0
\$\begingroup\$

I have a microcontroller in deep sleep and I want to send it a reset pulse every week to do some periodic task.

I have a power budget of about 1uA.

I’m thinking of using a circuit like this but inverted, using a 75nA TLV3691 and with some kind of RC combinaison which gives me about 1 week delay:

enter image description here

I don’t need a precise week, I could be off by a day and it would still be fine.

Do you think the 1 week delay is achievable?

\$\endgroup\$
3
  • 1
    \$\begingroup\$ does your microcontroller have a watchdog timer? \$\endgroup\$ Commented Nov 28, 2022 at 15:23
  • 1
    \$\begingroup\$ does it have to be precisely a week? \$\endgroup\$ Commented Nov 28, 2022 at 15:44
  • 1
    \$\begingroup\$ What microcontroller are you using? \$\endgroup\$ Commented Nov 28, 2022 at 20:27

3 Answers 3

2
\$\begingroup\$

There are dedicated (very niche) circuits to achieve your goal given your power budget: The S-35710 from ABLIC https://www.ablic.com/en/semicon/datasheets/rtc/wakeup-timer/s-35710-i/ is a low-power periodic timer, and seems to fit you needs. Quoted from the datasheet: "Settable on the second time scale from 1 second to 194 days". They also have a version with build-in crystal.

An external low-power RTC with programmable alarm can also work. NXP has the PCF85063A https://www.nxp.com/docs/en/data-sheet/PCF85063A.pdf, with a nominal 220nA consumption, well within your power budget.

In both case, 1 week is not an issue, the accuracy will by set by a quartz oscillator and will be quite accurate.

\$\endgroup\$
1
\$\begingroup\$

It's impractical to get a 1 week delay strictly with analog components.

Use some reasonable frequency and a CMOS digital counter/divider.

You could use a watch crystal oscillator or a MEMS oscillator and a divider rather than an RC oscillator and a divider.

\$\endgroup\$
2
  • \$\begingroup\$ But would I get the 1 week high 10ms low pulse on only a square wave? \$\endgroup\$
    – Francois
    Commented Nov 28, 2022 at 14:29
  • \$\begingroup\$ ? You can generate a pulse at a clock edge by a variety of means. But if your MCU triggers on an edge, that is sufficient. \$\endgroup\$ Commented Nov 28, 2022 at 14:38
0
\$\begingroup\$

The circuit shown has a period of 1 second. To get it longer you have to increase the resistance, capacitance, or both. The resistors are already quite high in value so you don't have much you can do there.

Keeping the resistors the same you'd have to scale the capacitance by the time period. One week is 604800 seconds. Multiply 0.082 uF by that and you're right around 50,000 uF, so you can probably see the problem there. And that's not even taking into account that the pulse would now be 60k times longer than you need, around 1 hour 40 minutes, so you'd have to have a way to edge detect it. I believe the 100k resistor controls the pulse time, so that could be reduced but there's probably going to be a limit to how low you can make it.

Simply not practical. You'll probably need some kind of counter/divider, but when it comes to this sort of thing a week is a long time and 1 uA is a low current.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.