TPS766: Voltage reverse during turn off

Hi Manu,

That's odd. It looks like the device attempts to start back up momentarily as the input voltage ramps down. This device does not have a UVLO circuit and the input voltage ramps down slowly, so I wonder if the internal circuits are behaving unexpectedly when the input is less than 2.7V (the recommended minimum operating voltage) but is held there for hundreds of milliseconds. 

What is the load on the device? Is there any way they can ramp the input voltage faster than is shown in the scope shot? 

This part was released in 1999, before low-ESR ceramic caps were super popular, and when aluminum electrolytic and tantalum capacitors were very popular. This is the reason our older devices were not designed to be stable with low COUT(ESR).

Regards,

Nick

Hi Nick，

The loads are all supply power of IC and the Cout is already 47uF in this case, which means it is not that necessary that the unexpected power release from load cause this scenario.

Also, ESR of Cout cannot be measured if it is higher than 300mohm,could this be a possible reason and if there is any other action that customer can do to solve this problem?

Thanks for your help!

Manu Chang

Hi Manu,

You can observe in the scope shot that the input dips at the same moment that the output rises, which indicates that the energy surge comes from the input supply, which means that the device attempted to turn back on briefly. Since this device doesn't have UVLO, I think the input voltage falling so slowly and sitting under the minimum recommended 2.7V is causing issues. Can they speed up how quickly the input voltage falls during shutdown? 

I don't think ESR is an issue here. If there were an ESR issue, the output would be unstable during normal operation, but we can observe in the scope shot that the output is stable and well regulated before this shutdown event.

Regards,

Nick

Hi Nick, 

Thanks for your response!

Our customer have tried to decrease the input capacitor to decrease the time of input voltage falling, even they use a external 5V DC source as input to test.

But they found out that the input voltage waveform in these cases seems no differences and the problem still exists. 

Are there any other methods that may speed up the input falling?

Since the problem doesn't get fixed even with an external source, could it caused by the load or other reasons?

Appreciate for your support, thanks.

Regards, 

Manu

Hi Manu,

Can they share a scope shot of the test with an external DC source as the input? It would help to further evaluate the cause of this behavior. 

What is the load when the device is powered down? It seems unlikely to me that the ferrite bead would cause a voltage spike like this because if the load were to disconnect quickly I would expect a voltage to develop across the ferrite bead with a polarity as below. With that said, you mentioned that when there is no load there is no voltage spike. For this test is everything else the same?

Regards,

Nick

Hi Nick,

First sincerely appreciate for your support, about the waveform with DC source, customers said that it was almost the same with the one I showed you before.

I am also very odd about this problem, customer also have tested with and without the ferrite and there were also no differences between the waveforms according to them. For the tests with and without load, everything else are the same according to my customers. 

Besides, they also have tested their TPS76633 in their other projects/cases, they also seemed to have this problem.

Do we hear from other customers about this kind of problem of TPS766 before? Or are there any other possible reasons? 

Thanks again!

Regards,

Manu

Hi Manu,

What I wanted to see in the DC source scope shot is the input waveform. I am still reasonably confident that this is an issue with the input voltage staying under the recommended minimum voltage for ~300ms. I think a good way to test my theory would be to use a function generator with a buffer amp so that they can bring the input voltage down quickly instead of letting it slowly discharge. If they did this I don't expect that they'll see the issue persist.

I was not able to find any documentation for this issue in the past, so it seems like this is the first time that this was brought up. 

Are they able to change to a different newer LDO? There are many newer devices that have UVLO circuits and will not exhibit this kind of behavior. 

Regards,

Nick

Hi Nick,

Sorry that our customer don't have the function generator, and due to the customer's strategy they cannot bring the board outside to do the tests.

Do we have EVM or demo using TPS766 so that I can test and try to recurrent this problem or try to fix this? I have searched in TI.com and was not able to find any EVMs. If you have some demo maybe you can send to me to do some test, or it would be much better if you can help to do this simple test to recurrent this problem, thanks.

I have also sent an email to you base on this case, may be we can keep communication through emails, thanks again!

Regards,

Manu