LM3100 Output Voltage

We use the LM3100 for dc/dc conversion from 24 V to 5 V (see schematic, calculated output voltage: 0.8 V*(1+6.81/(2.61/2)=4.975 V). It works, but the output voltage is not as constant as told in the data sheet. The output voltage is stable (very little switching noise to see) but it changes its dc value depending on the load:

Little load (approx. 250 mA, discontinuous mode): output voltage 5.09 V
approx. 1 A load: output voltage: 4.89 V.

So we get a difference of 200 mV between light and normal load. The feedback voltage at the feedback pin changes in the same way. What's wrong in our design? The layout is done the way described in the data sheet (resistive feedback divider close to the voltage feedback input pin, 4-layer with ground plane, no ground voltage difference at different points in the layout to measure).

Best regards,
Edwin Krasser

Here is the schematic:

5 Replies

  • Hello Edwin,

    Although I can't read your schematic (please re-attach it in another format) I'm pretty sure I understand the problem.

    The control loop in this family of parts uses the falling inductor current information to maintain regulation. Therefore, when the load decreases into DCM mode the output voltage tends to rise. This behavior gets worse at higher input voltages. That is why you see worse behavior compared to the datasheet curves which show typical 3.3Vout behavior.

    I have 2 suggestions. First, try using a larger inductance value which will decrease the DCM threshold current. Second, we have improved this behavior in some of our newer products. I would suggest using the LMR24220 which is almost identical to the LM3100, but has much improved load regulation. In fact, the load reg curve is on the front page of the datasheet.

    Let me know if you have any questions.

    Best regards,

    Allan Fisher

  • In reply to Allan Fisher:

    Hi!

    Thanks for your reply. I hope that adding the schematic works this time (last time I used JPG, this time PNG).

    Best regards,
    Edwin Krasser

  • In reply to Edwin Krasser:

    I hope it works this time.

  • In reply to Edwin Krasser:

    Edwin,

    Your schematic looks fine and should not be contributing to variations in load regulation.

    Regards,

    Allan

  • In reply to Allan Fisher:

    I just tested a small board with the LMR24220 instead of the LM3100. This part is much better!
    * Difference of output voltage between light load and 1 A: <20 mV
    * Much more stable. The jitter of the period time (switched signal) is better than 5 ns (LM3100: approx 50 ns).

    Hence, we will use the LMR24220 instead of the LM3100 and everything is fine.

    Best regards
    Edwin Krasser