Power is also fed through D5 and D6 to a 2 part sensing circuit. D6 feeds one half of a LM393 comparator, U3B, to detect when the sunlight drops below a point set by D3, a 5.1V zener. U3B turns on Q2, the positive side of the output. D5 charges C2 and fed into U3A used to detect when the sun is shining, set to approximately 5.7V (D2+D3). This helps to insure U3A (5.7V) is on only after U3B (5.1V) turns off for the day. The capacitor forms a time delay to prevent the output staying on indefinably if the level detectors fail, setting a maximum run time. This also arms the system each morning. U3A turns on the low side driver Q3 through Q4 only when Q2 is turned on (after dark) and helps to provide a high enough voltage to turn Q3 on fully.
U2A is used as an oscillator to send a signal to the moisture probe through Q1. This helps to prevent a capacitive charge from building up on the probe and limits the current draw through the probe due to it’s duty cycle.
U2B is used for turning the system off, either with a level sensor or moisture probe. When activated, the output pulls the input to U3A low, discharging C2 and turning Q4/Q3 off. A level sensor simply pulls the input to ground. The moisture probe set point is set with RV1 and the osculations are smoothed out at the output using C3. The moisture probe may not turn the system off instantly, but since it is active even during the day, it prevents C2 from charging thus preventing the system from ever arming.
The water level sensors are normally open magnetic reed switches with a magnets mounted in floats to activate them. There is one used for the water level to dispense, and one to detect if the barrel is empty.
The moisture probe is two stainless steel wires about 2 inches long and 1/4-3/8 inch apart.
A note on the 9.6v NiMH battery life:
The pump runs for about 5 minutes MAX at about 2 amps. or a total draw of less than 160 mAh (.16 Ah). In reality it takes about 2 min. to move 10 gallons. The pump is rated at 1000 gph with zero head,(36 sec!!) but raising about 6 feet slows it down. The battery I am using is ratted at 1600 mAh so I only draw about 10% of it’s capacity. The solar cell is rated at 1.8 W or about .15 A at 12V.(it’s actual unloaded output is 18v+) Charging the battery at slightly less than the suggested 10% rate,(.16 A) it will take about 10 hours of average sunlight to fully recharge, or an average day. That still leaves enough capacity to run for a week with less than average sunlight, and a few really bright days will make up for it. If the battery should go flat, it comes with a wall charger to freshen it back up if needed. The electronics are basic very low power op-amps and efficient MOSFET drivers drawing very little additional power.

With the water supply solved, I looked at the automation part. Anything I built needed to be solar/battery powered do to the remoteness. The main design goal was to water each night a set amount, reset the next day, and keep itself charged. The first version was built around two items from Harbor Freight. A battery powered back pack sprayer, and a solar trickle charger. The sprayer had a 12v gel cell battery and quite heavy duty pump and was used as a 5 gallon reservoir. The charger was used to charge the battery and to trigger the system each night at dark. The sprayer had a niffy little charging circuit that sensed over charging and shut down on low battery. With a little modifying, and a latching relay, I was able to get it to sense when the sun went down and latch the relay shut turning on the pump. I reversed the pump to suck water from the rain barrel and fill the backpack. The backpack was hung about 5 feet above the garden and 2 feet about the rain barrel. This kept the barrel from siphoning, and provided some pressure to water with. I put a reed switch/magnet in a float in the backpack so when it was full, the pump shut off. The system would not run again until the next day when the solar panel would once again power the circuit. The water would then drain out the backpack to water the garden through a drip feed system. The water ran out o backpack at the same time it was filling so since the backpack held about 5 gallons and about a gallon ran out while it filled, about 6 gallons was supplied each cycle. I could adjust the amount by the position of the float switch. Last thing was another float switch in the bottom of the barrel so if it ran dry, the system would be disabled. Of course, during the day the solar panel charged the battery. Since the pump only ran for about 4 minutes, trickle charge was enough to offset the draw each day. Towards the end of the season there were a lot of overcast days and the charger finally fell behind, so I used and extension cord to use the charger that came with the sprayer to let it charge for a few hours.

The original set up worked well the first year. The tomato plants were over 6 feet tall! Went on vacation with no worries. There were a few dry spells during the summer when the barrel did not get filled, so I ran out the hose and filled it and was good for another week.
There were a few draw backs I wanted to address. Cost, overkill, and reproduce-ability. Cost: I got the sprayer on an open box sale, but it would have normally been quite expensive. Overkill: The pump was way, way bigger than was needed as well as the battery needed to drive it. Reproduce-ability: Modifying the charge circuit was not a good solution, and the normal cost of the sprayer was too high to duplicate. The way I had the whole system installed would not work for most people, the reservoir was hung inside the shed. I started looking for cheep available solutions.
Here are the main components I chose for the second version.