ADP2302 self-destructs

Hello,

I would appreciate some assistance in debugging my ADP2302 design.

I am using the fixed 5V version of the device to provide a switching power supply that will accept between a 9 and 20 VDC input and provide 5 VDC to my board at up to 2A.  However, in my initial testing phase my load is somewhere in the 50mA region.

The problem is that the ADP2302 doesn't seem to be switching at all. The output voltage seems to track the input voltage. As I increase the input voltage from zero to 5V, the output voltage tracks it as expected, and the current consumed from the supply also increases. However, above 5V in, I would expect to see the SW output start switching, and the current consumed from the supply start to drop.  Instead, it contines to rise. If I increase the input voltage to more than about 9V, the ADP2302 starts to overheat, and self-destructs.

I have copied the reference design and layout from the datasheet pretty faithfully, with a few substitutions for components I already had in stock.

As part of the debug I tried putting a scope probe on the SW output and then connecting Vin to an already-running 7V power source (any higher and I risk damaging things).  The waveform looked nothing like the one in the datasheet, as there was no switching.  Infortunately my scope does not have the ability to save waveforms, so I have hand-drawn it and attached to this post.  The time/div setting was the same as the diagrams in the datasheet, but instead of seeing the SW net switching on and off as the output voltage ramps up, there is a smooth curve with an initial boost up to 7V before falling back and then rising again up to the static 7V.

Obviously the output voltage should not be as high as 7V.

I will attach a screen capture of the layout of the PSU section of my board.  It has been rotated by 180 degrees so that it matches the one in the datasheet.

The bottom copper layer is blue, and the top copper layer is red.  The bottom layer power plane is GND, and I have added labels (black text in white boxes) to label the power planes on the top layer. The big green box over the centre of the ADP2302 is a solder mask keepout to create a nice large heatsink pad.  There are also two vertical rows of vias under that pad to conduct heat to the GND plane on the underside.

I will also attach a schematic of the circuit.  In addition to the components in the datasheet, I have also added RFI filters on the input and output.

The VCC5VPSU on the right of the schematic is connected through an ideal diode to the main 5V power plane of the board (to allow the board to be powered by another power source without competition if necessary)

The components I selected are as follows:

C2: Samsung CL32A106KA9LNNE Ceramic X5R 25V 10uF

C3-5: Samsung CL32A226KAJNNNE Ceramic X5R 25V 22uF

L1: NIC NPIS27H 10uH 0.0216DCR 5.4A Isat

D3: STPS2L25U: Vf=0.375V, If(av)=2A

I can't see any obvious mistakes I've made, so is anyone able to notice anything that I've missed?  If not, how would I go about debugging a problem like this?

  • David,

    I have looked at your layout and it looks good. Schematic looks good as well. I am wondering if you are having a shorting problem on the board. make sure that your components are installed correctly and soldered good. I have a feeling your problem may be a soldering issue.

    Also, I am wondering if the input inductor ( input filter) may be causing some spikes on start up causing the IC to shut down. Have you tried to remove the input inductor from the circuit, replace with a short and see what that does?

  • Hello "power-layout-guy",

    Thanks for your advice.  Yes, I'd already checked the boards for shorts and dry joints with both a microscope and a multimeter.

    I've replaced the input filter with a hard short, removed the output filter to ensure that the rest of the PCB can't be causing any problems, replaced the ADP2302 die and checked the Schottky diode at the same time.

    When I powered it from my standard bench power supply, it worked fine, and produced a nice 5 V output - even when I connected up a 2.5 R dummy load to stretch it to its full 2A capabilities.

    However, I then tried powering it from the PSU I intend(ed) to use in production - a standard laptop-style 19.5 V SMPSU. It was only connected for a few seconds, and the die got noticeably hot. I went straight back to my linear PSU, limiting the Vin to about 7 V, but the ADP2302 die was already in "suicide" mode - the Vout now tracks Vin. If I connect it back to the 19.5 V PSU, I'm sure it'll violently overheat again, like the previous dice did.

    I've checked the voltage on the SMPSU, and it's always just 19.5V - it doesn't go over the 20 V max Vin of the ADP2302.

    I'm somewhat disappointed at how incredibly fragile the ADP2302 appears to be.

  • David,

      Get a fast scope, say 1 GigHz, 10ns/div, and set the trigger level for 20V, in single sweep mode.

    If it ever triggers, you have a problem.  A 10-20 ns spike a volt or two above 20V can cause E-Field

    breakdown.

    Harry


  • A few questions, do you have the belly pad of the IC soldered to GND?

    Okay, so I understand this correctly, when you have it connected to your bench supply it works fine. can you take your bench supply up to 20V?

    Can you check to see,( with a scope) what the output of your 19.5v supply looks like.

    Apps max on the ADP2302 IC is 24V. you need to see if your 19.5 volt supply sends out a large voltage spike on start up. Also, you need to check the ripple on the 19.5V supply during operation.  let us know what it looks like.

    What we think is happening is this.

    there is a long cable between your 19.5V power supply and the ADP2302. That means there is a ton of inductance in that cable. When you turn on the ADP2302, there is a massive voltage spike, maybe 30v or more. that will kill the IC real fast every time. you need to set up your scope and trigger on the input of the 2302.

    you Might be able to use a bulk cap on the input to help..