For example. My fridge is starting to have some issues. It wants to keep freezing everything in the fridge. It dials the temperature setting all the way up and we don’t know why. I’ve taken the fridge apart as best as I can the last time we defrosted it just to wiggle all the wiring connections.
I wanted to really see what was going on with the fridge. I grabbed a Raspberry Pi Zero W and a couple of 1 wire digital thermometers and put together a Fridge monitoring system in a couple of hours. I can now see what happens with my fridge. It does a defrost sequence and comes out of it cold, freezing up the fridge.
A quick dashboard configuration, and I now have a view of the current temperature and a graph of the temperature history.
Add a couple of more nodes and I now get email notifications when the fridge is too cold. Another node, and the Google Home announces the too-cold temperature.
My fridge now complains that it is cold.
How epic is that? Under $20 worth of parts to give my fridge a voice.
So I decided to put together a monitoring system. It will monitor my sump pump, letting me know when it is getting used hard so I know to pay it more attention. It will monitor my Aquaponics grow bed, telling me when my flood and drain / ebb & flow grow bed isn’t functioning correctly. It will also monitor my fish tank water levels and let me know when the water is getting low, or more importantly when I am filling it and it gets to where it ought to be.
This is put together with Raspberry Pis. An older original B model for the fish tank because I have it and it’s close enough to Ethernet that I can run a wire to it. The Sump Pump is getting a Zero W as it is further away, and I needed to buy something and it was the cheapest option ($10).
I am measuring water height by 2 methods. An ultrasonic distance meter and a differential pressure setup.
I coded up the project a couple of different ways, learning as I went along. I ended up starting with the hardest methods first, and moving towards easier methods. Starting at OS level triggering of shell scripts, moving through python programming, and finally landed on Node Red. I am happy for the path I took as I now have a solid understanding of what a Raspberry Pi can do for me and how to control it at multiple levels. Node Red is how I will be building most of my projects going forward as it’s easier for the kids to understand.
Node Red is a graphical programming environment that you use with a web browser. This means a quick tweak can be made from your cell phone! Not the best experience, a cell phone, but doable.
The core concept of Node Red, is you drag ‘nodes’ or blocks onto your screen and set them up with the particular details that node needs. Configuration settings such as the specific pins on the raspberry pi you have a sensor plugged into, a login for an online service, etc.
You then connect the different ‘nodes’ together with lines, and the whole thing just starts working. Amazing, really.
You program a computer using the same methods you would use to explain a process to another person. Draw a bunch of boxes saying this box does this thing, and connect the various boxes together with lines showing how different events are chained together.
When you use the Node Red menu in the Raspberry Pi, it opens up a text window, with a bunch of stuff on the screen. In amongst that text, is instructions on how to set Node Red up to turn itself on automatically when the Pi starts. Now you have automatic monitoring even if the power goes out and comes back on.
Direct reading of water level via sonar
Ultrasonic distance meters turn to out to measure the distance to a water surface fairly well. The water needs to be reasonably flat & calm for it to work reliably. The thing basically beeps at a high enough frequency that we can’t hear it, and listens to see how long until it hears it’s echo back. A little bit of math, which computers happen to be good at, and you have a distance measurement!
I picked up a bunch of HC-SR04 sensors for cheap from eBay. You can get them from reputable sources for around $5 each.
The HC-SR04 sensor tutorial I followed when writing code is found at https://www.modmypi.com/blog/hc-sr04-ultrasonic-range-sensor-on-the-raspberry-pi
If you want to learn about all of this, it is good to work through the tutorial. I ended up dropping the tutorial method and used Node Red.
Node Red needs an add-on node to ‘talk’ to the sensor. The one I found is https://flows.nodered.org/node/node-red-node-pisrf . Install it according to the instructions, restart Node Red (or the Raspberry Pi if you haven’t figured out how to restart Node Red) and reload your browser window for it, and you can now start taking distance measurements.
Differential Pressure water level method
Have you ever noticed that if you hold your finger over the end of a straw, stick it in your glass, the water goes up the straw only a little bit? When you do that, you are increasing the air pressure inside the straw.
If you compare that air pressure inside the straw, to the air pressure outside the straw, you are working with differential pressure. We can use this to simply see the cycle of water rising and falling, or calculate the actual height of the water inside the pipe. I don’t know what physics principle to use to do the math for calculating actual water height.
I used a BMP280 temperature and pressure sensor. The adafruit library didn’t work well for me. I did however find https://github.com/ControlEverythingCommunity/BMP280/blob/master/Python/BMP280.py which worked well.
The Node Red library has a bug in it at the moment. When you try to use it with the BMP280 module, it crashes Node Red. If you see this happen, the fix is simple, you need to call in the bigNumbers.js library in the right spot. Once you do this, things work correctly.
The BMP280 had some issues with longish wires. I ended up using some Cat5 with the tip from https://www.raspberrypi.org/forums/viewtopic.php?t=82049 for how to pick the wires to get the best performance. This worked well, if a bit time consuming to pigtail the doubled up wires so I only had 1 wire to solder onto the printed circuit boards.
Seeing the data
I logged the data to io.adafruit.com using MQTT. The library I used is found at https://github.com/adafruit/io-client-python for coding things the hard way. Node Red has a built in MQTT node as well.
I used Adafruit’s IO tool because it’s cheap (free) and easy, and is great for learning how to do all of this. There are other options available from Amazon, Azure, Google, IBM, and many many more. Adafruit’s tool is great to start out with.
I just came across an interesting use of a mobile phone for sciency things… a audio frequency analysis to count RPM.
Another great video by Matthias Wandel, or what I new him from for many years before I found his youtube channel, woodgears.ca
The premise is you find the audible frequency spike, combine that with the knowledge of what is actually compressing the air to make the noise, and do some simple math.
Simple, really. All great things are simple.
I thought I would give it a try with my small battery Dremel tool. A bit of tape to compress the air and get it to make noise as it spun, and I get some numbers. They are less than the rated speed, but that is to be expected as the tape flap is pretty big and slows things down a lot.
Testing an app on my cell phone
Notice in the following images in their top right corner, there is a red and green labeled numbers. Those are the peak frequencies. Frequency of the loudest noise it hears, plus the ‘local’ or whatever is on the screen. So you can zoom in on something and get the number of just what you are looking at. Red is the ‘Peak’ and green is current.
The low speed setting on my Dremel with tape reads at 97.85Hz times the 60 seconds per minute gives me 5871 which is a bit less than the 6500 listed on the label for low speed. Battery drain, load, etc. I think this sounds about right.
Trying the high speed, I see 170.89 or 10,253 RPM. Much lower than the 14000 listed.
I remove the tape and tried again now that I have an idea of what to expect.
Low Speed, the first spike is at 103.36Hz which works out to be 6201RPM, pretty close to the 6500 claimed RPM.
Looking beyond that first spike, we can see the harmonic frequencies. The first big harmonic is a multiple of 3. If I rotate the shaft by hand, I feel 6 ‘clicks’ so this makes some sense, but I don’t yet fully understand this. Something to look into.
Here we see the the 2nd spike at 311.46hz.
Zooming way out, we see all sorts of spikes in frequencies, fairly evenly spaced out.
Lastly, lets look at that high speed setting without the tape.
The high speed is coming in at 208.1Hz or 12486 RPM. Much closer to the 14000 listed on the label.
A different App
Trying a different App I get slightly different numbers. They are all less, so it could be that I drained the battery a bit and it’s running slower. Or I could be seeing a difference in the maths under the hood. Not sure, more investigation will be required.
I see 102Hz
I tried kicking up the FFT ‘bucket’ size as recomended by Matthias in his video and don’t see much difference here.
Looking at the higher speed I see 204Hz.
With the higher FFT size I see 207Hz instead. I think the higher FFT sizes gives smaller ‘slices’ of the data so it can better find the peaks.
Things to try
I am wondering if basically covering the microphone on the phone with something like a stick and touching the stick to the frame of whatever you are trying to count RPM might be a way to ‘hear’ the frequency. Kinda like a stethoscope.
I think this is a really great technique. Thanks to Matthias for creating an exceptional video explaining what is going on.
This is certainly a great way to use a cell phone in creative ways. Multiple apps are available to accomplish this cool trick.
I’ve been giving myself a crash course refresher on light over the last new months. It started when I picked up a used Laser Cutter and wanted to figure out how it cuts with light. What I’ve written here is my understanding of things. I may be wrong, if so, please let me know in the comments.
My simplification of a CO2 laser is that it’s a poorly designed Neon light that gets way too hot and produces a heat ray that we can manipulate with mirrors to vaporize things. Magnifying lens on a sunny day style. I fear if I ever find an ant in my laser cutter whatever project I was working on will be a total loss as I will be chasing the ant with my laser beam. https://en.wikipedia.org/wiki/Carbon_dioxide_laser
So I now have this really sharp cut thing that I can’t see the blade on. A CO2 laser beam is invisible. If you see a red dot on a laser cutter, that is a separate cat-toy style red laser put in place so we can guess about where the actual invisible laser will do it’s thing.
My laser cutter’s light is 9.4-10.6 micron wavelength. This is the same wavelength that Humans glow in the dark. Well, everything room temperature glows at this light frequency or ‘color’.
A thermal camera can see light in this ‘Deep Infrared’ zone. I have a Seek Thermal camera that plugs into my cell phone which allows me to see effectively heat. I can walk around and find things that are plugged in that are doing a bad job of being off, and give off heat because they are still on. I can tell the temperature of anything, just by looking at it with my cell phone. I can also look for things that are supposed to help keep me warm and are failing at their job like doors and windows.
Except nothing is ever that simple. Materials have a property called ‘emissivity’. This is how well they emit light at a certain wavelength. Things that emit light well, tend to absorb light well. Things that don’t emit well tend to be reflective in nature.
Humans have a pretty high emissivity about .98 (with 1 being perfectly emissive and 0 being perfectly not-emissive) so we need clothes to stay warm as we would glow all of our heat away without them. But because we are highly emissive, we can absorb heat well too, so this is why you can feel heat being given off by things like hot pans, light bulbs, and turtle heat lamps, and sitting in a sunny spot.
Things such as shiny metals have a low emissivity, so they tend to reflect heat like a mirror. This is how camping space blankets work. The thermal ‘glow’ that us humans have gets reflected back at us. We absorb a lot of this reflected heat, so space blankets feel warm to us. Because, you know, science and stuff.
But, what this means, is that my fancy thermal camera can’t take accurate temperatures of shiny metal things. What I am really taking the temperature of is the things reflected on these ‘heat mirrors’. To do a good job using thermal imaging for temperature reading the more expensive equipment has material tables that you can assign to spots that have a lookup to a emissivity table so it can calculate the proper temperature based on what it sees.
I am going to carry some electrical tape which has a pretty high emissivity number around .96 and just stick that on things I want a proper temperature of. Because, you know, lazy and stuff.
Materials have some pretty funny ideas about what is ‘clear’ and what is ‘opaque’ at wavelengths other than what we can see with our eyes. Thin plastic bags that we can’t see through are transparent to deep infrared. Stick your hand in a bag, you can see your fingers as clearly as if the bag wasn’t there. Windows, glasses, things that we see through all the time are as black as night to thermal. “Low E” windows are not only black to thermal, they are reflective as well, so you can see your heat reflection in a ‘good’ modern window. CLICK. OH, that’s why “Low E” windows are better, they reflect heat. I get it now.
Another portion of light, called Near Infrared, has some interesting properties as well. First off, things that aren’t metallic (reflective) are rather transparent. Things look kinda like jello at these wavelengths, the light can see into them a ways. A couple of centimeters often times.
The Near Infrared has another interesting ability. Oils, fats, sugars, alcohols, and proteins absorb certain frequencies of light – they have colors (for lack of a better word) in this range. Click here to Geek Out on Near Infrared. This means that a camera that uses Near Infrared is very useful around the house. We can look at something, and judging by it’s ‘color’ in Near Infrared, we can make a good guess as to what is made of, or at least major components of it. We can’t see near infrared, so we tend not to manipulate the colors in that range.
There is a gadget that takes advantage of these useful properties. The SCiO which is a Near Infrared ‘scanner’.
This little device is even cooler than the thermal camera. It’s small, and can tell you the interesting bits about your food like how many calories and of what type (fats, carbs, protein) are in it. You don’t have to guess at a restaurant if you are tracking your diet anymore.
This doesn’t work like a camera, it is instead a spectrograph. It doesn’t take a picture of stuff, it instead looks at all the colors that are present like how a prism works. You scan something with a SCiO and it breaks apart the intensities/brightness of different wavelengths of light (those would be the colors if this was visible light) and looks up what it sees against a database of stuff that it knows about and when it finds a match, tells you what it is looking at. If we were to present only pure substances it would be able to tell us what things are easily. However, we don’t have much of anything that is truly pure. Table salt, sugar, baking soda, for the most part tap water are about all I can think of commonly around the house. Most of the stuff we interact with is made up of a variety of things.
This is where we get clever with the SCiO. Instead of needing to extract out the stuff into individual bits (imaging taking a baked cake and separate it back out to it’s flour, sugar, eggs, milk, water, etc) we just capture what a thing looks like in different bits of Near Infrared light and correlate it to things we’ve told the SCiO what they are previously.
The thing that makes this work is that we LIMIT THE DOMAIN of what things are so the SCiO has a chance at making a reasonable guess. For example, there are a lot of things that are Red. Lego, fancy cars, strawberries, some apples, etc. If we showed you a particular shade of red, and asked you what was that color, you can come up with a lot of wrong answers that are that exact shade of red. But if we said we have a berry that is this particular color, you would be able to tell very easily what it is most of the time. Especially if you can look back against other color samples and compare what you have now with what you have seen in the past.
So for the SCiO to work well, we need to train it. We get together a bunch of things that we want to tell apart if it’s not properly labeled. We then teach the SCiO this thing is X, that thing is Y. We can than ask SCiO what is this stuff, it’s something that belongs to this group that we trained it on.
An example could be clear liquids. Clean water, vodka, strong vodka, watered down vodka, rubbing alcohol, denatured alcohol, clear soda, vinegar. These things all look similar to our eyes. They will all look different from each other in Near Infrared. We can train the SCiO about all of these clear liquids, and when we find a glass of one and we don’t know what it is, we can check with the SCiO.
There are some ways in which the SCiO can fail.
Shiny mirror like surfaces tend to reflect all light, regardless of the wavelength. Metals for example. We also see this in Visible light as well as Deep Infrared.
Things that are black – they absorb light – tend to absorb all frequencies of light including Near Infrared. Things that have been colored black will likely be black to the SCiO as well, and it can’t get a good read on them.
Only the major components of something can be read by a SCiO. If there isn’t enough of something to make a strong ‘color’ influence, it simply can’t be read. A SCiO can ‘see’ stuff that is more than 1% or so of the overall item.
Understanding how a fancy new tool works ‘under the hood’ helps me manage my expectations of what the tool can and can’t do well. I can ‘hack around the edges’ of it’s capabilities because I understand what the edges of capabilities are and why they exist.
I added more items to the Mobile Science Lab. The main new item is a WeatherFlow WEATHERmeter. This is a pretty neat bluetooth device that captures wind speed and direction, temperature, humidity, and barometric pressure.
New stuff means I needed to update the case. I fiddled a bit to get everything to fit. I think the next iteration will end up including layers. I will need to find some laser safe foam core or something light like that. The cardboard won’t hold up all that well when there are removable sections.
I ordered a much larger Pelican case tonight. The SCiO I am ordering at the end of the week should fit into the new case.
I’ve also added some NFC stickers a few places to make using the bits a little easier. The case has my contact info embedded in it. The WEATHERmeter is now set up to just tap the phone against it and the correct App will load.