• State
  • Locked Locked
  • Replies 6 replies
  • Subscribers 62 subscribers
  • Views 231 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.

LM34914: High Percentage of Buck Converter failures

Alistair Gadd
Prodigy 40 points
Part Number: LM34914

Hi there,

I've recently done a design of a 12DVC fan power supply which draws about 75mA at full power and I can only assume a startup current of possibly up to double that. I've used the LM34914 for the Buck converter to pull an input voltage of 22 to 36VDC down to the required 12VDC. The first batch of 7 boards have been in service for a couple of months now, but I have had 2 failures from this batch where the outout voltage has disappeared.

It seems that in both cases, the internal Vcc regulator is not generating any voltage (normally about 7V) - does this indicate chip failure because I cannot see that any other factors influence Vcc production. If so, would anyone know why I might have had such a high failure rate?

My design, which was an exact extract of a WebBench design, looks like this:

and the PCB top layer looks like this:

and a picture of the PCB segment looks like this:

 Because of the low current requirement, I haven't used vias and the back layer for cooling, but the exposed pad is soldered to the centre strip between the pins.

Can anyone see where this design might be lacking or shed some light on where I might improve reliablity?

Many thanks to anyone who can help, best regards,

Alistair.

  • 0
    TI__Mastermind 23825 points

    Hi Alistair,

    Do you know which pin got damaged? Is the High Side FET diode still there?

    One major thing that i saw from the layout is that the path for C34 and C35 is obstructed by a giant hole. Normally we try to keep the C34 and C35 closer to pin and minimize inductance in that path.

    You can probe the SW node at full load and see how high the spikes go on the SW node

    Thanks

    -Arief

  • 0 in reply to Arief Hernadi
    Prodigy 40 points

    Hi Arief,

    Thanks very much for getting back to me.

    If you mean the regulator high side FET, I get Vsd = 0.48V and Vds = 1.7V with board unpowered (multimeter diode voltage), whereas on a good one I get 0.74V and 1.8V.

    The hole that you mention is between C34/C35 and Vin (pin 10), but I didn't really have an option there so I made sure that the hole was plated through all round which gives a lot more copper to the path.

    Here are waveforms I get at full load (1A) on the SW pin:

    and at no load:

    Personally, I don't see anything sinister here, but maybe you would see something you don't like.

    Thanks again for the help, best regards,

    Alistair.

  • 0 in reply to Alistair Gadd
    TI__Mastermind 23825 points

    Hi Alistair, 

    The waveform looks ok for me. You mentioned that on the bad unit the VCC does not come up. Have you done a swap between bad unit on good board or a good unit on a bad board just to check any mechanical/soldering issue.

    Thanks

    -Arief

  • 0 in reply to Arief Hernadi
    TI__Mastermind 23825 points

    Hi 

    I hope you manage to resolve the issue. For now i will close the thread but feel free to reopen it when you have further questions

    Thanks

    -Arief

  • 0 in reply to Arief Hernadi
    Prodigy 40 points

    Hi Arief,

    Sorry for the long wait, but I had other urgent issues I had to resolve.

    I did as you suggested and took the LM34914SD from the working board and placed it on the non-working board, and then the chip from the non-working board and placed it on the working board (swapped the chips). The issue then followed the chip - the previously non-working board came to life but the previously working board didn't. The "bad" chip still had no output from the regulator on pin 9.

    The datasheet specs 40V as the max input voltage and we are using an 8S LiPo which gives a maximum of 33.6V, so I'm sure that is not a problem, but if there is any connection sparking I would expect that C34 and c35 would handle it.

    I'm now going to try replacing the chip with a fresh one and introduce some severe sparking on the input to see if I can blow it in the same way. I'll give you feedback.

    Worse news is that it seems that there has been a failure of exactly the same supply on an aircraft there in the US - there is dual redundancy built in for CPU supply so the CPU is still running, but there is no chance of cooling and it is worrying to the client. They are attempting to replace the unit, so that the bad one can be sent back here.

    So the bottom line is that we must resolve this issue - for me to move to another design would be a huge problem.

    Best regards,

    Alistair.

  • 0 in reply to Alistair Gadd
    TI__Mastermind 23825 points

    Hi Alistair,

    I think your experiment makes sense. Normally the VIN pin impedance to GND will be different from a damaged unit compared to a good unit when there is an overvoltage happens. Seems like it is safer to use a 60V device especially if there is a spike in the VIN. 

    Or you may need some damping capacitor or TVS diode to protect the VIN pin. 

    Are the failure for the other supply on an aircraft is using the same layout? 

    The large hole between VIN and C34 and C35 still a concern for me too since it add inductance in a high di/dt loop on a buck converter. 

    Thanks

    -Arief