This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM3478: mosfet and inductors overheating

Part Number: LM3478
Other Parts Discussed in Thread: LM5156

Greetings, I hope somebody could help me. I'm trying to design a SEPIC converter based on LM3478, with input voltage range 12-36V, and output 24V 2A. I'm calculated all relevant values of components, according to "AN-1484 Designing A SEPIC Converter" and LM3478 Datasheet, and the results was quite similar with examples provided, also I tried to use Webench, but its values we're far beyond mine and example's, so I ignored it. Anyway, after the assembly, I'm stuck with the problem, that converter holds output voltage at +/- 24V, but under load of 1.8 A (LED driver at the output holds current), both of inductors and mosfet quickly overheating, in, like, 30 seconds. In this way I burned two mosfets, because I haven't noticed, that diode became shorted in some reason too, so, I changed mosfet, and burned second one. Also, I noticed that in case of input voltage is 24V and more(buck mode), with no load, mosfet heats in pulses (like, sharply became hot, and then cooling down and repeat), and current on my lab power supply synchronously jumps to 0.2-0.3 A and goes back to zero (I havent's noticed this effect in boost mode, when VIN lower than 24V). So, here is the question, is there root of problem in poor PCB design, or there is some fundamental miscalculation, that I don't see? I re-read different datasheets and appnotes, and many topics on forum, but I still haven't clear solution.

Here's schematic, L1 and L2 have L=10 uH, DCR=30 mOhm, Irms=7.8A, Isat=8.5A, MOSFET is Infineon BSC0702LS.

Schematic

And PCB (all layers)

PCB's bottom (SEPIC located here)

PCB's top (here located LED driver, opamp and MCU)

Waveform of L2 (input inductor) at 21.7V. If I'm lowering the input voltage on lab supply, quantity of pulses rising, and if I'm increasing input, pulses count becaming lower until 24V, and after switching to buck mode, I assume, increasing again (and in buck mosfet starting to overheat, with no load).

Waveworm of L3 (output inductor), same behaviour.

MOSFET gate, same behaviour

Gate on 24 V input

Gate on 26 V input

Gate on 28 V input

Gate on 30 V input

I'm wondering, why in boost mode, when VIN < 24 V, quantity of pulses on gate increasing and decreasing slowly, but after VIN > 24 V it became superfast, like increasing VIN for 0.1 V doubles the pulses count.

Thank you in advance,

Ihor Ziuba.

  • Hi Ziuba,

    Thanks for using the e2e forum.
    The schematic itself looks okay to me.

    Even at max load of 2A, the peak inductor currents should be below 5.5A, so there should be no saturation.
    The FET is only rated for 60V. At worst case conditions (max Vin 36V), the voltage at the FET could go above the 60V threshold, so please consider using a higher rated FET like 80V or 100V.
    But this should not be the root cause for the overheating.

    I would take a closer look at the layout.
    Can you also show the GND planes? Right now, it looks like all ground connections are just floating.
    The ground planes are very important for heat dissipation and main current paths, so the root cause for overheating at higher loads might be related to this.

    Best regards,
    Niklas

  • Yeah, sorry, I thought drawing will be more readable without ground plane. I've updated drawings in post. 

    Hope for your help.

    P.S. also, I've tried to change inductors to 22uH, but without recalculating compensation circuit, and nothing changed ((

  • Hi Ziuba,

    It might be a problem that there are high current flows trough rather narrow traces without any way of heat dissipation.

    I draw lines for the main current path for primary (K1) and secondary side (K2).

    The secondary side loop (K2) is short and compact, which is good. K1 is a rather long path, so you can consider changing the placement of the input caps for the next revision.

    The input voltage line VIN is also rather slim for currents of ~5A.

    Regarding the overheat issue, do you have a thermal camera to check the heat generation around the inductors and the FET?
    The SW plane is small and compact, which is great for EMI, but on the flipside has very limited ways of spreading the heat away from the components.
    Therefore this is always a trade-off between board size & EMI and overheating hazards.

    One more question for clarification, does the overheat only occur during high load conditions, or a light load conditions as well?

    Thanks and best regards,
    Niklas

  • Mr. Shwarz, thank you for your advices! Currently, I'm playing around changing switching frequency and compensation loop, and MOSFET overheating with no load only when input voltage more than 24 V. On the scope, when VIN lower than 24 V, all duty cycles (MOSFET gate, inductors) changes smoothly, but in case of VIN higher that 24 V, it changes very sharply, and for me, it looks like problem with regulating or something, but I can't understand, why it occurs even without load. 

    If it possible, could you answer a few more questions for me?

    - for my case (VIN=12-26V, VOUT=24, IOUT=2A), will it be easier to change topology for something another, like flyback, or sepic simpliest one?

    - if I stick with sepic, should I change IC to LM5156, because it modern and has excel-format calculator from TI, so it more trustful than my mathcad calculations?

  • Hello Ihor,

    Niklas is out of office today. He can reply to your questions on Monday.

    Best regards,

     Stefan

  • Hi Ihor,

    Thanks for your patience.
    It is very interesting that the issue occurs also without load at Vin > 24V.
    Then this might be a stability issue.
    The current design uses two single inductors. When not using a coupled inductor, it is recommended to place an additional RC filter in parallel to C11.
    Good starting values would be e.g. 1 Ohm & 40uF. This should increase the stability.
    For additional calculations on compensation and current ripple, I can recommend our power stage designer tool:
    https://www.ti.com/tool/POWERSTAGE-DESIGNER

    I you prefer to create a completely new design, you can definitely consider switching to LM5156. This part is much newer than LM3478 and therefore has better availability and pricing. SEPIC topology is still a good choice, but you can consider switching to a coupled inductor. 

    Here is also an app note regarding single inductors vs. coupled inductor benefits for a SEPIC design:
    https://www.ti.com/lit/an/slyt411/slyt411.pdf

    Best regards,
    Niklas