4. Electronics production
For this week's assignment, I ordered some Quentorres boards and worked on making one myself.
Sending a PCB out to a board house
There's a lot of PCB vendors to choose from. I read a few reviews and liked this one, which provided comments on five PCB vendors, OSH Park, DigiKey Red, JLCPCB, PCBWay, and Aisler. I decided to try OSH Park first.
I hit the Browse for Files button on their home page and selected the quentorres_v2.kicad_pcb file from a directory in which I had modified the KiCad project to include "Feb 2024" on the board.
Initially, the app processed the uploaded file and then presented images for the top and bottom of the board and a price for 3 boards, telling me that "3 boards will cost $15.75". I didn't mind the minimum 3 board quantity since it's always good to have spares on hand in case you mess up soldering.
On the next screen, each separate layer was shown so that it was easy to check if the different layers were correctly read. A partial screenshot is shown below.
Once I verified that the design was correct, I was shown a cart which summarized the ordered options. I just took the default options.
After entering payment information, the order was in the queue!
The OSH Park ordering process was super easy and clear, and the UI was beautiful. Although it wasn't cheap, for my purposes the price was acceptable. For a larger run order, I'd be more concerned about cost.
In-house PCB production process
At Artisans Asylum, the E&R (Electronics and Robotics) Shop has an old version of the OtherMill PCB mill. In order to use it, you need to be "tool tested" on it, meaning that either the shop leads or one of the shop's tool testers has to clear you to use the machine. I stopped by during their weekly Wednesday "Circuit Hacking Night" to learn how to use the machine. Luckily for me, their co-shop lead Mike Beach was there to guide me through the process.
Boards. I had purchased some single-sided and double-sided blank PCB boards from Amazon, but it turns out, I didn't buy the right kind for the machine. Their mill can only use FR-1 boards. I didn't know what kind of board I had purchased (they turned out to be FR-4), so they let me use the stock they had on hand instead.
Software. The Bantam Tools Milling Machine Software is used to load the Gerber (or EAGLE) files and communicate with the PCB mill. They have a desktop computer set up and connected to the mill, so I used a USB drive to transfer my files over to it.
I was shown how to load my files into the software, choose bits, start/stop the machine, change bits, load the PCB board and start the job. For loading the PCB board, we used Scotch® brand double-sided tape on the back of the board to keep it in place. A partially covered board is shown below.
The board was placed in the mill, making sure to press it against the front left corner and pressing down on the entire surface to adhere the board to the mill's platform.
Shortly after starting the job, the board came loose and the red emergency STOP button on the right side of the machine was pressed. It turned out, the board was quite bowed so the double-stick tape was not strong enough to keep the board on the platform when the endmill was spinning on it.
After that attempt, I watched someone else mill their board (on a new, flat PCB board) and learned to be careful in specifying whether to mill the top or bottom side of the board. For the Quentorres board, the top side is milled since the traces are on the top for a surface-mount board. For the board that this other person was milling, it was using through-hole components and the board was designed to have the components on top and the traces on the bottom of the board. The software assumes by default that you are milling the top of the board, so they had to restart their job to specify that what they wanted to do was to mill the bottom of the board.
After watching that board get milled for a bit, I was cleared to use the mill on my own.
Milling the Quentorres board
I returned the next afternoon to mill my board. I loaded up the files in Bantam Tools by first selecting the front copper file, quentorres_v2-F_Cu.gbr, then the outline file and the plated through-hole file. A bottom file was not needed for this board. The name of the files depend on the PCB editor. For KiCad, the outline file ends in Edge_Cuts.gbr.
The program warned me that some traces were too small for the default 1/32" Flat End Mill, so I added the PCB Engraving Bit 0.005" to the tool list. This was what had been recommended to me the prior evening. A 1/64" Flat End Mill could have also been used instead.
The PCB board from the night before had been left for me to use, as there was still space to mill out more boards on it. I taped the back of the board, put it on the mill platform, and hit the button to "Start Milling". The software prompted me to be sure things are cleared for homing the bit, then I pressed the button to let the mill home the bit. Then the software prompted for the engraving bit to be installed. I put that in, and then hit the button to continue the job.
After a short period of time, the mill stopped and prompted for a bit change. I installed the flat end mill and let the mill resume. After completing most of the traces, the mill started on the QT design. It sat there for a long time in the middle of the Q, and the line number of the g-code was not advancing. After about a minute, I hit the emergency stop button, as I did not think that was correct behavior. I checked the bit, and it looked fine. I don't really know what happened. I didn't see a way to resume the job, especially since I didn't write down what line number of g-code the bit was at when I stopped it. Next time, I'll keep track of that.
Before making another attempt, I opened up the KiCad file and deleted the QT from the front copper layer. I then exported the files. When I loaded them up into Bantam Tools, the edited front copper layer was fine, but the other layers were not, so I used the other layer files from the set that Fab Academy provided instead of the one I had generated from KiCad. When there is more time, I should look into why the other files I exported weren't being read properly by Bantam Tools.
I followed the same process as before and didn't run into any problems this time. What was interesting was that the estimated time of 22m 7s was a bit off. The mill started the outline with about 10m to go, and I didn't think it would take the mill 10m to do the outline. In fact, at around 6m, the mill was done and the time dropped to 0s.
The mill completed the job successfully (in the front left corner)! You can see my earlier attempt in the top left corner of the picture below.
Board Assembly
To be continued...