search

G6EJD / ESP32-e-Paper-Weather-Display

An ESP32 and 2.9", 4.2" or 7.5" ePaper Display reads Weather Underground data via their API and then displays the weather
Other
959 stars 206 forks source link

[FYI] 5 months on single charge #120

Closed marekorok closed 3 years ago

marekorok commented 3 years ago

Hi, seems no better place so would like to share here my results with the project.

Elements: LolinD32, Waveshare 2.7" 3C (although used as 2C, refreshing was annoying), some random Li-Ion battery 2330mAh

I have added mqtt reporting for battery voltage and percentage to get nice graphs (mqtt -> Home Assistant -> Influxdb -> Grafana), also calibrated battery factors with a multimeter to get more accurate measurements and finally 3D printed case.

Refresh period 20 minutes with 7 hours sleep time. Whole set was working straight 5 months and few days on single battery charge.

image yes, there is a gap due to broken mqtt...

image

Thank you @G6EJD for this great project!

AWSW-de commented 3 years ago

Hi! Interesting to hear some values to that. I’m using a 7.5” ePaper with a Lolin D32 too plus a 5000mAh battery and charged it on 1st of January 2020. I am refreshing the data every 30 minutes between 07:00 and 23:30 o’clock. At 23:30 it takes a longer deep sleep timeout until 07:00 the next day. The battery actually has 61% of charge after 11 and 1/2 months now. I also thought of sending the data via MQTT to a server but this might drain the battery to much I think, but would be a nice try. 😁 Kind regards and from me also thanks again for this nice project.

G6EJD commented 3 years ago

Thank you both for sharing your experiences, your results are very similar to mine, some days I have to check that the battery really is doing well after many months of operation, there’s no doubt that sleeping over night makes a huge difference to battery life. I’ve also switch off all serial printing and gone for an event actuated start up of WiFi to save as much time awake as possible.

Yes colour displays are just too slow. The 7.5” version does well as it’s relatively quick.

Great to see the pictures too.

ElwoodSC commented 3 years ago

How on earth are you able to reach such a long runtime?! :-) I'm failling miserably with my setup, using a Wemos Lolin D32, a Waveshare 4.2 ePaper display and a 3000mAh LiPo battery. The longest time I was able to make was ~3 days with wakeup every 30 Minutes and sleeping from 23:00 until 6:00. Baiscally I use the 4.2 example with my own redentials. Is there anything I'm missing? I would be highly interested in @marekorok 's code with MQTT reporting as I failed to implement this by myself... Sorry for digging up that old issue but this here seems to be my best chance to finally also reach some better battery runtime... :)

G6EJD commented 3 years ago

Suspect a battery that is not actually 3000mAh but more likely 300mAh, it's a typical Chinese trick to add a 0. The 4.2" display is about the best you can get for speed and power efficiency so you should be getting really long battery life. First thing to check is the serial port and see how long it is awake for?

ElwoodSC commented 3 years ago

The batteries in deed charging aroun ~2500 mAh according to my USB power-meter. They might not have 3000mAh but certainly more than just 300. I just connected to my serial monitor and rebooted the D32, that is the result:

_PowerOn : 39001 _Update_Full : 1548001 _PowerOff : 20001 _PowerOff : 1 Entering 118-secs of sleep time Awake for : 3.923-secs Starting deep-sleep period...

Looks reasonable to me.

Edit: I yet have to measure deepsleep power consumption - I know for a Lolin D32 it should by in the uAmps but my runtime suggest differently...

G6EJD commented 3 years ago

Yes all looks correct. I've just used my mAh calculator and with that size battery it shows a de-rated value of 2450mAhr, your getting 2500, so that's OK. With a 30-min update rate it predicts battery life of 261-days assuming 145mA awake current and 0.07uA sleep current, a value to be expected. If I change the sleep current to 44mA to simulate an ESP32 that is not sleeping and I get 2.6-days, so that is the problem. For some reason the code is not entering sleep mode, has the code been changed in anyway?

ElwoodSC commented 3 years ago

Nope, I just flashed the D32 with the 4.2" example code and did only edit the owm-crentials.h with my data.

G6EJD commented 3 years ago

Do you have any means to monitor power drawn via the battery terminal? DVM?

ElwoodSC commented 3 years ago

Unfortunately not, I can just use my multimeter in serial with the battery... this hopefully gives at least an indication whether deepsleep is reached or not... BTW, I just recognized that I compiled with ESP32 1.04 - I'm currently updating and will recompile and flash with the 1.06 environment. But beforehand I will measure power consumtion to have a comparison.

Edit: So I measured with my DVM, unfortunately I have to use the 20A terminal, otherwise the ESP does not start. I measure between 145mA on reboot, going down to around 55mA when Wifi is powered of and then going down to 0,04 uA when deepsleep is reached. This looks totally correct for me - so the runtime should be so much better... I have to confess: I hat some code modifications (MQTT for exampe) running before I flashed your original code this morning. I will observe how the unmodified code is running now, maybe it was just my unability to code... :-)

ElwoodSC commented 3 years ago

Power consumption & battery life is much better now - unfortunately now the ESP32 dies around every 1,5 days. When I reboot the ESP32 it'll run for another 1,5 days until again it sleeps forever. But at least the battery does not drain as fast anymore. Still, it is plain the code from the 4.2 example with no additions at all... Ah, and it is still compiled with ESP32 1.04, I will try to compile with 1.06 next.

G6EJD commented 3 years ago

You should if you can monitor the battery voltage with a DVM to confirm it is not battery related. It is possible you have a faulty ESP32 or e-paper module as with the 4.2” variant you should be getting battery life of much greater than 1-year. I have 3 x 4.2” Displays running and many ESP32 devices in sleep mode and they have never failed. One thing you should check is your WiFi service as this may be creating the problem.

dreamy1 commented 1 year ago

https://github.com/G6EJD/ESP32-e-Paper-Weather-Display/issues/120#issuecomment-728185542 ...I’m using a 7.5” ePaper with a Lolin D32 too...

Just a question @AWSW-de : what kind of driver board for the 7.5" e-paper do you use?

Thanks in advance, Stefan

G6EJD commented 1 year ago

The one provided by Waveshare. Some early variants caused the display to fail, so they added a buffer on later variants.

AWSW-de commented 1 year ago

#120 (comment) ...I’m using a 7.5” ePaper with a Lolin D32 too...

Just a question @AWSW-de : what kind of driver board for the 7.5" e-paper do you use?

Thanks in advance, Stefan

Hi, I am using these newer ones for a while now too: https://www.amazon.de/HAT-Resolution-Electronic-Controller-Communicating/dp/B075R4QY3L/ I am running 3 of these now with the included cable connected to the ESP32 and with the blue board connected to the eInk… I think I charge them once per year with 5000mAh Lipo batteries in one device close to the WiFi router and the 2 others run on 10000mAh Lipos more far away from the router…

I also have one of the predecessor 7.5” eInk with (I think 640x320 pixels) and this one outperforms the 3 other displays with a 5000mAh with 3-4 months more run time…

HTH!

dreamy1 commented 1 year ago

Thanks @G6EJD for your feedback!

And thanks @AWSW-de for your detailed help too!

I have checked the schematic for the waveshare hat and some other driver boards...unfortunately the waveshare hat seems to include an additional (unneccesary) LDO with ~90µA extra current and a level shifter with 3.5...10µA extra current, both are implemented to get a 5V compatibilty. So there will be ~100µA continously wasted in a 3.3V_only environment...not much, but this current is 24/7 and this will take ~900mAh alone within 365 days.

The LDO on the Lolin D32 itself draws only ~40µA current and would be sufficient.

I will have a closer look and maybe i will design a small pcb to get this solved and have the advantage to get most out of the batteries :-)

Thanks again for your help an for your great work @G6EJD! Stefan

dreamy1 commented 1 year ago

Addendum: the board DESPI-C02 seems to be the best solution so far, but it is very difficult to get here in germany:

https://de.aliexpress.com/item/1005003600416933.html

G6EJD commented 1 year ago

I think that quiescent current levels are not the driver of battery life, say the sleep current with display was 150uA and a 2500mAh battery that’s a battery duration of 1.97years, so it would be better to concentrate on optimising the wake-time as that’s the dominant energy consumer.

dreamy1 commented 1 year ago

Sure, you are right - the wake-up-time is the most significant part. I will do some tests when the hardware is here, maybe there is some kind of optimization with varying the transmit level of the ESP32 and using an external antenna (with "free" sight to the router, the internal antenna onto the pcb is not ideal when using the Lolin D32 behind the e-paper because of shielding).

My goal is to get the best out of the hardware with minimizing standby current as a start point.

I will also implement a hardware switch to manually activate wifi and start a webserver to place new firmware OTA. This would be a great goodie to load new firmware into the station...