An energy harvester with advanced monitoring capabilities.
This project was created with Autodesk EAGLE
In order to download the source files in a zip archive, click here. Alternatively you may also clone this repository using git bash terminal commands. Substitute "John Doe" by your own username and use your own email address.
$ git config --global user.name "John Doe"
$ gif config --global user.email "john.doe@example.com"Next up, enter the following command to clone the repository into the folder in which you opened the git bash terminal. Choose your location conveniently, for example C:\Users\<username>\Documents\EAGLE\repos might be a good place.
$ git clone https://github.com/tinstructor/EHM.gitFrom here on I'll assume all directies to be relative to whatever directory you chose to clone the project in. Also, I'll be using Windows style backslash notations because I just happened to be writing this on a Windows machine. If that offends you, you should really grow up.
Once you've cloned the repository and thereby obtained the necessary files, a free version of Autodesk EAGLE may be downloaded from here. This is not some trial version that prompts you to spend money you really don't have, instead it is a limited version of the software meant for hobbyists. For our intents and purposes the features offered by the hobbyist version are more than adequate.
Note: if you're a student or an educator you may apply for an educational license here, which grants you access to the full version of the latest EAGLE release, albeit for non-profit use only.
When opening EAGLE for the first time, navigate to the Options > Directories tab in the top-left corner. A pop-up window will appear. It is here that you must now add the location of the project files in order to open them in EAGLE. Go to the Projects field and add the absolute location of the EHM\Project folder, seperating it from the default project location with a semicolon ;.
The project files will now appear in the control panel view. Notice that there are 2 files present, i.e., a schematic and a board file. Double clicking either of these files will cause both of them to open simultaneously in a seperate window. It is imperative that you always have both file windows open at the same time when making changes of any type. Otherwise you may run into horrible inconsistency problems. No worries, if you where to accidently close one of these windows, EAGLE will prompt you with a bright yellow and black banner.
Learning to use EAGLE efficiently may seem difficult in the beginning but it's not actually that bad. In fact Autodesk has some nice tutorials that'll get you going in no time.
In case you've made changes to the board layout, you'll need to generate new Gerber files. These files tell the machines of your PCB manufacturer of choice how to lay out all traces and other stuff exactly how you designed it. In order to do this, open the board layout window and navigate to the File > CAM Processor tab.
A pop-up window now appears. The process of generating gerber files is clearly demonstrated in this video. However, I've spared you some extra steps and created a simple cam job file gerb274x.cam (located in EHM\CAM) for you to load as follows:
After browsing the file explorer and choosing the appropriate CAM file, all you need to do is press the Process Job button in the bottom-right corner of the CAM Processor pop-up window. This will generate a new zip archive called Gerber.zip in the EHM\Project folder. If you didn't make any changes, a handy Gerber.zip archive is already located in the repository folder itself (EHM\Gerber.zip).
Now, when it comes to ordering PCB's there are a stupid amount of manufacturers out there and they're all claiming some perceived advantage over their competitors. Let me tell you, for the most part it's all marketing b*llsh*t. Which manufacturer you choose typically depends on price, manufacturing capabilities and how fast they can deliver. For hobbyists like you and me, the latter is typically of less importance while we tend to focuss more on low pricing.
These days, when I order relatively small dual-layer PCB's for prototyping purposes I'll typically order from JLCPCB. For the price their manufacturing capabilities and quality are outstanding. In fact, you'll have a hard time finding better prices regardless off manufacturing capabilities. I mean, €7.06 (before shipping) for 10 PCB's and a stencil is somewhat unreal.
Note: choosing a lead-free finish will add €5.30 to the total price. However, since most hobbyist solder with leaded solder anyway, you must ask yourself whether shelling out extra money for a small batch of prototypes is worth it.
A great way to save some cost in the long run while making your job easier is to use a stencil printer like the CYBRES SP2421 or the NeoDen PM3040. The purpose of a stencil printer will become clear in the section on Soldering. What you need to know at this stage is that there are two types of (manual) stencil printers: while the SP2421 requires no stencil frame (where the PM3040 does) it does require a set of precision cut holes in your stencil. If you've got the capabilities to cut these holes and do it precisely, it's a great way to save some money on both the printer itself as well as the additional shipping costs associated with framed stencils.
Note: stencil jigs typically accept stencils with a maximum size. Make sure to set the size of the stencil accordingly when ordering! Setting a custom stencil size comes with no additional cost, at least on JLCPCB it doesn't.
First of all, let me state that sourcing components for absolute bottom dollar (or euro in my case) may well be considered a black art. My advise would be to get what you can on LCSC and have them ship your order together with your PCB's from JLCPCB. That way you can avoid additional shipping fees.
Since the component selection on LCSC is somewhat limited you could go to places like Mouser for the miscellaneous stuff like the BQ25570RGRR energy harvesting IC and the MAX9934TAUA+ current sense amplifier.
Anyway, at the time of writing this, I wasn't sure yet as to which type of energy I'd be harvesting (solar, thermal, ...) so I couldn't really give you a Bill Of Materials (BOM). However, since most critical components are marked in the schematic with their full partname I'm sure you'll have no trouble obtaining the correct components if you really wanted to. My advice would be to always make sure you get the part with the correct footprint.
Note: make sure to order y14870rxxx00b9r type current resistors with a CSM2512 footprint as illustrated in
EHM\Datasheets\y14870rxxx00b9r.pdf.
Oh yeah and before I forget, sites like Banggood are amazing for getting pre-assembled cables like this one and this one. What's more, Banggood typically charges completely nothing for shipping. The only downside is that stuff will take some weeks to arrive at your doorstep. That's what you get for being so cheap.
While we're on the subject of Banggood, they have a pretty awesome selection of soldering supplies. I myself am a big fan of offerings from Mechanic and of their XG-Z40 solder paste syringe in particular. If you have no morality issues with leaded solder and you want to spend less money and time on soldering, XG-Z40 is the way to go.
Note: when buying solder paste in a syringe (like the XG-Z40) you'll also need some blunt tip needles. Make sure to choose tips with a sufficiently large diameter. Otherwise you'll have a hard time pushing out solder paste. My advice is to go with 0.9mm diameter (typically pink) ones. You'll also need to plan a trip to your local pharmacy to get some 10ml syringes because you'll need the rubber plunger thingy in order to push the solder out of the syringe.
While XG-Z40 is fine for small rework purposes (as shown in this video), when using a stencil you'd really need some solder paste in a jar. Numerous tutorials are available on how to start using stencils for solder paste dispension. For example this video from Sparkfun is a great place to start when you don't have the funds for an expensive stencil printer. If you do however (have such funds), this video shows how its done.
After putting solder paste on your board and carefully sticking the component to it, you have several different options for heating the solder paste so that it may eventually solidify. For example, you could hit it with a hot air station. Hot air stations are relatively cheap compared to other solutions and besides, they are an irreplaceable shop tool. Another cheap-ish option is a hot plate. The problem with it is that it really only works well for one side of your PCB, which would require you to manually solder components to the pads on the opposite side of the PCB with a hot air gun and/or a conventional soldering iron (with some special tips). Nonetheless this may prove useful if you've placed the most difficult to solder components on the same side. A much more powerful and vastly more expensive solution is something like a vapour phase reflow oven.
Note: some people even use conventional electrically heated skillets or toaster ovens. While these methods are inexpensive, they tend to heat your PCB unevenly, potentially causing spotburns and/or tombstoning. The same can be said about the IR reflow ovens of yesteryear.
According to the design principles and other useful information laid out in EHM\Datasheets\bq25570.pdf and EHM\Literature\sluuaa7a.pdf, the BQ25570 IC allows for a number of configurations to be made pertaining to its operation. But before we can start you might be wondering, what exactly is the function of this chip? The following quote should start to make things somewhat clear:
"The bq25570 device is a highly integrated energy harvesting Nano-Power management solution that ... is specifically designed to efficiently acquire and manage the microwatts (µW) to milliwatts (mW) of power generated from a variety of DC sources like photovoltaic (solar) or thermal electric generators. [It is] targeted toward products and systems, such as wireless sensor networks (WSN) which have stringent power and operational demands."
Texas Instruments
In order to do this, the capabilities of the BQ25570 are twofold. First of all, it contains a boost charger whose intended purpose is to boost the low voltage output of harvesters such as small solar cells or a Thermo-Electric Generator (TEG) (connected to J2) to a constant voltage level that is sufficiently high to charge a storage element connected to the battery terminal J4. It does this as efficiently as possible by means of a mechanism called Maximum Powerpoint Tracking (MPPT). The purpose of MPPT is to match the input impedance of the boost charger to the output impedance of the energy harvester which in turn ensures a maximum transfer of power.
Note: if you want to know more about how MPPT works I suggest you have a look at
EHM\Literature\erbay2014.pdfandEHM\Literature\kong2012.pdf.
More specifically, in this case, the integrated MPPT block continuously measures a certain user configurable percentage (i.e., the MPPT percentage) of the energy harvester's open circuit voltage and uses this information to control the switching frequency of the boost converter to match its input impedance to be equal to the harvester's output impedance, thus assuring maximum power transfer. Choosing the MPPT percentage depends heavily on the type of energy harvesting device. If its output impedance is purely resistive, setting the MPPT percentage to 50% (VOC_SAMP = GND) ensures maximum power transfer. However, the capacitive nature of a solar cell's output impedance requires an MPPT percentage closer to 80% (VOC_SAMP = VBAT). For your convenience I've also added the possibility to set a custom MPPT percentage by means of a resistor divider consisting of (R20 + R21) and R22.
Switching the MPPT percentage between 50% (VOC_SAMP = GND), 80% (VOC_SAMP = VBAT) or a custom value can be done by setting a jumper (see picture below) to the corresponding position on JP1. The same connector (JP1) also provides the ability to either enable or disable the entire chip by tying the NEN input of the IC to GND or VBAT respectively.
That being said, certain requirements need to be met before the main boost charger (with the aforementioned impedance matching capabilities) is operational. More specifically, when VSTOR < VSTOR_CHGEN (= 1.8V) the IC operates in cold-start mode. The cold-start circuit is essentially an unregulated, hysteretic boost converter with lower efficiency compared to the main boost charger. None of the other features function during cold start operation.
Note:
EHM\Literature\kong2012.pdfexplains in great detail why a cold start cirquit might be necessary.
The cold start circuit's goal is to charge the voltage on CSTOR (i.e., VSTOR) higher than VSTOR_CHGEN so that the main boost charger can operate. When the VSTOR voltage reaches VSTOR_CHGEN, the main boost charger starts up. Once the VSTOR pin voltage goes above VBAT_UV (1.95V) plus the VBAT_UV_HYST threshold (15mV), the VSTOR pin and VBAT pins are effectively shorted through an internal PMOS FET. The switch remains closed until the VSTOR pin voltage falls below the VBAT_UV threshold. The VBAT_UV threshold should be considered a fail safe to the system and the system load should be removed or reduced based on the user-configurable VBAT_OK threshold which should be set above the VBAT_UV threshold.
To prevent rechargeable batteries from being exposed to excessive charging voltages and to prevent over charging a capacitive storage element, the over-voltage (VBAT_OV) threshold level must be set using external resistors (see following equation). This is also the voltage value to which the charger will regulate the VSTOR/VBAT pin when the input provides sufficient power. The VBAT_OV threshold when the battery voltage is rising is given by:
The overvoltage threshold when battery voltage is decreasing is given by VBAT_OV minus VBAT_OV_HYST (24mV). Once the voltage at the battery exceeds the VBAT_OV threshold, the boost charger is disabled. The charger will start again once the battery voltage drops VBAT_OV_HYST.
The charger allows the user to set a programmable voltage independent of the overvoltage and undervoltage settings to indicate whether the VSTOR voltage (and therefore the VBAT voltage when the PFET between the two pins is turned on) is at an acceptable level. When the battery voltage is decreasing the threshold is set by:
When the battery voltage is increasing, the threshold is set by:
The logic high level of this signal (i.e., the VBAT_OK output) is equal to the VSTOR voltage and the logic low level is GND. Figure 21 taken from EHM\Datasheets\bq25570.pdf shows the relative position of the various threshold voltages discussed so far.
Anyway, apart from boosting the voltage of an energy harvester, the BQ25570 also contains a buck converter. The buck converter input is internally connected to VSTOR and steps the VSTOR voltage down to a lower regulated voltage at the OUT connector (J1 and J3). It employs pulse frequency modulation (PFM) control to regulate the voltage close to the desired reference voltage. The voltage regulated at the OUT pin is set by a user programmable resistor divider according to the following equation:
Connector JP2 allows you to configure the behaviour of the buck converter output. According to the BQ25570 datasheet (EHM\Datasheets\bq25570.pdf), if you tie VOUT_EN to VSTOR, the buck converter is disabled when the voltage on VSTOR drops below the VBAT_UV condition. The buck converter continues to operate in pass (100% duty cycle) mode, passing the input voltage to the output, as long as VSTOR is greater than VBAT_UV and less than VOUT. However, it fails to make clear the true functionality of the VOUT_EN input. You see, when reading EHM\Literature\sluuaa7a.pdf, it becomes evident that connecting VOUT_EN to VBAT_OK makes it so that the buck converter is turned off when VSTOR < VBAT_OK (when VSTOR is decreasing) or if VSTOR < VBAT_OK_HYST (when VSTOR increases). Connecting VOUT_EN to GND turns off the buck converter entirely.
The EHM\BQ25570_Design_Help_V1_3.xlsx spreadsheet is a handy tool that calculates a set of resistor values that fits best within the operational constraints provided by you, the user. For your convenience I've provided you with some values that reflect a common scenario wherein a single cel Li-Ion battery or a dual cell NiMH battery are used as a storage element in conjunction with, e.g., a small solar cell (with an output voltage ≤ 3.5V) such as this one or this one (when space is limited) or this one when device footprint is no issue.
Note: the MPPT percentage (40%) is just for demonstration purposes in this case. Solar panels work best when the MPPT percentage is set to approximately 80%.
Choosing an appropriately sized solar panel and storage element (most likely a 3.7V battery) is not an easy feat and depends heavily on your use case. Luckily I've provided another handy spreadsheet (EHM\bq25570SolarAppDesignExample_V1p3.xlsx) which accepts a number of parameters and calculates the required size of your solar panel as well as the minimum battery capacity required (given a nominal voltage). I've loaded the spreadsheet with some parameters that reflect the absolute worst case scenario for a sensor node that uses every possible function of the hardware (including chips other than the BQ25570) to repeatedly send a message every 60 seconds.
Note: seriously, even if you tried you wouldn't be able to design a sensor node setup with a worse performance than the one laid out in
EHM\bq25570SolarAppDesignExample_V1p3.xlsx. Except maybe if you used IEEE 802.11 links or something funny like that.
All resistors are E96 series 1% resistors with an 0603 footprint. The following table provides a summary of the proposed configuration and the passive components required:
I strongly recommend you to read EHM\Datasheets\bq25570.pdf for additional design considerations, e.g., regarding the appropriate sizing of several capacitors, etc.
If the purpose is to monitor both input and output power, you'd need to measure both input and output current (as well as voltage). Hence, a shunt resistor was added to both the input and output, each in combination with a MAX9934TAUA Current Sense Amplifier (CSA). This device amplifies the voltage drop across the shunt resistance before it may be measured by an Analog to Digital Converter (ADC). If I hadn’t use a CSA, the voltage difference would either be too small to measure or you'd need to make the shunt resistor very big, leading to unacceptable power losses.
The MAX9934TAUA is an interesting CSA in that it behaves as a current-source with a transconductance . Thus, by pulling its output to GND with a sufficiently large resistor (let's call it
for now) we are able to adjust the voltage gain
according to the following equation:
Let's say you'd want to use an ADC with a resolution of 10 bits. At this resolution there are levels ranging from 0V to the VOUT of the buck converter (2.7V if you used the proposed configuration). The following equation shows how one could calculate the measured voltage given an output value x of the ADC:
In our case, the smallest measurable voltage delta (by the ADC) equals . Now, assume we want to measure the current through the input shunt resistor with a resolution of 200µA, then the smallest possible voltage gain may be obtained from the following equation:
However since the ADC input only ranges to VOUT of the buck converter, there's also a maximum possible voltage gain. Let's say the maximum input current of the boost charger equals 200mA (which is well within the specs stated in EHM\Datasheets\bq25570.pdf), then the following equation puts an upper limit to the voltage gain:
Note: you could easily calculate that in fact, the shunt resistor is not the limiting factor when it comes to resolution. In all cases
must hold true. By assuming the resolution is unknown this comes down to:
A trade-off now presents itself. The larger the shunt resistance, the lower the voltage gain and the less influence small (tolerance) deviations have on the measurement. However, a larger shunt resistor also equates to more power loss. Let's say that a shunt resistor of 50mΩ does the trick. Assuming the measurement resolution equals 200µA, the voltage gain must be somewhere in the range of . The corresponding output termination resistor (
) can then be calculated as follows:
Therefore . I strongly encourage the use of a 10.7kΩ 0.1% 0603 resistor like the one found in
EHM\Datasheets\era-3aeb1072v.pdf. This offering from Panasonic is extremely precise. In that case the voltage gain equates to:
Note: calculating the current is now simply a matter of dividing a measurement obtained from the ADC by the known gain and shunt resistor values (in software).
The output shunt resistor employs the same principles. However, according to EHM\Datasheets\bq25570.pdf, the maximum output current of the buck converter equals 110mA. Therefore, the following equation must hold true:
Again, in all cases must hold true. By assuming the resolution to be an unknown variable x this comes down to:
Let's say a resolution of 110µA is adequate for our purposes and a shunt resistor of 100mΩ does the trick. Then, the voltage gain must be somewhere in the range of . Thus
. I strongly encourage the use of a 9.76kΩ 0.1% 0603 resistor like the one found in
EHM\Datasheets\era-3aeb9761v.pdf. This offering from Panasonic is extremely precise. In that case the voltage gain equates to:
Note: due to footprint requirements, the only possible shunt resistors are specified in
EHM\Datasheet\y14870rxxx00b9r.pdf. For example, in the above case, you'd need to order the Y14870R10000B9R and Y14870R05000B9R variants.
For your convenience I've added a spreadsheet in EHM\CSA_help.xlsx that combines all these calculations. This allows you to properly configure your current sense amplifiers based on the precision of your ADC (expressed as N bits) as well as the transconductance of the CSA and the output voltage (i.e., the positive rail of the ADC) of the bq25570 buck converter output. For example (see image below), with a 12 bit ADC one could increase the current measurement resolution significantly, at the cost of increased power consumption inherent to a more precise ADC that is.
The input resistance and capacitance of an ADC input are typically chosen to accomodate most common scenarios. However, when dealing with particularly small currents and voltages it is common to buffer these inputs for all sorts of reasons. Moreover, it is considered good practice to filter out spurious high frequency compoments at the ADC inputs because otherwise they might be replicated in the frequency spectrum after sampling and distort your measurements (think Shannon-Nyquist theorem). According to the principles laid out in EHM\Literature\tidu390.pdf and the information obtained from EHM\Datasheets\ads7041.pdf, a buffer-filter combination consisting of an EHM\Datasheets\opa314.pdf op-amp and low pass RC filter with and
was chosen.
Note: both
EHM\Literature\tidu390.pdfandEHM\Datasheets\ads7041.pdfstress the importance of using an NOG type ceramic capacitor forover a common XR7 type capacitor since these may yield spurious frequency harmonics due to the fact that their actual capacitance value is typically temperature dependent.
Note: since
, the sampling rate should be at least 1 MSPS in order to comply with the Shannon-Nyquist sampling theorem, and ideally even higher since no analog filter is infinitely steep.
All signals that are to be sampled are multiplexed onto a single line just before the buffer-filter input with a EHM\Datasheets\dg2034e.pdf low rds-on 4 channel MUX. Since the input impedance of the buffer is very high, the additional resistance introduced by the MUX shouldn't result in significant attenuation of the signals. The following thruth table lists which input (denoted with S) is switched to the output according to the logical signal level on the A0 and A1 inputs.
When properly configured (by means of the appropriate solder bridges) S1 is connected to the (divided) input signal of the bq25570 boost charger, i.e., the positive terminal of connector J2. In similar fashion, S2 is connected to the (divided) positive output of the connected storage element
while S3 is connected to the output of the CSA measuring the bq25570 buck converter output current and S4 is connected to the output of the CSA measuring the bq25570 boost charger input current. The following image shows the location of connector J7, which contains the A0 and A1 inputs.
Because the battery and input voltage may be greater than the MUX, buffer and ADC supply voltage, I've added the possibility to divide the corresponding MUX input voltages by means of a resistor divider. It is advisable to go with large resistance values so as to minimize power losses. Something like the 10MΩ 0.5% 0603 resistors from EHM\Datasheets\crcw060310m0dheap.pdf are an exellent choice.
Coming soon
Coming soon
Coming soon
This project and all original material included with it are released under CC BY-SA 3.0
