Carvey PCB Machining Characterization
EE201: Network Theory I Honors Project
Fall Semester 2023 | Dr. Adam Harris
Fall Semester 2023 | Dr. Adam Harris
This project aims to characterize the Carvey machine and understand its capabilities for rapid-prototyping printed circuit boards (PCBs). It also seeks to simplify and document a workflow for simple PCB prototyping. The Carvey machine is a desktop CNC machine currently most commonly used for milling plastic or thin metal. Dr. Adam Harris, the professor for EE201, provided me with a touch probe for the Carvey, a variety of bits for milling, and several online resources for completing this project.
The deliverables for the project involved this documentation as an explanation and summary of the project, documentation for the workflow of using the Carvey (see the Carvey Workflow Instruction section below), as well as a 5-minute video to further detail my work (see the Video Summarization section below).
The Carvey Machine by Inventables
Following the steps that the Fab Academy used for their Carvey characterization, I found and downloaded their test design to test the limits of the trace sizes that the Carvey can accurately mill. As per their instructions, I converted their PNG of the trace test into an SVG file in Inkspace for importing into Easel (see top left image).
Upon measuring the FR-4 blank, I input its size - 3" x 2" x 0.6" - into Easel and changed the material to PCB. I set the cut depth of the test profile as 0.3mm and positioned it in a horizontal orientation toward the top of an FR-4 blank (see bottom left image).
On the physical machine, I placed the FR-4 blank underneath the Smart Clamp in the bottom-left corner of the workspace. One setback I noticed for PCB milling on the Carvey is that the use of the Smart Clamp prohibits milling in certain areas so that the machine does not accidentally damage the clamp. For larger pieces, this is likely not a problem, but since these blanks, and most PCBs, are quite small, it is inconvenient. I am unaware if this is a setting that can be disabled, but from my understanding, it is built in Easel. A workaround could be using a universal g-code sender rather than Easel.
In their example, the Fab Academy used a 45-degree V-bit and several tests to accurately reproduce each trace in their test design. I, however, attempted to mill their trace test PCB with 0.4mm endmill bits. In Easel, I input this by saying I am using an "Other" bit with a diameter of 0.4mm. I was unsure of what feed rate, plunge rate, or depth per pass values would be appropriate for this bit, so I went to a forum here. After looking through this forum, I set the feed rate to 150mm/min, the plunge rate to 30mm/min, and the depth of 0.3mm per pass. When I ran the Carvey, the bit immediately broke upon trying to create the first hole (see first image).
The reason the bit broke was that the depth per pass parameter was too large. After the bit broke I used the troubleshoot menu in Easel and learned the depth per pass should be less than half the bit's diameter. I then set the depth per pass to 0.1mm. While re-running the program, the bit broke again, this time when the machine began to mill a slot.
I continued to break bits, adjust parameters, and rerun the program, each one running a little bit better than the previous one. This continued until I had broken five 0.4mm bits and the calculated milling time for the test trace design had inflated to over three and a half hours (see image 2). This method seemed extremely unreliable for repeatability and efficiency, so I advanced to testing with the 45-degree V-bit.
I tested the 45-degree V-bit with the same trace test design from Fab Academy. Erring on the side of caution, I set the feed rate to 163mm/min with a plunge rate of 30mm/min. This test took fifty-five minutes to complete. In comparison with the Fablab's parameters, this was 120% slower than their settings. After the first successful trace test was milled 0.3mm deep in the horizontal orientation (see first and second images), the speeds were increased to a feed rate of 650mm/min and a plunge rate of 250mm/min. These speeds were then used for a vertical orientation test, which took only 14 minutes to complete.
After two very successful tests, demonstrated by the visibility of each trace on the test piece, I turned the test pattern forty-five degrees and milled with the same settings as in the vertical test. Since the Smart Clamp covers much of the board when it is clamped in the bottom-left corner of the workspace, I used double-sided tape on the back of the board (see fourth image), and situated it in the middle of the workspace (see fifth image).
These forty-five-degree tests did not prove as successful as the initial two tests. Despite having the same depth as the previous tests, the boards were obviously milled deeper, leading to a loss in resolution for the smaller traces. It is likely that the workspace for the Carvey is slightly unleveled. Since I was milling with 0.1mm parameters, a very small amount of skew can lead to these results.
Further tests were done with depth in different areas. In the bottom left corner, which seems to be the most level, a depth test involving three boxes of different milling depths was conducted. Then in the middle of the workspace, three more trace tests were milled, each at different depths to one another. Once again this test proved that the bed wasn't entirely level. The 0.1mm depth in the bottom-left corner of the workspace was very shallow, while the 0.1mm depth test in the center of the workspace was as deep as the 0.3mm test in the bottom-left corner.
One way to counteract this would be to use an autoleveling plug-in found in several g-code sending software - see the Future Plans section for details.
After thorough testing with the forty-five-degree V-bit, I hoped to test the capabilities of the sixty-degree V-bit. After calculating the width of the bit at a depth of 0.1mm - the same value as the depth per pass parameter - I lowered the feed rate and plunge rate to 250mm/min and 200mm/min, to once again err on the side of caution.
The test ran with no problem, but it was going to take over an hour to complete. After analyzing the completed test piece, The 60-degree V-bit struggled to create small traces, even when set to the same depth as the forty-five-degree V-bit.
After running a simulation in Easel, it was obvious that the 1mm bit would not be able to create over half of the traces in the trace test design due to the large diameter. Therefore, I used the 1mm bit for drilling through holes and the milling of the PCB outline.
To test this, I created a circle from Easel's shape library to replace a mounting hole in a PCB. I then edited the parameters of each shape so that each hole had different-sized tabs.
Throughout these tests, there were several things that I noticed with the Easel software and Carvey machine itself. The items and their solutions are as follows:
Despite my tests having a fairly short duration (around 13 minutes each), I noticed that the collet would feel warm when I switched out the bit.
Much like the systems used with 3D printers, we could create a control system that controls the temperature in the Carvey enclosure. Small bits - such as the 0.4mm bit - are far more heat-sensitive due to their size. This heat can lead to breakage, so if we were to control the temperature of the bit (often done with coolant in machining), we could preserve and lengthen the life of the bits used.
The Carvey would incessantly lift between milling passes on the board, lengthening the time of fabrication - especially with bits with minimal plunge rates, such as the 0.4mm bit.
After checking the Easel general settings and the Carvey settings, I found no solution to editing pathways. To counteract this inefficient milling path generation, we could use different software such as Chilli Peppr, Candle, or the Universal Gcode Sender, both of which can generate code for a given circuit.
To further improve the process of rapid-prototyping PCB designs on the Carvey, the following items will be explored or enacted on in the future:
Further testing on optimizing feed and plunge rates for different sets of bits.
Test repeatability by removing and reinserting the bits between each test.
Implementing auto-leveling software such as Candle, Chilli Peppr, or Universal Gcode Sender.
"Hacking" the Smart Clamp so that thorough probing of workpieces can be achieved.
Test efficiency of fabrication with a custom PCB.
Implementing a custom gcode generating software to interface with the Carvey.
There were promising results with the 45-degree v-bit and 1mm endmill bit. Using these two bits, it is possible to fabricate a reasonably complex single-layer PCB out of FR-4 board. Further improvements - as listed above - should be made to make a seamless and efficient process for the user.
Araujo, M. (2022, June 27). Electronics Production - Marius Araújo. Retrieved December 9, 2023, from fabacademy.org website: https://fabacademy.org/2020/labs/fct/students/marius-araujo/assignments/week04/
Fab Academy 2018 - Fablab ULB. (2018). Retrieved December 9, 2023, from fabacademy.org website: https://fabacademy.org/2018/labs/fablabulb/ga_pcb_process.html
Harris, A. (2022, January 9). How to Export to a CNC from KiCAD and Fab Mods. Retrieved December 9, 2023, from SheekGeek website: https://sheekgeek.org/2022/adamsheekgeek/how-to-export-to-a-cnc-from-kicad-and-fab-mods
Inventables Inc. (2023). Press Materials - Carvey. Retrieved December 9, 2023, from site.inventables.com website: https://site.inventables.com/press/materials/carvey