Hello,
We experienced several TPS40170 controller and high side Mosfets failures during load transient test recently. The DC/DC buck converter design is based on the TPS40170 reference design, with 20V to 36V input, 16V/18A output. Steady state testing is good, pumping output all the power required and regulating well. The failure happened when we start testing the prototype with 0 to 100% load step, with load resistors. After the board failed, there were no more gate control signals output. Replaced high-side Mosfets and TPS40170, board worked normal again. This kind of failure repeated 3 times on two different boards, with same repair results.
The removed High-side Mosfets all had same G to S short, with resistance 1 to 2 Ohms.
Does anyone familiar with this failure mode?
Thank you for your time.
David
Hi David,
Which reference design is your project based on?
Which pin failed on TPS40170?
Regards,
Na
Hello Na,
Thank you for the reply. The reference design we mainly based on is PMP6680 (48Vin, 12Vout@20A).
The pin failed is pin 17, Hdrv for high side drive. After the failure, TPS40170 did not have any gate driving output signals from pin 17 or pin 14.
Are you sure your MOSFETs can handle the peak current from the sudden 0-100% load step?
What about a short period of transient overvoltage for the high side?
What changes did you make compared to the PMP6680? The minimum change would be output capacitors with higher voltage rating and UVLO_on threshold lower than 20V. If you do change other components, please let me know. Is the PCB layout following the reference design?
What exactly is the test condition when the failure happened? For example, load steps up or down? Input voltage of 20V or 36V?
About the high-side FET failure, only G to S short, no D to S short or open, correct? What is the impedance between D to S? In the 3 failures you have seen, both high-side FET has the chance to fail or always one of them failed? If always one of them, which one is it?
Thank you for the detailed questions. I will re-check and measure the failed parts and let you know the results, as soon as I can access the lab.
I did a little more measurement in the lab, and checked my lab notes, here are descriptions of the three failures, sequenced:
Testing condition, Vin: 26V, Vo <= 17V; Load resistor, 1.0R. High/Low side fets, two (BSC057N08) in parallel.
1st board, high side fets, off the board, one G to S 13.4 Ohm, the other one G to S 1.1 Ohm; both S to D body diode measure around 0.5V; D to S resistance in mega Ohms; TPS40170, pin18 to pin16, 0.9M, and increase; pin17 to 16, aound 200k; pin 14 to gnd, aound 200k;
2nd board, high side fets, off the board, one test normal (NOT damaged), the other one G to S 1.6 Ohm; D to S resistance mega Ohms, S to D diode 0.5V; TPS 40170, similar result as 1st one;
3rd board (1st board repaired), high side fets, off the boards, one G to S 1.5 Ohm, other one G to S 1.8 Ohm; body diode normal, D to S in mega Ohms;
Regarding schematic changes, I have to do a comparison with PMP6680 in order to report. This is our second PCB for TPS40170. 1st PCB is a smaller prototype board, with only TPS40170 related parts, we tested at steady state with very good results, and also tested with load transient upto 60% without any issue. The 2nd PCB design is an integrated design with a lot more functions and parts, and a MCU on board, with TPS40170 related PCB only occupy 1/4 of the total PCB. The 2nd PCB layout, because of the PCB size limitation, is not optimized for TPS40170 DC/DC only, it had to deal with all the other functions as well.
The TPS40170 based DC/DC converter is used as a load shedding device in our design, it's not used as a "normal" DC/DC converter. If requried, we can test the TPS40170 DC/DC converter seperately for trouble shooting. TPS40170 is very difficult to check or measure once taken off the PCB, we have to install it in circuit for testing.
Thanks for the details!
Based on the failure description, I am not convinced that HDRV of TPS40170 failed. Does the replaced TPS40170 not work with good FETs?
I would suspect that there is overvoltage stress on the gate-source of the high-side FET at load transient, either positive or negative, mainly due to layout parasitics. Can you capture the Vgs voltage right across the FET if you have differential probe? If not, you can probe both HDRV and SW with respect to the same adjacent ground with short probe lead. In order not to break the part again, you may try slower load transient and smaller load step and compare the difference at different conditions. To trigger the scope at load transient, output voltage or load current should be measured too.
Another concern is how the 1-ohm load resistor switched on/off. Does it connected to the circuit manually or by solid-state switch?
Na Kong Hi David, Thanks for the details! Based on the failure description, I am not convinced that HDRV of TPS40170 failed. Does the replaced TPS40170 not work with good FETs? I would suspect that there is overvoltage stress on the gate-source of the high-side FET at load transient, either positive or negative, mainly due to layout parasitics. Can you capture the Vgs voltage right across the FET if you have differential probe? If not, you can probe both HDRV and SW with respect to the same adjacent ground with short probe lead. In order not to break the part again, you may try slower load transient and smaller load step and compare the difference at different conditions. To trigger the scope at load transient, output voltage or load current should be measured too. Another concern is how the 1-ohm load resistor switched on/off. Does it connected to the circuit manually or by solid-state switch? Regards, Na
Thank you for the detailed response.
I will try to test the replaced TPS40170 with new Fets when I have access to the lab next week. The assumed TPS40170 high side drive pin failures are based on the datasheet information, Mosfet Vgs is rated at <= 20V, the TPS40170's HDRV pin to SW pin differential voltage is rated at 8.5V, so the drivers are not likely to survive. Out of the 3 boards, first one of the board was tested with replacing the Fets only, it was not working, so TPS40170 was replaced as well, then it worked. The other two boards, high side Fets and TPS40170 were replaced at the same time.
Regarding the scope signals, I will see what I can do to get more info.
The load switches are mechanical relays, not solid state switches. Yes, we know there will be arc spikes, so we have transorbs at both input and output of the DC/DC, which show no signs of stress.
Thank you for all your time and effort.
Thank you for the tech support. With the system requirement changes, we will have to redesign the PCB, so TPS40170 related layout will be reviewed and optimized for better grounding. We will start to test the new boards in a couple of weeks.
Thanks for the update!