A crude schematic. Most of the components are pre-built modules, readily available.

This is a design for a 12V PWM solar charge controller. It is not the simplest of controllers. This honor could arguably be claimed by Julian Ilett's PWM5 design or by his alternate even smaller design. An open-hardware charge controller does have some advantages.

• Very simple construction: Most of the components are readily-available modules. There is no need for surface mount or high-frequency work. Any hobbyist with even rudimentary electronics and soldering skill can manufacture this project with ease.
• Data-logging capability for battery voltage and current, which can be used to monitor the state of the battery bank and detect imminent failure or tell when replacement is needed. This is a feature found only on very high-end commercial charge controllers. These readings may be output via opto-isolated current loop serial (Electrically identical to MIDI) or an HC-11 radio module.
• Compatible with all battery chemistries. The only restriction is that the voltage during charging may not exceed 26 volts - this does rule out use of 24V panels. Other than this, the controller has access to continuous battery voltage and current readings and controls an entirely software-defined charge regime. This means that with only software modifications the controller can be used with almost any rechargable battery chemistry. Lead-acid, li-ion, NiCd, NiMH, even supercapacitors. You can use 18V panels and batteries, but don't forget to stick a zener on the charging MOSFET gate or you'll go over max Vgs and blow it!
• Ease of modification. The design is fully open, so it's easy to modify. Do you want to use a dump load? Do you want to hook up an impressivly large status display to show your friends? Perhaps you want to switch things off in stages, saving the last dregs of battery for a radio relay even if it means sacrificing lights? Do it. This is your controller, customise it to your needs.
• Effectively unlimited current capacity: Just add thicker wires and more MOSFETs and heatsinks.
• All parameters configurable in software. Charge current limit, saturation and float voltage, over-current cut-out, load disconnect and reconnect voltage.

Note that the shunt does not need to be at all precise. In my prototype I used a shunt of too low a resistance for accurate measurement, and fixed this by simply drilling a hole in it. You can easily calibrate via software - you need only hook up a sizable load in series with an ammeter, note the shunt voltage reading and apply Ohm's law.

The PWM function itsself is almost identical to (and inspired by) Ilett's approach, though different in implementation. Both use an N-channel MOSFET, but we differ in our approach to driving it. Ilett's approach is to switch the panel on the high side, requiring the use of a charge pump to provide the required gate voltage, while I instead used a MOSFET on the low side driven via a PC817 optocoupler (a 300Ω or 330Ω resistor is suitable for current limiting, a 6k as a pull-down). This alternative design has a slightly lower component count. Note that this is a PWM-only design, not MPPT, but I intend to experiment with designing a seperate 'power optimiser' project for MPPT.

Only battery current is monitored - there is no provision to monitor solar panel or output current, as this data is of no importance when monitoring battery charging or condition. If this function is desired it would be trivial to add via either hall-effect current sensor or additional INA219 modules. They are I2C devices, and addressible.

Readings from the prototype when connected to a lead-acid battery bank with appropriate charge programmed. The battery is drained overnight by my garden lighting. The following morning the charge controller maintains (erratic cloud coverage permitting) a voltage of 14.4V on the battery until such time as the battery requires less than one amp to maintain this voltage, then switches to a lower-voltage 13.6V float mode. As well as showing the charge cycle, these readings also measure a voltage drop on the batteries when drawing even three amps that indicates an internal resistance many times greater than specification, excessive recovery times and a capacity many times less than rated. As these batteries had been sitting unmaintained for close to a decade and occasionally abused to power science projects demanding multi-hundred-amp surge currents, such poor condition is unsurprising and the batteries should be recycled.

The above readings came from this prototype. This is a cut-down version without a low-voltage load disconnect.

The current software includes only support for lead-acid batteries. Lead-acid battery charging is managed using a constant-voltage system - current limiting could be done entirely in software, but as the test-setup panel I used is only 100W there was no need for this. The target voltage is determined by a three-state machine:
- The 'init' state is entered when the controller powers up, and simply waits for one minute before transitioning to the 'top-up' state.
- The 'top-up' state maintains a voltage of 14.2V (sun permitting). It transitions to the 'float' when the current required to maintain this voltage falls below one amp, subject to some filtering to prevent unwanted transitions due to clouds blocking sunlight.
- The 'float' state maintains a voltage of 13.9V. It transitions to 'top-up' if the battery voltage drops below 12V, or after one week has passed (An occasional full charge being important to prevent sulfation).
- Both of these voltages are default for flooded lead-acid batteries, and should be lowered when using SLAs.

Other battery chemistries could be used with firmware alterations alone, no hardware modification required. There is no limit on the current the design can handle so long as sufficient MOSFETs are installed to share the load, but the voltage at the battery cannot be allowed to exceed 26V or the INA219 will no longer measure correctly. This is a serious and fundamental limitation, as it means the controller is not suitable for 24V or higher lead-acid systems unless the INA219 is replaced with an alternate form of current sensor.

The low part count is achieved in part by performing noise removal filtering in software - this eliminates the need for hardware low-pass filtering to remove noise from the PWM charging process or from the load devices when measuring battery voltage and current.

There are a few practical things to remember during construction:

• After construction the shunt resistance must be calibrated.
• The arduino light will illuminate when the battery is charging.
• If you inadvertently connect the INA219 sense leads in backwards, you can simply negate the reading in software.
• You'll need to desolder the shunt from the INA219 module in order to use an external one.
• You can also substitute the 3.3V switching regulator module for a simple linear regulator to simplify the design further, but current consumption will increase.
• Many MOSFETS are suitable. I used an IRFB4710. An IRFZ44N would also be ideal, and is easier to obtain.
• A blocking diode is required. In the controller, in the panel, in between, it doesn't matter so long as you have one somewhere. It you put it in the controller though, you can modify the design to use an active bypass MOSFET controlled by the INA219 reading to create an 'ideal diode' for an increase in usable power of as much as 6%.

Source is available.