For all of my projects so far, I have used vero-board cut to the right size. But this time around, the circuit used some RJ-11 sockets which don't fit into the normal 0.1" veroboard matrix. Besides, I thought it was time I tried my hand at making PCBs. A recent thread on the PIC mailing list provided some additional information. The circuit was for a DCC Turnout Decoder. I had a deadline of mid-Feb 97 which was one and a half months away.
The process of PCB development is highly dependent on the characteristics of the material at hand - this page describes the process using materials obtained in Australia.
The PCB pattern was drawn using TurboCAD which is a cheap, general purpose CAD package. A PCB layout package may have been a better option but I was accustomed to the CAD package. I used the layers feature to put the donuts, the holes, the tracks and the component legend on different layers. This allows me to print just the PCB pattern, the components and holes or any other combination. Any component being used was first added to a library and then used.
The pattern is drawn looking from the component side of the board. So when it's printed out, the toner side can be put against the PCB. I use a font called PRMirror which is a mirror image font. I also have to write the text backwards to achieve the final effect. The pattern was printed on a laser printer(HP Laserjet 5MP) on overhead transparency. The printer is fairly new and the printed pattern had no discernible gaps or other problems.
I bought pre-coated positive resist boards from R.S. Electronics. The stuff they sell is good quality in my opinion and quite affordable in the opinion of the Sultan of Brunei. They also supply data sheets to go with it. I would expect the product to be uniform across countries and across batches. The boards cost AU$ 18 and are 100mm x 220mm x 1.6mm(single Eurocard size). You can also get double sided and double Eurocard size PCBs. If you need thinner PCBs, Digikey has 1/32" PCB material.
I built my own UV Box out of particle board. I put a single UV tube in it along with the ballast and starter that go with a 40W fluorescent fitting. The tube is a UV-C tube and is not particularly suitable for PCB development but it can double as an EPROM eraser. The tube type is Philips TUV 15W G15T8 UV-C. It cost AU$ 22. It is slightly shorter than a standard 40W tube. The box has a tight-fitting lid with a glass sheet above the tube. It is roughly 15cmx15cmx50cm. I haven't yet put a microswitch on the lid to shut off the tube when the lid is opened or a timer. The inside surface of the box was left unfinished.
The optimum exposure was found to be 1 hour 15 minutes. This, in hindsight, was my biggest stumbling block. All resources I had access to suggested values of a few minutes. There are a few ways of decreasing exposure time - use 2 UV tubes in the box, use a more appropriate wavelength or paint the inside of the box white or silver. If putting the board in the developer for a while just turns it purple instead of some of the copper on the board being exposed then the exposure time is too short.
I used sodium hydroxide which is available as Digger's caustic soda in most super-markets. The concentration of the developer was not as critical. I used about 1 teaspoon of sodium hydroxide in a glass of water(200 ml). I did try sodium carbonate(washing soda) initially but the other parameters were way off to judge its effectiveness.
I used ammonium persulphate dissolved in hot water boiled in a kettle. The board was small enough for the etching to complete before the solution cooled down too much. The proportion is 1 part etchant to 5 parts water. The etchant was bought from Dick Smith and costs AU$ 9.95 for 600 gms.
I used HSS drill bits bought from Jaycar for AU$ 1.75. The hole size need not be very accurate as long as it is slightly oversize. However, holes for connectors are quite critical - an oversize hole may put excessive strain on the soldered connection. I measured most leads with a micrometer. Here are the measured sizes, suggested hole sizes and the ones I actually used.
My electric drill could not take bits of 1mm or smaller. If you are buying one, see that the jaws of the chuck can grip finer drills. I had to borrow one from my friend(Thanks Sands). I put it in a drill stand. Drilling the holes was the part that required some skill and I had trouble centering the bit. The holes in the middle of the donut are indispensable. I also gripped the bit in the chuck so that a very small part of it actually sticks out - about 5 mm. I could see that the drill flexed a little to get to the centre of the donut. The PCB was glass epoxy but the drill bit did not seem to have a problem getting through. I did try to find tungsten carbide bits but they cost around $10 at R.S. Electronics and you got to buy five at a time. They are brittle and breakages could be a little expensive. R.S. also has normal HSS drill bits with a common shank diameter of 2.4 mm. A set of 7 bits cost about AU$22.
Having designed the board, I went out and bought a couple of blank PCBs. Full of optimism, I tried to etch the entire PCB right away and failed miserably. Twice. Then I bought a few more and followed Steve Hardy's(Thanks Steve) more scientific approach described below.
I started off by cutting a test strip from the board, in my case 10cm x 2 cm, with a jigsaw with a fine blade. I also created a test pattern repeated four times, each of which consisted of a few donuts, lines of 5, 10, 15, 20, 25 mil widths. The four squares were exposed to the UV source for 2 mins, 4 mins, 8 mins, 16 mins. This is easily done by covering all but the last square and exposing it for 8 mins, covering only two squares and exposing it for 4 mins, covering the first square and exposing it for 2 mins and exposing the whole board for 2 mins. Four developer solutions at concentrations of D, 2D, 4D and 6D are prepared. Several initial efforts were complete failures and the deadline grew closer and closer. The board would turn purple but the resist was never removed to the extent of exposing the copper. Higher concentrations of developer did not seem to make an impression. Finally I tried exposures of 15 mins, 30 mins, 45 mins and 1 hour. Even at the lowest developer concentration this resulted in a fairly well developed board. This was then etched. Watch out for the red logo on one side of the board. This can be seen from the other side during etching and looks like unetched copper. I overetched one of my boards because of it.
After etching, patch exposed for one hour was perfect - the donuts and the holes were sharply defined, the 5 mil line was even and did not have any breaks, The 45 minute patch had some unetched copper left. It also left behind small, nearly transparent patches which were not conductive. This may be the glue that bonds the copper to the board. The 30 minute patch was largely unetched. A grubby fingerprint could be seen in copper. The 15 minute patch was completely unetched - the test pattern was faintly visible in the resist.
After the successful test, I did the entire board using an exposure time of 1 hour 15 minutes. It had more than one circuit on it so as to use up the entire board. It turned out perfect. I cut it up with a jigsaw and drilled it and voila! I had my first home made PCB. Since then, I have assembled the board and tested it. It works - the PCB that is, the software still needs work. Did I make the deadline? Not really. Before I could finish assembling and testing the PCB, my wife gave birth to our first son. He was born on 10th Feb and weighed 8 lbs. We called him Rishik. My efforts with the circuit have continued albeit at a much slower place.
Five years after my first PCB, I decided to make one for the updated version of the DCC Command Station. R S Electronics had meanwhile replaced their PCB with a newer version. The only change they suggested was a lower concentration of developer. Neverthless, I thought most of the old settings will still apply and made the PCB straight away. The developer dissolved all the resist - exposed and unexposed. Bugger. Back to the drawing board.
I reduced the developer concentration to about 1/4 teaspoon in 200ml. I used a test strip with exposures of 20, 30, 40, 50 minutes. The board came out OK with not much difference between the different exposures. The 5, 10 and in one case the 15 mil vertical lines were gone. All the horizontal lines were perfect. This is probably because the toner cartridge is not brand new and laser printers are not very good at printing vertical lines. Holes of 25 mils dia in donuts were not very clear. The 30 and 40 mil holes were fine.
I used an exposure of 40 minutes for the final board and used an aspect so that all 15 mil lines were horizontal. The board came out fine except for small patches of copper at one edge. Even fine text came out OK. The following are the parameters in the PCB design.
This PCB was for the IR Analyser. This time, instead of the laser printer, I used an Epson Photo810 inkjet with Celcast inkjet transparency(Part no. IJ25). Unlike the laser printer, the printout was slightly out of scale. I set the scale to 1:1.025 to compensate. The PCB material was the same as before. I exposed the PCB for 40 minutes. I used a developer(NaOH) concentration of 1/4 teaspoon in 200ml. The board developed cleanly. I used etchant(Ammonium Persulphate) concentration of 3 teaspoons in 200ml boiling water.
The results were satisfactory. All tracks and holes within donuts were clear. The fine text did not do as well with some unetched copper left. The thinnest lines were 15 mils and all of them turned out OK. There were small patches of copper at the edges. There were none at newly cut edges. I suspect the pressure during storage at the edges results in the resist hardening and subsequently, it does not dissolve in the developer.