LunaLight
Abstract
There are 1.6 billion people in the world without reliable access to grid electricity. Many of these people illuminate their homes using kerosene lanterns, which cause respiratory illnesses or burns. About one-third of these people have a cell phone subscription and no simple way to charge their phones. The LunaLight provides solar-rechargeable LED lighting and cell phone charging to these people in a product that is sleek, compact, and bright. The objective of this project is to redesign the PCB by fully integrating the electronics on a single board, improving manufacturability and reducing assembly costs.
Project Description
The first LunaLight prototype had wires leading to the batteries, LEDs, and ON/OFF switch (Figure 1). Measuring, stripping, and soldering these wires took time and skill, which inspired the idea for a fully integrated PCB that incorporates the battery, LEDs, and switch onto a single board. Another goal of the project was to reduce the weight by replacing the NiMH batteries with a Li-ion battery. Lastly, the revised circuit has a slide switch that offers two brightness settings; the “HI” setting is the same brightness in the original LunaLight which provides six hours of light on a full charge, and the new “LO” setting offers a low-power option that lasts about four times longer on a full battery charge.
Fig. 1 The LunaLIght Rev 1.0 charging an Apple iPod Touch
Circuit and Schematic
The circuit board for the LunaLight contains three sub-circuits with the following functions:
1) Safely recharge the Li-ion battery from solar panel input power
2) Drive the LEDs from the battery
3) Boost the battery voltage to charge a cell phone
The ICs selected to perform these functions were the LT3652 for the battery charging and the MC34063 for driving the LEDs and boosting the battery voltage for cell phone charging. The schematic for each sub-circuit was designed from the manufacturer datasheets. A double-pole triple-throw (DP3T) slide switch controls the sense resistor on the LED driver, providing the HI, LO, and OFF brightness settings.
The schematic for the LunaLight Rev 2.1 PCB is shown below (Figure 2).
Fig. 2 The LunaLight Rev 2.1 schematic, created with DipTrace.
Component Selection
Size and manufacturability considerations drove the decision to use surface-mount devices (SMD) for the majority of the board components. Due to the way the board will mount in the LunaLight housing, the larger components had to be pin thru-hole (PTH) devices that could be soldered onto the back of the board. Right angle PCB mount devices were selected for the USB port, barrel jack, and slide switches so that they can be aligned along one edge of the board (Figure 3). These components will then lay flush with the exterior of the housing once the PCB is mounted inside the LunaLight. The orientation of the slide switches was critical to ensure that the “OFF” position was the lowest switch position.
Fig. 3 The right-angle PTH components that must align along one edge of the PCB.
A battery holder was selected that holds a Li-ion 18650 battery and mounts directly to the PCB (Figure 4). The holder snaps into place, and the electrical connections to the PCB are made by soldering the positive and negative PTH leads to the board.
Fig. 4 PCB-mount battery holder for single-cell 18650 Li-ion batteries.
Dimension limitations of the PCB were important because the board had to fit inside the relatively compact LunaLight housing. I measured out the absolute maximum board dimensions to be 2.96”x4.3”, but during the design process I was able to trim the board outline down to 2.75”x3.65”. I created a CAD model of my board plus the large components in SolidWorks to ensure that the system could fit within the LunaLight housing (Figure 5). I inserted screenshots of my DipTrace board design as decals to double-check that my components were placed precisely where I planned them to be.
Fig. 5 SolidWorks CAD model of the LunaLight board with the large components.
The board has four mounting screw holes for 4-40 screws. These screws will secure the PCB to the reflector, and the board will be oriented with the battery on the bottom so that the LunaLight is bottom-heavy instead of top-heavy. The LEDs are placed on the board to align precisely with the hole in the center of the reflector.
Manufacturing
Once the board design was completed in DipTrace, the Gerber files and N/C Drill file were assembled and sent to Imagineering, Inc. (www.pcbnet.com), a company in Chicago that consolidated the student designs and sent the bulk order to a manufacturer in Taiwan. The unpopulated boards were shipped to Cal Poly for the students to assemble.
For the SMD on my board, I decided upon reflow soldering technique using the Heller 1500EXLMS SMT reflow oven in the lab set up with the “Marc_Thesis” reflow profile. To avoid the cost of buying a solder paste stencil, I applied a layer of Kester SnAg3.0Cu0.5 solder paste onto the SMD pads using a toothpick and sent the board through the reflow oven (Figure 6).
Fig. 6 The unpopulated board through the reflow oven to reflow the solder paste.
After reflow of the solder paste, I manually removed solder bridges between pads or excess solder from larger pads. After making these corrections, I applied Kester TSF Tacky Flux onto all the pads and carefully placed the tiny components using tweezers. The tacky flux helped to hold the components in place as I walked over to the reflow oven to insert the populated board (Figure 7).
Fig. 7 Populated PCB being sent into the reflow oven.
With the SMD components soldered into place, I manually soldered the remaining PTH components with a soldering iron and SnPb37 solder. I cleaned off the tacky flux residue with 91% isopropyl rubbing alcohol and an old toothbrush, and I checked all the electrical connections with a continuity tester to ensure that all the intended connections were made and that there were no shorted devices. Content with the finished product, I was ready to test the fully assembled board (Figure 8).
Fig. 8 Front and back of the fully assembled LunaLight Rev 2.1 PCB.
Thermal Modeling
High-brightness LEDs are about 30% efficient, which means that 70% of the power (W) used to drive the LEDs is converted to heat. The LED junction temperature must be kept below 150°C, because high temperatures will kill the LEDs by either mechanical fracture or diffusion within the junction. To combat these issues, designers normally mount high-brightness LEDs on metal-core PCBs (MCPCBs), which are then attached via a thermal path to the outer enclosure where the heat is dissipated by free or forced convection. Because my LEDs are directly mounted onto FR-4, a poor conductor of heat, I had to design additional features that would help to dissipate heat from the LEDs.
I created a simplified model of my design to help visualize where heat will flow (Figure 9). My calculations for thermal resistance assume unidirectional heat flow, and a finite element analysis program such as FloTHERM or COMSOL would better model the thermal behavior of the PCB.
Fig. 9 Simplified diagram showing the thermal design of my PCB.
The LEDs are soldered to the top Cu pattern with SnAg3.0Cu0.5 solder, and underneath each LED are 10 thermal vias (15-mil diameter) filled with solder. The top and bottom Cu planes were made as large as possible to spread the heat over a larger area and increase convection. The plastic reflector will make a secure connection to the board near the LEDs, providing a thermal path to the exterior of the housing. I created a thermal resistance circuit to approximate the major heat flow paths (Figure 10).
Fig. 10 Resistor circuit modeling the heat flow from the LED junction to the ambient air.
I computed all the thermal resistance calculations in MS Excel so that the inputs can be adjusted and the outputs can be returned instantaneously. The user can adjust the LED voltage, current, efficiency, ambient temperature, and many other variables to calculate the thermal resistance and resulting LED junction temperature for each scenario. One sample calculation is shown in Appendix A.
The variables that play important roles are the via fill factor and the thermal conductivity of the reflector. These variables were estimated on a worst-case-scenario basis, and the resulting junction temperature was 111°C, about 74% of the rated maximum. Increase the ambient temperature to 35°C (a hot day in Kenya) and the resulting junction temperature is 81% of the maximum rated value.
Future design iterations should pay attention to the thermal design because overheated LEDs that should last 10 years will only last a few weeks or months. An example of a good thermal design is where the LEDs are mounted on a MCPCB that is securely attached to a metal enclosure. However, this increases the material and assembly cost and may be over-engineered for the application. At the maximum brightness setting of the LunaLight, the heat that must be dissipated is only about 0.84 W.
Design Verification and Testing
Without doing an extensive analysis of the board performance, I performed a few tests to make sure that the board was functioning as planned and see how well the circuits were performing. One test was to test the illuminance levels of the LEDs at the different brightness settings. A test rig with a light meter was set up to record the illuminance (in lux) from 12” away, normal to the board (Figure 11).
Fig. 11 Testing the LED illuminance from 12” away at the LO and HI brightness settings.
The illuminance on the HI brightness setting is 473 lux from 12” away. By comparison, a 40-W incandescent light bulb’s illuminance from 12” away is about 450 lux. The LO brightness setting uses about one-fourth the power of the HI setting, and the resulting illuminance is 94 lux.
The cell phone charging tests were simply verification tests. I plugged in my cell phone using the USB micro adapter and the charging indicator turned on. I also plugged in my iPod Touch to verify that the circuit could charge an Apple product (Figure 12).
Fig. 12 LunaLight Rev 2.1 PCV charging an Apple device.
To test the solar battery charging circuit, I took the PCB and a 2.5-W solar panel outside to measure the charge current into the battery (Figure 13). After orienting the solar panel to face the sun directly, the illuminance on the sunlight was measured at 124.8 klx. The open-circuit voltage of the panel was 10.14 V and the short-circuit current was 299.1 mA. The resulting charge current upon plugging the solar panel into the barrel port of the PCB was 212.3 mA of current into the Li-ion battery. At this rate, the 2200 mAh Li-ion battery would take about 11 hours to charge.
As the temperature of the junction increases, the forward voltage of the LEDs decreases. According to the CREE XP-E datasheet, the temperature coefficient of voltage is -4.0 mV/°C for white LEDs. Because I have two CREE XP-E LEDs wired in series on my PCB, the combined temperature coefficient of voltage is -8.0 mV/°C. I turned the LEDs on and recorded the instantaneous voltage over time until it leveled out. The LED forward voltage over time is shown below (Figure 13).
The starting voltage was approximately 5.95 V and leveled out to 5.75 V after 30 minutes. Assuming that the LED junction temperature reached thermal equilibrium, with the surrounding ambient air, the total change in voltage was -204 mV. The temperature increase is calculated as follows:
Assuming an ambient temperature of 25°C, the LED junction temperature reached 50°C, which is 33% of the maximum rated junction temperature. However, the temperature coefficient of voltage is not necessarily linear, and under more extreme conditions and longer durations it is still possible that the LEDs could fail from overheating.
Conclusions and Discussion
I set out to see if I could improve the LunaLight internal circuitry by fully integrating the electronics onto a single board. This first iteration of the fully integrated board is aimed to determine if this would be a feasible design solution in future revolutions of the LunaLight design.
Accomplishments
I believe the board was successful because it performs the basic functions properly and there were no major errors. The LEDs produce the brightness I expected, and I think the HI and LO settings would be useful to the consumer. The mass of the populated board with battery is 88.5 g, which is 60% of the 147.6 g that the original PCB and equivalent components weighed. The solar battery charging and cell phone charging circuits work according to their design. The thermal model approximates the LED junction temperature to be about 74% of the rated maximum, but the thermal voltage drop test indicates that the actual junction temperature of the LEDs reaches 33% of the maximum temperature.
What I Learned
There were three “mistakes” on the PCB, but luckily I caught them early and they were easily fixable. The first was with the DP3T slide switch where I miscalculated the distance between the parallel rows of leads as 3.5 mm instead of 2.5 mm. This was fixed by pressing down on the switch as I soldered it in place, which will unfortunately lead to some excess stress in those solder joints. Also, the datasheet for the DP3T switch indicated that the common ground pins (2 and 6) were electrically connected, when in fact they were not. I caught this mistake when performing the continuity tests and was able to solder a jumper wire to permanently connect pins 2 and 6 of the slide switch. Lastly, the LunaLight logo was supposed to show up on the board but instead showed up as a white rectangle. Apparently, the background of the image must be transparent in order not to show up in the silkscreen layer. DipTrace only accepts BMP and JPG files, which is a problem since neither of these file types seems to support transparency. Another PCB design software such as Eagle might have better support for adding images.
On the electrical end, I believe there is much room for improvement in the component selection and circuit design. My limited electrical background has only allowed me to make a board that performs the basic functions. Optimizing the circuit design and improving the reliability would require help from someone experienced in electrical engineering. For example, when I first turn on the LunaLight I can hear a high-pitched sound of a capacitor as the brightness slowly reaches its maximum point. There might be too much of a load on this capacitor because the LEDs should turn on instantly. The LunaLight project would greatly benefit from the help of someone experienced in analog circuit dsign.
Recommendations for Future Students
I would first recommend that all future students use a continuity tester after soldering to check the PCB’s electrical connections prior to providing power to the board. I was able to catch the error with the switch early on because I painstakingly tested the board connections before inserting the battery.
Another recommendation would be to have future students select an enclosure for their PCB before they can start designing in DipTrace. Dealing with actual size limitations is a great challenge, but the result is that you have an enclosure to protect your board. I had to deal with real dimensional limitations for my PCB, and by modeling in SolidWorks I was better able to understand what my finished PCB would look like. I had to carefully analyze the datasheets for all my larger parts to obtain the exact dimensions. Future students–maybe for bonus points–could design and rapid-prototype a small housing, build their own enclosure with available materials, or order one from the extensive selection of off-the-shelf enclosures available online.
The tour of GateWorks was quite informational, and I would recommend continuing this planned tour for future IME 458 students. Learning about PCB assembly in lecture was important, but seeing the actual factory setting gave more insight into the practical considerations that are important when designing for manufacturability. I liked hearing about the quality control methods they used to ensure that each board was manufactured according to the engineered designs.
Downloads
- Presentation (PDF)
- BOM (PDF)
- Schematic (PDF)
- Thermal Model (PDF)