• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 64 subscribers
  • Views 4763 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.

LM2596: Looking for cause of damaged L1 Inductor

Adam L
Prodigy 30 points
Part Number: LM2596
Other Parts Discussed in Thread: LM2595, LM2575

Hello!

I have been working on fixing a problem with a design where the L1 Inductor in this switcher is failing due to over heating / over current.  I am trying to figure out what could cause the Inductor to fail, and I need to be sure I have covered every possible reason in the fix on my next board revision.

The input V (on the left, but not shown) is roughly in the range of 10V to 14V from an LiFePO4 battery.  There is also a battery charger that can charge the battery and put a very slight ripple (at about 1MHz) on the ~12V rail when it is plugged in.

The current draw (3.3V side) on this design is in the range of about 10mA (while the unit is asleep) to about 500mA. 

All caps on shown in this schematic are currently ceramic.

The layout for this part looks like this (there is a ground plane not shown that is located immediately below the top layer) :

After getting some units in the field, we are probably seeing a 2% to 3% failure of the L1 inductor where it will overheat and either completely fail short, or overheat and lose some inductance.  When partial inductance is lost, the output gains a little bit of ripple and chop (not shocking!). 

After I went through the design, I found many issues:

* The output inductor appears to be too low given the sleep mode of the unit.  Even though the actual output voltage is very clean (almost no noise) at the low current draws, it sure seems like the Webench tool is suggesting a larger inductor

* The ouput capacitance is far too low and should be a greater value. 

* The input and output caps should not be ceramic -- The inputs should likely be Tantalum Polymer (not solids or the inrush could kill them when a battery is plugged in), and the outputs should likely be Tantalum Solids (or polymer if I want to spend the extra money).

* The feed forward capacitors appear to have not been calculated and are too high.

* I also found that there are some ceramic caps placed on the board in poor locations (on the edge of the board, near mounting holes, or both), and it is possible that some caps are failing short circuit right across the 3.3V rail (either intermittently or permanently).

Solutions:

* I have changed all of the ceramic caps to automotive flex safe caps, and moved their locations to be away from the board edge and mounting holes.

* The input and output Cap types around the switcher have changed to Tantalums

* The output cap value has increased to match the WeBench Design Suggestions for input of 11-15V, output of 500mA, 3.3V output, etc.

* The feed forward cap value has changed to approximately match (+- a few nF) the value provided by following the rules in the data sheet

A little more info:

* I have TRIED to make the inductor fail by decreasing output capacitance, increasing the FF cap value even higher (making it even worse), putting a short on the 3.3V rail, etc., and I cannot figure out a way to duplicate the overheating inductor on my bench.

* If I replace the inductor on a failed unit, it fixes the issue.  The chip itself never takes any damage.

Questions:

* I'm hoping someone can give me some smoking gun type logic as to how I was smoking the inductor in the original design

* I'm hoping someone can tell me if I still have any design issues in the schematics and layout I am about to show for the "fixed" / "Improved" version

* I also want to confirm the input and output cap values, and the inductor values -- Should I really design them solely based on the max output current, or should I also take into account the really low current draw while the unit is sleeping??

* Is there something outside of this chunk of schematic I should be looking to? 

* Are there any major layout issues that could be causing this problem (or others)?

* Am I asking missing any questions that I should be asking?

Here is the new "fixed" schematic:

Here is the new board layout for what I hope is the "fixed" design:

Thanks so much for any input!!!!   I'd love to make that silly failing L1 inductor issue GO AWAY!


Best regards to all of you,

Adam

  • 0
    Prodigy 30 points

    The first message appears to have dropped my images!

    I am trying to attach them to this message...

    Failing SCH:

    Layout of above:

    Hopefully Fixed Design?:

    And the layout for the "fixed" design:

  • 0 in reply to Adam L
    TI__Genius 11085 points
    Hi Adam,

    The first thing to check is the inductor current rating. Note that the LM2596 is a 3A part. In cases of output voltage short to ground fault condition, the inductor is going to see much larger current than the 500mA max load. The part has thermal shutdown when it is over heated. So it is possible the inductor was damaged in a short circuit fault conditioin.

    When you say the output capacitance was too low and Cff was not correct, they both lead to marginal loop stability. It is also possible that the inductor, or other components, have tolerance and a small percentage of the boards were not stable. Instability will cause current oscillation, and the peak current will be much larger than a stably operating loop. That could also damage the inductor, if the current rating was not enough.

    When you say if the damaged inductor was replaced, the damaging condition cannot be reproduced. That sounds like the inductor part to part variation plays a role here.

    Yes, the inductor should be selected for 0.5A load, not 10mA. Since the part is a 3A part, the 47uH inductor works for the part. But it will have more current ripple than a larger inductance. But the current rating is usually lower with higher inductance within the same family of inductors, or case size.
    Would you consider using a lower current rated part: LM2595? They have the same package, but not pin to pin compatible. The LM2595 is a 1A part, so it has much lower current limit to protect your inductor. Following Webench design should be a good start.

    For the PCB layout, there are a few general rules:
    1. The power traces (connecting VIN, SW, VOUT, GND of the diode and CIN, COUT) should be as short and wide as possible. Thin traces from CIN ground to IC ground should be widen.
    2. For Cin's, the lower value ones should be closer to the IC. Since the lower value ones are more effective in bypassing high frequency switching noises, putting it closer to the IC prevent the high frequency noise spread to other parts of the board. It also applies to Cout's, not Cin is more critial in buck converter layout.

    We can review your layout once you are done.

    Regards,

    Yang
  • 0 in reply to YangZhang
    Prodigy 30 points
    Thanks very much for the response Yang!

    I was wondering about using the LM2575 since it is pin for pin compatible. It appears to be switching at a much lower frequency, so it might require bigger components (I haven't actually looked into that yet). In general, that makes sense as to how I could be smoking the inductor -- since the current limit is so high (3A is high relative to my 1A inductor), it is likely smoking my inductor when a bad condition hits my board.

    I'm guessing the LM2595 is almost exactly the same as the 2596 in terms of switching frequency, and most other items but just has the lower switching current and the different pin out.

    I will look into both of the other parts to protect my inductor, along with possibly just increasing the inductor size.

    Thanks for the feedback, and it would be great to know what you think of me possibly just dropping in the LM2575 in my current design.

    Have a great rest of your day,
    Adam