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.

LM25119 design board bringup procedure - diagnosing why voltage is correct but wont source current

Other Parts Discussed in Thread: LM25119

LM25119 Webench generated design:   Vin 30v to 40v, Vout 27v (max), Iout 12amp

Laid out the first board (4 layer) with signal on the outer layers and with 0805 passives to aid in debug.   Board assembly house built first two and both have the same problem.  

This design has provisions for adjustable output voltage, however I've removed the eePot and have fixed resistors.

The problem I'm having is output voltage is as expected, 27v, but when it is loaded with 30 ohm resistor the voltage sags to just above zero.

I'm looking for feedback on the board bringup procedure and if others have seen this kind of problem.

I've gone thru the components and they look correct - there could be an assembly error since this is first article.   There could also be a layout or design error but I did a thorough check plot prior to fabbing the PCB.  

Notes on Webench improvement suggeestion:: it would be good if Webench generated a 'dot' on intentional junctions and crossovers had no junction dot in the generated schematic.   Also, the eagle cad generated output is a good start, however the *.sch and *.brd have consistency errors so I could only use the generated *.sch.

OK, so here's the board bringup procedure I'm following:

0) Go for it - connect 32v to the board set a current limit of 2amps, and apply power.   Voltage output is 27v, however with 30 ohm resistor, the output voltage folds to near zero.   Do the same on the second board.   Same result.   suspicions:

 a) RT is set wrong and the switching frequency is way off

 b) current sense resistor is wrong value (should be 0.008 ohms)

 c) under voltage lockout is being activated.

1) verify design and layout against that generated by Webench

2) check component values as stuffed against BOM vs Webench generated, check part orientation of FETs

3) Share the design on TI e2e community in the hopes someone spots an error or has a helpful suggestion 

4) measure voltages on UVLO, look at waveform on CS1, CS2

5) drink a beer...

I'd appreciate suggestions and I'll try to post what I find.

  • Hi Thom

    You can check the board following this quick test procedure.

    1. Check VCC. If no VCC,no switching .
    2. Check SS. If SS is grounded, no switching.
    3. Check COMP, if COMP is grounded , no switching.
    4. Check RES, If RES is in hiccup, the output will be almost zero even if there is a switching.
    5. Check gate waveforms.

    Regards,
    EL
  • Thanks for the guidance.   Oddly enough, it seems there a two modes.  One with the correct Vout and can't supply current and the other with Vout at 6.1v.   It seems like when I was probing around I could get Vout at designed voltage, but have not figure out the magic procedure.

    Here's my measurements from last night:

    Vin: 34.4v

    Vout: 6.1v


    1) VCC1 is 7.5v, VCC2 is 7.7v

    2) VSS1 is 0.124v

    3) Vcomp1 is 0.085v

    4) Vres is 0.00v

    5) Vuvlo is 1.85v

    6) Vfb1 is 0.21v

    I'm not sure what hiccup mode is, but what I saw on gate waveform is a small pulse (<1us) at 5Khz rate.

    Here's a version of Webench

  • I've ohm'd out the connections on a bare board - there are no shorts in the PCB. It looks like there is some error condition being detected by the LM25119 but the error is not obvious to the Webench designed part. I've verified the 0.008 ohm resistors so I doubt the fault is a current limit. My guess is a wrong part has been stuffed in the compensation network. I will remove the suspect components and stuff with a verified value. If that doesn't work, not sure next step. Any one out there want to make some money debugging a Webench generated design?
  • In case anyone is interested in pictures of the board (apologies for the blur)

  • Hi Thom
    Please check if RES is in hiccup mode. If it is in hiccup, please try a RC filter at CS-CSG.
    Regards,
    EL
  • I only have the data sheet to go by and it says that hiccup mode is due to current limit.  I've tried shorting the current sense resistors however this does not change the mode - how can the part sense current if the resistors are near 0 ohms.  

    I have a 0.008 ohm resistor across CS CSG this is inline with the hgh current path thru the switching FETS.   I assumed that RC filter would be a cap in parallel across the 0.008 current sense resistor.  I've tried 0.01 uF and 0.1 uF with no change in results.   As noted above a short across the resistor does not disable the current limit either.   Oddly enough the design is pulling 11mA at 37v input - not what I'd expect with current limit.   Is there some other secret fault that can cause hiccup mode?

    I AM NOT ABLE TO GET THE SWITCHING converter to emit a switching frequency and I've not yet captured the 256 cycles of switching that are supposed to occur prior to hiccup mode..   Perhaps I should cut the traces going to the FETS and feed with a function generator to 'bump start' the design into switching, then perhaps I can see what is wrong.   Before cutting the traces,   I'm going to try and disconnect the second converter as perhaps they are not actually able to operate in parallel (has this been tested at TI?).  Maybe if I can get one half working at half the current that would be a start.

    I wish there was some way to disable the fault detection in the device, but it looks like I"m stuck.    The layout matches the schematic generated by Webench, the components are stuffed correctly as best I can verify.  

    Perhaps it was silly to assume that Webench could generate a working 10amp buck converter.    If someone knows how to disable the fault detection that is preventing the converter from operating it would be appreciated.   Also, if there is another part that might do the job in a simpler fashion I'd be willing to sacrifice efficiency for something that has a chance of working.

    Bounty offer:  I will pay shipping and $200 for someone who can get this design to function - send me a note with your contact info if interested.

    Thanks