TPS7A33: Regulator failing for unknown reasons

Hi Justin, 

The datasheet has noted that the HIGH-ESR cap may affect stability. Have you tried to remove the two tantalum caps on the output? 

I see you have provided the AC-coupled measurements on the outputs if you put it on DC, how much change have observed? Have you monitored the input signals at the same time? 

Regards, 
Jason Song

Regards, 

Jason Song

I have tried removing the tantalum caps and this did not make the wandering output voltage go away. My understanding is that if the tantalum caps were causing instability, it would manifest immediately on powerup and then go away immediately once they were removed, correct?

What's happening to me is when I power on the regulator the output looks fine, but then after a few hours it starts wandering. Subsequently, any time I powerup the board again the wandering starts again.

I haven't tried DC coupling the scope because the voltage is so high (32VDC), but I will. I also haven't looked at the input voltage when the output starts wandering; will do that too and get back to you.

Ok, here is input (yellow) and output (green) of bad regulator:

To contrast, here is input/output of a good regulator:

Here is the output of the bad regulator, DC coupled:

Finally, here is the output/input of a bad regulator with all the tantalum caps removed (from the input and output):

Hi Justin, 

Thanks for providing the additional scope-shots. I understand some of the TPS7A33 on the boards show slightly higher long term drift than the others. By looking at the DC-coupled scop-shots, the Vpp is around 150mV which is drifting from 33.6V to 33.75V and percentage-wise, it's about 0.4%. 

Have you tried to do the ABA swap to make sure the drift is not caused by accessory circuits like caps and resistors? As you mention, you have not seen problems when setting it to -18V, is it possible that once the output voltage is approaching to -32V, some of the characteristics of the components on your boards have also been changed, which could potentially affect the ratio of the divider and have a larger leakage? Have you ever tried to isolate those other factors? 

I am trying to understand the power rail needed by the op-pams that are powered by the TPS7A33. This LDO's accuracy specification is +/-2.5%, even with the 0.4% long term drift, I believe it's still within the accuracy range. Is that right? What's the tolerance on the rail and what's the issue have you observed that is a direct result of this 0.4% long term drift? By answering this question, I am trying to see if I can help to propose some solutions. Unlike the precision voltage reference, a linear regulator could see a long term drift and this is not something we have in the specification. 

Regards, 
Jason Song

Here is the -32V regulator measured over 3.3 minutes, DC-coupled:

Here is the same measurement on the -18V regulator (also a TPS7A33):

And the same measurement on the +32V regulator on the board (a TPS7A4701, the companion part):

I want to note that the -32V regulator actually looks like the 2nd and 3rd plots when I first power it up, but over time it slowly degrades to the first plot.

I have to admit you're right: the drift range (271 mV in the first plot) is less than the accuracy spec of 2.5% (0.8V for -32V). But I thought that spec meant the output voltage could be a steady -32.8V instead of -32V, not that it would be drifting randomly between the two values. I have a hard time believing that the noise in the first plot is below the spec'd 16 μVRMS (10 Hz to 100 kHz), but I have not measured it.

If you're saying this is expected behavior, I guess I have no way to refute that. But the drift seen here is much worse than the input voltage, meaning putting this regulator inline is actually worse than not having it at all. If this really is expected behavior I may just replace the regulator with a wire.

I have not tried isolating the regulator from its load. Unfortunately I didn't put a component in series with the output so it will be a little hard to set that up but I will try it.

Hi Justin, 

Have you compared if the difference of the layout on the NR pin and FB pin connections among the -32V, the -18V, and +32V? Those two pins are critical to the accuracy and noise level of the output. Do the -32V and -18V  outputs have exact same layout with same passive components? 

Regards, 
Jason Song

Yes, the layout is pretty much the same for all the regulators. The resistors that set the output voltage are obviously different. The input/output capacitors are the same for the +32 and -32V regulators, but the -18V regulator uses a 47uF tantalum in parallel with a 10 uF ceramic on both the input and the output. As described in the E2E forum post I linked to before, I have removed the NR/SS capacitor from the -32V regulator, but that had no effect on this issue.

If you would like to take a look, I would be happy to share my design files with you if you could provide me with your email address. We already have an NDA set up with TI which I think would apply.

It just occurred to me that it can't be the layout because if it was, a new regulator would have the problem immediately on power-up. But when I replace the regulator, the output looks fine for a few hours and then slowly degrades. On subsequent powerups, the problem is there immediately. This has to be damage occurring to the regulator (specifically to the reference circuit, I think).

Any ideas here? I'm running out of time and options.

Hi Justin, 

Hi Justin, 

I sent you a "connect" request, if you accept the invites, we could exchange email or setup direct phone call. 

Regards, 
Jason Song

Sorry for my late response; I got pulled onto another project for a week. I isolated the load but this had no effect. The output looks the same even when there's nothing loading the output.

I changed the resistors on the -32V regulator so it would generate -18V, like my other regulator. Surprisingly, this did not make the drift go away either.

I decreased the input voltage to the regulator slowly from -34V to -20V. Somewhere between -26 and -25.5V, the problem disappears. It's like flipping a switch, the output just immediately becomes stable.

I tried increasing the input voltage to my other, working, -18V regulator. I took it all the way up to -30V, but did not see the drift start. However I know this effect doesn't happen immediately on a new/working regulator, so I may just need to wait.

Does this provide any clues as to what could be going on?

Sorry, I missed your last message about the connect request. I will accept it and we can move the conversation to email/phone.

Hi Justin, 

I have emailed you, and let's plan to have a WebEx meeting to discuss the problem. 

Regards, 
Jason