• State Resolved
  • Locked Locked
  • Replies 10 replies
  • Subscribers 64 subscribers
  • Views 644 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

LM22672 getting stuck in a weird regulation cycle

Kevin McKee
Prodigy 100 points
Other Parts Discussed in Thread: LM22672

I'm using a LM22672 to step down 24V to 5V. Occasionally (1 time in 200-300) the LM22672 will come up stuck in a limit mode at about 2.2 V with 140ns wide pulses. It will stay in this mode until power is cycled.

This is a related problem to an earlier discussion (LM22672 into ideal diode weird behavior). The suggested solution of adding a reversed bias schotky diode from output to imput seems to improve how often the problem occurs, but doesn't eliminate it. What I am seeing now is that the LM22672 can enter the limit state without the other supply being present.

The 24 V power supply doesn't always behave nicely. It can be hot connected and there can be a lot of contact bounce when first connected. This can last up to ~100 ms. However I have found no correlation between the contact chatter and the LM22672 going into the limit state. Or for that matter between anything & having it go into the limit state.

Can a similar latchup occur with other pins (besides Vout) being higher than Vin? The soft start and the sync pins are possible candidates since they could be at higher voltages then Vin during the input transients.

  • 0
    TI__Genius 10095 points

    Hi Kevin,

    Let me talk to some of the other applications engineers and I will try to think of what could cause this.

    Can you post a schematic and board layout for the part?

    Regards,

    Marc

  • 0 in reply to Marc Davis-Marsh
    Prodigy 100 points

    The schematic is easy... There is a reverse biased 1n5819 added from C108 to C104, per my earlier discussion with TI.

    I'm having a bit more trouble with the layout.

    BGA244_IND_Rev D SCH.pdf
  • 0 in reply to Kevin McKee
    Prodigy 100 points

    Since it won't take gerber files & it is a 4 layer pcb. How would you like the layout file?

  • 0 in reply to Kevin McKee
    TI__Genius 10095 points

    You should be able to zip the gerbers files and attach them.

    Regards,

    Marc

  • 0 in reply to Marc Davis-Marsh
    Prodigy 100 points

    Dooh!

    Here you go.

    thanks,

    Kevin

    SRS PCB.zip
  • 0 in reply to Kevin McKee
    TI__Genius 10095 points

    Hi Kevin,

    I don't see anything blatant.  Here are a couple possibilities.

    1) On test point 103 the switch node comes very close to the Vout sense trace.  Maybe this is injecting noise into the system.  We could try cutting the switch node trace near the inductor.

    2) You currently have enable floating.  With Vin being hotplugged, you might want to add a UVLO circuit.  Say a 13.5K top resistor and a 1.6K bottom resistor divider from Vin to EN.  This would disable the part when it is below approximately 15.1V.  This would help to keep the part from trying to start up when Vin is not yet stabilized. 

    I'll keep looking.

    Regards,

    Marc

  • 0 in reply to Marc Davis-Marsh
    Prodigy 100 points

    Marc,

    Some more info:

    I made a "power on transient test" circuit to recreate my +24V input voltage behavior. After a bunch of tests I created a waveform that fairly repeatably causes the LM22672 to lock up in the fashion I previously described. About 1 in 2-4 trys fail.  I've attached 4 graphs all using the same stimulus signal: one short pulse, a brief off period and a much longer pulse (which mimics steady state).

    The green trace is the control signal. The dark blue trace is the +24V into the LM22672. The pink trace is the sync pin (TP104). The lite blue trace is the +5V out.

    Graphs 1 & 3 shows successful power on. Graphs 2 & 4 are locked up in the "bad" state. Graphs 2 & 4 have the +24V, sync & +5V all at the same scale & origin. There are small differences between each measurement & there isn't an obvious difference between the good & bad ones.

    I hope this will give you some more ideas. I'm going to try out your EN pin circuit & will report on its success.

    Kevin

    Graphs.zip
  • 0 in reply to Kevin McKee
    TI__Guru 63005 points

    I tend to agree with Marc about the EN pin.  You may want some kind of UVLO on that.

    Also, the dV/dT on the Vin pin might be causing some trouble; if possible you may want to try to slow it down some.

    FD

  • 0 in reply to Frank De Stasi
    Prodigy 100 points

    Using the EN pin as UVLO seems to solve the problem. Changing the dV/dt on the input pin isn't a good option since the design is pretty much completed, but is something I will explore for future designs.

    One mystery is it looks like the enable threshold for the enable pin is around 2.1V. From the data sheet I would expect it to be around 1.6V. This means that 1.58k/13.3k parts gives a turn on voltage of ~20.6Vdc. I changed the parts to1.58k/11.3k and got a turn on voltage of 17.9V.

    Do you have an idea of what on & off threshold voltages I expect and how much does it vary part to part?

    thanks,

    Kevin

  • 0 in reply to Kevin McKee
    TI__Guru 63005 points

    The OFF threshold is 1.6V typ; as found in the data sheet under "VEN". 

    The ON threshold is 1.6V + 0.6V = 2.2V; the 0.6V is hysterisis; in the data sheet under "VHYST"

    Also, there is a 6uA pull-up current that may effect your calculation if you use high value resistors.

    FD