TPS62932: URGENT: Request support for suspected circuit instability and PCB failure

Hi  Roberto Di Magno,

My colleague Bella will reply to you later.

Thanks

Colin

Hi Roberto

We recommend selecting inductor with Isat> IHS_LIMIT(3.4A for TPS62932).

During start up, the converter need to charge the Cout so that Vo rise to the set value in the tss, the iL can be large and may hit the IHS_LIMIT, especially when Cout is large.

The 2.3A Isat of the inductor is too small, the datasheet didn't specify the effective inductance under >2.8A,  ferrite core can have sharp saturation curves. Please replace L1 with a inductor with Isat>5A, and see if the issue can be solved.

The load profile has steps and spikes, is this a programed load profile? What's the load of the TPS62932?

Thanks and best regards.

Dear Gui He 

thanks for your reply. The load profile has steps and spikes because it consists of: microcontroller, digital sensor with an active heater system, 2x relay, and a pulsed IR lamp with duty cycle of 2Hz that by itself alone requires 115mA in active mode, etc...

I would like to ask the following:

1. Can the use of the LSXND4040TKL3R3MDG inductor in this configuration damage the DC/DC or the load in a short time? It's very important to have an answer to this question because we have to stop the delivering of those products to our costumers in case this risk can exist.
2. As for devices already manufactured and tested (full test completely succeeded), what is the risk of failure over time? Is the MTBF affected? On a few samples we tried to operate a stress test on the power supply (plug-in and disconnect repeatedly the power supply with standard load) and we didn't notice any anomalies but we have three cases of failure with the DC/DC in short circuit on a batch production of 50 units. In two cases the failure of the DC/DC happened at first power up, while it seems that one case happened on field, but we are still waiting the unit to return for inspection. Can you advice us how to deal with the rest of manufacturing that is around 150 units that are already assembled with LSXND4040TKL3R3MDG?
3. We are working to replace the inductor with one having an equivalent footprint. Is the DC resistance of the inductor a critical parameter? The WEBENCH tool recommends using an inductor with a maximum DC resistance of 41.67 mΩ, while we have some options ranging from 55 to 75 mΩ max. How to  proceed, what you suggest?
4. We selected a possible substitute for the inductor. Could this P/N: MPL-AL4020-3R3 be a good replacement?

I report the main parameters of the inductor MPL-AL4020-3R3:

Thanks and best regards.

Hi Roberto,

Bella will reply to you next Monday.

Thanks,

Colin

Hi Roberto

For 1 and 2, I am not sure if the inductor is the root cause for the damage. Tests and investigations need to be done to identify this. But inductor with higher Isat is recommended anyway. 

3. higher DCR inductor will have higher power loss under same load condition, it doesn't matter if the efficiency / inductor temperature is not out of spec.

4. The inductor is ok.

The failure rate now is 3 out of 50, do you know under what operate condition did the damage happens? Can you reproduce the failure ?

I think you can try adding an overload /short circuit/ hot-swap.  You can record waveforms of Vin, SW, Vo. 

"once the power supply is given the DCDC starts to get hot and probably damaged"---Did you use a thermal screen to monitor the temperature of the device? what temperature would it be?

Thanks and best regards.

Dear Gui He,

thank you for your reply.

Let me give you more details about the failure of the 3 units:

1. It always happened with the minimum load (around 100mA in continuous) on power plug in. This means that after normal use with load for several hours, leaving the units unpowered for few hours and plugging the power in with the minimum load, the DCDC fails to supply power and starts to heat up until, after several plug outs and plug ins, it fails.
2. Measuring the input impedance of the unit it shows around 150mOhm. Thus, powering with 24V let the PTC protection to open.
3. In this situation I'm sure at 100% that the load is ok since by powering with 5V on USB auxiliary port of the device it works normally.
4. There is also one case upon three where the input is close to open circuit, but no power at the output at all.

At this point I think the failure can be due to an instability at start-up. Can be this due to the limited inductor saturation current? Do you think that we solved the issue replacing the inductor?

Actually we are testing 3 units with the following inductor part number:

1. P/N: MPL-AL4020-3R3
   • Inductance = 3.3uH
   • Current Rating = 4.4A
   • Saturation Current = 5.2A
   • Rs = 34.5mOHm
     
     
2. P/N: TMPF 0403LR-3R3MN-ABD
   • Inductance = 3.3uH
   • Current Rating = 5A
   • Saturation Current = 6.2A
   • Rs = 28.6mOHm
     
     
3. P/N: WPN4020H3R3MT
   • Inductance = 3.3uH
   • Current Rating = 3.45A
   • Saturation Current = 4.7A
   • Rs = 72mOHm

Thank you for your support and BR,
Roberto

Dear Gui He,

I'm trying to use your WEBENCH tool to redesign our schematic. I'm experiencing some difficulties in use it:

1. WEBENCH sometimes automatically sets the "Configuration options" diffrently from my choises, so I can't analyze the system performance and do simulations.
2. I noticed that, once I set the "Switching frequency", when I recalculate the operating values for different Vin, the Fsw listed is not the same as the one set in "Configuration options". Is it normal?

Can you help me to use this tool? This point is becoming critical because we are about to produce a new batch of units and need to find a solution to the problem reported in the previous messages.

I'm trying to costumize the design with this configuration:

• Vin: 8-28V
• Vout: 5V at 0.8A
• Small Footprint Design optimization
• Soft Start Time: 5ms
• Output Cap Type: Ceramic
• Cff disabled (we can't add this capacitor because actually we can't modify the PCB, only do a component replacement)

Can you kindly let me know if you can support me in this task?

Thank you and BR,
Roberto

Also, can you measure the Vout and SW  during normal load profile and check if it works normally? The Cout is one 0805,22uF,10V ceramic cap, I didn't find the effective capacitance curve under 5V DC bias, but it might have > 50% capacitance derating under the 5V bias. You can try adding more capacitance.

For the webench simulation, I see in you schematic the RT is connected to GND, the fsw is 1.2MHz.  The Cff can't be disabled in the simulation to gain enough phase margin for the loop stability. 

https://webench.ti.com/power-designer/switching-regulator/customize/91?VinMin=8&VinMax=28&O1V=5&O1I=0.8&base_pn=TPS62932&AppType=None&Flavor=None&op_TA=30&origin=pf_panel&lang_chosen=en-US&optfactor=3&Topology=Buck&flavor=None&VoltageOption=None 

Hi Gui He,

We have done a few test, here below the recap:

1. Load with 10Ohm resistor at 5V and 7.5Ohm resistor at 5V to simulate 500mA and 650mA steady load, providing multiple fast connect and disconnect of input with several voltage including 24V, 12V, 7.5V and 5V. Nothing happened, the circuit didn't show any problem. 
2. Minimum load close to zero (can be least than 1mA) providing multiple fast connect and disconnect of input with several voltage including 24V, 12V, 7.5V and 5V. Nothing happened, the circuit didn't show any problem.
3. Load with 10Ohm resistor at 5V to simulate 500mA with power at 7.5V and 24V with a 7.5Ohm resistor in series with the power supply to simulate bad external power supply condition. Nothing happened, the circuit didn't show any problem.

On monday we will do the measurements required by you with a proper instrument and we will try to inject signals to get the Bode diagram and we will post you back the results. 

In the meantime, since the time is short to solve this problem and we need to confirm production no later than 21th August, we have the following options:

1. Keep the current configuration with inductor having higher saturation current (MPL-AL4020-3R3, TMPF 0403LR-3R3MN-ABD or WPN4020H3R3MT, listed in order of our preferances). You have already the main characteristics in preavious post.
2. Replacing the output capacitance of 22uF with one of the following suggested by WEBENCH tool:
   1. GRM21BR61A226ME44L (22uF, 3.0mOhm, 10.0V)
   2. C2012X5R1A476M125AC (47uF, 2.9mOhm, 10.0V)
3. We have the option to increase the bulk input capacitance actually there is space for 2x1210 ceramic caps and the actual design is using 2x10uF 50V X7R (CL32B106KBJNNWE). How far should we go is something that can improve the stability, considering that we don't know wich power will be provided by the user?
4. Move the limit Fsw from 1.2MHz to 2.16MHz to overcame the possible instability with high load and low input voltage according to the design suggested by the tool:

This configuration is stable according to WEBENCH tool in all the possible input and load conditions. Also the load is limited to max 800mA by an active current limiter, so there is no possibility to go over this limit in any case.

Regarding the PCB layout you can have a look here:

The only possible ilimitation should be the one highlighted in yellow:

1. 20 mil input track with 37mm length and 38 mOhm resistance separating the 2x10uF bulk capacitors from the input pin. The 0.1uF capacitor instead is closed to the pin.
2. By mistake of gerber revision, there is a separation between the SW pin and the inductor with a 8mil track with 0.2mm length, see the Top layer artworks.

Also, please, have a look to the input protection circuit. We found a note on datasheet that TVS on the input side with "snap-back characteristics" is not reccomended. 

------------------------------------------------------

Considering the above situation and the measurement test we'll do on monday, we will need your support to:

1. Decide if we can use the PCB already manufactured with this layout to go in production
2. Define the BOM with the improvement proposed above (inductor with higher Isat, replacement of Cout capacitor, increase of Fsw to 2.16MHz, increase of the bulk input capacitor if needed)

We don't have other solutions and the deadline for us is 21th August to close this case.

Thank you in advance for your support and availability.

Best regards,
Roberto

Hi,

Bella will reply to you later.

Hi Roberto

The layout has some issue and must be optimized if you got chance.

• The Cin should be placed as close as possible to the IC Vin and GND pin. The Cin in your layout is not close and the Vin trace is thin, there can be larger parasitic inductance on the Vin trace and this can cause voltage spike across the mosfet during switching, especially when Vin hot-plug. You can test the SW voltage and there should be voltage spike. And you can add a 0.1uF cap close to VIN and GND manually and the spike should be able to see the spike becomes smaller.
• Please connect the GND in completely, do not leave this hollow.

• The SW trace is too thin, which will also cause larger parasitic inductance and higher voltage spike.

• The feedback resistors ,R3 and R2, should be placed close to the IC FB pin. R3 should also be close to the IC GND pin.

Attach the layout example of the EVM, and there is pcb layout guidance in the 12.1 section in the datasheet.

Will you be able to optimize the layout before the 21th August production? I think this is quite important.

Recommend replacing the inductor and the capacitor as you listed above(item 1 and 2). Others I think is not so important, but you can update it as well.

For the TVS,  if the TVS broke and clamp the Vin at a low rate voltage, but the Vout is higher than this voltage, there is voltage across the HS and there can be large current flowing through the body diode, that can cause damage. the diode you selected do not have a snap-back characteristic, it should be ok. 

Thanks and best regards.

Dear Gui He,

here below a few measurement, with a 125Mz digital oscilloscope:

1. Vsw without load
   1. Test conditions: Vin = 24V, Iout = 50mA, BW limit = none
      
         
2. Vout without load
   1. Test conditions: Vin = 24V, Iout = 50mA, BW limit = none
      
      
      
3. Vin without load:
   1. Test conditions: Vin = 24V, Iout = 50mA, BW limit = none
      
      
      
      
4. Vsw with 10R load:
   1. Test conditions: Vin = 24V, Iout = 500mA, BW limit = none
      
      
      
5. Vout with 10R load:
   1. Test conditions: Vin = 24V, Iout = 500mA, BW limit = none
      
      
      
      
      
      
6. Vin with 10R load:
   
   
   1. Test conditions: Vin = 24V, Iout = 500mA, BW limit = none
      
      
      
7. Vsw without load:
   
   
   1. Test conditions: Vin = 24V, Iout = 50mA, BW limit = 20MHz
      
      
      
      

In the meanwhile, we had other 3 units got broken in the following conditions:

1. 2x units connected to a 12V Lead/Acid battery until full discharged, then when reconnected to the battery fully recharged did not work
2. 1x unit stopped working during the oscilloscope measurements while the probe was connected to the SW pin and the nearest ground. This happened at first start-up

During the measurement session we have tried to overcame the layout issues in the following way:

1. Adding a short wiring connection between the SW pin and the inductor with a solid wire with 0.3mm thickness, also tinned to increase the current rate. Nothing changed in the SW pin voltage measurement.
2. Adding 2x 10uF bulk capacitors close to the Vin pin with a solid wire track of 3mm length, 0.3mm thickness also tinned, and short wired to GND. Nothing happened.
3. Adding a bootstrap resistor of 5Ohm and the only effect was a smoother rise of SW pin voltage, nothing changed on the undershoot.

Considering the above testing and adjustement seems that we cannot find the root case that is leading the units to be permanently damaged.

We need to take a decision: 

1. Replacing the inductor with one with higher Isat (please note that we have done also measurement with a reworked unit with an inductor having an Isat of 5.3A, but the result is not visible in the measurements. We can still keep the bootstrap resistor and replace the 22uF capacitor with a 47uF one, but we noticed that the undershoot seems to be worsened by increasing the output capacitance, but not damage to the circuit happened with this change.
2. Review the entire layout design, but this means stopping the next batch of production putting the existing PCB in the garbage. Can this layout improvement resolve the issues at 100%?
   

Considering lot of sensitive information in this post, can we continue the conversation in private?

Thank you and BR,
Roberto

Hi Gui He,

any news?

Hi Roberto

We didn't see SW voltage spike from the waveform, Can you measure the SW during start up? Better to have Vin, vout ,SW on the same page.

For the damaged devices, are these pcbs with new inductors? Can you measure the damaged device and see check the Vin , SW ,GND pins and see if there are short-circuit between these pins?

• 2x units connected to a 12V Lead/Acid battery until full discharged, then when reconnected to the battery fully recharged did not work

          ---------What's the Vin voltage for a fully charged battery? Can you measure the SW when powering up? you can trigger the vout rising stage.

• 1x unit stopped working during the oscilloscope measurements while the probe was connected to the SW pin and the nearest ground. This happened at first start-up

          -------Did you start-up the circuit by hot-swap? Can you catch the waveform?

For below actions:

1. Replacing the inductor with one with higher Isat (please note that we have done also measurement with a reworked unit with an inductor having an Isat of 5.3A, but the result is not visible in the measurements. We can still keep the bootstrap resistor and replace the 22uF capacitor with a 47uF one, but we noticed that the undershoot seems to be worsened by increasing the output capacitance, but not damage to the circuit happened with this change.

        ---------Did you test with new inductance and see if the damage still happens?

        2. Review the entire layout design, but this means stopping the next batch of production putting the existing PCB in the garbage. Can this layout improvement resolve the issues at 100%?

         ---------cannot say for sure now without test verification. But the layout does need to be optimized.

This thread is internal only. You can also send email to bella-he@ti.com 

Thanks and best regards.

Hi Gui He,

as told you before, we need to find the root cause of the damage of the IC, any other attempt is just not compatible with this product and application. We cannot release in production one product that still have, even residual, risk of failure.

The main problem is that we cannot reproduce the failure that seems to happen randomly. We had failure with 24V, 12V and mostly with minimum load, around 50mA, in unpredictable situations, and whenever we tried to reproduce exactly the same failure conditions we never succeded in breaking the DCDC.

The only consideration that I can do after all the time we spent to investigate is that the failure seems to happen in hot-plug, at least for those happened in our hands. I'm really worried about this situation as I can see that we are not the only one experiencing this frustrating issue without a clear path toward a solution. We found the following threads describing exactly our problem:

1. https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/1289998/my-tps62933-is-failing-exploding-when-under-normal-use
2. https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/1389956/tps62932-properly-spec-d-vin-is-causing-the-device-to-fail

We don't see any positive solution to both threads, just everything left as it is. May I ask you if you can provide an MTBF report for this component? 

We have urgency to solve this issue and take a decision for go/no go in production, and we don't have time to iterate for fix and try... Even with the inductor replacement we need a significative number of units to reproduce statistically the possibility to have a failure and this is the same for any attempt for improvement including the layout revision.

I'm wondering if it's not the case that TI open an internal investigation to understand if anything related to the characteristics of this component may be the root case of failures, I think 3 threads describing this problem are enough to procede urgently in this direction. We work to release stable products intended to last for years (at least 10). 

I will contact you by email to understand how we can move on.

Best Regards,
Roberto

Hi Roberto

I have replied you email. We see damage case thread for other products also since we support for all customer's that are using the devices but most are application issues.  We will try to find the root cause, but might need some test/measurement results for debugging.

Thanks and best regards.