Timer1 during sleep high consumption

[gabirelms]

Hi all, I'm working in a simple idea: any time I press a button, MCU has to wake up and write current timer in the memory.
Device will be in sleep mode 99% of the time, maybe only press the button 4 or 5 times in a day.

I'm using PIC16LF1459 with a button cell, and my problem is power consumption during sleep.

I don't need real time, just a counter, when I start up the device it starts counting from zero, I don't need an accurate counter.

I checked sleep mode, power consumption is only 0.1uA or less, and it can interrupt when I press the button.

The problem is to calculate time during sleep. I'm using timer1 during sleep with an external 32khz crystal.

But the device wake-up on an overflow and this consumes a lot of power, I checked power during sleep in this case, one second 0.1uA, one second 700uA, changing continuously. Is this normal ?

Please I need help with these questions:

1. Can I use timer1 during sleep, without using interrupt on overflow ? I only want my device to wake up when I press the button.

2. If device must wake-up on an overflow to keep timer1 counting, what power consumption is good ? I think 700uA is very high.

3. I'm thinking about another option: Not to use sleep mode. Then I will try to disable all peripherals, and leave timer1 running. Can I use LFINTOSC in this case ? Or is it better to use LP mode with external crystal ?What power consumption can I expect in this case ?

Data sheet, page 1: Timer1 Oscillator: 600 nA @ 32 kHz, typical. Is it possible to disable everything, and leave only Timer1 running with less than 1uA ?. This is much better than sleep mode for me.

Any help would be appreciated

Thanks

[PCM programmer]

[gabirelms]

Ok, I will post my code on monday (not at home now)
I'm reading power consumption using a multimeter, it is hard to read because it changes. But before I enable timer1 power consumption was always 0.1uA.
Is it normal that power changes due to timer1 interrupt ? What should I expect ?
I will be very happy if I can get a continuous 5uA during sleep

[Ttelmah]

Even 0.1uA is rather high....

Critical thing with 'sleep' is to make _sure_ you set everything to reduce power. Things that are often missed are that (for example), a pin left floating as an 'input', will draw more power, than the same pin set as an output!. A floating pin, if it floats into the 'transition region' (where the internal gate is switching levels), results in a spike of power every time it switches. The output FET set to drive, provided there is no load, draws almost nothing.
However you do say 'or less', so things probably are OK.

The 700uA though is silly. It suggests something is happening like a pin being driven in your 'I'm awake' code. Key is that your sequence should be to just handle incrementing the timer overflow register, if this is the cause of waking, and then go straight back to sleep. It possibly sounds as if the code is going straight into the 'do something' part of your code, assuming the button was the reason for waking, and only after a lot of time, 'realising' this wasn't the case. If handled correctly, the extra power will be under a couple of uA, for just a handful of uSec every couple of seconds.

You want the overflow, or the timer can't give you a 'time' beyond a couple of seconds, or a 'count' beyond 65535.

Best Wishes

[gabirelms]

Thanks for your reply.

Sometimes I get 0.1uA, and sometimes 0.0uA, my multimeter cannot read less than 0.1uA.

What you said give me a lot of hope, that means power should be 0.1uA during 2 seconds, and 2uA during 1u second.

I'm very new to microchip, and I'm not sure of my results.
During wake-up, if PIC16 HFINTOSC run at 4Mhz, that means power should be only 100uA ? Datasheet page 1: Operating Current: 25 uA/MHz @ 1.8V, typical.
I got around 2mA, but Im using 3.6v battery.

By the way, I have 2 load caps with the external crystal, but I see some designs using only crystal, without caps.
How can that be done ? If I use a crystal with only 4pF load cap, can it work without caps ?
For example this one:
www.abracon.com/Resonators/ABS06-107-32.768kHz-T.pdf

Edit: Thoses design are with TI mcu or Atmel mcu, maybe those mcu have built-in load capacitors. Anybody tried with PIC ?

[gabirelms]

Compiler 5.015

[Ttelmah]

First, start to use CCS!....

For example, the timer setup, is one line, using the 'setup_timer_1' command in CCS.

Then 'rethink' slightly.

Basically, get rid of the global interrupt enable. Then work like:

[gabirelms]

Worked! thanks so much.

With SPI flash connected, without timer I got 6.6uA.
Now, with SPI flash, and timer working during sleep, I got 7.9uA.

Do you think it is a good result ?

[Ttelmah]

The flash is obviously the big 'drinker' here. Question perhaps whether there is a lower power version/equivalent, or whether you can fully disable this (FET to turn off it's supply!...)?. However the 1.3uA sounds a much more sensible figure from the PIC.

6.6uA, from something like a 240mAHr coin cell, would potentially give over 4 years operation....

Best Wishes

[gabirelms]

Ttelmah, I really appreciate your great support, thanks again :-)

By the way, do you know anything about this:

[Ttelmah]

There are chips specifically designed to do this. For instance the DS1307 RTC. It has slightly higher than normal pin capacitances on the oscillator pins, and specifies a crystal load capacitance (12.5pF).

Now, you have to think carefully, since the capacitance seen by the crystal, is not the pin capacitance. It is the capacitance _between_ the pins that is seen as the crystal load. With just the pins, and PCB, these two capacitances paralleled together, in series with each other. This is why (for instance), you will typically see something like 30pF capacitors being used on crystals with 20pF specified capacitance. It sees the 30pF, paralleled by the track capacitance (perhaps 5pF), and the pin capacitance (another 4 to 5pF), all paralleled together (40pF), in series with the same capacitances on the other pin, giving 20pF across the crystal.

Most crystals will happily oscillate without the correct capacitance, just be slightly more 'in error', than they are specified to be. With a crystal designed for a low capacitance like 4pF, provided your pad layout is reasonably careful (so only gives perhaps 4 to 5pF of capacitance), you will be very close indeed to the correct value. The unit is designed for exactly this reason.

[gabirelms]

I'm already placing my order in digi-key. Thanks