Help needed to design Swithmode step-up converter, this time not so Nixie-related

[Per Jensen]

Hi folks. Usually i build and tinker with Nixie tubes, so a Step-up converter is nothing new to me.

I usually use the MC34063A-based MK1,5 supply design from M. Moorrees which works wonders!

Now i got a job set for me - I need to retrofit a head-mounted lamp with LED's - I got a whole bunch of 5050-style warm white LED's from some scrapped lamps (the original smps power supppled blew up) so i got a good supply of these LED's. There is 3 dies inside each LED package, and i reckon they can handle about 40 mA without any active cooling.

On the original boards, these LED's are mounted in 5 chains of 6 LED's each - no resistors mounted, i guess the power supply was current limited.

I milled a new PCB with room for 20 of these LED's - put them all in series, so i have a load that likes around 70V and 40 mA. I milled a pcb with a MK1,5 Power supply, replaced the voltage divider with a 33 ohm resistor for the low side, so the power supply is current-regulated, and it works just fine.

The problem I am having, is the heat loss from the Mosfet and inductor - Packing this inside the bulb renders these parts at 100 degrees C, which i don't think they will last very long being that temperature...

I am using a 2SK3592 switching transistor and a 10 uH low loss SMT inductor.

I replaced the SMPS IC with a NCP3064B which is just a descendant of the venerable MC34063A chip, so that won't change anything really.

Now, the question is, any advice on getting the efficiency bumped up a bit, and would i be better off rearranging the LED's so i have a lower voltage and higher current ? I could parallel all the LED's, but that is a big no-no as I am concerned.

Any tips or advice would be welcomed ;-)

// Per.

[John Rehwinkel]

> The problem I am having, is the heat loss from the Mosfet and inductor - Packing this inside the bulb renders these parts at 100 degrees C, which i don't think they will last very long being that temperature...

I can think of basic way to make the inductor run cooler - use a higher current version. This may well not be practical due to space reasons.

I can also think of four basic ways to make the mosfet run cooler:

1) heat sink - again probably impractical due to space reasons

2) bigger mosfet - yet again probably no room

3) reduce switching frequency - this would need a bigger inductor, for which there is probably no room

4) turn it on and off harder, keeping it out of the linear region - look at Mike's MK2 supply for one way to do this with a couple more components. There are also ways to use the boosted voltage to get higher gate drive, and/or use a dedicated mosfet driver chip.

- John

[MrNixie (UK)]

Hmm, I think you need to give us a bit more info - your proposed circuit, supply voltage, etc. The mosfet and inductor won't complain much at 100C, but your LEDs might start to! And any electrolytic caps wont be winning any Long Service awards at those temperatures... Is the Mosfet running at this temp barbecue it is dissipating power, or is just in a hot place to begin with?

Make sure you are indeed switching the Mosfet cleanly - check the Vgs curves in the spec sheet. Remember that at high frequencies, the gate of a mosfet just looks like a capacitor - you really have to DRIVE current in to it - are you relying solely on the output of the SMPS IC?

Re series and parallel LED combinations - general rule of thumb is to arrange your LED load closest to the supply voltage, so that your SMPS is not having to step up or down too far from the supply rail, in either direction. You will see losses if you go for a very low voltage but high current output, or vice-versa

My two milliamps' worth -

[MrNixie (UK)]

BARBEQUE?? That came out of my fingers as "Because"...

[John Rehwinkel]

> Is the Mosfet running at this temp barbecue it is dissipating power, or is just in a hot place to begin with?

Reminds me of a license plate I saw a while back reading PNP BBQ:

http://www.vitriol.com/images/humour/PNPBBQ.jpg

- John