The idea of this project is to try making device, capable of collecting RF-energy, particularly that leaks beyond the microwave oven's protective shielding on front glass.
As result, special PCB was made. When sticked to mW-oven glass from outside, it is capable to collect enough energy to light-up an onboard LED, which can be used to visually detect the leakage of radiation.
The first goal of this project was to build a prototype, to check following assumptions:
1) Some energy gets out of microwave oven in form of radiation (2.40GHz), although there is protective glass. 2) This energy is enough to collect it and power MCU
At this moment I can say, that p.1 is true, and energy is enough to lid a LED!
The other application of this PCB could be measurements of radiation beyond protective shielding of oven - just connect output of PCBA to ADC circuits (preferably buffered with opamp) of MCU, and you can see in real time the transformation of radiation into capacitor's charge.
p.2 - false: Some extra attempts have been taken, to make some energy-harvesting PCBAs , capable of transforming collected energy into something useful, to power MCU. However, at the moment, practice shows, that such circuit efficiency plus tiny quantities of collected energy are not enough for practical implementation.
The last hours ticks, before HackadayPrize last deadline will be met, and here is my video-presentation of the succesful part of project.
I have structured and posted all documentation. You will find in file section not only the final version of this project, but also all my previous attempts to build extra boards, designed for energy harvesting and transforming to power MCU - I believe that they will find their application too.
So, there is everything available to build your own mWessenger! - it's quite simple, - there are only 20 components, so just try it! ;)
(If there will be enough people, who shows the interest in it, I could manufacture some dozens of it).
Before proceeding to final video, let's "arrange all dots above all "i" letters", as we say in Latvia, because any result - whether positive or negative, still is the result.
Week ago I've received all components and PCBs, required to make energy harvester, based on ADP5091 chip.
This time there were no problems with launching all circuits - everything worked with first attempt, even no debugging was required :)
The good news are: as expected, ADP5091 shows much better performance than LTC3108 IC (the last one is described in some previous logs)
However... the bad news are: efficiency of ADP still is not enough for our application =(
when 1000uF capacitor, charged to 1V (equivalent to 30s of operation of mW-oven with glass of water and detector-PCBA) is placed at input of PCBA, measured voltage at output rises significantly (up to 1.5-2 volts), but this is not enough even to start up the BLE module. I even don't speak about it's MCU intialization sequence and sending messages...
So... At the moment we see, that the only successful outcome of this project is more techno-art-like - the detector-PCB with LED, which glows, when some radiation gets behind the oven's shielding.
Oh, I remembered, that I promise to post the schematics, I designed for Voltage switching: when capacitor is charged up to some level, P-Fet, connected to next stage (i.e. MCU, or other capacitor), turns on, thus powering next stage, or multiple such stages (remember idea with many caps and their quantity depending on accumulated energy?).
I didn't use this circuit in last design, because this switching is already provided by power management chip ;)
I've made simulations in LT-Spice, (the best tool ever!!!) - see "File" section of project for LT-Spice files. Looks like idea should work =) and power consumption should be low enough too.
In one design as voltage reference for "turn on voltage" I've used tvs diode in other - LEDs in forward connection (which gives lower voltages than TVS, and lower leakage currents); different color LEDs provides different "turn on voltages" :)
I tried to recreate mW-oven as pulsed current source, so it charges capacitor periodically (when food is in "correct position" in oven)
switching stages are implemented two options: one, connecting diode-resistor circuit directly to N-FET, other - through 74AUP1G79 or similar D-type flip-flop. As seen from simulation, behaviour of circuits is different: in first scenario, second stage charges smoothly, until reaches voltage of first stage. In second scenario, when D-type flip-flop turns on P-FET, all energy "excess" from cap before P-FET transfers to cap after P-FET, until their voltage is equal.
I also implemented one more stage, that simulates LOAD, to see, how everything operates, if it discharges all caps.
I haven't implemented this design in PCBA ; Therefore, at the moment there is only mWessenger_v2.0 PCB (which I reviewed in previous post), ordered, and I'm waiting for it now. Btw., part of components, ordered from Digikey already arrived :)
After some final polishing of schematics and board, they're ready - see the project's File section ;)
Shortly, about new design features and intent:
This PCB is based on two big components: 1) ADP5091/5092 - Ultralow Power Energy Harvester PMUs with MPPT and Charge Management, which I am going to use with previously built detector PCBA. The features of this block:
1.1) input turns on, only when voltage on capacitor reaches 0.7V (set with resistor on MINOP pin). After that, MCU can pull that pin to GND via FET, thus allowing to operate IC, until input capacitor is empty.
1.2) All components (electrolytic capacitors, LDO) are selected, so that their leakage or quiescent current are not higher that few microamps.
1.3) there is possibility to use PCBA with other energy harvesting methods - from solar cells, TEG, etc... - for this purpose I've added 2.54mm headers for most useful pins.
1.4) With resistors I've tried to configure IC, so that it's outputting around 2.25V. There is also regulated output at this IC, which I configured on 2.0V . I'll try to use it by default.
1.5) If idea with using integrated LDO, mentioned in p.1.4 won't work, I've foreseen on PCB extra LDO - (TPS78218DDC with ultra-low quiescent current) - for powering BLE module
2) As you noticed, the second major component is BLE module ( I decided not to make too much iterations in PCB design, therefore I placed module already now - even if mW-oven energy idea won't work, still, I can attach solar panel and LiPo battery to this board, and everything will work for some other nice home project, or... even for measuring mW radiation quantities - because I still can measure voltage using expansion header =) ).
I've searched BLE, considering two major criteria: it has as low as possible operating voltage. And it has ultra-low power modes.
As result, my choice was BGM111, because it can operate starting from 1.85V (!). In sleep mode it consumes just a few microamps. And, it has lots of nice peripherals inside: ADC, Timers, UARTs, etc... some of them I decided to use - see schematics ;)
So, after my vacation, and some urgent issues at work I'm finally back to my project. :)
I was trying to understand, which approach should I use, to find the reason of non-working combination of PCBAs. The best result gave approach "do everything that could help, and see what happens" :D and then structurize results, adding some calculations! . Fortunately, previous experience in electronics and some brainstorming gave me lot of useful thoughts.
I checked datasheets for LTC IC one more time:
Actually, the result should be more obvious for me at selection stage of this IC, if I made more calculations. I was worrying about that difference in minimum Vin Voltage for different ICs - for LTC it was 20mV. For others it was 80mV. I thought, that this makes a huge difference ( bad intuition!!) in energy. However, I was wrong, and I'll show you how should the correct way of thinking should have been looking like:
1) I desolder the LED from first PCBA, so there is totally no useless energy loss. the output of second PCB still was "quiet"
2) Put more capacitance at first PCBA, because voltage multiplier on it is able to provide up to 2.5 Volts in open-circuit mode. As result, I was able to charge 1000uF capacitor to 1.0 Volts in 30 seconds (!); This was proof, that detector-PCBA is able to collect enough energy without losing it
3) Now I could put previously charged 1000uF capacitor to the second PCBA input. At the same time I was measuring output voltage, where I expected to see more than 2 Volts. Result: Voltage raised a bit, and that's all. Input capacitor was empty too.
4) So.... what was the actual efficiency, when both PCBs were connected ? - obviously, because there was not much current from detector-PCBA, input capacitors all the time were pumped out by LTC chip. If we watch curve above - efficiency at Vin below 200mV is only around 20% (For other transformer rations - even lower, down to 5%!). This correlates with practical results.
5) So, the next question now would be - what efficiecny could provide other ICs?
After analysing TI and Analog Devices ICs, I liked BQ25570 and ADP5092 the most. The first one due its integrated buck-converter, which is beautiful solutions, when there is higher Vin/Vout conversion ration. However, assuming, that Vout should be as low as possible (Bluetooth IC operating Voltage + LDO dropout voltage), calculations showed, that LDO will have same efficiency.
One more feature, that I liked in ADP chip - there is already implemented circuit, I discussed previously: input turns on only after input voltage reaches certain level! (actually, I already modeled one interesting low-power circuit, which activates output, when certain voltage is reached - I'll share it later ;) )
The efficiency graph of ADP5902 was much more promising.
Conclusion: the higher initial voltage of capacitor, when voltage conversion starts, the better efficiency of conversion!!
If we assume, that energy harvesting IC switching starts only after capacitor reaches certain level, we see, that we can reach efficiency up to 90% !!! (Of course, average efficiency during cap discharge will be lower, because Voltage will drop down till 80mV)
6) At this point it's good to know, to which Voltage level 1000uF 1V capacitor is able to charge smaller caps (with, let's say, 70% efficiency); results are following:
330uF => 1.46V
100uF => 2.65V
33uF => 4.61V
Conclusion: we should place as small as possible capacitance at voltage harvester IC output!
7) At this point I started thinking from other side of project: if this energy is enough, to power BLE radio? Some Google-ing showed great Application Note from Texas - http://www.ti.com/lit/an/swra347a/swra347a.pdf, where some numbers were provided. From there I was able to extract, that 1Tx event requires about 75uJ of energy
Calculating the energy, collected by cap: E=C*U^2/2; in our case it's 500uJ.
As you've seen in previous logs, I managed to get done 2 things:
1) get RF-detecting PCB (further in this text - PCBA-1) working, so that even LED is flashing, when sticked to mW-oven.
2) get energy-harvesting (further in this text - PCBA-2) IC working, so when there is ultra-low voltage power source, the buffer capacitors are charging.
I thought, obviously, now I just connect these two PCBAs together, and everything works!
Well.... I was wrong! :D - nothing worked as expected! - the capacitors at PCBA-2 are not charging at all.
I've made various assumptions, why this could happen (and what can be done to check if it's true):
1) When PCBA-2 with it's headers is inserted into socket of PCBA-1, the leakage radiation, which powers PCBA-1, also influences power inductor, placed at PCBA-2, or other it's components, therefore it malfunctions. (Then I've tried to connect PCBA-1 to PCBA-2 using two long wires, so I can place PCBA-2 outside mW-oven's window... Nothing changed - same problem....)
2) adding extra wires and/or PCBA-2 to PCBA-1 detunes it. (Then I've tried to connect only wires to PCBA-1- LED is still glowing... then wires have no influence... I've connected PCBA-2, only GND-wire, to check if it's GND plane detunes PCBA-1.... the answer was "no" - LED was still glowing)
3) Considering two previous points, I conclude, that energy-harvesting circuit has such a low efficiency, that tiny amount of energy, detected by PCBA-1, is wasted almost totally.
As result, I have following action-plan for next 2 weeks:
1) It's obvious, that if there is enough voltage and energy, generated by PCBA-1, to light up a LED, then there will be enough voltage and energy to power MCU+BLE.
2) Then, the only challenge is to store ALL possible energy, without losing it, before using it. When stored energy is enough to send message, start using this energy (maybe even using ultra-efficient buck-converter). This could be achieved, by increasing output capacitance at PCBA-1 and adding special circuit, that activates main, more "energy-hugnry" circuits after some threshold. Similar approach is perfectly described here - https://hackaday.io/project/159691-electron-bucket-extreme-power-management-module . mWessenger-s circuit has same energy budget, but much higher power budget, therefore "triggering"-circuit may be done simpler (more leakage allowed). =)
3) The drawback of bigger capacitance add-on would be lower average voltage of system (if it's supercap with tenths of Farads, voltage probably would never reach more than 1V). Anyway, I see, that boost-converter would be still required, after voltage at supercapacitor drops below MCU's minimal VDD. In this case I would redesign PCB, responsible or energy harvesting, by replacing LTC3108 with some more power-efficient IC. Some examples: BQ25570 / BQ25505 / BQ25504 from TI. I'll study later also other companies' products and select the most efficient and feature-rich.
So... plan is ready, - action!.
P.S.> Oh!... one more idea: use one main capacitor for energy storage, and some extra "switchable" capacitors - when the first cap overflows, some logic connects next capacitor, using MOSFET, therefore reducing system's voltage by some amount and allowing system to store more energy....
Last week, I've been searching for the problem in energy harvesting IC. The schematics and layout looked very simple, however, I managed to make mistake here =).
I suspected that issue is in component values, because I tried to use the closest nominal values I had at the moment (i.e. instead of 4.7uF - 10uF, instead of 330pF - 470pF, etc,). But, after I resoldered the most critical components , the situation was the same: the circuit just didn't want to start!!! - Pointing with oscilloscope into different nodes of circuit didn't help to understand "WHY?!".
Then I googled pictures for the evaluation kits and devBoards with this LTC chip, to see - what's difference there (that's the bonus side of designing circuits that already exist! ) . And at some moment, I noticed there, that inductor is placed with the 1-st pad in opposite direction. Later, when I rotated it 180° on all my 3 boards, everything started working like a charm! - I'll fix this in documentation.
Here is the screenshot from oscilloscope, which made me happy :)
Yellow line is voltage on Vout node, red line - voltage on Vstore node - everything just like in datasheet! I used ordinary power supply as power input, where I set 100mV voltage and current limit around 10mA)
The conlcusions are: 1) The reason of mistake was: I didn't pay attention, that coil is asymmetric, and one side have just a few turns, but the other - hundreds!! So, If I want it to increase voltage, then I need place side with few turns to low-voltage source :) 2) It's much simpler to solder 0603 component on 0805 footprint, than vice versa. Therefore if space allows, and I have not only 0603 but also lots of 0805 in my stock, for first prototypes better to use 0805 footprints.
All this week the PCBA with detector circuit was attached to my microwave oven. And guess what... at some moment I noticed, that sometimes LED is emmiting light!!! After some observations I could say, that energy leakage beyond the glass is dependant on food type, quantity, shape and rotating plate angular position. I managed to capture some video with the help of le croissant:
in the dark LED is visible much better:
At this point, IMHO, LED can be desoldered from PCBAs.
Moreover, I managed to assemble the PCBA part with energy harvesting IC, finally!
I made three prototypes with three different turn ratios of transformers: 1:20, 1:50 and 1:100. Unfortunately I forgot to order some of ceramic capacitors, therefore I used slightly different values or combined some other values in series / in parallel.
Don't know, whether due to that deviations from nominal (because there is strict oscillation frequency range for the circuit required), whether due to some other bug, but none of these PCBAs worked as it should :\ - using voltmeter, there is no increase of voltage on output of IC.
also I noticed the bug in silkscreen - GND and VOUT2 text near header is swapped - need to fix this in next version.
So the next step will be debugging - I'll try to change component values to their designed (already found some 0805 capacitor mix, which I'll manage to solder in). If this doesn't help, I believe that power supply and oscilloscope will be enough to detect and solve the problems.
...aaand, after long waiting, two PCBs of detector circuit were assembled. They are identical, except inductor at Pi-network near antenna.
Today I am satisfied with result of work. However, I still have to solder energy harvesting IC part of project - components and PCBs are in the box. and then we'll see, how much energy can we squeeze out of it....
Finally, the PCBs arrived! The breakout feature works perfectly - already tested on some boards. I would say, that breaking happens even too easy - in next version I should increase the distance between holes.
How well designed other part I'll see, when solder the prototype