Solar panel and supercapacitors

Projects General Guidance
1

I'm doing the first tests for a project to power an ESP12-F with a solar panel and supercapacitors, without batteries. The ESP will be in deep sleep most of the time.

For my first approach I built this, still incomplete but a first proof of concept:

imagen
imagen1920×772 95.5 KB

It works but only when the capacitors are almost full, I'm not sure why.
The idea is to replace the four capacitors by one or more supercapacitors.
The goal at the moment is to power the ESP, it has the blink test loaded so I can see the led blinking.

The idea is that the first boost converter charges the capacitors up to 5V. Then the second converter takes the charge from the capacitors and panel and delivers 3.3V to the ESP. When there is no sun the converter should take the power from the capacitors until some low voltage level.

The solar panel is:

  • 1.5V 0.65W
  • max 400mA

The first Boost-up converter:

  • Input: 0.9V to 5V
  • Output: 5V

The Boost-buck converter:

  • Input: 0.9V to 6V
  • Output: 3.3V

This way I can use more charge from the capacitors, in the range of 5V down to about 1V, or the minimum level to run the ESP once more when it wakes up.

At the begining it didn't work. I measured the volts and current produced by the solar panel and was fine, 1.5V and 300mA. The first boost converter in the input had 1.5V but in the output only 0.9V. And the capacitors were not charging beyond 0.9V. I disconnected the ESP and it was the same.

But if I disconnect also the 2nd converter then it works and the capacitors charge quickly up to 5V.
Then I connect the 2nd converter and the ESP with the capacitors charged and everything works fine, the led starts blinking. If I disconnect the solar panel the ESP works for a few seconds until the capacitors discharge.

I don't know why the 1st boost converter is not working when the second is connected. Maybe it's because they are set in serie and that can't be that simple.
In the specs it says that input power should be higher than the output power, but when the ESP is phisically disconnected it should be fine.

What could be the reason? and what can be done? Maybe separating the converters somehow when the capacitors charge is low?

1 Like
2

Did you actually study how super capacitors are charged? Obviously not. Until they are nearly fully charged, they are a SHORT on the output of your converters. You MUST charge a super capacitor in series with a current limiting resistor!

3

I think I would try the diode after the first buck convertor to see if it is more efficient . You might even be able to not have a diode - is there any reverse current through the buck ?
Or try with just one buck convertor upto 5v and put the caps on the output .
Lots to try …
What wrong with batteries …

4

Therefore it's better to charge the cap with a (5-6volt) solar panel directly (= current source).
And use a shunt regulator across the solar cell (not across the cap), set to 5volt.

If you use a buck/boost converter, then think several hundred Farad (of a good brand).
Idle current of the converter could drain 18mF in minutes.

I had a project like this before, and had to change to a LiFePo4 battery setup.
Which I charge with a 4volt solar panel with LM431 shunt regulator to 3.5volt.
That battery can connect directly to the ESP.
Leo..

5

Solar panels act mostly like current sources, not voltage sources, and do not work well or at all with boost converters.

As mentioned above, charge the cap with a higher voltage panel, as described in this fun solar powered/supercap Arduino tutorial: https://www.gammon.com.au/forum/?id=12821

Be careful not to exceed, or even closely approach the absolute maximum voltage on the supercap.

6

CitaDid you actually study how super capacitors are charged? Obviously not. Until they are nearly fully charged, they are a SHORT on the output of your converters. You MUST charge a super capacitor in series with a current limiting resistor!

I'm just a newbie but I actually checked it. What I understand is that solar panels are like current sources and can deliver current into a short circuit or into a discharged capacitor.

Actually when I test with only the solar paner and the 5V boost converter it works and it charges the capacitors smoothly in a few seconds.

7

Your boost converter is also a current limiting device.

8

Ok, I can try it. The diode is to avoid that the current goes back when there is no sun, but I don't know if the buck could block reverse current.

But then I would loose capacity. When there is no sun with the boost-buck converter accepting input from 5V down to 1V from the capacitors I can extract much more power. If I put the capacitors on the output the voltage range is smaller.

Nothing :slight_smile: It's just a challenge, maybe not easy but supercapacitors have also some advantages.

9

Yes, I plan to try also with a solar panel of 5V. And the idle current can be a problem indeed. My plan is to use supercapacitors of 50F or more, but anyway idle current should be as low as posible.

Yes, I will try also with a small 5V solar panel. I know that it is a bit extreme with so small voltages and current. Anyway the consumption of the ESP will be very low, micro amperes in average, as it will be sleeping most of the time.
I know the tutorial, what I try to do is similar but I try to get more voltage range of the capacitors with the boost converters, from 5V to around 1V. And the operating voltage range of the ESP is narrow: 3.0-3.6V
I don't know if it's posible or worth.

10

You will spend a lot of time/effort/money 'till you come to the conclusion that a LiFePo4 battery is much easier/better/cheaper. LiFePo4 has an almost constant discharge voltage of 3.3volt and a much longer life than a 3.7volt Lipo. And you don't need a power-wasting buck/boost converter.
LiFePo4 is used in 'better' garden lights, and shops that sell them have the batteries.
Leo..

11

It is not possible if your boost converter tries to draw more current than the 1.5V solar panel can produce, which (if you believe the spec) is at very most 650 mA in full sunlight at noon, in July on the Earth's equator.

Less to much, much less, elsewhere and at other times.

12

I've doing some tests and I think that the problem is that I can't connect two converters in series. I have even replaced the solar panel by an stable source of 3.3V and it's the same.
But with only one of the converters in place everything works fine. Both converters work well with the small solar panel.

I have read somewhere that you can't connect 2 switch converters in series except if both have separated GND for input and output. In my case only one have separated GND.
I'm considering to replace the second one by a LDO converter, from 5V to 3V. This would give a marging of 2V of the supercapacitors that could be suficient. If I reduce the clock frequency maybe I could go even to 2.5V.

The good news is that the boost coverter (0.9V-5V to 5V) can charge the capacitors up to 5V even with bad conditions, completely cloudy.
In that conditions with a load of 4Ω the solar panel would producte only about 0.5V and 80mA, but connected to the capacitors it's almost like a short circuit and then it has 1.5V, that is more than the minimum of the boost.

With supercapacitors it will take longer to charge, but will have plenty of time during the day. The goal is to survive the dark hours. I've doing some calculations and with 10F it could last about 20 hours, in practice I supose than less.

I could also monitor the voltage of the caps with the ESP. When it goes below lets say 4V it would enter in a 'low' mode where the sleep time would be longer, 1 hour instead of 15 min, and would never connect the wifi to send the data until the charge goes up again.

I know that with batteries it would be a lot easier, I will do it also with this project or another. But now I'm trying it with the supercaps just for learning. I find it quite interesting.

13

Only if you built your own development board. Development boards have extra components that draw current when the ESP32 is in deep sleep mode.

14

You could also use the ULP for as a deep sleep processor. The ULP can collect hours of data then when its full, wake up the main processor, hand the collected data to the main processor, the main processor do its thing, and then it goes back to sleep till woken again by the ULP, which has gone back to collecting data.

Or the ULP can monitor the charge on the power source and only wake the main processor when the charge is sufficient.

https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/system/ulp_instruction_set.html?highlight=ulp

1 Like
15

It is the ESP-12F (ESP8266). But this ULP coprocessor of the ESP32 looks interesting. I don't know if there is something similar for the ESP8266, maybe external.

The ESP12-F in deep sleep mode should consume less than 40uA, but I have not tested it. What I don't know is the consume of the booster and LDO when idle. And the waste of power of the LDO in heat when working.
Thank you

16

Do you know what the leakage currrent of your supercaps is.
The ones I bought were quite high. In the mA range.
Leo..

17

No, I don't have them yet. But you are right! even if it's lower will be significant.
I have seen in the specs for a similar one that the leakage for 5F is about 40uA. That would double the consume when idle. I will add it to my calculations.

Some people comment that at the begining the leakage is higher but it improves after serveral charge cycles.

Thanks!

18

In my experience leakage only dropped when I exposed the cap to (almost) it's rated voltage for a much longer time. Not after more charge cycles.
Leo..

19

Hi,
I think you had better read up on Dc-DC converters.

If you have say a 1.5V converter to 5V.

If you draw 100mA at 5V that equals 5 x 0.1 = 0.5W output.

If the DC-DC converter is 100% efficient, which they aren't.

The the input power to the DC-DC converter will have to be 0.5W.

The input current at 1.5V will be 0.5 / 1.5 = 0.33 A

The input current in this small example has 3 times the input current required to get 0.1A.

Tom.. :smiley: :+1: :coffee: :australia:

20

Mine too, as in 24 to 72 hours.