Custom I.MX28 Power Supply Hanging Problem

[paulhilton]

I have a line powered, inaccessible, I.MX28 board, no batteries, working well, running Linux.

I have discovered a 'hanging state' by interrupting power for short periods.

I have 'PSWITCH' tied to VDDXTAL via a 10k resistor  to turn the thing on when power is applied, which usually works.

I believe that this was suggested as appropriate in documentation.

There is no other external dc load on VDDXTAL, and the 32.768kHz oscillator is disabled.

By interrupting line power for brief periods I can get the power supplies into a stable condition where:

VDD5V = 5.2V

VDD4P2 = 4.7V

Switching supply is stopped

VDDXTAL = 0.97V

PSWITCH = 0.97V

Various 3.3V supplies at 0.94V

By momentarily pulling PSWITCH up with an external 10k resistor to 5.2V the whole thing starts up just fine.

Changing the 10k connection to VDDXTAL to VDDIO33 would seem unlikely to fix this.

Any suggestions for a fix that does not require an external watchdog?

[igorpadykov]

Hi Paul

there is no full good software solution and VDD5V  interruption may

hang board, recommended solution is given in attached schematic as

"RESET CIRCUIT OPTIONS".

Best regards

igor

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

Note: If this post answers your question, please click the Correct Answer button. Thank you!

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

[paulhilton]

Thanks for the reply Igor,

Is the mechanism of the problem understood?

I am not in a situation to add new circuitry, these devices are already in production and installed.

Are there any software solutions that are not "full good" ones?

The suggested FET 5V switching is very close to what my 5V supply is already doing, it senses a brownout and removes the source of power to the 5V for about 150ms, rather than the 200ms of the RT9801A.

The 5V line, of course, doesn't get to 0V, as it wouldn't with those circuits, because when the FET turns off the capacitcances on the various rails support the voltages.

I am skeptical that these circuits would prevent the problem.

Unfortunately I can't lengthen that 150ms to ensure a full power down (which takes about 10 seconds when the only load is the 24MHz oscillator).

I have managed, by inexpert fiddling with power_prep.c , to make the state much harder, but not impossible, to enter.

Now it seems that to get there i need two power drop outs, one to cause a reboot, and one again before the kernel is fully up, most commonly while the kernel is being decompressed.

Among the recommended applications for this part are several autonomous ones which could not possibly tolerate this sort of hangup, as ours can't.

The cost of one of these hangups is about a hundred times more than the cost of the device.

Best Regards,

Paul Hilton

[igorpadykov]

Hi Paul

if custom board power supply senses a brownout and removes the

source of power to the 5V then it is good solution and if there are

no hang-ups observed, then one can leave it as is.

Reason of hang-up is that PMU has analog circuits which can

lock-up when 5V sags.

Best regards

igor