I kept all of the big currents (charging and SHED LEDs) on the add on PCB. To make the A/D readings as stable as possible. I added a unity gain OP-AMP on all inputs, a 16 sample rolling average, some thought on the ground connections and ground pour on both sides of the add on PCB. But the OP-AMPS have to powered by the development board, so minute current is flowing between the Curiosity HPC board and the daughter board. Also during development the whole system is powered by USB, since the development board has the debugger built into the board (PIC24 16 bit). When installed in the shed, the battery runs everything with a 5VDC LDO replacing VUSB. That was enough of a change in the ground currents to cause the A/D readings to move a little. Fortunately I coded all of the trip point values (Battery over voltage, Solar Input over/under voltage, etc..) as constants derived from a spread sheet. This program takes almost 64K of FLASH because of the embedded fonts for the OLED. So adding table look ups for converting A/D readings to voltage did not add much in comparison.
A major concern now is that the whole setup draws about 75mA when just idling. The OLED only displays when a button is pushed. And one LED is cycling at 50% to indicate that the main loop is running. So I am guessing that the PIC24 implementation of the debugger is the main current draw. I looked at the schematic and there is no clean way to cut the power. Right now I get about 4-6 hours of some charge. The sun is too low and the house next door is 5 feet higher and blocks the morning sun and the remainder of the day the angle is very low. In about 6 weeks, the sun should be high enough to give more charge time and by spring it will be over 10 hours a day and very direct. For now I will just watch the battery voltage to see what the trend is.
No comments:
Post a Comment