voltage drop over the purple LED on any boost converter I attached it to).

Through experimentation I find I get a better voltage output if I attach the legs of the diode to either side of the inductor, but still not getting the output I want and a very hot MOSFET, so I remember the autotransformer and think eureka and came up with this

enter image description here

EDIT: This is actually an Update of the original schematic that worked quite well once I added the MOSFET driver. However there should be a shunt resistor between the source pin and ground to measure MOSFET current and a comparator with a reference voltage at one input with the secondary comparator input tied to the source, to shut down the oscillator once the desired current has been achieved to avoid inductor saturation as well as the destruction of the MOSFET. This was a cool, basic, (yet can be DANGEROUS) experiment to see how a boost converter operates, but there should have been be a bit more regulation implemented.

At this point, I wouldn't design such a simple circuit to power a load of any real value. Though it did operate well for months as a desk lamp until I retired it.

No more nasty voltage drops, MOSFET stays a lot cooler loaded with a white LED it came up at 35V (32.3V on purple), but wasn't so bright, so I turned it up and it got brighter, but it was running at 37V.

It seems I was getting the opposite of a voltage drop? I replaced the MOSFET for one better suited for the job STP80NF70, that one let me go as high as 45V with a 60% duty cycle. I like the MOSFET, decide to test it with a 20N60S5.

I went with the auto-transformer because I figured if I was only charging a portion of the inductor during the on cycle I could send some of that stored energy in the other direction, and what is a boost converter really? If you add a secondary winding attach the diodes there and its a flyback supply. So I figured Id make a step up auto-transformer flyback.

At 50% I get a semi dim light at 33V but when I turn up the duty cycle marginally, the light gets extremely bright and I get a reading of 150-180V this freaks me out. The light is extremely bright, the crappy heatsink on it isn't getting hot at all, but the MOSFET gets hot. that mosfet can handle 20A I have a 10A breaker switch on the line, so what gives? why the freaky voltage and the hot mosfet?

One more Question, how do I stop the thing from literally desoldering connection from the drain pin and killing the FET when there's no load?

  • \$\begingroup\$ Can you see what your gate drive signal looks like? Slow gate drive can result in your FET overheating. \$\endgroup\$ Jul 12, 2016 at 2:34
  • \$\begingroup\$ A multivibrator is not the best way to drive mosfets. Custom IC's provide a strong source and sink current path, while resistor based sources and sinks slow down rise and fall times, causing the mosfet to be less efficient and run hotter. \$\endgroup\$
    – user105652
    Jul 12, 2016 at 3:21
  • \$\begingroup\$ The FET overheating occurred on both my boost converter and the purchased one which is what lead me here. Now I'm only having temperature issues depending on the FET's rating, which is what can be expected because from what I understand, during normal boost converter operation the FET has tp be rated high enough to handle the inductive current as well as the current of the source supply. Which is why I added the auto-transformer believing that the booster winding would absorb some of the excessive current and seems to have worked as far as FET temp and voltage output. \$\endgroup\$
    – iuppiter
    Jul 12, 2016 at 9:00
  • \$\begingroup\$ You're not using an SMPS controller at all but instead some kind of resonator circuit.. That's not going to end well. You cannot decide arbitrarily "I want 2.7A" , it depends on the voltage and LED vf. LED efficacy goes down if you increase the voltage (and current). \$\endgroup\$
    – Barleyman
    Jun 26, 2017 at 11:42
  • \$\begingroup\$ @Barleyman Dude, you're way too late for this post. First of all, that IS NOT a resonator(look up the definition of resonators), it's a BJT astable multi vibrator which is an oscillator that generates square waves. I've learned so much since then and FYI, that bizarre circuit did work quite well for months. I made a desk lamp with it. I'm not sure if you noticed, but there's a trim pot in that circuit to control duty cycle. That was my method of controlling the voltage output beyond the normal tripling of input voltage. \$\endgroup\$
    – iuppiter
    Jun 30, 2017 at 8:55

1 Answer 1


Your not paying attention to the 'dead-time' in the transformer. As the pulse width increases the OFF time decreases, thus more output voltage and current. What you need to do is adjust R2 or R3 so that a usable maximum brightness is achieved-but no more than that. Your paying a penalty as it is by using pull-up resistors which slow down the rise time at the gate of the mosfet, causing it to be less efficient as a switch.

The transformer core MUST have a minimum OFF time so the magnetic field can collapse before the next drive pulse. Maximum safe ON time is likely about 95%, but your current design may push it close to 99% or more. Chances are the transformer may be getting warm as well, yet it can run cool while the mosfet gets very hot.

With no load the transformer can have large eddy currents, so overheating of the mosfet is no surprize. One thing that works for me is to find out the voltage across the output capacitor when LED's are at maximum brightness, then insert a string of 24 to 50 volt 5W zener diodes across the capacitor so that if no load is present, the zener diodes act as a dummy load and clamp the voltage to just 5 or 10 volts higher than if the LED's were hooked up. This way the circuit always has a load, either the LED's or the diode string.

  • \$\begingroup\$ Thank you for explaing the capacitance, resistance(resistor/potentiometer) ratio, I realized I had to follow that would either shut down the oscillator if too low or the murder the FETs if I set the duty cycle too high. If there were no pull down resistance (I'm guessing R1 & R4) on the supply it would kill the transistor because they are connected directly to the 12V rail if the Power supply which if I'm calculating correctly are supplying .054A to each transistor. Could I eliminate 1 and ride both of them off of the one resistor? But it still doesn't explain the 100-150V I see on my meter. \$\endgroup\$
    – iuppiter
    Jul 12, 2016 at 10:01
  • \$\begingroup\$ That happens just over a duty cycle of 50%( a hair over half way over the pot ), it's as of the voltage increases exponentially but the current drops enough that the light is a lot brighter than usual but the heat sink Isn't getting anywhere near as hot as usual. If this helps at all I'm using cat 3 solid strand wire wrapped around a toroid 12 winds on the primary and 24 on the secondary. \$\endgroup\$
    – iuppiter
    Jul 12, 2016 at 10:34
  • \$\begingroup\$ You seem to know that you're talking about so I have another question, why can I run the oscillator at different frequencies with different PWM loads, for instance, the FET more suited for this application (STP80NF70, but I covet like the holy grail because I only have a set of four for now and I want to make a full bridge isolated step up to play with) that I can run it with .56nF caps, which if I'm calculating correctly runs at 154kHz but I have to drop it to 134kHz for it to start up with the 20N60S5? \$\endgroup\$
    – iuppiter
    Jul 12, 2016 at 11:13
  • \$\begingroup\$ Your answer wasn't a direct path to the answer (Well the Zener Diode load was quite direct), but you heat in the transformer or the FET, I've experienced that, which made me inspect the transformer and I noticed the primary coil was quite hot so I rewound one with a heavier gauge primary winding and the secondary with the phone wire and now i'm getting a more controllable output. As I was writing this I thought why not switch back to the 80NF70 and Bingo! a nice smooth DC output..well, i'll see when the oscilloscope arrives. Thank you. \$\endgroup\$
    – iuppiter
    Jul 12, 2016 at 14:56
  • \$\begingroup\$ For a 12 volt primary you normally need just 6 turns of heavy gauge wire, but it is 6 turns minimum. Both primary and secondary windings must wrap around the toroid evenly to avoid 'dead' zones where the magnetic field is zero. That is due to eddy currents, which create magnetic peaks and dips in field strength. \$\endgroup\$
    – user105652
    Jul 12, 2016 at 16:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.