Possible TPS65217C/Beaglebone Black Issue

New to BBB but experienced with embedded systems.

I’m working on a project using the BBB. Supplying +5V (up to 3A) directly to the pins on P9 from a quality bench supply. I’ve found that briefly switching the +5V supply OFF and then back on can pretty reliably leave the BBB in an odd state characterized by…a) No LEDs on
b) Very little current drawn from supply (10mA or less)
c) +5 present on P9.5 and P9.6
d) 0.687V on P9.7 and P9.8 ( should be SYS_5V ).
e) P9.9 = 3.57V
f) P9.10 = 0V

I’ve found that once the system is in this mode no amount of pressing/holding the momentary BBB pushbuttons will get the system working again. Removing input power, waiting 10 sec or so, then restoring power will get things working again.

Has anyone else seen this? It seems sort of like an issue with the TPS65217C chip but I’ve not found any reported errata on that part.

Thanks
Jim

Obviously the PMIC is not liking what you are doing. The 3A current is not a factor in this scenario.

Check the timing diagrams in the datasheet and make sure you are not violating the power up ramp requirements No lights means there is no power to the rest of the board and the PMIC has basically shutdown either due to a lockup of the sate machine or an over current condition The only way to reset the state machine, is to pull power and start the process over.

It could also be seeing excess current if you have anything else plugged into the I/O pins. If you are careful you could try grounding the reset pin on the PMIC and see if it recovers when in this state. I don’t recall how many states are affected by the reset line.

No, I have never seen or heard of this before.

Gerald

Gerald

Watching the 5V supply on a scope I see that there is a sort of 5V “brownout” during the time that the power is switched off but it does not go down much below 2.5V. This appears to leave the PMIC with most of the outputs latched OFF. I will add some external stuff to switch the 5V all the way off if the BBB hangs.

Jim

Is there anything else plugged in when the power is switched off?

Gerald

When I first observed the behavior I was testing a custom cape (sort of an LCD4 clone w/ some mechanical changes). To eliminate the cape as a factor I did the same test without the cape and powering the BBB directly from the header pins. The results were the same either way.

Regards,
Jim

Then you should be aware of brownout. If the voltage goes below operating level you must
recover by making sure that the voltage goes below a threshold which is application dependant.
It could be below 100 mV. Only very advanced circuits will recover from brownout in a desirable way.

Hey, did you try another power supply? Probably the PS has problem supplying the current and drops voltage? Or there is a current limitation.

I tried to reproduce by only connecting P9.5, P9.6 and P9.1 (GND) , but could not.

As I said in the original post, the bench supply is capable of more than 3A…far more than the BBB takes. I’ve duplicated the problem on 2 difference bench supplies and with multiple BBBs (ie the problem is not specific to one particular BBB board).

J-

Just to confirm this.

From the TPS datasheet:

OFF In OFF mode the PMIC is completely shut down with the exception of a few circuits to monitor the AC, USB, and push-button input. All power rails are turned off and the registers are reset to their default values. The I2C communication interface is turned off. This is the lowest-power mode of operation. To exit OFF mode one of the following wake-up events has to occur: • The push button input is pulled low. • The USB supply is connected (positive edge). • The AC adapter is connected (positive edge). To enter OFF state, set the OFF bit in the STATUS register to ‘1’ and then pull the PWR_EN pin low. Please note that in normal operation OFF state can only be entered from ACTIVE state. Whenever a fault occurs during operation such as thermal shutdown, power-good fail, under voltage lockout, or PWR_EN pin timeout, all power rails are shut-down and the device goes to OFF state. The device will remain in OFF state until the fault has been removed and a new power-up event has occurred.

When the brownout occurs, the unit goes in the “off” state and happily stays there. Apperently, the voltages subject to brownout recovering does not define as a “positive edge”.

If anyone has a pretty solution for this, I’d be interested. Obviously, I can implement some kind of a watchdog that cuts the 5V altogether at some point but is there a SW or other ‘easy’ solution?

Just to confirm this.

From the TPS datasheet:

OFF In OFF mode the PMIC is completely shut down with the exception of a few circuits to monitor the AC, USB, and push-button input. All power rails are turned off and the registers are reset to their default values. The I2C communication interface is turned off. This is the lowest-power mode of operation. To exit OFF mode one of the following wake-up events has to occur: • The push button input is pulled low. • The USB supply is connected (positive edge). • The AC adapter is connected (positive edge). To enter OFF state, set the OFF bit in the STATUS register to ‘1’ and then pull the PWR_EN pin low. Please note that in normal operation OFF state can only be entered from ACTIVE state. Whenever a fault occurs during operation such as thermal shutdown, power-good fail, under voltage lockout, or PWR_EN pin timeout, all power rails are shut-down and the device goes to OFF state. The device will remain in OFF state until the fault has been removed and a new power-up event has occurred.

When the brownout occurs, the unit goes in the “off” state and happily stays there. Apperently, the voltages subject to brownout recovering does not define as a “positive edge”.

If anyone has a pretty solution for this, I’d be interested. Obviously, I can implement some kind of a watchdog that cuts the 5V altogether at some point but is there a SW or other ‘easy’ solution?

Unfortunately no, there is no software solution since the processor has no power. You have to use a power supply monitor/controller with a state machine to deal with this issue. This type of circuit is normally included a small energy reserve (battery or supercaps) so that the OS has time to close open files and prevent file system corruption during power fail issues. Normally, any power supply interruptions initiates an orderly shutdown of the OS. When the processor finally halts, power is removed from the PMIC. When power is available, power is applied to the PMIC and everything powers up normally. There are several corner cases that must be considered, such as power interruption during power up phase or power available during power down phase. A simple state machine takes care of these corner cases. Overall, the circuitry includes several voltage regulators (input buck convertor, output boost convertor), energy balance (supercaps), battery charger (battery), and a state machine (8 bit micro or my preference - GreenPAK).

Regards,
John

Thanks for confirming my original observation! I think John raises some important points regarding power handling…some of which go well beyond the context of my original post. As he suggests a software solution is not an option …unless you have some other platform to run the software on! For the application about which I originally posted, we are less concerned with power fail warnings, writing system state to nv memory, etc so the only real issue for us is that when power is available we want the Bone to run! Unfortunately, if there is a power disturbance that is enough to trigger the TPS65217C into shutdown you can have perfectly good power for a month and the system will never come back on. I worked around this by doing something like the following…

  1. Find a watchdog chip that will run off the raw 5V supply available and connect the watchdog ICs input/tickler to some digital signal which tends to be available an an early stage in the boot process. The horizontal/vertical sync lines for an external LCD are available on the P8/P9 headers and usually start toggling pretty soon. You also need to make sure that the watchdog IC can be configured for a fairly long timeout period and that its reset output is also longish (seconds).
  2. Wire the RESET output of the watchdog IC to a power switch so that if the watchdog times out it will cut power to the BBB. There are some fairly inexpensive USB host power switch ICs that can handle 2A so these are sometimes appropriate. Another option, if you are stepping down a raw input voltage to 5V for the bone and the dc/dc converter IC has a shutdown input, you can use the watchdog IC to turn off the dc/dc converter. In this case you have to generate an independent 5V supply just for the watchdog IC …but a simple linear regulator can do that for low cost given the small current required.

With this configuration if your TPS65217 gets stuck in the shutdown state the watchdog will eventually time out, cut 5V power, and (hopefully) kick the TPS back into normal operation. One thing to watch for however, is that with the TPS in shutdown mode, there is very little current drawn from the 5V supply so it will take some time for the 5V supply to decay to 0V even with the power source removed. This is why you need a watchdog IC with a long reset output pulse width.

Maybe the TPS65217D will have some additional internal logic which makes all this unnecessary!

Jim-

I just foolishly posted pretty much the same question…

I’m seeing very similar behavior, I also noticed that it can still be powered from the USB when in this state but not VDD_5V.

Can you please elaborate on the brownouts that you’re seeing. You said they’re occurring when “the power is switched off”, is this in software, i.e. “shutdown -h now”? I haven’t been able to see this on my scope but I also can’t reliably recreate the situation in which it occurs.

I know the PMICs are the same on RevA and RevB, but I believe I’ve only seen this on RevB boards.

Thanks!
jay

Hi Jay,

I think this was covered in my original post…

“I’m working on a project using the BBB. Supplying +5V (up to 3A) directly to the pins on P9 from a quality bench supply. I’ve found that briefly switching the +5V supply OFF and then back on can pretty reliably leave the BBB in an odd state characterized by…
a) No LEDs on
b) Very little current drawn from supply (10mA or less)
c) +5 present on P9.5 and P9.6
d) 0.687V on P9.7 and P9.8 ( should be SYS_5V ).
e) P9.9 = 3.57V
f) P9.10 = 0V”

The command line is not involved.

I was using a lab supply and just switching it off for about 500ms then back on. I have also been able to cause the problem using an adjustable output supply by lowering the input voltage to around 2V then going back up to 5V.

Regards,
Jim

So you generate a Brown-Out condition, which means that you operate the part outside the spec.
Dont expect to get any electronic to work after you enter Brown-Out…
Once Brown-Out occurs, you typically have to recover by totally removing power
until you are below certain thresholds which are chip dependent.
This can take a number of seconds.

To protect agaist Brown-Out you need a backup battery (or a SuperCap) and electronics
which shuts down the Beaglebone gracefully and keps the Beaglebone off power until
nomal power is restored.

Best Regards
Ulf Samuelsson

28 apr 2014 kl. 16:39 skrev James Littlefield <jms.littlefield@gmail.com>:

Brown-out conditions are handled quite robustly by many devices including many micro-controllers and at least some of the PMICs that I’ve used in other systems… but apparently not the one used on the BBB.

The particular behavior about which I posted is not related to “graceful shutdown” but rather to coming back up when power returns. The TPS65217 part will apparently never re-enable its outputs after a brownout …no matter how long the valid input voltage persists.

There are a number of mitigation approaches which all require additional external circuitry but don’t need additional backup power sources.

Cheers,
Jim

Could this problem be a result of leaving the Reset pin (44) of U2 (TPS65217C) unconnected in the BBB design?
Would resetting U2 cause a recovery?

Well, as I understand it, resetting the PMIC resets the PMIC. That would in some cases cause the destruction of the processor. But, feel free to give it a shot.

Gerald

I just posted a similar (same?) problem. Funky voltage measurements on TPs similar or the same as yours. Just tried the system again this AM after reporting problem to rma just to discover that BBB booted up perfectly and is now operational again.

I saw that.

Gerald