LM63635-Q1: Chips are burning out

Hello

Your schematic should be OK.

It looks like the device is on the bottom side.

You will need to flood the bottom with ground plane to connect all the grounds.

It is hard to see where AGND (10) is connected; it should have a direct connection to the DAP.

Also, try to avoid the use of thermal reliefs on the power components.

Please ensure that the inductor is rated for the current limit of the regulator.

Do you have another ground plane on the layer directly under the regulator ??

Can you monitor TP1 during a power-on cycle to ensure there are no voltage spikes.

How much load are you using during the power-on cycle ??  And what type of load ??

Thanks

Hello Frank,

Thank you for your quick response.
-You will need to flood the bottom with ground plane to connect all the grounds
-It is hard to see where AGND (10) is connected; it should have a direct connection to the DAP.
-Do you have another ground plane on the layer directly under the regulator ??

The bottom is already flooded with a ground plane (it is set to translucent in Altium to view the other layers). In addition there is an uninterrupted ground plane on layer 3 (directly below the components, layers 1 and 2 also have been flooded with ground plane.
Pin 10 is connected to ground directly.Image: https://imgur.com/a/5IrMWBE
-Also, try to avoid the use of thermal reliefs on the power components.
I agree that the thermal relieves should not be around the power components, this will be updated in future versions. Although we do have another PCB with the same chip (and schematic) that also burns out at random times. In this circuit the power components have been connected directly to the ground plane without thermal relieves.

-Can you monitor TP1 during a power-on cycle to ensure there are no voltage spikes.
The board is powered by a bench power supply, so no voltage spikes should occur, here is a measurement that has been done without a load:Image: https://imgur.com/a/UAsilvW

-How much load are you using during the power-on cycle ?? And what type of load ??
There is a small capacitive load on boot up of around 46.22 uF, including the converters own output caps, after that there is an active load, the rest is resistive.
The active load consists of 7 power isolator circuits, all using the SN6501 IC.
This circuit uses a transformer, the transformer's data sheet states that it has a minimum inductance of 475 µH, Test conditions for this value are specified as "N1+N2/ 10 kHz/ 100 mVAC".
Some tests have been done with an electronic load in constant current mode at 400mA. This is above the highest measured current draw of the system as a whole (with a simple multimeter). See the results:

Image: https://imgur.com/a/JZvENxw

Let me know which other tests can be done to further investigate.

Kind regards,

Gino Krol

Hello

Can you monitor VCC, Vout, SW, and VIN during your power-up sequence.

Also, can you supply a schematic of your load.

Thanks

Hi Frank,

These  are the measurements results, this is measured on a PCB with only the converter circuit on there, an electronic load was used as a load here (siglent SDL1020X-E) :
https://imgur.com/a/9DXHfyX
yellow is Vin, green is Vcc, orange is SW and blue is Vout.

Within the actual circuit, there are seven units with isolated power:
There are five isolated RS422 PulseDir Modules and two isolated RS485 Modbus Modules.
https://imgur.com/a/B68fbRm

Aside from those, there is a PIC32MZ2048EFH100 MCU and a UART to USB port IC:
https://imgur.com/a/E4W5EUi

Again, none of these schematics provided here were attached to the load during the provided test results. But they were connected when the chip burned out.

Hello

You can try to re-connect one load at a time and see if it changes the start-up response, or causes damage.

Thanks

Hi Frank,

We have performed the same measurements with the full load attached, including 7 of the isolated supply circuits based on the SN6501 and a PIC32MZ micro controller:
https://imgur.com/a/CuTmYQi

The first two images are of the previous measurements using an electronic load.

There seems to be some ringing in the SW signal, could that be the cause of the IC failure? If so, would it help to place bigger capacitors next to the SN6501 isolation circuits? Or would a bigger change likely be necessary?

Hello

The ringing on the SW node is just the normal DCM behavior at lite loads.

It may be a random voltage spike on the input to the regulator.  You may want to use 

an aluminium electrolytic capacitor on the input and a transient suppressor.

Then you can test again and see if you get any failures.

Thanks