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.

LM5118 startup issues

Other Parts Discussed in Thread: LM5118

Hi,

I am trying to debug a LM5118-based power supply:

In most cases, it works perfectly fine. However, it does not reliably start, if the output voltage is not zero.

For example, imagine that the power supply is turned off at time 0. Ten seconds later, the output voltage has decayed from 12V to 3V. The unit is sturned back on and one of two things happen:
- Not much happens. The output voltage stays at approximately 3V.
- The output voltage slowly creeps upward, and a couple of seconds later suddenly jumps to 12V.

The exact behavior seems to be dependent on the particular unit being tested. In most cases the output voltage "problem range" is 2-3.5V. Either less or more than that, and the power supply starts fine. But, I have seen several units that won't start until the output voltage is 0.5V or less.

Any ideas/suggestions?

Thanks!
Bart

  • Bart,

    What is the input voltage when you are trying to start-up with the pre-biased load?

    Please check the voltages on the VCC and SS pins to assure that the voltages are correct.

    Thanks,

    Garrett

  • Garret,

    With input pre-biased (with an external resistor divider, for consistency) to 2.25V, the power supply would not start at the following input voltages: 12V,15V, 20V, 25V and 30V.

    VCC is 6.9V. SS voltage rises linearly to 400mV and seems to hit a hard stop there. (In normal operation it rises all the way to 1.37V).

    Is there a way to "subscribe" to a topic so that I get email notifications? It took me a while to notice that you responded :-(

    Thanks!
    Bart
  • Hi Bart, 

    Can you provide some waveforms of the start-up issue? Please include SS, VCC, COMP and VOUT in one picture and then VIN, SS, VCC, VOUT in another.

    Also when you are pre-biasing the output voltage is there a diode in the path? I see that you mention that you are using a resistor divider but there is no mention of a blocking diode for when VOUT is above the pre-bias voltage.

    To get email notification there is a check box that say "Notify me when someone replies to this post". This will be available when you are writing a response to this post.

    Thanks,

    Garrett

  • Garret,

    There is no diode, just two 1k resistors.

    All traces below are BW limited to 20MHz.

    SS (yellow), VCC (magenta), COMP (blue) and VOUT (green):

    Hmmm. It seems that COMP goes to 5V and stays there for some reason.

    SS (yellow), VCC (magenta), VIN (blue), VOUT (green):

    And here are the same waveforms for a normal start-up, with no pre-biasing:

    Thanks!

    Bart

  • Bart,

    Please tie the supply that you are using to pre-bias the output with through a diode as shown below. Without the diode you are clamping the VOUT to the pre-bias supply voltage without any way for it to be disconnected when the output voltage of the diode is 

    It is normal to see the COMP pin clamp at 5V. This just means that the controller wants more current to got to the output, but as discussed above this isn't happening because VOUT is clamped to the pre-bias voltage.

    Let me know if you have any questions.

    Thanks,

    Garrett

  • Garret,

    I added the diode. It made no difference; the waveforms did not change.

    Why does the diode matter? Without it, the pre-biasing voltage source will sink (12V-5V)/1k=7mA, and my power supply can handle 8 amps.

    At a pre-bias voltage of less than 1.5V or more than 3.5V the power supply starts fine, diode or not.

    I have just noticed something interesting. With pre-bias of 3V, I get the "bad" set of waveforms first, and then the "good" one after a delay of approximately one second. SS (yellow), VCC (magenta), COMP (blue) and VOUT (green):

    Also, in your post, is there something missing after "the output voltage of the diode is..."?

    Thanks!

    Bart

  • Bart,

    Please remove C28. This might be causing an the FB pin the rise slowly resulting in the SS pin getting clamped.

    Above you mention that the you only see the problem when VIN is above 12V. Does it start normally at lower voltages?

    In the last waveform:

    1. Can you zoom in on the section where you turn off the converter and the output voltage rises.
    2. Are you using EN to toggle the LM5118 on and off?
    3. How does this change with input voltage?

    Thanks,

    Garrett
  • Garrett,

    (It seems that I have been misspelling your name - sorry!)

    I removed C28 and observed no improvement.

    There is an external undervoltage protection that will prevent the EN signal from bein asserted below 11.5V - that's why I started my testing at 12V.

    With C28 removed, the last waveforms changed somewhat (and I am not sure is there is a cause and effect relationship here). Again, SS (yellow), VCC (magenta), COMP (blue) and VOUT (green):

    Note that VCC does not drop - actually, it rises.

    And here is the same waveform, zoomed in. You requested a closer view of the moment when I "turn off the converter and the output voltage rises". Just to clarify - I do *not* turn it off. It does it all by itself. I verified that the EN signal stays high after the initial enable.

    Yes, I use EN to turn the IC on. The DC input appears first, and the EN signal is asserted a couple of seconds later.

    The "spontaneous restart" seems to only happen below 25V DCIN. Other than that, I do not see any dependence on input voltage.

    Thanks!

    Bart

  • Bart,

    I'm looking into this right now.

    Can you try one more thing and decrease the SS capacitor value? Try lowering it by a factor of 10 and see what difference it makes.

    Thanks,

    Garrett
  • Garrett,

    With the SS capacitor value lowered to 0.1uF, I could not get the power supply to fail. I do, however, remember trying this before and still seeing failures (might have been a different unit).

    What does it tell us? I should be able to set the SS cao to an arbitrarily large value, right?

    Thanks,
    Bart