I’m looking at solar energy to power IoT device that will be used in a car. I did not found a lot of documentation on what we can do with solar panel and what kind of energy we can expect to get with a such system.
This post will summarize the information I got on solar panels and measures done.
Solar panel technology
I uses IXYS panel for my tests as they are the one I was able to find easily with Internet distributors and that look industrial ; I was looking for small one I could mount easily on a PCB. You can also find larger one from china but in my case they did not really fit my need. By-the way if I get some time I’ll test some.
According to this document, the solar panels are based on three technologies :
- Polycrystralline – they have a spectral range of 500nm to 1100nm (visible to IR) and a performance of 13%. They are mainly use outdoor.
- Monocrystalline – they have a larger spectral sensitivity from 300nm to 1100nm (UV to IR) ; thank to that they have a better efficiency and make sense indoor as outdoor. Performance can reach 22%
- Amorphous – the spectral range is 300nm to 600nm (visible) ; they feet well indoor but the efficiency is about 5%
The test I have made are based on a Monocrystalline cell with a 22% efficiency ; the main pain point I’m trying to evaluate is the impact of being behind a windshield in a car where the sun IR are filtered to avoid temperature growth. Meaning that the available spectral width will be limited.
The solar panels are composed of one or more cells, the surface is generally the same for multiple devices, so depending on the cells organization you can choose between a higher voltage capacity or a larger current capacity.
Basically a cell delivers 500mV and 15mA ; depending on package you can have 3 cells (KXOB Series) or 8 cells ( SLMD Series). Cells can be serial or parallel, so this can be something like 3×0.5V = 1.5V with 15mA or 3x15mA = 45mA with 0.5V capability
Then, you can plug multiple modules in parallel or Series to get what you want. Voltage and Current indicated are the one you can expect in the best conditions of sun exposure. Depend on this the result will decrease (really quicky…)
Select the right chip for my test
My first choice was a SLMD600H10-ND ($7 unit price) as my need is to power a battery for loading it, this module offers 5.01V/22.3mA in the best situation ; 2 or 3 of these module in parallel could be a good source of energy.
Depending on sun exposure the voltage is varying ; also, depending on the load it will vary, based on U=RI formula.
I’ll do some immediate measure by evaluating the current passing over a 10 ohm resistor and do some long term measure by evaluating current passing over a 2×10 ohm divider bridge and measuring voltage out of this divider with an ADC. The results will be collected, aggregated and transmitted over sigfox network.
I made test with this POW11D2P a 5.5V – 80×100 mono-crystalline solar panel also $3.95 unit price.
First set of measures have been made a light cloudy day in May in France @ 3:30pm
Inside Car Outside Car SOUTH 2.5mA 5mA NORTH 1.8mA 3.5mA EAST 3mA WEST 3mA EAST-Horizontal 0.7mA 1.4mA WEST-Horizontal 0.9mA 2.2mA
In these test, SOUTH / NORTH / EAST / WEST are in the sky direction, the windshield giving the angle. E/W-Horizontal means that I was using side windows of the car ; the sensor is in this case not pointing sky.
The car windows is basically cutting the efficiency of the solar cell by a factor 2 in the tested car. To make a comparison, I did a test in a old C15 vehicle not having thermal windshield, the results are fully different :
Inside Car Outside Car WEST 2.12mA 2.20mA
This factor 2 division have been confirmed with the second mono-crystalline panel (POW111D2P)
Inside Car Outside Car South 52mA 104mA North 19mA 42mA
What have to be noticed is that the cell is able to provide 22mA and that day in the best condition, it provided only 6mA. The previous day when I was calibrating the system, I’ve got some sunny period and got a production up to 20mA. The impact of the weather is really important to be taken into account.
Different solar Panel
I have tested different solar panel offering low cost solution like these one :
This panel is basically a 1$ solar panel based on mono-crystalline component. My morning test gave me a 49mA capacity (at 9am) outdoor.
This panel is basically a 1$ solar panel based on a poly-crystalline component. My mornin test gave me a 40mA capacity (at 9am) outdoor.
As my test purpose is to use them inside a car, as was interested in the impact on the technology in this case. As poly-crystalline have a larger use of IR it was supposed to be less efficient behind a wind-shield.
outdoor windshield ratio poly-crystalline 43 18 42% mono-crystalline 39 14 36%
This test finally goes at the opposite of what was expected… due to precision of my test it seems that we do not have a real difference between these technologies in term of ratio.
Impact of Temperature
The temperature have a negative impact on the solar panel efficiency. Behind a windshield the temperature can be really high, up to 65°C (you will notice than if like me you try to use a PLA printed box for your device behind the windshield and get a kind of gloubiboulgabox at end of the day) … So according to some website, the impact is about -0,5% per degrees over 25°C. It means for 65°C, the impact can be about 20%
Quality of 1$ solar panel
The price is one thing but the quality is another. I received different lot of solar panel with different visible quality for China. I can confirm it is better to verify panels are working correctly as on a lot of 55 I have been tested 3 were not working (5%).
During this test, I measured the shortcut current on a constant light for each of the panels. The result is shown in the above graphic showing the number of panel delivering the current slot (23,5 means delivering between 23 & 24mAh) :
The performance variation is up to 12% for a working panel to another. The average is 24,5 in the test conditions.
Mounting solar panel
The first way to power my circuit was to use an existing solution based on a LiPo charger circuit MCP73831T. This circuit is requiring 2,7 to 5.5 volt as input. The best way was to connect two solar panel in parallel to get more energy from them.
To get a higher current, it is possible to mount two (or more) solar panel in parallel; to protect them against current fall-back from circuit when solar panel goes to the dark, a diode is added in series ; to avoid current peak over a limit accepted by the circuit next to the panel, a Zener diode is added like in the above schema.
This solution is not efficient enough, my measure gave me bad results as getting only 8mAh going to the battery by a sunny day at noon being right in front of sun… This decreasing to 5.5mAh once the windshield temperature grown. There are different reason for this :
- MCP73831 is a linear regulator, not efficient to transform the solar energy to refill the battery
- MCP73831 is not designed for solar panel and is not able to adapt its impedance to optimize the solar panel
- When I did the measure I do not exactly know what was the battery state but charge current depends on battery charge level – by the way < 10mA is not a normal charing level for something like 50% charged.
After looking ate the Serial Solution described above, I did a try with a Texas Instrument BQ25504 ultra low power boost for energy harvester application. This chip implements a MPPT system to optimize solar harvesting.
The second solution was so to use a specific solar panel charger like sparkfun sunny buddy based on a LT3652 device. This device requiers move than 5v as a power supply to be efficient. For this reason the solar panel mounting will be serial :
This circuit have 3 diodes, the one on top is a blocking diode to protect energy loss in the panels when in the night. The two other are bypass diode : if one of the panel is in the shadow, the second one will be able to produce something the power level will be lower (only one panel). For a use with LT3652, we do not really need them as with 5V the circuit is not working.
I made a small change : removing the Timer pin capacitor : this is stopping battery charging after about 4 hours. So I replaced C6 by a 0 ohm resistor as indicated in the documentation to stop charging when C/10. This mean, the charging is finished when the overall load is < Isense / 10.
By default Isense = 450mA so C/10 is too high. As my battery is 300mAh, default value is too high and as the maximum current I can get from the solar panel I’m using is 50mAh all of this is really too high. So, I also changed Rsense to have a 150mA Isense.
Rsense = 0.1 / Isense = 0.1 / 0.15 = 0.66 ohm
With this solution, here are the results I got
Cloudy summer day Blue sky summer day peak current produced in the car 20mAh 50mAh
This sounds really nice but in the real world it did not gave me the expected result: the LT3652 have a boost mechanism to extract the maximum power of the solar panel even with a small luminosity as a consequence it consumes battery energy for this. The matter it consumes about 10mA even if the load does not need this. And basically the battery is discharging more than it is charging if the solar exposition is not good. This solution was not the good one for my application with a small solar panel. I recommend to test it if you have large > 12V solar panel with a good and constant exposure.
Long term measures
In this test, the sensor is put in a car and we don’t care about the sensor when driving and parking the car to get something like a real use case, even if not especially representative.
The first thing noticed is that energy is produce (> 1mA) from 8am to 5-7pm not more in a light-cloudy day. This is 9-11 hours of production only in good weather condition to refuel a battery.
On a sunny day where I got about 40mAh per cell cumulated during the day. A dark rainy day has given about 12mAh cumulated the day.
As an example, here is the energy captured during a winter period on a week from 2 solar cells :
We have large variation depending on weather and for sure exposing time.
—- Will be completed soon when data will be gathered —