This is the sixth part of the meta-tutorial, where I talk about designing a cheap plant watering sensor. If you did not already read the first, second, third, fourth and fifth part please do it now. These parts contain a lot information which lead to this point of the tutorial.
The fifth part ended with step 24, where I talked about calculating the total bill of materials. This part will focus on preproduction of a small batch of sensors to solve some final details.
Just note, I obviously do not follow these steps in a perfect sequential way. Often I start with some tasks earlier and things are running in parallel. There are various dependencies and it would make no sense to wait with some task just to follow a strict sequence. 🙂
If you follow my blog you may already read some details about ongoing tasks. I will just briefly talk about them in this article. You will find more details in the other blog posts.
Step 25: Build an Alpha Series
Everything looks very promising, so its time to build a small batch of the final devices to see if they work as expected. This is also a test to see how a larger number of these devices can be produced and what kind of tools are needed for this task.
Order the Components
First I order the components. This is very important, because the availability of electronic components changes all the time. It is nice to have all required components, so you can order the boards with the correct footprints. If you order the boards first and are unlucky, an important component is suddenly unavailable and you have lots of boards with wrong component footprints on it.
The components for the plant sensor are really cheap, so there is no huge risk. Even it turns out a huge issue requires a component change – it will be a small loss. SMD components also do not take a lot of space, I can easily store all of them in a very small box.
After ordering the components I wait until I actually get them. In my case all components were available, expect the green LED where I only got ~30 and had to wait more than two months for the rest. I was confident the rest of the LEDs will arrive, and the initial batch was enough for the alpha series.
Prepare the Components for Production
If you do not own a pick and place machine and plan to do the assembly by hand, it is really helpful if all components are in small boxes with labels.
In the photo above you can see components for 200 plant watering sensors neatly prepared and labeled. I write the reference designator on the lids, which is the fastest way to pick the components and place it at the right locations on the boards.
The SMD boxes you can see here are from Wolfgang Warmbier. They have a very good quality and can be joined to a setup as shown.
Order the Boards
Now its time to order the final boards. The amount of boards you order depends on your confidence in the final design. In my case I ordered just one panel of the final head part with 8 boards in it, but all foot parts – because it is unlikely there will be any huge change.
Eurocircuits delivered boards in an astonishing perfection. You can see how precise all layers match and how exact the pads and traces are produced. It seems they share my perfectionism! 🙂
I ordered the foot parts panelised and separated by Eurocircuits. As you can see, there is no trace of any bridge left on the boards. This saves a lot of time, because I can use them directly with no additional work required.
Prepare the Wires
Each head part is connected to the foot part using two strong 0.8mm diameter wires. Looking back I should have ordered them cut and stripped from a Chinese supplier – but I had no idea how much work it takes to do it by hand.
First I created a small device to put the spool with the wire on it. The device had a board with the required length of 20cm. So I could just pull the wire from the spool to the end of the board and cut it.
I spent hours cutting 400 wires to the right length. Luckily I got help from my girlfriend, so I could focus in placing there wire while she cut it at the right place. Next I had to strip the insulation from the ends of the 400 wires — just 800 ends to remove the insulation. 🙂
You can see all 400 wires in the box in the photo above. Half of them with removed insulation.
Soldering the Wires to the Foot Parts
To solder the wires to the foot parts I created a small holder from cardboard. I embedded strong Neodym magnets to hold the board and wires in the right position for soldering.
You can see the flux blobs everywhere. One reason for the narrow window for soldering, was to prevent flux blobs on the lower part of the board.
One hour later, 200 soldered foot parts in the box.
Searching for the Best Insulation for the Foot Part
It took many weeks.
My original idea to seal the foot part using a hot shrink tube, PLA and a silicone material was a complete failure:
The idea was to use a PLA material, as used for hot glue, on the wire end of the board and shrink the tube with hot air. The PLA will melt and seal the wires and contacts perfectly.
You can see in the photo above, this worked well and reliable. The PLA material is like wax and repels water.
To seal the hot shrink tube on the bottom end of the board I used a special silicone material (Dow Corning 3140) which is especially made for electronics. The plan was to dip the tip of the board into this material which seals off the tube and tip of the board.
This silicone paste is a wonderful material, and if used directly on a board to seal electronics it produces really great results. For my use there were a number of problems. First it did not flow between the tube and the board and the material did not stick on the hot shrink tube.
The curing process works with the humidity in the air. It caused some air pressure from the trapped air under the hot shrink tube which caused air bubbles at the edge. You can see them in the photo.
The Quest for a Coating
After this failure I searched for a replacement for the hot shrink tube and experimented with different coatings. The next picture represents all failed coatings:
To make things short, here a list of coatings I tried and why they did not work.
- Acrylic (various): There is a wide range of acrylic coatings. I found ones with a nice viscosity to cover the board perfectly. The problem with all acrylic materials is, they eventually absorb water. If you put the probe into the soil, the coating will absorb water from the soil. If the soil is dry, this absorbed water takes days to leave the coating again.
- Varnish: I got the nicest looking results using regular varnish for wood. It will never cure completely on the board.
- Polystyrene: This would be an ideal material because of its insulation characteristics and very low water absorption. Using it as a coating does not work, because the required hardener reacts with oxygen. If you pur larger blocks, this is no issue, but very thin layers never cure perfectly. So you end with a very sticky surface.
- Plastik 71 Spray: This works well, but it is very hard to apply it in a consistent way to several boards.
Experiments with Epoxy
I got the best results using various combinations of epoxy resins and hardeners. It is a very nice material, because using different combinations, you can get any kind of viscosity, curing time and chemical property. Contrary to other materials, epoxy creeps into smallest cracks, similar as water.
In the first tests I tried to coat the whole board with the contact and wires in epoxy, but there were always small errors in the coating around the solder joints. I assume, this is because of small flux blobs. I also feared cracks in the coating if the wire was bent too close to the board.
Next I tried to add a very short piece of hot shrink tube in this area and got the first promising results:
The only problem were the air bubbles in the hot shrink tube. To solve this issue I just used a longer piece of the tube which formed a small “neck”. If I dipped the part in epoxy, this neck was filled with resin and the excess was creeping slowly into the free space, filling the hot shrink tube completely.
Here you can see the first perfect results. To achieve this I had to experiment with different epoxy resins and hardeners.
Some notes and warnings about using epoxy:
- Make sure you use good protective gear! Your health can be permanently and irreversibly damaged if you ignore the following warnings.
- Epoxy is a really dangerous material until perfectly cured.
- Absolutely no material for kids to use – not even let them in proximity of it.
- Wear eye protection.
- Wear thick Nitrile gloves. Be aware they do not protect you hands from the epoxy hardener. They just delay the exposure for a few minutes. If you get hardener or mixed epoxy to your gloves, immediately remove them, wash your hands and use new ones.
- Wear minimum an A2 gas filter if you work indoors.
- Make sure no skin is exposed and at risk to be touched by epoxy. If you get mixed epoxy or hardener on your clothing – remove the exposed clothing – wash the skin under it.
- The boards have to be perfectly oil free. I used gloves to solder and prepare them to avoid any finger prints on them.
- The room temperature while you are working with epoxy is very important. It influences the pot time, hardening and final result. If the epoxy is cured at a low temperature it does not cure perfectly.
- Fast hardeners can easily create very high temperatures in the epoxy which can start fires.
- Never use polystyrene cups to mix epoxy.
Prepare Foot Parts for the Alpha Series
This are a small number of final foot parts for the alpha series. You can see each foot part has a serial number.
I used this setup and work process:
- To avoid epoxy waste, I prepared a small cup with aluminium foil, to get a mould with a minimal shape which fits approximate five foot parts.
- I prepared a plate with aluminium foil used to pour the excess epoxy on it. This is required to make sure the excess epoxy does not overheat while curing.
- I prepared a box for the coated foot parts to dry. On top of the box there are wires to hang the foot parts on. At the bottom of the box I use aluminium foil to make sure the excess epoxy stays in the box.
- Next I mix a small amount of ~60g epoxy, use the mould in the cup to dip five foot parts into the epoxy and wait until all air bubbles are gone. This is a quick process, and after 20 minutes 50 pieces are done.
- The pot time for the used epoxy is 40 minutes, this is enough time that even after 20 minutes the viscosity is low enough to work.
- I dispose the excess epoxy on the flat surface of the prepared plate to avoid overheating.
After 20 hours I check the parts and add some extra epoxy with a syringe, if there are pieces where the epoxy did not fill the tube up to the edge.
Prepare a Small Number of Head Parts
Next I prepare a small number of final head parts.
I use the special registration system from Eurocircuits. The benefit of this system is some standardised sizes and locations of holes on each panel and on the stencil. Eurocircuits provides different kind of tools to precisely align the panels and the stencils.
The panel is aligned on the tool using two small metal nobs.
Two additional metal nobs align the stencil on top of the board. I used some tape to fasten the stencil, but I found out this is not necessary. The stencil is aligned perfectly and does not move.
Adding solder paste on the stencil.
Spread the solder paste across the stencil until all gaps are completely filled.
Carefully removing the stencil from the board.
The solder paste is now applied on all pads on the board.
You can see the perfect alignment of the stencil. The solder paste perfectly covers all pads.
Next I place the prepared components on the boards.
The populated board before hot air soldering.
The final board after hot air soldering. Now I just have to add the battery holder and solder the foot part to the boards.
Step 26: Start a Long Term Test
I started long term tests relatively early, but the first one I did was a failure. I used one of the prototypes with one of the final foot parts. Because of problems with the insulation of the foot part I got really strange and useless results.
For this reason I first finished my experiments with the best coating for the foot parts, before I started a second long term test.
This long term test is still running, using the final board connected to the logger using an optocoupler.
This setup provided very good and accurate results. I already analysed the first watering cycle recorded by the logger.
The curve looks great and shows successful measurements. There is this strange jump between 120kHz and 130kHz which seems to repeat with each watering cycle. I have no clue what the cause for this jump could be. It will have no influence on the usability of the sensor, so it seems this is no huge problem.
I will continue the long term tests, until I collected data for some different plants and at least five watering cycles.
Step 27: Start Writing the Manual
Start writing the manual for the sensor early has some benefits. You have to recap the whole usage of the sensor again. You also have to think about the user and what he would like to know about the sensor in order to use it.
Usually you need at least this parts in your manual:
- Table of Contents
- Safety Statements
- Elements of the Device
- Mode of Operation
- Signals / Display / Errors
- Maintenance and Cleaning
- Technical Data
I created a first version of the manual for the plant watering sensor which contains the complete text, but no images yet. This should be enough for test persons to work with.
Step 28: Get Feedback from Test Persons
As soon you have the devices from the alpha series and a first manual you should start handing the devices to test persons. Never explain things, just hand the device with your manual and let them work with it.
The feedback you get from this persons is very important. If they do not understand something, you will have to explain this aspect in a better way in your manual. Sometimes it is even necessary to go back and make changes on the design or on the firmware to address the problem.
In my case this is also an ongoing process. Because of the nature of a plant watering sensor, I will get feedback after weeks, not hours. The test persons will first have to setup the sensor and experience two or more watering cycles before they can provide a meaningful feedback.
Thank you for reading this article! I hope the details of this article series will give you inspiration to start your own project.
In a few weeks I will publish the final part of the series. Here I will talk about the last tasks to do, like packaging and about insights and feedback from test persons if I get any.
Update: Please see this post with details about the published files.
I will publish all sources from the hardware and software under an open source license starting from August 2017. These files I will publish are exactly the same I used for my version of the plant watering sensor. They are made for hot air soldering using a stencil and do not work for hand soldering. The board also requires a very precise manufacturing process which not all board houses can provide.
The idea behind this article is to motivate people creating their own devices. 🙂 This delay will let me finish the project and get some of the development costs back. Starting from August 2017, everybody may produce clones of the project or sell kits. I will choose a permissive license which also allows commercial use of the sources.
If you have questions, miss some information or just have any feedback, feel free to add a comment below.