Minimize power consumption

[mortenfyhn]

The entire device (measured on battery input at 4.1 V):

• Active: 50-60 mA
• Sleep: 0.5 mA @ 4.1 V = 2.05 mW

The load cell + amplifier (measured on 3.3 V line):

• Active: 10 mA
• Sleep: 0.0 uA

The displays (measured on battery line at 4.1 V):

• Active: 50 mA
• Sleep: 190 uA @ 4.1 V = 0.78 mW

The push button uses a small amount of power when pressed and nothing when not pressed.

So where is the remaining 1.27 mW going?? An 328P in deep sleep shouldn't use nearly that much.

Standby: 0.5 mA gives around 100 days.
Usage: 50 mA gives around 24 hours.
So if I use it, say, twice a day, 5 min brew + 5 min timeout each time, that's 20 minutes usage per day + 1420 minutes standby. That's 16.7 mAh (usage) + 11.8 mAh (standby) = 28.5 mAh -> 42 days, assuming 1200 mAh battery.
Very back of the envelope.

[mortenfyhn]

• Do the displays use more power in standby because they're hooked up to the battery line instead of the 3v3 line? -> They're on the 3v3 now

[mortenfyhn]

The TM1637 driver has a parameter bool on. I wonder if the power consumption is changed by this, when all segments are off:

https://github.com/avishorp/TM1637/blob/82214be06742bf187fab4f29aaabb64c289eaae0/TM1637Display.h#L49-L51

[mortenfyhn]

Maybe switch to an ATtiny1614 eventually

[mortenfyhn]

Some profiling of the active state loop of today's version (235c0ac):

task	description	approx. duration (us)
1	check for button press	8
2	wait for hx711 ready	(irrelevant)
3	read from hx711	800
4	compute grams [1]	80
5	add value to filter	80
6	get filtered value	1000
7	compute hysteresis	60
8	format scale segments	450
9	set scale segments	23 150
10	update timer display	34 250
11	check for steady state	700
12	check activity timeout	8

[1]:

coffee-scales/firmware/src/main.cpp

Line 183 in 235c0ac

const float weight_in_grams = hysteresis_.compute(filter_.getValue());

[mortenfyhn]

However by using

https://github.com/bremme/arduino-tm1637

instead of

https://github.com/avishorp/TM1637

the write to segment duration is much shorter:

(I've also added some exp decay smoothing with factor 0.1)

task	description	approx. duration (us)
1	check for button press	8
2	wait for hx711 ready	-
3	read from hx711	746
4	compute grams	76
5	add value to filter	70
6	get filtered value	963
7	compute hysteresis	53
8	format scale segments	481
9	set scale segments	1015
10	update timer display	1503
11	check for steady state	700
12	check activity timeout	14

Here's the diff: dbe9ff6

[mortenfyhn]

So that means we for each 100 ms iteration (HX711 is configured for 10 Hz output), we work for 5-6 ms, and then do nothing (busy loop) for the remainder.

Note that the above is all with -Os.

Next:

• Actually implement sleeping and interrupt wake for when the HX711 is ready
• Try to squeeze even more speed improvements?

[mortenfyhn]

Anyway come back to this after standalone MCU PCB without arduino board.