Charge Pump failure, although recommended fixes were applied

Hi there,

We are using a TMC2160 Stepper Motor Driver.

We design the PCB according to the recommendation ( well my HW colleague missed the Filter recommandations for the charging pump but designed part of the circuit according to the datasheet) 

The driver has been used  several days without problems with 24V voltage input ( Lab DC power supply ) 

The initial design is in the picture here. 

While trying to increase the voltage of the Power supply to 48V ( which is our voltage target) we experienced troubles as the driver became hot, drew more current even disabled state, and the motor could not longer be controlled. 

We tried the recommended changes  (Datasheet Section 3.1, p14) which is an RC Filter on the charge pump input. 

With power supply of 24V, the signal on CPO and CPI were 2 Square signals, Amplitude ~12V , F= 750KHz and an offset of +24V on CPI. The square signal was also visible on both sided of the signal once the failure occurred , but the 24V  offset was missing.

I had no time to make an scope screenshot, as soon after , the same failure appeared on the driver ( not sure if I might have made a short with my probe or if it was the same problem, but I didn't even had the time to try the motor this time) 

We have 3 x 220uF Caps on the 48V , as near as possible to the power bridges.

Did anyone come across a similar  problem ? 

Is there any additional recommendation concerning what that issue, that we could try before burning any more chips ? 

Thanks

[Edit] 

After checking our PCB, the 3 x 220uF Caps for 24V/48V buffering/decoupling capacitors seems be connected to PGND and not GND.

PGND is used to measure the motor current / calculate power ( PGND and GND are connected through a 20mOhms Shunt)

The driver however is connected to GND and 24V//48V

Not sure if that could be one of the cause of the Charge Pump failure in the Driver.