• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 63 subscribers
  • Views 859 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

LM25085: Odd input snap across P-CH FET that prevents regulation with a Vin above 12.5V

Nicholas Price
Prodigy 100 points
Part Number: LM25085

I am debugging a set of LM25085 based buck power supplies.  They are both doing the same thing, but let's start with the +5V version.

Here is the schematic:

This had originally been generated in WEBENCH, but I've made a few changes on the board to improve possible issues that there may be in the layout, design, etc.

Following a few posts I've read, I increased the input capacitance to around 50uF.  C8 is also DNP.  In an attempt to rule out a bad kelvin connection in the layout, I increased R12 to 3.32k (imagine that).  

This power supply works very well without load.  It regulates slightly above +5V with little to no spikes (with a +15.5V input).  My issue  is that any kind of load or increase in Vin will crash the supply down to an improper voltage and nothing will fix the output until the load has been removed and the power supply has been power cycled.

That is pre-load (CH1).  The next image is what happens when the load is introduced (CH1 output,  CH2 gate):

This is a zoom in of the ring I have been observing when the chip enters this state.  I assume this is a parasitic issue, but have tried my best to address what I can.  What concerns me is that there is no recovery once the load is removed.  My primary concern is that diode D4 could be affecting regulation by resonating with the switch cycle per 8.2.2.9 of the datasheet.

If I decrease the supply VIN to +12V, then I have a fairly ugly output, but the power supply remains in regulation with the load  out to 2ish Amps no problem.  That particular test seemed to rule out my kelvin connection concerns.  The connection is not ideal, but also not too bad.

My leads for measuring are longer than normal for safety reasons, but when I do the measurements right at each node manually, the ring is still there (with slightly lower amplitude).

I am going to run through adding some gate resistors to slow down the P-CH fet and slow down the turn-on time of the FET.  Any other advise may help, I imagine a ring of this magnitude will be detrimental to downstream components, even with CMOS..  This power supply is also supposed to run up to 24V, so it will need an update, for sure.

This problem is similar to this post:

I suspect they had similar issues, but did not address the FET turn-on time issue they may have had.  

Remember, this problem is directly related to VIN.

Nick

  • 0
    TI__Prodigy 360 points

    Hello Nick,

    Thank you for your detailed explanation. Assuming VIN is really the problem, one solution would be to place a Bypass Capacitor (value = 1uF) from VIN to Gnd as close to the VIN pin as possible. This will help eliminate any ring possible.

    Attached is an example circuit from the datasheet section 8.2.LM25085 example circuit.docx

  • 0 in reply to Gautam Hari
    Prodigy 100 points

    Gautam,

    I played around with input/output capacitance for quite some time.  That helped marginally so.

    In order to eliminate the ring, I added a gate drive resistor to slow down my P-CH FET.  I was avoiding this because it involved some lead forming on my FET to look pretty.

    See below.  With a 100 ohm resistor, the parasitic ring is nearly gone.  CH1 is the output and CH2 is the switch node.

    Ripple at load is increasing the regulation voltage so I suppose I will tackle that next.

    Here is the same set of wave forms with a 3 ohm load on the output (purely resistive).

    Transient response is good and ripple is not so good.  I imagine for the final rev, I will need to bring up the inductance a bit with the output capacitance.

    My efficiency is also in the dumps (77%) so I am likely to walk back the gate resistance to save some switching energy.  The FET is now at a 30C rise.

    Ignore the crosshairs, my FLIR is slightly out of cal.

    Based on the original design, how would you calculate output capacitance and inductance?  The equations and calculator that I have used, put me here as a sweet spot and it seems as though I should be increasing most of these values across the board.

    Nick

  • 0 in reply to Nicholas Price
    TI__Prodigy 360 points

    LM(2)5085_Quick_Start 5Vout 600kHz.xlsHello Nick,

    I still think placing a parallel input bypass capacitance (<1uF) in addition to the input capacitance you have will help reduce the ringing.

    Attached is a Quickstart tool which will provide guidance on your ripple injection circuit:

    Regards,

    Gautam