Chip hanging in Powersave(?) even with watchdog

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

Thank you for the schematic, O-Family. There are many similarities to mine.

I think that the answer is, as you said, to leave out the BOD. It seems to be causing an unwanted reset on power up in my circuit and I'm not sure that it's worth having. Without it, the power cycling behaviour is fine.

I'll continue to think about this issue and about ways to improve the recovery from Powersave.

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

EDC - the program flow as I see it is this.

Start: Initialise and enable Watchdog
Main loop: Reset WD every loop.
If power goes off, disable WD to save power
When power returns, re-enable WD
Continue resetting WD every loop.
If chip hangs for whatever reason, the WD isr is called and program flow is directed to address 0000
The instructions Enable Wdt and Start Watchdog are executed again before the main program loop.

By resetting the WD many times per second in the main loop only, it ensures that the program is flowing correctly. If the WD reset was in the sectic isr, the program could be stuck in some infinite loop in the main loop but the WD would never overflow to save the day.

MWS · Joined: 22 Aug 2009 Posts: 2262

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

That makes sense. Your suggestion of the Schottky diode is good and I'm adding it to the redesign that I'm finishing now.

I suspect that part of the problem is the choice of voltage sensing point. If, instead, a point upstream of the 3V3 regulator is chosen, suitably divided and clamped to the 3V3 rail, this would provide an earlier signal which would make Powersave happen while the mains supply is still dropping but still holding up the 3V3 rail.

Combined with your delay suggestion I expect a better result with BOD enabled.
[img]

[/img]

O-Family · Joined: 23 May 2010 Posts: 320 Location: Japan

I do not recommend stopping BOD.
Not all products have BOD enabled.
I ship many products with my hobby works.

If you need backup with supercapacitors, BOD consumes a lot of power and needs to be stopped.

If his circuit design is such that supercapacitors are frequently emptied, it is better to use batteries such as lithium ion.
You have the choice of adopting BOD safety measures or ensuring backup time with supercapacitors.
Backing up with supercapacitors is meaningless if the design is such that BOD occurs frequently.

If it is important to protect the EEPROM from corruption due to an AVR runaway near 1.7V, you can either review the backup circuit to enable BOD, or choose to duplicate the EEPROM or protect the data with a checksum.

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

My intention is to have BOD enabled in future. There have been some instances of EEPROM corruption in the past and micro-controller freezing which is a huge hassle as the devices are not readily accessible. There is no way to reset a microcontroller that you cannot access other than by disconnecting power and waiting for the supercapacitor to discharge over many hours. It is dangerous to re-apply power too early, because then the supercapacitor is recharged and you have to restart the waiting period, so you have to err on the long side.

By using the Picopower version, we can disable the BOD during Powersave to reduce current consumption, then re-enable it when power is restored.

The reliability of the product is more important than the duration of the battery back-up, and if necessary, the size of the supercapacitor can easily be increased (currently 0.22F).

Paulvk · Joined: 28 Jul 2006 Posts: 1257 Location: SYDNEY

O-Family · Joined: 23 May 2010 Posts: 320 Location: Japan

If you disable BOD during power save, I think you can't prevent the AVR from running out of control and damaging the EEPROM when the power supply is around 1.7V.
The built-in power-on reset of the AVR works, and the AVR may run unstable.

If reliability is important, it is a good idea to install a low power consumption reset IC externally.
Since the transient phenomenon of power supply rise and fall is also affected by changes due to the environment such as temperature, the problem may reoccur even if it is temporarily escaped by the program Wait.

MWS · Joined: 22 Aug 2009 Posts: 2262

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

Paulvk - Partly through stubbornness, partly because I want to improve devices that have already been made and partly because the board might sit on a shelf for months before being installed.

O-family - I agree that the BOD is necessary for reliability. I originally believed that the BOD only protected spurious EEPROM writes when values were being written at very low voltages, but it seems that faulty operation can happen at other times too.

I would prefer to keep the reset function in the micro partly because the WD isr can save the current date and time which, although it could end up being a few seconds or minutes out, is still better than starting the clock at midight on Jan 1st 1970.

MWS - that looks like a neat solution, but the thresholds that I saw in the datasheet were, Low - 1V (0.3 x 3V3) and High - 2V (0.6 x 3V3).

IC1-VO = 3.54V (measured) when Vcc is 3.3V, giving the Vf of the Schottky as 0.24V

Assuming a Vf of the 1N4148 of 0.7V, then Vpower_on will always be 0.46V below Vcc, so for a low Vpower_on threshold value of 1V, Vcc will have already dropped to 1.46V

In other words, Vcc would drop below the BOD of 1V8 before the Power_on sense line reached 1V.

Two 1N4148 diodes in series might do the job.

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

To me, the great advantage of having the Vpower_on sense line come from a 5:1 voltage divider at the input to the 3.3V regulator is that there is early warning of a power cut, allowing the micro to go into Powersave without draining the supercapacitor of any charge. This is because the input to the voltage regulator is still at 5V when the 1V low threshold is reached, telling the micro to go into Powersave.

At 5V input to the regulator, the micro is still being supplied with 3.3V mains power.

When power returns, the supply to the regulator will have already reached 10V when the signal to come out of Powersave (2V on the Power_on input) is triggered.

So, under all conditions, the system is well behaved, being able to shut down early and start up late.

MWS · Joined: 22 Aug 2009 Posts: 2262

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

So many variants! So many datasheets!
But OK, even with those thresholds, I still say that it performs well.

MWS · Joined: 22 Aug 2009 Posts: 2262

sentinel · Joined: 02 Feb 2009 Posts: 141 Location: Tasmania

My apologies, I might not have been clear that the point in the circuit that connects to the micro-controller Power_on sense pin is the junction of R10 and R14 in my schematic (perhaps confusingly labelled Vraw). It's the junctiom of R2 & R3 in yours.

Your D1 (my D?) Schottky is just there to stop that Vraw voltage from exceeding the maximum permitted port input voltage.

We may have been talking at cross purposes.