Popular Post esaj Posted August 23, 2018 Popular Post Share Posted August 23, 2018 (edited) Part 1: Preface This is a "tutorial" I've been meaning to write for a good while (like... 1.5 years or so ). I don't claim what I will describe (in the later posts, this is just the preface) is the only or "best" way to mill out PCBs, but I've done way over a hundred boards in total (maybe something like 70+ different designs, largest "batch" has been 40 boards of single small design, the IR-led / photodiode transmit/receiver boards for the robots) with the CNC and have developed my own "method" which makes good boards almost all of the time, and when it doesn't, it's usually either due to dull V-bit, trying to "save time" by skimping on the autoleveling grid size or a mistake in design, or a combination of these . I got a cheap Chinese CNC 2418 about two years ago, in the fall of 2016, and I haven't regretted it. There are other models too, usually named after the work area they have (X- and Y-planes, Z-plane movement is usually not indicated in the name), for example, CNC 2418 has a X/Y -workplane of about 240 by 180 millimeters. CNC3040 has (roughly) 300 by 400 etc. You get the idea. To get an idea what to expect, here are some (more or less randomly picked) examples of a few boards (nothing really complicated here though, and probably not in chronological order): This one was very early on, some number of days after getting the machine and slowly learning to use it, the same design seen under the ruler was a total disaster, not using autoleveling, but this one was nearly perfect when it comes to the isolation milling: If you look at that close, it might seem that the lines are not sometimes cut totally, but that's actually just some leftover copper/substrate dust the vacuum didn't pick up that gets washed away after drilling & cutout and before you start soldering the board. The first time I was dealing with SOT-23-6's (the soldering tip there is 0.8mm, SOT-23's are roughly 1.5 x 3mm without the pins, and pin-pitch of 1mm, although there's some very small variation from chip to chip): An example of a larger board with SMDs maybe about a year or 1.5 ago, this is actually a "reinforced" testing board for Arduino Nanos, with current limiting resistors, separate regulators and output pinheaders for high currents, LED-bars showing GPIO-states that are triggered by logic level mosfets connected to the IO-pins, polyfuses and whatnot. Basically something that prevents me from doing stupid mistakes and killing off a bunch of microcontrollers when trying things out with them : You can also see slots cut for a DC barrel-connector in the middle right: Flashlight LED-board with constant current sinks for holding the light output steady independent of battery voltage (still a wasteful design, as "excess" voltage is dropped off by current sinks; I've since replaced it with a 3W led & step-up switching), probably the first and only time I paid that much attention to component positioning to get the leds at equivalent distances and the board looking "nice" (like it really matters ): Distortion/delay guitar-pedal I designed (well, the delay portion is mostly from the "Small time: Delay with tails" by Valve Wizard: http://www.valvewizard.co.uk/smalltime.html ) and built as a gift for a friend: The cut isn't that good on that one, especially seen on the text engraving, but it works. Also some of the ground plane-thermal reliefs for some components are really close to the edge, but I had designed the the size and shape (the cutout; we'll get to that at a later part) specifically for the 1590B-enclosure, but had to move to a larger enclosure, because of the amount of potentiometers in the front plate. Also, at the time, I didn't know a pedal named "Shredmaster" already existed. This design has nothing to do with the original "Shredmaster", but I might have done a slight, unintentional breach of trademark rights there... Self-oscillating led-flasher, powered by a 1.5F super capacitor. Don't remember the exact dimension, I think the entire board is about 20mm across. The led blinks brightly at about 1.5-2Hz, but the average current consumption is still in the microamp-range. The 1.5F capacitor can keep it running for 3 days and charges up in seconds. First time I was soldering 0603's (or was it 0804's? I forgot...), I did a few of these mostly just for practicing soldering smaller component sizes and trying out milling very small boards (this was before I made the even smaller IR-led/receiver panels for the robots): Motor driver for an older robot, board being inspected in a light table for possible short circuits (I had slightly wrong settings, you can see that the milling left thin copper "slivers" around the traces that are partially loose, you can scrape those off, or just make a new board): A bunch of IR-LED transmitters / IR-photodiode receivers with amplifiers being built for the embedded course -robots, SOT-23-5 op-amps, 0603 resistors etc., relatively small stuff, although I still to this day haven't entered the 0402 -territory (but already got a sample book of 0402 resistors, everything in time...): Prototype of the robot board for the course; first time I designed a step-down transformer including calculating the capacitor and inductor values: I've also used it for cutting holes in plastic enclosures and some light metal engraving, but the motor's nowhere near powerful enough to actually cut metal (the copper-layer of PCB's is so thin that it's not an issue). The holes in the aluminum case were done with a "normal" battery-powered drill remembered wrong, I actually used a friends' drill-press for those: This one's an old picture; you can tell because the ER11 precision chuck isn't there yet. Cutting a hole for a 120mm computer fan to an encasing: The ability to create my own boards in my home at (relatively) fast rates and cheaply has really helped me "zoom along" with my electronics hobby as well as made it possible for me to use SMD components (do note that you can just as well create throughhole boards, and I do combine both THT and SMD components in many boards, partly because I have lots of THT components and also because many large power MOSFETs and such are only available in large THT-cases, obviously because of the high power dissipation needing large cooling area). For example, this week I created a step-up DC/DC-converter board with SMDs from the schematic design and layout to milling, soldering and testing it in something like 4-5 hours including breaks (granted that it is a small board with simple design, mostly just following the DC/DC converter chip datasheet and calculating some feedback resistor values). The "traditional" method of etching boards is still viable, and does have its upsides vs. milling boards, but personally I don't want the hassle of having more tools (printer for films, if using UV-masking, or transfer paper, hot bath for the chemicals) and storing and getting rid of the "spent" chemicals, which are nearly toxic waste (the acid solution like ferric chloride isn't really a problem as far as I know, but once you got dissolved copper in it, it becomes toxic waste that shouldn't be dumped into toilet or elsewhere). A quick list of pros I could think for when it comes to etching is: It's fast: the larger the board, more faster it is vs. milling, or if you need create a larger amount of boards It's accurate: assuming you've got a good printer for making the masking film/toner transfer, you can get 0.1mm traces (or so I'm told) once you know how long to hold it in the bath (which might require some trial and error with time, heat and the solution "strength", but so does learning to use the CNC) It's easily repeatable: assuming you're using films and UV-lighting, making another board is just a matter of masking a new copper-clad with the existing film and etching it (while the repeatibility is also otherwise good with CNC, you need to do a full grid depth-probe for "auto-leveling" each time you start or change the copper-clad, this can easily be 10-40 minutes alone each time, depending on size and how small grid you want/need) It's (relatively) easier to create 2-sided boards, at least with film The process is more or less silent Cons of etching: You need a printer for making the masking (toner transfer or films), of course if you already have one anyway, not an issue (I don't) If using films, you need a UV-machine for masking the boards (and copper clads with UV-sensitive masking surface) For better controlled process, you need a hot bath Storing and getting rid of the chemicals used You have to drill holes separately for THT-components / screws holding the board or whatever, imagine trying to drill holes for something like a 1.27mm pin-pitch TO-220-5 or for a DIP at exactly the correct spots Cutting slots or V-grooves is really hard manually, so basically you don't get these without a CNC You have to cut the board to size manually, if you want round edges, you need to file them down All the PCB-factories (to my knowledge) use etching for creating the boards. It would take them AGES to mill out tens, or hundreds, or thousands of boards, whereas by etching, they can dump something like a square meter copper clads into the baths and get consistent and good quality with their refined and "infinitely" repeatable process. Of course they use CNC's for cutting slots and cutting out the boards, and the drilling machines they use are pretty impressive (like compressed air driven drills running at 100k rpm ). Not something a hobbyist could really dream of getting (unless you're a millionaire or billionaire and can build a separate house for the equipment, although then you might just as well set up an entire PCB factory ). List of pros for CNC: It's (relatively) accurate: the cheapo CNC 2418 I'm using has a <0.1mm repeatibility (the accuracy at which the milling head will return to same position if told to move some amount of millimeters to left and then that same amount of millimeters back to right), I've made boards with 0.3-0.4mm traces, going further down it becomes risky, the milling bit does have some "runout" (fluctuation of the milling bit while it's turning, although precision ER-11 chuck will keep this to minimum assuming the motor axle is straight) and can "snap" a really thin trace or if the board needs some light sanding (I use 1000-grit water sanding paper if need be), the thin traces can get scraped away by accident The Z-axis (up/down) accuracy is even higher; remember that we're running isolation milling in boards with 0.035mm thick copper layers It can drill out all the holes at exactly the right spots during the process It can mill out slots for example for DC barrel-connectors It can mill V-grooves for snapping boards off (although usually I just cut them out straight away) It can cut out the board at the shape you want Regarding cutouts, check this (the shape in this case has no real functional purpose, I was just messing around and testing cutting more irregular shapes, it would have been enough to leave room for the fixing holes of the secondary board near the corners ): There are cases where the cutout matters, I have one on-going project where a battery case-"door" will be replaced by a circuit board, and it's not just a rectangle, but has this stretching out rounded part through which a screw goes and needs to be the exact shape and size, and the general size needs to be down by about an accuracy of one millimeter (which is no problem for the machine, mostly it comes down to myself getting the measurements correct). List of cons for the CNC: Making large boards is slow (the isolation milling for a lot of traces with wider isolation and/or lots of copper removal can take hours) Making lots of boards is slow (depending on size, a big batch of small boards can be done fast too, it all comes down to the total length of trace edges running through the boards, amount of different sized drills etc) When a copper clad is changed or you start a new job, you need to do the auto-leveling grid from scratch, which can take a long time for large boards with small traces/components (10-40 minutes usually, for very large board and/or small grid, this can take over an hour) Making 2-sided boards is a hassle, I generally just don't do those (although I will show examples of a couple of different ways how one can be done later on) FR1 (bakelite) and FR4 (fiberglass reinforced epoxy) dust is NOT healthy to breathe, get a vacuum with HEPA-filtering While the machine itself is relatively quiet (unless you put a 20k rpm motor on it, like I did at one point ), large drills and cutout will make noise, as well as vacuuming the dust/chips In my case, the CNC-router seemed a much better option vs. etching, but depending on your needs, etching might be a viable and/or better solution. I'll probably fill out the pros/cons further if/when other things come to mind, these were at the top of my head right now. With that out of the way, next posts: or how the hell do you get good boards out of that? Getting the milling gerbers and drilling excelons "right" from KiCAD (or just in general what to pay attention to, probably the "knowledge" is transformable to other EDAs) FlatCAM: machining paths for different types of isolation, drilling and cutouts, details and things to pay attention to (this is where you easily get it wrong) bCNC: The utmost importance of depth-probing and autoleveling to get high quality cuts; milling, drilling and cutout procedure More special cases: Slots, text engravings, V-grooves, 2-sided boards, making multiple copies of a single design in one go Edited September 1, 2018 by esaj 4 2 Quote Link to comment Share on other sites More sharing options...
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.