The next images show the structure for the top and bottom of the PCB. All the components and the majority of the traces and copper pours are on the top; the bottom is primarily a ground airplane.
The microcontroller is an EFM8 Sleepy Bee from Silicon Labs, and the (relatively) large connector that you write in the cue section provides a direct connection to the SiLabs USB debug adapter. This connector consumes quite somewhat of PCB real house and makes the overall design look larger than it really is.
These image shows the dimensions of the PCB (in inches). The shorter side to side dimension is my try to estimate how small the board could be if the debug connector were removed (and the other components rearranged).
So my guess is the fact that a two-layer board with all the components on one side could be less than 1. 5 square inches. I’d say that’s pretty good, especially due to the fact we’re talking about a two-layer PCB.
Furthermore, I don’t think that I’m losing any performance by using two layers instead of four, because underneath part is almost a strong ground plane and there is plenty of room on the top for wide power traces and generously sized ground connections (and also because the microcontroller will be working at very low frequency).
Small, Nonetheless it Could Become Smaller
Here are some other ways in which the size of this panel could be reduced:
I selected larger passive components (0805 and 1206) because they are better to assemble. If you plan to have the board put together professionally, you might consider using 0603s or even 0402s (you might be able to find acceptable 2. 2 µF caps in an 0402 package, but you could definitely go with 0402 for the 0. 1 µF capacitor and the resistors).
I chose a larger package deal for the microcontroller as well; it’s a being unfaithful mm × 9 logistik QFP32. The 32-pin no-leads package is significantly smaller (5 mm × 5 mm), and there is also a 24-pin no-leads package that is even smaller (4 mm × 4 mm). It seems in my opinion that most programs built around this power supply wouldn’t need more than the usual handful of I/O buy-ins, and so the 24-pin package is probably the most suitable choice. I actually used a 32-pin device only because this microcontroller doesn’t come in any other leaded (i. at the., non-no-leads) package.
I integrated a high-precision 32. 768 kHz crystal for current clock applications; it’s about the dimensions of an 0805 part. The microcontroller has an internal low-power oscillator with terrible accuracy (±10%), so if you have no need for precision timing you could omit the crystal.
The charge-pump transitioning regulator currently has four 2. 2 µF result capacitors, but only one is necessary.
The LED and its accompanying resistor are for debugging only; they could be omitted in the final design.
An individual might be convinced that you could eliminate all the circuitry associated with the debug power (a switch, an LDO, and two capacitors). I wouldn’t recommend that, because the solar-powered supply is not really a convenient power source for firmware development and testing.
The Two-Sided Option
The final product on the how-to-make-it-smaller listing is to have components on the top and bottom side of the board. ?nternet site was writing this article I started out to wonder if the complete circuit could fit in an area corresponding to the size of the solar cell, such that you could design a board with the particular solar cell on the top and everything else on the bottom. I determined to eliminate some unnecessary components from the schematic and then give this idea a try, and here is what I found (dimensions are in inches):
This is a rough estimation, but nonetheless, you can see that we’re pretty close to the purpose of cramming all the circuitry into the PCB real estate entertained by the solar cell.
To create this element placement I eliminated 3 of the four end result capacitors, the crystal, the LED, and the LED’s resistor. I also turned the microcontroller package to a QFN24. The couch potato components are still 1206 and 0805, but these larger packages compensate for the fact that you would need some means of connecting the microcontroller to the debug adapter. Presently there certainly wouldn’t be much space for routing, but I don’t think that’s a serious impediment if you can use a four-layer board (and there is also plenty of empty space on the top side beneath the solar cell).