Based on earlier solar harvesting projects I wanted to design a solar harvesting gadget
This Saturday I have started selling my boards on Tindie. https://www.tindie.com/products/jaspersikken/aemblink/
I have poured the PCBA with this epoxy into a silicone piramid and heart shape. The pouring is a delicate process. You need to work very clean, have everything prepared before actually starting. I have cleaned the silicone mold with water and soap and used a can of compressed air to blow out the water, then I have put the mold in the oven at 100 degrees celsius to dry further. On the table I have put couple sheets of protective baking paper. And put everything on it, the bottles of 2 parts epoxy, the bottle of UV stabilizer, a couple disposable cups, a couple wooden stirring sticks, a roll of kitchen towel, a kitchen weighing scale. And then I started mixing the 2 parts epoxy, 50g of part A, 30 grams of part B, 1 gram of UV stabiliser. To avoid bubbles stir slowly and in 1 direction, after 2 minutes pour the stirred mixture slowly in another cup, and mix again, this makes sure that the two parts are well mixed. Then to reduce amount of bubbles I slowly poured the mold 3 quarters full with epoxy and put the electronics slowly in. Then I filled the mold full. Using the stirring stick I tried to remove the bubbles. The curing is slow, you have about 15 minutes to finish, after a few hours it feels hard and after 24 hours curing is done.
It was quite easy to remove the epoxy from the silicone molds.
Changes in AEMLOVE R9
Below is the schematic design
I had 10 boards assembled at Elecrow
I have created a simple pen shaped PCBA test tool. It is to test the PCBA after assembly and before the solar panel and supercapacitors are soldered. It's just a small perf board with four 0.1" pitch male headers, a 1.5V alkaline battery to simulate a solar panel and a 220uF capacitor to simulate the supercapacitor. When the LEDs on the AEMBLINK board start to blink I know that the AEM10941 and the TPL5110 circuits work correctly. It also has a LED on the 1.8V output from the board. I
The super capacitor is full in 5 minutes in full sun.
Blinking duration is at least 8 hours on a full super capacitor.
I have tested blink duration with with 20, 27, 33 and 43 ohm LED series resistance. All 10 assembled boards blinked for at least 8 hours with a 33R series resistance.
So far what I've learned
I only made a small change because I received a discount code from Eurocircuits and I thought let's try that out. I only moved the TPL5110 circuit to the side to make space for any circuit that an user may want to add. I did not even assemble the PCB.
What I've learned
In the previous versions I didn't like the LED blink interval (5s) and the ON duration (80ms) was fixed. I have read somewhere (sorry for no link) that the human eye/brain perceives a 10ms long pulse as full brightness and that flashes with an interval between 0.5s-5s strongly pulls attention. And so I can reduce average consumption while drawing more attention. I found that the TPL5110 "Ultra Low Power Timer with MOSFET driver" that is also for sale on Adafruit and Sparkfun was a great choice to make a bright and low power flash that draws a lot of attention. This is a timer IC ideal for low power applications. Normally it is used to power gate applications and power it ON at a programmable interval. In shutdown it consumes only 35nA. I can choose the blink interval between 100ms and 7200s by choosing a resistor value. This IC is intended to be connected to a microcontroller that sends a DONE signal to the chip when it is done performing its task, for example sending a sensor value wireless, and then it is completely powered off until the timer expires. In my application I don't have a MCU but a simple LED that is powered on. I wanted to feed back the power signal through a RC low pass filter to the DONE pin
Design changes in revision 7
See the schematic below.
On bottom right you see the TPL5110 circuit. The LED is power gated through a P-channel mosfet. The LED current flows through a 20R resistor and two LEDs in parallel. The power gated signal is fed back through a RC low pass filter to the DONE pin. The DONE pin is considered high when the voltage exceeds 0.7*VDD. With a 1M/22nF the RC time (63%) is about 22ms and close to the desired 10ms. A 6.8kohm resistor sets the interval to ~2seconds. I have added a 100k pull down to the feedback signal to make sure the RC filter capacitor is discharged before it is re-enabled after 2s.
The PCB wihtout any components assembled and a solar panel for reference
I have used the to measure actual LED pulse duration and current from the supercapacitor.
You can see a video of the device actually blinking here. https://twitter.com/jrsikken/status/1137466544555008000
And it shows actual LED was ON for ~14ms. I have experimented a it with the RC values and found shorter pulses were actually perceived as less bright and taking into account tolerances on the capacitor value so I sticked with 14ms.
The average current with the 47 ohms LED series resistor was 6.8uA acording to the Qoitech Otii Arc. I calculated that with a full 1F supercapacitor and 0.5V voltage drop the LED would blink for 20 hours which is far more than the desired 8 hours.
I have changed the series resistor to 20 ohms which gave 1.6mA peak current and 9.6uA average current. Then the LED flashes were nice and bright and the blinking would last 14 hours which is great.
Then I built 4 devices to test actual charge time in full sun and blink duration in the dark.
I found at 800W/m2 it takes only 5 minutes to fully charge the super capacitor and they kept blinking in the dark for ~10 hours! This is less than the expected 14 hours. So I need to check what is going on.
What I've learned and can improve:
The design changes are:
This is the schematic
And the assembled PCB looks like this.
And when it is worn as a necklace is looks like this
I also casted the PCB in two component epoxy resin
And it turned out like this
I made my mother wear it as a ear ring
I also casted it into a silicone mold
And then it blinks like this
What I've learned from Revision 6
In previous AEMLOVE designs I used the STATUS2 output from the AEM10941 to flash a LED which is high for ~80ms every 5 seconds. I think 80ms is actually pretty long and it can be shorter while keeping brightness perception. I learned about 1.5V powered LED flash circuits that have very low average current (~10uA) and wanted to make a similar circuit.
Normally the 1.5V is not enough to light the LED because red led forward voltage is at least 1.7V. So how does it work? First the capacitor is charged to 1.5V through the two resistors. On the left it is 1.5V and on the right 0V. When the mosfet is closed the left side suddenly changes from 1.5V to 0V the right side remains 1.5V lower than than left, and so it is -1.5V Then shortly there is 3V over the LED and it flashes until the capacitor has discharged.
And since the LED can be powered from a very low voltage that is basically doubled it was possible to try out LEDs with higher forward voltage, like a white LED. Since the AEM10941 has two regulated output voltages 1.8V and 1.2V I could design the circuit twice and experiment with it.
I read here that the eye perceives 10ms flashes at full brightness and after that it decays to 50% in 20ms. This means for a 10ms flash you actually perceive 30ms. This could really reduce average current
In revision 5 I made following design changes
Below is the circuit diagram
What I've learned from Revision 5
I was disappointed by the LED brightness and life time. So I made a simulation in LTSpice. Not perfect but it gives me an idea.
And I found that peak current is actually really high and the duration very low, something like 5ms.
That basically explains why it was not perceived as very bright. In addition it was not low power. So I decided this is not the circuit for me.
I skipped revision 3 because I ordered the board and very shortly after that I ordered revision 4.
In revision 4 I made following changes
I selected the KPTL-3216SURCK-01 LED from Kingbright. According to the datasheet it has 550 mcd at 20mA, so more than the Wurth 150060RS75000 which has 250 mcd.
What I've learned from this redesign
In Februari 2019 I designed revision 2.
I changed following specifications
What I've learned from revision 2 PCB
I also wanted to experiment with waterproofing because it must be wearable. The boards were poured over with super clear epoxy resin.
The epoxy was pretty thin. In the beginning it made a nice layer but after 6 hours most of the epoxy dripped off the board. It does not have same thickness everywhere. The epoxy does not fill the holes around the supercapacitor because it is too thin. The epoxy layer on the surface mount solar cells is actually ugly
It's December 2018 when I wanted to design a solar harvesting gadget.
Since I have 2 children and no time I designed a quick and dirty PCB and ordered from Elecrow. Below is revision 1 schematic.
Every 5 seconds the status2 pin goes high when the IC does MPPT evaluation (check solar panel open circuit voltage), I have no idea what applications would need this information but I can use it toe blink a LED with low duty cycle. The LEDs are switched through a DMG1012UW N channel mosfet.
I designed LED on the 1.2V and the 1.8V output of the IC. It was silly to design it on the 1.2V becaise there is no LED that works at such low voltage.
The battery can also be charged from the USB 5V. I used a voltage divider to make a 2.7V voltage for the supercapacitor and a diode to prevent reverse current.
In January 2019 I received the PCB
On the left is the supercpacator A 1F 2,7V through hole supercapacitor from Aliexpress (CXHP2R7105).
On bottom is the solar cell.
On left side is the AEM10941 IC and it's large boost inductor.
The other two large though hole capacitors I did some experimentation with that I won't explain here.
The PCB has 2.0mm thickness. That is not enough for the USB connector so I had to manually add more solder paste.
As you can see she solar cell is surface mount solderable.
What I've learned from revision 1 PCB
PCB revision 2 you will find in the next project log.