[time-nuts] Ultra low power RTC

[Attila Kinali]

On Tue, 6 Mar 2018 22:59:34 +0000
Mark Sims <holrum at hotmail.com> wrote:

> Sparkfun is selling an interesting RTC clock chip board. 
> It draws 22 nA.  It has a rather novel clock generator...
> a tuning fork crystal disciplines an RC oscillator every few minutes.
>  They claim 3 minutes per year drift.

*Sigh* There are a couple of things wrong with that description.

1) The part is from Microcrystal, so you can believe the specs
they publish with high confidence. This also means this is probably
the exact same chip you will find in Swiss quartz watches.
(Microcrystal is part of the Swatch group and the main supplier
of 32kHz oscillators and electronics for the Swiss watch industry)

2) The 22nA is the _typical_ consumption in auto-cal mode.
Max (the number you should design with) is 32nA. Keep in
mind this number is for the chip alone, no external connection.
It also includes the implicit condition that all input pins are
at valid voltage levels. If a pull up/down resistor is too weak
(because you tried to safe a few more nA) it will result in
pins being driven by leakage currents, possibly reaching invalid
voltage levels, which in turn will cause shot-through currents
through the input stages, increasing the power consumption 10 fold
at least, 1000 fold easily.

3) The current ratings are at 25°C. Going higher means also
an increase in power consumption. How much, is not specified.

4) The 3minutes per year number comes from the +/-3ppm first year
aging. This is, as usual, at 25°C and is on top of the typ. +/2ppm
time accuracy. So, the real accuracy is more like 5min after the first
year... when running from Xtal all the time!
Which has a typ. 60nA/max 80nA current spec!
Taking temperature into consideration, assuming something in the order
of 10 minutes per year is probably more realistic... if you stay
close to 25°C. The quadratic nature of the temperature dependence
for tuning fork X-cut crystals causes high deviations pretty quickly
(In one of the devices I designed, the 10ppm spec of the crystal
suddenly became 300ppm when taking the whole temperature range at
which it had to operate into consideration)

5) The precision of the auto-cal mode is not specified. It can be
litterally anything. Especially considering that the datasheet
talks about using it "several hours at Backup Supply Voltage," 
i.e. as an emergency measure when the normal power supply is lost.

			Attila Kinali

-- 
<JaberWorky>	The bad part of Zurich is where the degenerates
                throw DARK chocolate at you.