TPS62065 Failure during burn in

I'm sorry that you're having some failures.  This is a very robust IC.  Almost all failures can be attributed to exceeding one or more of the absolute maximum ratings.  We need to make sure your circuit is not violating these limits.

Have you tried testing a bad IC on a known good board?  If it behaves the same as it did, this would verify that the IC is damaged.

Could you send a schematic of your circuitry?  What is connected at the output during your testing?  What is the load?

Could you send your PCB layout?

The EN pin should always be terminated.  Having a look on your schematic could show a way to prevent it from floating.

What temperature does your testing go up to?  If the device is operated with a heavy load and with a thermally poor PCB layout, even a relatively low ambient temperature could cause the device to heat up enough to exceed its junction temperature rating.  Your PCB layout and test conditions will help determine if this could be the case.

Thanks for the reply. I'll provide a copy of the schematic and layout tomorrow night.

I have not tried testing a bad IC on a known good board. I can try that tomorrow. Since replacing the IC caused the circuit to work, I have trouble believing the PCB or rest of the circuit are the problem.

The circuits were unloaded during testing when the devices failed. Normal load testing is normally less than 750 mA.

Under normal operating conditions, the EN pin is terminated. Unfortunately, the terminating resistor is on the motherboard, so testing just the "daughter card" power supply PCB doesn't terminate the EN pin. This has been fixed in the replacement design. I've also added terminating resistors to my test carrier board.

The thermal cycling is from 10C to 70C and the elevated burn-in is at 60 to 65C. During this testing, the circuit is unloaded. The 24 mW disipated in the IC (forced PWM Mode) shouldn't cause the junction temp to exceed 125C at these ambient temps. I've run the circuit at 110C ambient in an unloaded state and it still continue to work. I was trying to reproduce the failure mode.

I'm not sure how or when I could be exceeded the absolute max specs, but I'll double check all of my numbers and see if anything new jumps out at me. Let me know if you need anything else.

Thanks for your detailed explanations.  I agree that at no load you should not be hitting a thermal issue.  A few more questions and things to look into:

During the testing, is there anything connected to the output of the inductor besides the output caps and the FB pin resistor divider?

What dielectric caps are you using?

How is Vin applied?  Many times, an adaptor is plugged into the wall first and then the other end of the adaptor (which already has the voltage on it) is plugged into the unit under test.  This creates a hit plug event where the voltage on the adaptor output oscillates and overshoots through the lengthy adaptor cable and on board capacitance that has little damping.  This could create overshoots that exceed the abs max on Vin.  I would suggest looking at a scope capture of Vin and Vout when power is applied.

Was the EN pin terminated during your testing?  I'm sorry but it was unclear from your respone which version of the design was used in this testing.

I would appreciate the chance to look at your schematic and layout.  You can also order the EVM for either IC and see how they perform under the same test.

During thermal testing, nothing is connected to the circuit output except a trace to a test header, but the header is unused. During hand testing this header is used to connect a variable resistor and multimeter for load testing.

I'm using X5R or better in the design, so I don't believe that would cause this problem.

Vin is applied with an Agilent power supply. The input voltage is ramped up in a controlled manner versus just applying it hot. My supply voltage is 3.3V, so the overshoot would have to be pretty large to exceed the 7.0V absolute max input voltage. Also, there is additional circuitry on the board that has an absolute max input voltage of 6.0V and it hasn't failed. I would expect it to stop functioning also if the input votlage was overshooting to the damage point. I can take a look at the input and output voltages at startup to see if there is anything out of the ordinary going on.

I moved a "bad" IC to a good board and it still didn't function. I would say this is a good indication that the ICs are failing.

The EN pin termination is a potential problem. As I stated, the TPS62065 ICs are on a daugther card and the EN lines are driven by the motherboard. During testing, I've got multiple daugther cards on a test carrier board. In order to test every supply separately, the input voltages are jumpered to each daughter card and the EN lines are jumpered to the input voltage. Unfortunately, I forgot to add terminating resistors to the carrier board. Therefore, if the jumpers to the EN lines are pulled, the EN would be floating. It's also possible that during elevated temp burn-in that the Vin could get disconnected and float if the jumper fails to make a good connection and the EN pins are still driven high. That being said, I've applied a voltage to the EN pins with Vin floating and everything continues to work. This was done multiple times at elevated temp with no problems. I've also run the circuit just driven off of the EN lines. It draws excessive current, but functions properly when returned to proper operating conditions. I've been trying to replicate the high current draw failure, but I haven't. I redesigned the board to add pull-down resistors to the EN pins, but I'm holding off fabrication and assembly till I have an idea why the ICs are failing.

The units where the IC draws almost no current (~400 uA), the enable line is tied directly to AVin, so as long as there is voltage applied to the board, the IC should be enabled. The confusing part is that there are 2 different PCBs using the same design and both have had failures. The only difference is connecting the EN pin to AVin versus controlling it externally.

I've included a PDF of the schematic and the layout. The schematic is reduced to just show the power IC circuits. The PCB is a 4 layer board and the views are as seen from the top. The thermal pads on the TPS62065 are filled with GND vias, but they don't show up very well.

Please let me know what your thoughts are on what could be causing the ICs to fail. Thanks.

Yep, schematic and layout look fine.  Thank you for posting those.

Just to make sure nothing weird is going on, I would take a scope picture of Vin when you apply power.  Also, do you cycle power repeatedly during your testing?  I've seen this cause oscillations and overshoots before.

These ICs don't need an RC filter on AVin.  I don't know of a customer that uses one as it requires two extra parts.  You might try shorting R5 and R6 and seeing if you still have failures.  Possibly you are seeing too much drop across them creating a differential between AVIN and PVIN.  Or possibly your AVIN input cap is not large enough.  Shorting those resistors would solve both of these possibilities.

Which IC has failed--U1 or U2 or both?  Sometimes forced PWM mode causes some issues as the IC is pumping energy from the output back to the input at no load.

If none of the above resolves this issue, then I would recommend contacting your local TI representative or distributor who can begin a failure analysis on your failed ICs.  This may take some weeks to finish, which is why I would like to try everything else first.

Thanks for reviewing the schematic and layout.

I'll work on getting you scope pictures of the startup. It may take me a day or two.

The RC filter on the AVin pin is added as standard practice on prototype designs. It's added in an attempt to reduce output noise by filtering any input noise spikes from the DC/DC regulator. We test with and without the RC network and if the noise performance is better with the RC filter, we keep it in there. As for the voltage drop across the filter, if the IC is working properly, it's only 40mV. I'll look at shorting out the resistor to see if that helps with our problems.

I've had both U1 and U2 fail in the design. I've already moved the forced PWM one over to auto-switching. 

I'm in the process of testing all of my remaining prototype boards. They've gone through all of the thermal cycling and are finishing the elevated temp burn-in. I'm hoping this will provide more information. So far, everything looks promising based on overall current draw. I'll let you know what I figure out.

Thanks again for the help.