TPS65218D0: Circuit design not starting at power up

Hi Ruben,

Could you tell me the exact measurement of the peak voltage you are seeing here on the 3V3 rail? 

It looks like it is exceeding 3.3 V/3.4 V and may be triggering an overvoltage event that may be causing the PMIC to shutdown.

Ruben Sarmento said:

A - SYS5V B - 3V3

nWAKEUP also appears to be using this 3V3 as the pullup source which may explain its behavior.
Are you able to use another rail as the pullup source for nWAKEUP and observe if it still has the same behavior?

Best Regards, 
Sarah

Hello Sarah,

I did measure about 3.44 V before I implemented your suggested change to the nWAKEUP.

Initially I tried pulling up to DCDC6 (VRTC) but the line was toggling as well, so I pulled it up to SYS5V. The 3V3 output didn't go above 3.36 V after that change but the circuit still doesn't start up...

B - 3V3 after nWAKEUP pull up to 5V

B - nWAKEUP after pull up to 5V

Hi Ruben, 

Are you able to increase the capacitance value at DCDC4/3V3A rail, (The maximum recommended value is 100 uF)

And can you measure the other outputs to make sure they are not triggering an under/overvoltage event as well?

Thanks, 

Sarah

Hello Sarah,

Something odd has happened...
I left the board powered for some time (about 30 min to 1h) and when I noticed the 3V3 was available and the circuit had started up!

I'm still trying to understand what happened as I made no changes to the circuit and was just measuring some outputs before leaving the board on without interaction for that amount of time.

I checked today after keeping it off during the night and it still starts up, but the DCDC6 has a 4.71 V output and I don't understand why. I figure I should also use the DCDC5 or maybe ground it but I don't know if that is a problem or even related...

Here you have the measurement of the outputs I got:

Hi Ruben,

That is certainly odd indeed! Is the circuit consistently powering on now?

Regarding the DCDC6 output, I do think grounding/using the DCDC5 pins would help to stabilize this output. 
It is unadvised to leave both L5 and FB5 floating, although I did not comment since it didnt seem to be causing an issue before.

Let me know if the DCDC6 changes after changing the DCDC5 configuration.

Best Regards,
Sarah

Hello Sarah,

Yes, it powers up consistently every time now.

I grounded the DCDC5 pins and now the DCDC6 is at 0V... They don't seem to be connected in the datasheet, but all the data mentions them together, so they may be...
I'll try to make a full circuit with DCDC5, similar to DCDC6 and see how it works. I'll update once I have results.

Best regards.

Little update, as promised:

Both DCDC5 and DCDC6 outputs are now at 4.4V...

DCDC5 has the 10uH inductor and two parallel 10uF capacitors in the same configuration as detailed in the datasheet.

Don't know if the PMIC got damaged for being powered for too long, I don't understand why it isn't working properly... And this behavior is confusing to me...

Another thing to add:

I have another board, completely assembled with the CPU (which is an AM3352) and I tried powering it up for an hour to see if it also started up after a while. It did and I was able to measures some voltages, however it had the added behavior of turning off after 20s.

I checked the datasheet and, according to the state machine, if the PWR_EN is low and the PB is high, after 20s of WAIT_PWR_EN, the PMIC will go to OFF state (Figure 5-35 on the datasheet) - this seems to be what happened, but on section 5.3.1.1 of the datasheet, the text says that " If both, the PB and AC_DET return to logic-high state and the PWR_EN pin has not been asserted within 20 s of entering WAIT_PWR_EN state, the power-down sequence is triggered", which implies that the AC pin should also have been asserted high for the state to count as WAIT_PWR_EN, or am I understanding it wrong? It must also have to do with the presence of the CPU, as it may be pulling the PWR_EN pin low, but I didn't check the state of that pin yet.

Unfortunately I damaged the board while getting the output measurements and can't tell you the state of DCDC6 or LDO1 before repairing it, but I got the following voltage values:

Hi Ruben, 

Thank you for the updates.
Regarding the first board,
it appears perhaps the internal FET may have been damaged since it is affecting both the DCDC5 and DCDC6 outputs. 

Is it possible the system voltage (SYS_5V, IN_BU) may have exceeded 5.5 V and caused this damage?

Best Regards, 
Sarah

Regarding the second board that turns off after 20s, 

I suspect the PWR_EN signal is not being pulled high within these 20s, and thus is triggering the power down after the timeout window. 
I would check the status of this pin to be sure.

AC_DET is ok to be low during the WAIT_PWR_EN state.

Since AC_DET is pulled LOW and PB_IN is pulled HIGH, only the last three bullet points apply to your design for a power-up event:

The device should move from OFF -> WAIT_PWR_EN when IN_BIAS is being supplied, since the AC_DET pin is grounded.
Since IN_BIAS is the signal triggering the power-up event, the PWR_EN may still be LOW at this point. 

If PWR_EN is not driven HIGH within the 20s of being in the WAIT state, then device will go into power-down. 

Ruben Sarmento said:

If both, the PB and AC_DET return to logic-high state

I believe the intended meaning of this statement is that the PB and AC_DET will revert to their original input states from before the power-up event (which has been assumed to be logic-high) but in your case, the original state is LOW, so it will remain LOW. 

This appears to be the path of operation modes that the device is following: (Datasheet Fig. 5-34)

Best Regards,
Sarah

Thank you for the reply Sarah.
The system voltage has not exceeded 5V, I'm using a separate power supply to make sure no other conditions affect the circuit.

However, I checked and there may have been a brief short circuit with SYS_BU and CC, as the capacitors are close to each other and during measurements I may have short circuited them unintentionally.

I have replaced the PMIC and kept all the previous changes to DCDC5. I'm currently waiting for it to power up after the first hour has elapsed.

I still don't understand what may cause this power up delay issue, do you have an idea? It will be difficult to have a product like this if it has no solution...

Regarding the second board:

Thank you, it makes more sense after your explanation. indeed that is the path the state machine is taking! I may need to change the circuit so the CPU doesn't pull the PWR_EN down before the firmware is downloaded to it. I'll try to confirm it and solve the issue.

Just wanted to give a little update regarding the first board, whose PMIC I changed: It never powered up even after I kept it powered for more than 3h.

I also tested the PB assembly with the AC pulled to 5V and it still doesn´t power up. I'm at a loss on what I can do to make it work now...

Hi Ruben, 

For the first board, I definitely agree it should not be taking any more than a few seconds to power up, the startup taking minutes/hours is greatly unexpected. 

Are you able to provide some more information on the behavior of the rails so I can maybe get a better idea of what could be causing the issue here?

• When power is initially applied, do all output rails start up and shut down like the V_3V3A scope capture, or was this the only strange output?
• Is PWR_EN being pulled high (and kept high), like discussed for the second board?
• Are you able to capture the waveform of nINT to see if an interrupt is being triggered?
• This may not be relevant to the issue, but just something I flagged as atypical from TI's typical applications: What is the purpose of the diode connection from V_3V3A to CC input here? 
  

Thanks, 
Sarah 

Thank you for the answer Sarah,

Sure, here are the answers:

When power is initially applied, do all output rails start up and shut down like the V_3V3A scope capture, or was this the only strange output?
All except V_MPU (DCDC2) show some type of similar output with insufficient voltage. V_MPU is always at 0V.

Is PWR_EN being pulled high (and kept high), like discussed for the second board?
Yes, I checked and it is kept high.

Are you able to capture the waveform of nINT to see if an interrupt is being triggered?
Here you go:

I also measured after pulling it to SYS_5V, but the signal is the same.
 

This may not be relevant to the issue, but just something I flagged as atypical from TI's typical applications: What is the purpose of the diode connection from V_3V3A to CC input here? 
The purpose of it is to charge the battery during normal functioning of the circuit (we are using an ML-1220 rechargeable battery). 

Hello Sarah,

Have you gotten any clue from the previous post? If you want I can make more measurements.

Best regards. 

Hi Ruben, 

Thank you for the answers and apologies for the delay. I am still unsure of the root cause here.
Can you clarify a bit more about your setup here, are you testing with the CPU attached?

I noticed that PWR_EN is not an open drain signal, so effectively PWR_EN = SYS_5V. This is not an issue for the enable signal,
but it renders the PMIC_PWR_EN signal useless, if this is intended to be the enable signal from the processor.

Are you able to capture PGOOD with DCDC4 and PGOOD_BU with DCDC6?

Best Regards, 
Sarah

Hello Sarah,

No problem, I understand.
It is quite odd, especially because the TPS6521825 was working fine with the same circuit...

The board I am using for testing at the moment, has the PMIC circuit fully assembled, but nearly no active components assembled, so no CPU, DDR or eMMC and nearly no other components at all.

I understand, indeed all the reference designs I saw don't have this pull up, however I kept it for testing purposes to ensure the bit is on if the PMIC didn't start. I was surprised when I tested with the second board I mentioned, that has the CPU assembled, as the pull up on PWR_EN may be insufficient, since the CPU seems to still pull this pin down at least before it is running FW.

Here you go:

A-DCDC6 B-PGOOD_BU zoom in at startup

A-DCDC6 B-PGOOD_BU zoom in

A-DCDC6 B-PGOOD_BU zoom out 2x PGOOD_BU

A-DCDC6 B-PGOOD_BU zoom out full

A-DCDC4 B-PGOOD zoom in (it is always like this even at start up)

A-DCDC4 B-PGOOD zoom out

Hi Ruben, 

Thank you for the info and photos, I understand.

PGOOD should not be capable of pulling to a negative voltage..
Are you able to double check all the ground connections on your board, including the thermal pad?

Best Regards, 
Sarah

Hello Sarah,

Your point is interesting and may be the reason for the difficulty in starting up. I checked the layout design rules on the datasheet and I'm missing more vias on the EPAD of the PMIC - that can justify the behavior I have observed.

However, we purchased the BOOSTXL development kit and wired it to the design in a way that the PMIC circuit would be replaced by the BOOSTXL kit. We found that, when DCDC2, DCDC3 or DCDC4 are connected to our product board, the BOOSTXL seems to not start up... The PGOOD LED never turns on and the PGOOD_BU LED starts flashing if DCDC2 or DCDC4 is connected, but is on if DCDC3 is connected.

Bellow are some images of the PGOOD (B) and DCDC4 (A) signals while the BBOSTXL output is connected to the product board.

Is it possible to have an evaluation of the circuit and layout design by TI? We are figuring it may be a design failure on our part that we are not figuring out. 

Thank you for the help.

Hi Ruben, 

I'm not immediately sure of what the issue could be with the BOOSTXL setup is, but as a word of caution, this development board is not specifically designed to handle heavy loads. It is intended for programming mostly. 

For schematic and layout review, are you able to send your schematic and layout files? 
You may send more details in a private message or email, if privacy is a concern. 

Best Regards,
Sarah

Hello Sarah,

Thank you for the reply.

We are paying attention to that and have powered the board using a power supply with current display, but the board doesn't seem to be using more than 100 mA all together.

I'll send you the schematics and layout files through private message if you don't mind, it is better for us.

Thank you.