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.

LM3481-Q1: Inconsistent switching and low output voltage with a Flyback design

Part Number: LM3481-Q1
Other Parts Discussed in Thread: LM3478, LM3481, LM5155

Hello !

After a first failed attempt with the LM3478, quite probably because of a bad assembling process on my end, I bought a CNC router and used WeBench designer to to craft a flyback transformer using the LM3481 that would in theory produce +15V at the output from a regular Li-Po battery.

Sadly, I still don't get +15V out but barely 2V under no load and I also noticed that I need to switch the circuit quickly off and back on to start some switching.

Now power electronics is not my strong suit and I can't figure out if it could be some Isense limiting, the Comp feedback loop that doesn't respond properly, or simply my assembly which is flawed.

So here's the design, straight from WeBench, and mine updated with the component I had :

LM3481 flyback PSU A0.pdfWBDesign26.pdfLM3481 flyback PSU A0 - Layout.pdf

You can also see my build here :

And here is some measurement I took (With about 4V steady at the input) :

- Output voltage :

- The Gate signal :

- The primary winding output :

- The secondary winding :

- The Comp pin :

- And finally the Isense (both peak to peak and a close up)

So there it is : I have several doubts of possible problems, but the fact that I tried to follow exactly the Webench design sort of throws me of. Do you have a clue of what could go wrong ?

Thanks in advance for your help !


  • Hello Xavier,

    Is it possible that you connected the secondary side winding a different way than given in the Webench design? It seems that you have the dots of the transformer on the same side, in a flyback you need them the other way around.

    Please check what happens if you switch the connections of the transformer to the diode and output cap.

    In general I recommend to order an EVM from TI and possibly use a newer device like LM5155 for a new design.

    Best regards,

  • Hello !

    You're absolutely right, I don't know where I had my mind at ! So I swapped the terminals on the secondary and now we do have a similar waveforms on both sides of the coil. But I'm still stuck at about 2v at the output (with a nice steady 4V input) and the device only uses about 20 to 30mA.
    I do wonder if the comp section is working as intended ? Or could it be that the spikes on the Isense pin trigger the current limit ?

    Thanks again for your help !


  • Hello Xavier,

    Please clarify with the manufacturer if the transformer is ok for a flyback. In a flyback the transformer needs to store the energy, so it needs magnetizing current and normally a gap.

    Best regards,

  • Hello !

    That might be it : according to the datasheet, they're designed for sepic designs. I thought that coupled inductorq might work, given that it is quite hard to find matching transformers, even on prolific sites like mouser. The Webench designer recommended a custom design, but that exceed my knowledge.
    Given that, in the end, I'm trying to obtain to lines : a +15V and a -15V with a very low ripple current (~10mV) : what would you suggest as the best option : two simple boost topology (but would the ripple current get out of control ? Or would a boost design with coupled inductors as the one I'm using would do the trick (I don't know really how the energy would behave between the two outputs though)

    Thanks again for your help !

    Best regard,


  • Hi Xavier, 

    Sorry for the delay. I would recommend a boost converter for the +15V line and a discrete charge pump circuit to generate the negative rail.  Minimizing ripple voltage is a function of your output capacitance. Please see the application note here.

    You should be able to do this with the LM3481, and its next-gen variant the LM5155 as well.

    Please let me know if you have any questions.