• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 3 answers
  • Subscribers 63 subscribers
  • Views 365 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.

LM5121-Q1: Negative to Positive Converter issue

Scott Rubenstein
Prodigy 80 points
Part Number: LM5121-Q1
Other Parts Discussed in Thread: LM5121

We are using the LM5121 on a pluggable board to convert -28V from a chassis backplane to +12V. The +12V feeds two other DC-DC converter  to produce +5V and +3.3V. The +12V output is fairly stable when the board is operating in the chassis, but when the board is pulled from the chassis while powered up, the +12V spikes to about +17-20V. This spike is causing catastrophic damage to the downstream DC-DC converters (internal short between +12V input and GND).

 

The LM5121 circuit was designed by a third party, and they are no longer available to help.

 

The area that seems to be causing the issue is in the feedback loop. The divider is connected directly between the +12V output and the -28V. So when the board is removed, the -28V jumps toward 0, causing the +12V GND to jump as well, which causes the spike on the downstream DC-DC input and eventually causes the failure. We think that we need something in the RFB2/RFB1 path to stabilize the GND reference for the +12V supply. Not exactly sure the best method for doing that.

 

Any help would be much appreciated.

  • 0
    TI__Guru* 78470 points
    Hi Scott,

    Thank you for posting the question. Could you share the schematic?

    Thanks,
    Youhao
  • 0 in reply to Youhao Xi
    Prodigy 80 points

    Thanks for the response Youhao. Please see the attached schematic page. Note that parts with values starting with the @ symbol are not populated.

    3731.-28VTo12V.pdf

  • 0 in reply to Scott Rubenstein
    TI__Guru* 78470 points

    Hi Scott,

    I reviewed the schematic and your feedback is not correct. When you want to produce a regulated +Vout from a negative input, and you need to employ a level shift circuit to sense the +output voltage, because Vout references to the system GND but  your IC references to the negative input rail.  

    You circuit works when plugged in because you actually set your boost regulation voltage to 40V wrt to the IC GND pin (which is your negative input rail).  When you suddenly remove the input, your IC GND will no longer stay at -28V wrt the system GND, but your feedback setting still wants to regulate to 40V, which leads to the Vout rise wrt to the system ground.

    If you use the level shift circuit for the FB, such issue will be resolved.  Attached is the level shift circuit for the Feedback in such system.

    Level Shift Sensing.docx

    Best Regards,

    Youhao Xi,  Applications Engineering

  • 0 in reply to Youhao Xi
    Prodigy 80 points

    Hi Youhao,

    Thank you very much for reviewing the schematic and for your suggestion. We were thinking the problem was in the feedback, but weren't sure how to go about fixing it.

    With regard to the level shift circuit, I am a little unclear of a few things:

    1. Ic = Vout / R1   Is Vout wrt IC GND or System GND?
    2. VFB = Ic x (1+HFE)  x R2   HFE seems like it can be a large range (typically between 60 - 400). A calculation using this variable will vary widely depending on the HFE value used.
    3. VFB = VREF = 1.2V, correct? And is this 1.2V wrt IC GND?

    Appreciate the feedback,

    Scott

  • 0 in reply to Scott Rubenstein
    TI__Guru* 78470 points

    Hi Scott,

    Sorry there are typos in the formula but the basic idea was correct in the drawing that I provided earlier.  I messed up Ie and Ic (emitter and collector current, sorry for the confusion).  I attached the update drawing below.  Now the answers to your questions:

    1.  Vout is wrt system GND.   Ie (the emitter current) is determined by Ie = Vout /R1.  The two NPN's Vbe can be considered cancelling each other.

    2.  VFB = Ie x HFE /(1-HFE) x R2.    HFE's effects is actually very small.  Please plug in 60 or 400, and you won't seem much difference.

    3. Yes, VREF is wrt IC GND.

    Hope this clarifies.

    Level Shift Sensing RevA.docx

    Thanks,

    Youhao

  • 0 in reply to Youhao Xi
    Prodigy 80 points

    Yes that clarifies my questions. Thank you Youhao.

    Using 12V for Vout, I used a 10K for R1, and a 1K for R2 (and the other resistor connected to the 2nd transistor) . That came out to just under 1.2V (1.18) for VFB.

    When implementing the modification on the board, the output voltage was not the 12V as expected however. It changed to about 18V using those calculated values. The VFB voltage measured 1.19V.

    Changed R1 to 47K, and R2 (and other resistor) to 4.7K. Again, Vout measured 18V, Vfb = 1.19V

    Changed R1 to 43K. Vout = 16.8V, Vfb = 1.18V.

    Change R1 to 39K. Vout = 14.8V, Vfb = 1.18V.

    At this point, we probed the Vout to see if it spiked when the board was removed from the chassis. It did not. So this modification method is solving our problem, but the Vout is not at the 12V output that we would like.

    We can keep trying to reduce the value or R1, but it seems something else (Rcomp, Ccomp values) is affecting the Vout value. 

    Any thoughts?

    Thanks,

    Scott

  • 0 in reply to Scott Rubenstein
    TI__Guru* 78470 points
    Hi Scott,

    The paired/matched PNP bipolars should be used to cancel the Vbe voltages of the two devices for accurate output setting. Otherwise the Vbe difference will introduce regulation errors.

    You can verify this by swapping the PNPs and you will see the VOUT would be reduced after swapping.

    Hope this clarifies.

    Best Regards,
    Youhao Xi
  • 0 in reply to Youhao Xi
    Prodigy 80 points

    Thank you for the suggestion Youhao. We removed the two single bipolar transistors and replaced it with a dual PNP (intrinsically matched) transistor. The Vout remained about the same using the above combination of resistor values.

    We experimented with a few combination of values, and settled on 12K for R1 and 2.7K for R2. This provided an output voltage of 12V and a VFB of 1.2V. 

    There was no spike in Vout when removing the board hot from the chassis or when other boards were inserted hot into the chassis while this board was running (which was happening previously without this circuit).

    Not quite sure why the math doesn't work out, but by all accounts this seems to have resolved our issues. 

    Thank you very much for all your help.

    Best Regards,

    Scott Rubenstein