search

melkati / CO2-Gadget

An advanced CO2 Monitor/Meter firmware for ESP32 with Android and iOS App for real time visualization and charting of air data, data logger, a variety of communication options (BLE, WIFI, MQTT, ESP-Now) and many supported sensors.
https://emariete.com/medidor-co2-gadget/
GNU General Public License v3.0
63 stars 13 forks source link

Strange behavior of battery indicator #101

Closed hapklaar closed 6 months ago

hapklaar commented 8 months ago

I have an 18650 connected to the battery connector of the TTGO T-Display and configured the battery to be full at 4.1 volt and empty at 3.0 volt in the menu on the device.

In other words, the indicator seems to have it's own life, separate from the voltage readout :)

Suggestion:

melkati commented 8 months ago

For a Li Ion battery, full voltage should be 4.2V and empty around 3.5V (3.0V is a dead battery, when at 3.5V the energy remaining in the battery is already very low).

The indicator is slow to respond because the battery measurement is done only every 15 seconds to, in the future, avoid draining the battery with too continuous measurements (but you can change it in variable timeBetweenBatteryRead):

https://github.com/melkati/CO2-Gadget/blob/70aaaa06ea40db3ad947cf2d6dc135d09808b091/CO2_Gadget.ino#L66

Have you adjusted the voltage reference? Maybe it's somewhat off.

If you want the battery indicator to be perfect, you have to spend some time to adjust the three values on the battery configuration menu.

Regarding the suggestions for battery indication:

The battery indicador has six levels (not five):

https://github.com/melkati/CO2-Gadget/blob/70aaaa06ea40db3ad947cf2d6dc135d09808b091/CO2_Gadget_TFT.h#L133-L158

The battery voltage is measured with an external library.

The battery level is not a percentage, it's the result of a Symmetric Sigmoidal Approximation (to approximate a Li Ion discharge curve, as it's not linear, these batteries are most of the time during discharge at about 3.7-3.8V).

https://github.com/rlogiacco/BatterySense/blob/147417342bff5cc91b3571f5dcfef91b06dc24c1/Battery.h#L86-L107

hapklaar commented 8 months ago

The voltage shown is correct, only the indicator is way off and even changes between three states (2 bars, one bar and 0 bars) on the same voltage. Right now it shows 3.9V and a red battery indicator, this doesn't make any sense.

Would adjusting the voltage reference be able to correct this when the voltage shown is already correct?

melkati commented 8 months ago

Before writing my previous answer, I connected my CO2 Gadget to a lab power supply and did some testing. Everything looks fine.

I'll reopen this issue and will review and try to reproduce it again.

melkati commented 8 months ago

Today I spent almost two hours trying to replicate your issue without success.

Maybe it's an issue with your particular hardware?

Please, double or triple check power connections and battery wellness.

I added sending of battery percentage and voltage by MQTT with other measurements (is now in the master branch and will be included on next release). Maybe it helps you logging the data to chase the issue.

I suggest you also spend some time to adjust the three values on the battery configuration menu. It's really important that these values are correct (I have max at 4.2V min at 3.5V and Voltage ref at 930, you must find out the right Voltage ref for your board).

Sincerely, I don't understand what can cause the issue on the firmware side (I don't say it's impossible, just I'm out of ideas). The ADC on the ESP32 is far from perfect and can cause the battery indicator to go up and down by one bar, but I think it can't make it jump all over the place.

As nobody told me until now about this issue, I'll close the issue until there is new information.

hapklaar commented 8 months ago

I'll try and debug some more on my side. Also with lab power supply and different batteries. I also have some more T-Display's to test with. Thanks for trying.

hapklaar commented 8 months ago

Battery indicator still acting strange. Battery sensor went straight from 92 to 14 when voltage was at 3.8v

Plotted behavior in a graph: image

melkati commented 8 months ago

Thanks for helping me debugging the issue.

I reviewed the code again, line by line (including the external library), and can't find where this issue can come from (if the three settings on the battery configuration menu are correct).

melkati commented 7 months ago

@hapklaar

Hi.

I think I fixed the issue with the battery. There were some issues in the code...

Can you compile CO2 Gadget? If so, please test with version in branch fix_101_battery_voltage.

If you can't compile, I hope this fix will be on the next release version in a few days, after we have the time to do some testing.

hapklaar commented 6 months ago

I don't use this on battery anymore as the screen fails to stay on on my T-Display-S3 on battery (as specified in https://github.com/melkati/CO2-Gadget/issues/96#issuecomment-1912695396)

It behaves a little different now, as it flashes on and off for a short while after disconnecting USB. When I raise the voltage ref it stays on longer, but readout is too high. When I lower the voltage ref to have a correct voltage readout, screen immediately turns off when disconnecting USB.

Maybe a vid explains better. You can see it starting to go off and on, this will increase in speed as the battery voltage drops a little and eventually after 5-10 minutes the screen will remain off. The battery will still be above 4 volt then.

https://github.com/melkati/CO2-Gadget/assets/7703320/f2bc0f9b-acde-4c60-8af8-38d24f7cbce9

melkati commented 6 months ago

Thank you very much, @hapklaar, for recording and uploading the video. Now I have a very clear view of what's happening:

The voltage reference is not right, it's showing 5.3V on USB power and 4.4V on battery. It should show max 4.2V on a fully charged battery and 4.8 to 5.2V on USB (5.3 is out of USB specifications).

CO2 Gadget assumes it is running on external power (probably USB) when the measured voltage is higher than the maximum battery voltage by 5% (4.2V + 5% = ~4.4V).

This is, probably, what is happening in the video:

How to solve it?

You must make sure (it's really important) that the battery configuration is correct, you MUST set the maximum battery voltage to 4.2V and minimum battery voltage to 3.5V. You must have a really very good reason to modify these values (probably I should remove them from the user configuration settings).

It's very important to set the right Voltage reference value (as the resistor tolerances are big and the reference voltage varies 1 to 1.2V depending on the chip), you must find out the right Voltage ref for your board). The battery voltage for a fully charged li-po or li-ion battery can never be more than 4.2V.

On my side I already modified the threshold for CO2 Gadget to assume is running on external power from maximum battery voltage + 5% to maximum battery voltage + 9% (~4.58V). It's available at https://github.com/melkati/CO2-Gadget/tree/fix_101_battery_voltage until changes are tested and can merge into the release version.

I hope that the explanation is sufficient to understand what is happening and that the problem should be solved with the indicated solutions (if there is not something else still undetected).

I look forward to hearing from you.

hapklaar commented 6 months ago

@melkati I think your analysis of the cycle with voltage dropping and rising is correct. When I change the voltage reference to 900mv, I have correct voltages, but the display is always off. Your explanation tells me this is by design, the screen should always be off when on battery?

I have some questions which to which the answers still elude me:

  1. Why can't I have the display on when on battery on the T-Display-S3 ?
  2. What exactly do the maximum and minimum voltage settings do? I thought these values only determined the behaviour of the battery meter bars.
melkati commented 6 months ago

@melkati I think your analysis of the cycle with voltage dropping and rising is correct. When I change the voltage reference to 900mv, I have correct voltages, but the display is always off. Your explanation tells me this is by design, the screen should always be off when on battery?

It depends on the Display Config:

image

From the blog:

https://emariete.com/en/co2-meter-gadget/#Display_Config

I have some questions which to which the answers still elude me:

  1. Why can't I have the display on when on battery on the T-Display-S3 ?

The display should remain on when on battery if Time To Off is 0.

That said, you are very right: I connected a battery to my T-Display S3 to understand once and for all what was happening and... the screen turns off when I disconnect the USB or as soon as I exit the menu.

  1. What exactly do the maximum and minimum voltage settings do? I thought these values only determined the behaviour of the battery meter bars.

These values are used for a few things:

Please, let me look at the issue now that I have a battery connected and, I hope tomorrow or the day after, I will come back to you with the fix.

Please don't install the fix_101_battery_voltage branch until I finish with fixing as we are doing testing on that branch and probably it will be in an inconsistent state. Right now, you have the best version installed.

Thank you.

melkati commented 6 months ago

Hi @hapklaar.

I just uploaded a new version (0.10.13) to development branch (not to fix_101_battery_voltage branch as it maybe unstable) with many fixes related to battery and display on Lilygo T-Display ESP32 S3 board.

I hope everything is fixed now.

Please test it thoroughly and if everything is OK, I will include it in the new release as soon as you confirm it.

Thank you very much for your help.

hapklaar commented 6 months ago

Testing it on battery now. Module stays on and battery indicator seems to behave much better.

I'll report back with battery voltage and level graph as soon as the battery is empty, which takes longer than I expected. It's already running since 00:18, now at 3.3v and still going strong.

melkati commented 6 months ago

Thank you @hapklaar

I just uploaded v0.12.000 as general release with many fixes and improvements.

It's mainly a maintenance release as it doesn't include any new big features.

Source code is at master, and it's available for easy installing from web browser here.

I'm closing the issue for now. Please, feel free it the is still any issue or if you want to give some additional feedback.

Thank you very much for your help reporting and testing all things related to battery and display. You helped make OS2 Gadget a better firmware. :-)

hapklaar commented 6 months ago

It stopped working reliably below 2.9v. CO2 values were no longer correct. If you check the voltage graph, you could determine battery can be considered empty a bit later than current 3.5v. It still ran for 5 hours after reaching 3.5v until it was at 3.0v. (which is 27% of total runtime)

I would suggest battery empty should be at 3.0 volt.

image

Other than this, haven't been able to find any more issues. Well done! I'll keep testing, so might find more later ;)

melkati commented 6 months ago

Right now, the battery minimum is at 3200mV. v0.12.000 automatically updated the minimum voltage to 3200mV (only if the minimum voltage was still at the default 3500mV not to change customer settings some users can have). You can still modify this value to suit your needs and the specific battery you are using (different 1860 batteries have different discharge curves, especially when close to 0%).

Anyway, I have on my long to-do list to test new discharge curves...

This is my discharge curve, right now on a T-Display S3 (also different circuit designs have different discharge curve). At the end this must be a one size fits all solution.

image

There is still a bug when changing usb/battery/usb/battery that affects menu and timing. I'm waiting for a board I ordered with a built-in hardware debugger to investigate more as I was unable to find the cause until now.