Post Go back to editing

Query - LT3759 SEPIC Converter Output regulates to 23V instead of 28V for an load current of 3.5A.


We are using SEPIC convert LT3759 in one of products to generate 28V/3.5A output for an input voltage range - 20 - 33Vin.

For an input voltage of about 20V the SEPIC output regulates to 23V for 3.5A load instead of 28V.

Enclosed schematic of SEPIC section. Please review and share your feedback.

Sadham S.


  • Hello,

    without going into detail: if your Vout decreases with decreasing Vin i would suspect your current limit setting is not sufficient for the demanded Pout. Check Vou behaviour over varyiing Vin to proof my assumption.
    You can check your circuit possible/available output power with a LTSpice simulation.

  • Hello Markus,

    As per datasheet, the switch current will limited by the Rsense value, which is set as 3mOhms.

    When Vin=20V, for an 3.5A load current, the maximum switch current arrived by calculation was 8.5Amps (IL1(5A)+IL2(3.5A) = 8.5A). In such a case, a 3mohm resistor will allow up to 12A switch current (by considering current limit threshold with a 20% margin. Rsense = 40mV/3mohms)

    Further, as per datasheet statement, if Vsense reaches a maximum threshold of 50mV, then comparator A6 will reset SR1 and turn-off the switch immediately.  but converter output regulates to 23V. I assume that problem is not with current limit resistor.

  • Hello,

    the mentioned 3mOhm seems ok for desired output power. Are you sure the signal from Sense reistsor is clean and well routed in the pcb layout?

    Does the circuit achieve the desired 28V if you increase your input voltage?

    As I cannot see the value of external components (resolution too bad) I refer to my 2nd hint: Please 'rebuilt' your circuit in LTSpice and check simulation results and compare with your real results. Maybe simulation unveils your problem

  • The desired 28v output is achieved when the input voltage is above 24V.
    When the voltage is reduced below 24V,the output voltage begins to dip.
    In simulation,the output is perfect for 20V input also.
    PCB Layout are well routed.
    We have made below changes,
    Coupling capacitor value = 20uF (previously 2.2uF)
    Independent Inductor value = 20uH (previously 10uH)
    Rsense Resistor value = 1mohms (previously 3mohms)
    Switching frequency was reduced from 600KHz to 300KHz.
    Still the output voltage dip was there when the input voltage given was below 24V.
    Refer the attched design scheme (initial Scheme) and provide your feedback
    We are suspecting voltage compensation for this instability. Could you please share the calculation sheet for selecting RC for VC pin of LT3759?

  • Further to the above comment, we have tested with the following conditions,

    Coupling capacitor value = 20uF
    Rsense Resistor value = 3mohms
    Switching frequency = 600KHz

    Compensation Pin Resistor Value = 9.76K

    Compensation Pin Capacitor Value = 10nF

    SEPIC output voltage was configured to 28V output for 20-33V input.

    But at 20V input for 3.5 A load, the output voltage is only around 23.9V.During this condition the duty cycle was operating at around 80% due to this after ten minutes of continuously ON, the mosfet was failed. But as per the calculation the maximum duty cycle is only around 58%.

  • Team,

    Awaiting for your reply and suggestion.


  • I need to ask a few more questions to get more information:

    Does the output oscillate when viewed on an oscilloscope? 

    If you were to look at the Switch node (on oscilloscope) at the input voltage (above 24VIN), when the output is correct, does the duty cycle of the converter match what is expected from simulation? The 80% duty cycle at low VIN looks very wrong, and could lead us to what is wrong.

    If possible, do you have access to oscilloscope current probes, can you take scope shots of the inductor current waveforms to make sure they are operating properly and matching simulation? You could have inductor ringing from the SEPIC tank. It is quite difficult to find this without current probe. 

    This could be layout issue. If you send us the layout PDF or pictures, we can see if there is something wrong there.