TPS62136: Circuit Failure Analysis: Buck IC SW and GND Pin internally shorted

Hi Deniel,

Since you are testing with no load connected, the HICCUP current limit should not be triggered.

I have a few ideas as to how the device could've been shorted from this design.

1) Since you are using some external circuitry to change the output voltage there is a possibility that Vout is exceeding the setpoint and overshooting above 8.4V and causing a breakdown from exceeding ABS max ratings on the VOS pin (Vin+0.3V).  This could happen especially if there is ripple or noise getting injected into the FB pin. Also since there is no load connected it would be easier to overshoot the output voltage because there is no load to drain the current quickly.

Were you able to measure Vout at the time of failure to see if it exceeded the Max ratings?

Thanks,

Joseph

Hello Joseph,

Thank you for your response.

Unfortunately we were not able to see what the Vout was at the time of failure. 

Regarding our buck circuit and the external circuit to control it:

The external circuit to control buck output voltage is based on the https://www.ti.com/lit/an/slyt777/slyt777.pdf Application Note: Methods of output-voltage adjustment for DC/DC converters.

The voltages also in that circuit are 5V for the NPN transistor and PWM (0 - 3.3V) from the Microcontroller. 

The Resistor values on the Feedback Pin are based on Equation 1 in the application: The V,DAC is 0V - 5V and the Output Voltage would vary from 0.8V to 8.4V. Our Input Voltage to the circuit is 8.4V. The only time we get an 8.4V output voltage in the buck is when we have the input voltage higher than 8.4V.

Another note about the buck circuit in our system is that there are two of these exact same circuits on the same PCBA. Both share the same Input Voltage. The only difference is that each buck can be controlled independently (enable and output voltage) but in our program we control the both buck output voltage at the same time. The voltages of both bucks are almost exact but during the time of the failure only 1 out of the 2 buck IC got the shorted. Both were at the same conditions (max output voltage no load) during testing. 

 We also have a total of 5 PCBA with the exact same circuit and only 2 out of the 5 PCBA had a the single buck failure (only 1 out of the 2 buck IC on the PCBA failed). But all 5 were treated to the exact same scenario where the control circuit was set to 3.3V so that the Output Voltage would be around 8.4V.

We tested a known good board with the same conditions (max output votlage). Here are the observations we have found

Output_Control is set to 3.3V to have Max Output voltage

Input Voltage to Buck IC - 8.4V

Measured Voltage at the output of buck (after inductor before relay) - 8.39V

Noise/Ripple in the Output Voltage - 70mV peak ripple voltage

Feedback Voltage - 0.694V

Noise/Ripple in the Output Voltage - 14mV peak ripple voltage.

Do you think the issue is still related to exceeding VOS (Vin + 0.3V)? Do you think the issue that we saw is an edge case? Is there a way for us to replicate this failure by forcing VOS/Vout to be greater than 0.3V than Vin? 

Does this issue only happen when you have no load and go to the 100% Duty Cycle Mode feature of the TPS62136 Buck

Are there other possible causes for the failure given the observations we have made? 

Thanks,

Deniel

Hello Joseph,

Yes, the 1 out of 2 IC failure for both PCBA was in the same location. The layout we have on the board is the similar to the EV Board of the TPS62136. Both buck circuits on also have the similar layout. So far we did not see any voltage spikes in the output part of the circuit. The SW pin of the buck IC has spikes in the ripple voltage. Are the voltage spikes expected since the SW pin is connected to the inductor in the circuit.

We will try to repeat the failure condition to see if we can see any issues. 

We are also operating the buck converter from input voltages of 5.5V to 8.4V. The reason why we have 8.4V as the max output voltage from the feedback resistors is so that we can always have the output voltage near the input voltage (100% duty cycle mode of the buck converter). Is there an issue with doing this? is there an issue if of setting your maximum voltage above your input voltage? 

Thanks,

Deniel

Hi Deniel,

My colleague Joseph will give you reply in Monday!

BR

Zixu

Hello Joseph,

Thanks for the response. I looked back at our previous TI /postsquestions regarding the TPS62136 and found that we did ask a similar question last year about having Vout greater than max Vin. We were told that it should be fine since the 100% duty cycle mode. But back then the way we change the output voltage of the buck converter was via the SS/TR pin and we then change it to the external circuitry since that is what they also recommended in the forum

This is the exact Q&A from the forum question

"1. When the set Vout is greater than the max Vin, is that an appropriate condition for the chip to be in? The output voltages in this case become equal to the input voltage, correct? Does this stress the chip in any way?

This should be fine. Since you are using tracking functionality, the VFB pin voltage will be modulated as a function of the VSS/TR voltage. When VIN and VOUT are close enough, eventually, 100% duty cycle mode will kick in and keep the internal HS FET on. At this time, the VOUT value will be VIN minus the IR drop across internal HSFET and inductor DCR. You can check this out in the 100% duty cycle operation section in datasheet. "

Here is the link to the actual question/post for full context:

https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/986404/tps62136-using-ss-tr-pin-to-vary-output-voltage

The reason I brought this up it to clarify if operation (Vin < Set Vout) is acceptable or are you referring to something different to what was answered in the post/question we linked above. Or does it only work with the SS/TR pin method of controlling the output voltage?

If Vin < Set Vout  is find to use, Does it mean that there might be another cause for the buck IC to get shorted internally? 

We built the whole operation of the buck converter from the answer that we got from the previous question we posted last year .

Setting the max output voltage to 8.45 and utilizing the 100% duty cycle to always get the output voltage near Vin when we need it ( we can still control it to go below Vin) 

Thanks,

Deniel

Hi Deniel,

When you adjust the Vout setpoint from the FB pin side and Vout>Vin, the device is technically trying to regulate the voltage higher than Vin (boosting). Even though it hits the 100% duty cycle limit which prevents Vout from rising higher than Vin in steady state, it won't do anything to prevent overshoots less than the regulation voltage set at the FB pin (8.45Vmax  in your case). 

Basically, designing the FB pin such that you are relying on 100% duty cycle to limit your output voltage isn't the best practice because there can be edge cases where Vout my get higher than Vin (Load steps, sudden changes in Vin).

A possible workaround to add more protection could be using the SS/TR pin connected to resistor divider on Vin, this would allow the output voltage setpoint to decrease as Vin drops. This way you could still use the manual control of the FB pin and have the SS/TR pin provide protection as Vin drops.

Unless you are able to capture waveforms at the time of failure, this seems to me the most likely possibility for failure since you only saw a couple ICs fail.

Thanks,

Joseph

Hello Joseph,

Thank you for your response. We appreciate these detailed explanation/information regarding the possible issues resulting when Vin is below the set max Vout. It is unfortunate that this information was not clear when we ask about this test case in the previous post we asked last year. 

If the goal is to prevent a scenario where Vin is less than Vout during the 100% duty cycle mode, I would like to propose an alternate workaround without needing to use SS/TR pin for the following reason:

1. There is a possibility that we might not be able to do a current PCB layout change that has a resistive divider between VIN pin and SS/TR Pin

2. We do want to avoid using the SS/TR pin to adjust the voltage. We did ask this similar question in the previous post/question. We did see in the schematic and other application notes that the SS/TR is mainly used for sequencing and tracking. We have not seen any application where it is use the change the output voltage during a full load connection. Plus it also adjust the FB pin voltage which would give a wider tolerance for accuracy. 

Q: Is it appropriate to be using the TR pin in this way as the primary output voltage adjustment? In the application schematics I typically see the TR pin being used to synchronize two TPS62136's, or to adjust on-ramp voltage, but I can't see an explicit reason to not use it as a variable output control pin.

A. Yes normally TR pin is used for tracking - you can refer to this app note for that: https://www.ti.com/lit/an/slva470a/slva470a.pdf  If the voltage at the FB pin is below its typical value of 0.7 V, the output voltage accuracy may have a wider tolerance than specified. I believe this is specified in the EC table. Margining via the feedback pin avoids the additional inaccuracy of the gain between TR and FB. But if you have a closed-loop system, which monitors the actual output voltage, it should be fine. 

The alternative workaround we are considering is to limit the actual output of the buck converter by monitoring the input voltage as it goes down and limiting the maximum duty cycle of the PWM input to external circuit that changes the output voltage. 

In the current design (referencing the circuit in the post above), Output_Control goes from 0% duty cycle to 100% duty cycle and makes the output of the buck converter go from 0.8V to 8.4V. Via the software program we can limit the duty cycle so that the output voltage is set to what the current input voltage is. Example: with 8.4V input a duty cycle of 90% changes the output voltage to 7.56V, if the Vin drops down to 7.56V VIN, we limit the Output_Control duty cycle (via software) to only 90% (or slightly lower) there by still forcing a 100% (or slight lower) without forcing the device to regulate above the current input voltage. As the input voltage goes lower in time, the possible settable max output voltage is equal or slightly lower than the input voltage via the controlled output of the external circuit. Solution is similar to the SS/TR pin without needing to use that pin. 

As you mentioned, controlling the output  voltage limit (Vout < or = to Vin) might be a possible solution unless we are able to replicate the issue and see the waveform during a failure scenario. 

Let me know if this is a good alternate workaround or if there are also some issues that this might introduce (design-wise) 

Thanks,

Deniel