Making a Printed Circuit Board

1. Summary
2. Software
3. Transparencies
4. Photopolymer Plates
5. Cutting the PCB
6. Workspace
7. GC Pre-Sensitized Copper-Clad
   1. Exposure
   2. Developing
   3. Etching
   4. Results

Summary

Some useful tools and processes to help get you started easily and cheaply. The goal is a board with copper traces on it.

Return to Table of Contents

Schematic and PCB Software

Software Notes

Extensive links can be found at OpenCircuits.

Eagle is a great end-to-end solution for PCBs smaller than about 4x3". I dislike the schematic capture interface as being awkward. If you do a lot of fairly simple boards like these and don't need any fancy features, you might consider building a library of primitives with your favorite vector drawing package, as I have, to keep control over your process and data.

Autorouting tips for homebrew PCBs in Eagle can be found at Instructables. Short version: In board window, click on DRC icon in left toolbar (probably at bottom of right column) which brings up the design rule window. Select Clearance tab and change Different Signals entries for wire and pad, in particular, to some larger value such as the 0.8mm (31mil) value in the example. If board is double-sided, vias should be changed, if you use them. Select Sizes tab and changes minimum width and minimum drill entries to value similar to above. Select Restring tab and change the max and min values to bracket the above value for at least the top pads, and possibly the bottom as well as the outer regular vias. Select Shapes tab and change the top, and possibly the bottom pad to appropriate shape (perhaps round for through-hole components and square for SMDs.) A user comment mentions Drill-aid.ulp closes down the hole to a size that you specify in mm to steer drill bit better.

Better PCB checking was achieved with another PCB layout using ExpressSCH and ExpressPCB. (No longer supports Windows 98 in current version.) A couple of nice features are being able to link a schematic with the PCB file and have it automatically highlight each net so you can see what's supposed to be connected together in the layout. There is no autorouting or ratsnest, no export of bitmap schematic, yes to bill of materials. The downside is I had to screen-capture it--the bitmap export function in PCB produced only a black bmp file. I pieced together the captured screens in ACD FotoCanvas and converted it into a 256-color PNG using ACDSee 3.1. I then imported it into Draw 3.0 for scaling and tiling my small circuit boards, then exported the whole as a 2-color PNG at 600dpi.

I use Greenstreet Publisher 3 for The Bulletin newspaper at the Dickens' Christmas Fair. Print-to-file using a Postscript printer driver, and render as bitmap in GhostView and GhostScript.)

For these PCBs I used a vector file from Greenstreet Draw 3.0 exported as bitmap. It also directly supports EPS (Encapsulated Postscript will also open in GhostView and GhostScript.)

PCB123

PCB123 has net checking and integration between schematic and layout, also, but includes the powerful autorouting feature. (Version 2 does not support Windows 98.) Netlist goes to a .net file. BOM goes to a .prt file. Schematic and layout output to a Postscript print file both look very professional black & white in GhostView. Copies to clipboard will paste into a bitmap editing program at the visible size and colors.

PCB Transparencies

These are needed in any photographic technique, whether exposing a photosensitive resist applied to the copper, or exposing photopolymer to make a printing plate (for simple printing like silkscreens, or selectively applying resist to the copper.) The two requirements are:

• UV transparency >50%.
• High contrast.

Translucent Paper-- ~ $0.30 each, good UV transparency, high-contrast printing, but most products curl while exiting printer.

• UV/Ultra II Translucent Paper, manufactured by Kimberly-Clark, Neenah Paper Division, which will not curl exiting printer. Highly recommended.

  Failing that, go to your local office supply store and search for translucent in the name of papers. A paper that is functional:

• Foray Translucent Inkjet & Laser Vellum, 30lb, 8.5x11-in., which proved 70% permeable to UVA. The price is about US$10/50 sheets, and is about 0.004" thick and somewhat stiff with sizing. The black produced is very dense and the quality of the negatives quite good, but the paper will curl quite stiffly and must be firmly hand-held as it exits the inkjet printer until it is dry after a couple minutes. It will still curl terribly with a lot of inkjet ink, attempting to snap into a tight roll, but at least it won't smear when dry, ruining the print. It did not curl at all in my laser printer.

Plastic Transparencies-- $1 each, good UV transmission, but not high-contrast on inkjet printers.

Standard Vellum-- < $0.10 each, good for tracing and drafting, high-contrast printing, but many do not transmit enough UV to be useful. Test.

Inkjet Ink-- A significant cost can be that of the consumer inkjet ink. Ink for a negative print was $1 for 8.5x11" (used with photopolymer described below.) Ink for a positive print was $0.25 (used with GC pre-sensitised copper-clad board) will use only about $0.25 of ink. I have not yet checked the quality of the negative from a laser printer.

For artwork, you could try making a printing plate using water-washable, photopolymer sachets (used for flexographic printing or other rubber-stamp-type work) for laying-down the resist. JustRite is the reseller in the USA for the UK manufacturer, Polydiam, for their Imagepac product (some good reference material on the manufacturer's site.) Their patent covers a clever idea of putting the thick, clear, liquid photopolymer in a flat plastic sachet to avoid a lot of messy, expensive, and time-consuming preparation. The JustRite photopolymer sachet order numbers for the main stock sizes are (as of November 2008):

• PO-06690 Imagebox Starter Kit (selection of sachet sizes and related consumables)
• PO-06884 A4 CLEAR POLYMER, 10pk (approximately "letter-size" in USA terms)
• PO-06885 A5 CLEAR POLYMER, 20pk (approx. 1/2-letter-size)
• PO-06886 A6 CLEAR POLYMER, 10pk (approx. 1/4-letter-size)
• PO-06888 A7 CLEAR POLYMER, 20pk (approx. 1/8-letter-size)

You need an exposure frame to hold the artwork against the sachet and maintain an even thickness over the entire surface. I was fortunate enough to be given an 11x14 photographic exposure frame. The thick glass on the face cuts about 12% of the UVA, but acrylic will work fine. A double-sided, thick acrylic frame would allow both sides to be exposed without having to open the frame (JustRite sells these for small sachets.) I had to add another washer under the "latching" fender washers in the back to allow an extra .125-inch greater thickness than film to be placed within the frame. The thickness I needed was that of the final plate (.125-inch nominal, +.012 for current batch) plus the sachet top and bottom sheets (.003+.003 for A6 or A7; .005+.003 for A4 or A5) plus the negative (.003), or about .134 nominal. I made a stand-off insert frame cut from a single piece of 0.125" thick luan wood plus another 0.011" of heavy paper stock, which just fits the exposure frame.

I found their PO-06670 Imagebox Stamp Maker Machine a little pricey at $399, since it appears its main purpose is as an exposure unit. 18-inch, 15W UV fluorescent lamps (F15T8 BLB) make ideal exposure bulbs. The cheap, soft plastic fixtures can easily be opened and the ballast and lamp contacts repositioned on a board without the case, if a 2" lower-profile is desired and you have more time and skill than money. You can attach ballast circuit board it to your new mounting board with 5-minute epoxy or silicone rubber and insulate exposed metal leads with same (use judiciously so circuit board won't overheat). The lamp contacts often come with a single screw-hole for mounting on a block or the sides of your new exposure box, or you could cut away the portions of the old case which hold the lamp contacts and mount those instead. Several side-by-side in a white-painted mixing chamber with a frosted face-pane of glass or plastic which passes UV is ideal. New bulbs, I found, have 3X the output of my old one's output of 110 ft-cd at 2.75". For comparison, I measured sunlight UVA at only about 140 ft-cd. We want to use lamps because it is a more stable light source, and multiple lamps and/or a mixing chamber so large sachets need not be moved about to ensure even exposure. Unlike circuit board and photograph exposures, we actually want undercutting of the negative artwork by the UV to some degree so there will be a sloped shoulder to support the printing image on the face of our stamp--fine details will have shallow relief and be better-supported. If working with a single, exposed lamp, move the exposure about and turn it periodically so it mimics the effect of a stationary, wide UV exposure panel.

The back is exposed for 7.5 minutes without the negative to create a floor for the printing plate which is nearly three-quarters the thickness of the final plate. The sachet is then flipped and the negative is placed over it. The front exposure is about 5 minutes with the single lamp. If you have a caliper you can actually remove the sachet and measure the thickness that has hardened. The hardening proceeds from the side facing the UV source. Here is the current test results on a year-old photopolymer sachet stored at room temperature which appears to still work just fine (note this year's test needed about 4.5 minutes for a three-quarters floor):

Time(min)	Thickness(in)	Notes	Est. Uncompressed Plate	Delta(in)
0	.006	Sachet sheets only--no compression	.000	-
1	.038	Initial faster cure--caliper compression measurement	.037	.037
2	.058	Soft cure--caliper compression measurement	.061	.024
3	.073	Soft cure--caliper compression measurement	.078	.017
4	.086	Soft cure--caliper compression measurement	.094	.016
5	.101	Soft cure--caliper compression measurement	.111	.017
6	.113	Soft cure--caliper compression measurement	.125	.014
7	.123	Soft cure--caliper compression measurement	.137	.012
8	.123	Soft cure caliper compression=17%	.137	.000
25	.143	Hard cure--no compression +/-0.003 inch	.137	.000

The sachet is removed from the exposure frame and the top sheet is cut all the way around near the edge carefully with a hobby knife where the photopolymer has (hopefully) hardened to a firm, rubbery consistency. The top sheet is peeled away and discarded (it is messy with unhardened photopolymer). A pair of gloves should be used to avoid much contact with the liquid photopolymer. I use paintbrushes with medium-stiff bristles to help washout proceed quickly. The sachet is placed under water (running or in tray) not too much over body temperature with liberal amounts of dish soap. Then, work the brush tip quickly and fairly aggressively over the entire plate to lift the unhardened photopolymer from the soft, rubbery plate. Use plenty of soap on the brush end periodically to aid washout and help keep the brush clean. A final separate rinsing should be used to remove the soap and photopolymer residue. If the water is too cold, the photopolymer will wash-out too slowly, and if too warm, the plate with immediately develop a white film which is water being absorbed into the surface. This rapid absorption should to be avoided, though the plate will dry again if allowed to.

The washed-out plate is still too soft to use at this point and needs further hardening. The plate is placed in a tray with about a centimeter of tepid water and exposed a second time for about 5-min. on each side (not too time-critical). The plate is removed and allowed to dry. The texture is rubbery and quite sticky. The stickiness is ideal for clinging to smooth surfaces for making printing plates and rubber stamps (JustRite sells acrylic blocks for this. Go figure.) De-tack salts can be purchased which leave a salt coating which covers-up this natural stickiness, if so desired. Washing the plate again will remove the salt. The plate should be kept from too much light, air, and dust, when possible, to retain its stickiness and resiliency. A plastic bag, stored flat, is ideal (Yeah, yeah. JustRite has some, but any smooth plastic will do.) Plate will deteriorate if used with solvent-based inks, but will be quite useable for some time before it gets crumbly.

My 8.5x11 plate test plate varied no more than about .010-inch overall. A simple matrix was made using an 8.5x11-inch piece of quality plywood coated with a plastic varnish. The test plate was to be used for a page of my Dickens' Christmas Fair newspaper, The Bulletin, and included a mix of drawings and text. The natural tackiness of the plate allowed it to adhere nicely; over time, the bond increased, but was never too much to be comfortably peeled away. Printing on a letterpress gave clean results, providing at least 100dpi with fine lines fading completely by 150 per inch, which is about what might infer according to the manufacturer's description of their "finer" product, Imagepac Xtra, being able to hold 150 lines.

Non-Photographic Techniques

Other Techniques

Past experience with noxious and dangerous chemicals led me to immediately choose an environmentally benign approach. I once used the GC Electronics type D negative technique, but you must perform this outside with a light cross-breeze. Xylene and Xylol are volatile, poisonous teratogens, the slightest whiff of which can make you feel ill. The spray photo-resist and the developer must be disposed of as hazardous waste. It just wasn't worth it.

Toner Transfer

If you can get the process down, the transfer of laser/copier toner from a special sheet is good. Toner is basically just plastic and carbon, which makes an excellent resist. You need to get the copper sheet itself nice and hot to guarantee the toner will adhere to the copper--cold metal just doesn't do it. Another discovered that the non-stick sheets Avery labels come on makes a decent subsititute for the specially-coated toner sheets.

Plotter Printing

Another technique some were using involved using old HP plotters to directly draw the PCB artwork onto the copper. One fellow who tried many ink pens discovered the Staedtler Lumocolor 318-2 (the -2 means red) worked best. The line width is 0.8mm. The Lumocolor 313 (0.4mm) wore out too quickly for another person. Another fellow mixed his own ink from India ink and lacquer.

Printing on Copper

In principle, there are probably a couple laser printer models in the world which could handle a copper-clad sheet directly--ones which don't bend the board, or accept copper sheet stock which could subsequently be made into copper-clad board.

DIY Flex Circuits

Homemade flex circuits can be made by this clever technique with copper-clad polyamide (Kapton) film ("LF9120 has 1 oz Cu, 1 mil adhesive and 2 mil Kapton - seems to work best in the printer" he writes) printed with resist using solid ink (wax) printer such as the Tektronix or Xerox Phaser series.

Home-Made Copper-Cladding

It would be an interesting experiment to try and make your own, either using self- adhesive copper foils, metal tape, with components applied with conductive epoxy. Alternatively, try to find an adhesive which actually could withstand high-temperature soldering on narrow traces and apply it to a plain copper sheet after printing a resist upon the copper. It could be used with photopolymer plate printing, toner transfer, or plotter printing.

Copper Chloride Etchant

An interesting alternative etching solution using copper chloride, hydrochloric acid, and hydrogen peroxide is give at Instructables.

CNC PCB cutting

Using a CNC machine to directly cut the copper of the PCB. One small-scale commercial version used a teflon rider on the board to adjust the height of the cutter. It was a bit like a tiny plotter on steroids. This process can also cut multiple boards apart.

I have access to a 19-inch industrial paper cutter for cutting stacks of paper which I use for cutting the circuit board. I prefer the phenolic substrates over the fibreglass boards as they can be scored and snapped when a cutter is not available, and are less abrasive to a cutting blade than fibreglass. I'm not yet sure when the best time to cut is. Cutting before exposure can do some slight damage to the resist and can expose your board, however slightly, through the back. Cutting after etching risks the finished board. Cutting after exposure but before etching is not a good idea as you will risk damage to the resist and further weaken it through exposure. I lifted some traces on an etched board with the sloppy use of a scroll saw. Scoring and snapping works well, but you must allow for a bit more variation in the edge.

Space-- Requirements are minimal for small runs and scale proportionally. I need only a 1x2-foot square table space for small boards. Ventilation is poor in my work area, so a low vapor process is a must for me.

UV Source-- A UV spiral black light lamp was used in these test (with cardboard shade in picture), but it is a weak source. A long fluorescent black light has a good output, but there is noticeable undercutting of the negative. I much prefer sunlight when I can because the exposures are short and it is effectively a point source.

Exposure frame-- A simple contact printing frame can be used. You can also use a simple sheet of glass or acrylic plastic to press the transparency artwork against the board

Trays-- You will need 4: Developer, developer rinse, etchant, and etchant rinse.

My favorite is the Rubbermaid "Servin' Saver" container which is about 6x3x1.5 inches. There is also a polyethylene container of similar dimensions I purchased through Tap Plastics which is made as a molding frame. I also have some 1-quart flat-sided plastic milk containers with one side cut out and the lid screwed on. If heating a solution from outside of a plastic tray, remember that most plastics soften in boiling water so ensure your heat source doesn't produce more than this unless your container can take it. I recommend using polyethylene, or acrylic(?) containers. Polystyrene and polypropylene just melt at too low a temperature--enough to cause the bottom plastic to begin to stick slightly to the hot plate.

Heater for etching solution-- Chemical reaction speed up by roughly 2X / 10°C. The picture shows a serious hot plate for soldering work warming the etchant. I have since built a new heater from a surplus three-cup coffee warmer element to replace the hot plate in the picture. I cut the power in half by installing a power diode in series with the heating element and routed a frame from a piece of wood 6x10x3/4". The temperature of liquids on it is typically about 50°C.

Solution preparation-- To avoid wasting chemicals and make their preparation easier, the practical minimum quantity of liquid I have selected is 50cc for small circuit boards. I have a 2 oz. bottle handy for this amount. This is sufficient to get some wave action over the board in the etchant as you rock the container. For measuring convenience for the sodium persulfate, I made a measuring spoon from a bottle cap with a capacity of 11 ml which will sufficiently measure the 12.5 g of granular sodium persulfate I need to mix with 50cc of water.

The exposure can be accomplished by one of the newer 15W black-light spiral-tube fluorescent lamps with the screw-base (perhaps even a standard non-black-light one), which I've socketed pointing downward. I've placed an eye-shield in front of the lamp using some cardboard. Expect exposure times of around an hour at 12-inches, however, so you can work for several minutes in normal room lighting without affecting the board significantly. If your artwork is not directly against the copper (even to the side of the sheet it is printed on) you can get significant undercutting of thin traces. My preference is 30 seconds in bright sunlight. It's a narrower source, too, so there's less undercutting worries. The artwork and board are placed in a proof-printer frame which has support foam and a cover-glass. I understand an acrylic sheet of plastic will also work to help hold the artwork against the board. Typical full sun is in the range of 6000-10,000 foot-candles. I happened to have a nice old analog lightmeter with a foot-candle scale and intensity filters which solves the exposure-time variable nicely. I can now characterize the board exposure times meaningfully. Outdoor sunlight reads about 10,000 ft-cd here today--a bright spring day in San Francisco in early afternoon with the sun at maybe 60° from the closest horizon. We will use 10k ft-cd as our normalized standard intensity. Indoor lights show about 80 ft-cd at one foot (end view of spiral fluorescent). This bears out the 120X exposure time for indoor vs. outdoor and suggests the UV spiral fluorescent lamp is indeed putting out approximately the wavelengths the board needs. Later in the afternoon the sunlight was about 5800 ft-cd. A 30-second exposure (normalized) needed to be 30 * 10000 / 5800, or about 52 seconds. I found that the minimum 10k ft-cd normalized exposure time was about 20 seconds. Development time was rather slow at about one minute. The resist was best with this minimum exposure time. 30 seconds at 10k ft-cd was pretty good. There was some intrusion through the resist leaving it pocked but solid. 40 seconds at 10k ft-cd was definitely too much with copper noticeable through the resist. I don't know how much energy is lost in my overhead transparency artwork or in the cover glass for the proof-printing frame, but I will use 25 seconds at 10k ft-cd next time.

Getting the board in, out, and around these solutions is the job of the "sparge", some sort of dipping-frame in its best form, but a pair of etchant- and developer-resistant gloves work in a pinch, or a large pair of plastic tweezers. A foam brush is an excellent substitute for the tiny loose sponges which are often provided for this technique.

The GC developer contains sodium hydroxide (NaOH) at 4-10% concentration (per the label), rather caustic and supposedly touchy concerning temperature and concentration. I don't seem to be having problems, probably because developing is done manually with visual feedback as to when it is complete. An open tray of NaOH solution is only good for one day as it will react with atmospheric carbon dioxide (C0₂) and neutralize into a solution of sodium bicarbonate (NaHC0₃).

But, one person in the UK strongly recommends using sodium metasilicate pentahydrate (Na₂SiO₃*5H₂O.) I simply diluted the MG developer 10:1 (1% by weight) as per instructions and used it at room temperature (about 21°C). (One could mix 9g NaOH (solid) per liter of water or 0.4g NaOH (solid) per 50ml of water to get an equivalent usable solution. For convenience in dispensing measured amounts of the GC NaOH developing solution, I put some into a 20ml polyethylene dropper bottle on which I have marked 5ml graduations. For pre-mixing the developing solution, I use some 100ml polyethylene dropper bottles for plain water graduated in 50ml increments for .

Brush gently with a small sponge until the unwanted resist washes away. Remove and rinse quickly before it starts dissolving the resist which has not been weakened by the UV, as well. It's a good idea to have a rinse or stop-bath for each step, in this case one for the developer and one for the etch solutions. Do not get them mixed to avoid ruining any of your solutions. Be careful when adding more NaOH while the board is in the tray or you will find your resist suddenly gone in one area and your board ruined! Mix first! You can afford to waste the small quantities used here a lot more than your PCB. Naturally, appropriate protective gear is in order--goggles, gloves, et al, as appropriate.

You could touch-up damaged resist at this stage, but recognize simple damage to avoid wasting time on a flawed developing process.

For the etchant, sodium persulfate is far superior to ferric chloride, though more expensive.

• Fast etch-- 8 min. @ 45°C. (Reactions roughly double their rates for every 10°C increase in temperature.)
• Cleaner-- No ferric chloride brownish-yellow stains (try 10:1 oxalic acid)
• Colorless when fresh allowing you to see your work and gradually deepening in blue when exhausted.
• No fumes when heated the way ferric chloride (at 57°C) and many acid etchants do.

The standard solution is 250g solid sodium persulfate with water to make 1liter of etching solution. Gently rocking the board (or bubbling with an aquarium pump) rather than any violent vibration, may improve the detail of the etch. I have found the sodium persulfate will etch about one square-inch per gram of solid etchant. When the solution is exhausted, a precipitate will start forming. Discard immediately and use fresh solution as the crystalline sludge (copper II sulfate) will damage the resist and the longer time needed creates uneven etching. The bare board doesn't start becoming evident until the last minute or so, the etch being very even.

Board 1-- Initial exposure testing was done with negative artwork for convenience, and the results were superb. I had 0.010-inch (0.24mm) lines in the resulting test board (those which would be spaces between pads and pass-between traces on the positive artwork. (I suggest you cut a number of small unexposed boards no larger than a square-inch if you need to do some initial exposure testing.) Note a couple of pinholes in the resist. It would be a good idea to touch these up with a resist pen after development and before etching if you see any bare copper where you don't want it.

This board was exposed by lamp for 90 minutes (a bit long) and etched in 10 minutes. I recommend discarding the solution when etch time has doubled. The quality of the etch appeared so unexpectedly fine I would seriously consider attempting to put them between 0.05-inch-spaced pads. During the etching process, tiny bubbles will be seen to form on the resist. I believe this may be due to very small pinholes in the resist over the entire surface as I have noticed the general degradation and pitting of the copper under the resist under magnification when the board is exposed and/or developed a bit too long. The resist is a very thin, roller-applied, green layer. Essentially, the resist may or may not look thin, but the intrusion occurs fairly evenly over the area of resist.

Board 2-- The artwork was printed on UV/UltraII Translucent Printing Paper from Kimberly-Clark Corp. I had obtained it in the hopes of avoiding the puddling effect inkjet inks can have with various transparency media, and the requisite need for too-thin ink coatings (they have been improving over the years.) These were printed on a black and white copier directly from a computer file. The print looked very sharp, in spite of using finer traces. The exposure time was clear sunlight about 20-25° above the horizon in a glass proof-printing frame for 20 seconds. The developing took longer than before at perhaps two minutes--the resist seemed to dissolve slowly. This was insufficient exposure; it should have been a good 30 sec with clear sun at least 30° above the horizon (about 10000 ft-cd.) The etching time is nine minutes. The IC pads shown in the vertical row are .050-inch pitch. The pads are nominally .0275-inch, allowing .0235-inches between for the trace. The trace passing between the pads is .0084-inch (5 pixels) with .0084-inch above and .0067-inch (4 pixels) below. There is a hairline bridge of copper between the trace and the upper pad which would doubtless go with a little more etching time. At 600dpi, my pixels are .0017-inch. There was a one-pixel line around the PCB which showed breaks in the resist but not in the copper. The width was averaging close to .004-inch--another indication of slight under-etching. My best guess is I can expect typical variations of about ±.002-inch. One could possibly manage .004-inch features with care. That would be enough (barely) to get a trace between .025-inch pitch ICs. Whew!

Board 3-- These three pictures show another PCB exposed in 10,000 ft-cd sunlight for 30sec at 60X magnification. It was in the developer about a minute, but could have been shorter as I didn't have a sparge or good coordination with the sponge. This reflects my suspicion that the 20sec solar exposure weakened the exposed areas somewhat less to the action of the developer. The etch was 8 min at 46°C--complete in 7 min., the final minute being to test for overetching effects.

The first picture (pad, trace, border) shows the one-pixel reference line near board edge at 60X and the overetching of the rectangular land nearby. Print density was not ideal, but better on the UV paper than it is on a plastic transparency. You need to err on the side of underexposure with a low-density etching mask.

The second picture (trace, border) shows a bridge in a four-pixel space. The artwork under magnification does show a slight overspray of toner--a light dusting of about 0.003-inch (2 pixels). Normal variations in the print show the one-pixel line varying by a factor of two in width, so a lot of what you see here is normal 1200 DPI digital re-rendering noise from the laser copier/printer.

The third picture (2 pads, trace) shows another bridge, the same as the test board shown above due to a speck on the artwork. Another factor besides overspray in effect here is the 1200dpi rendering of the 600dpi artwork--there are very slight but visible differences in spacing in different copies of the PCB on the same artwork. So, the board would support .002-inch features, but the artwork print might not. Judging from the 5-pixel space below the 5-pixel trace in the third picture, it appears there is approximately .0023-inch extra around all copper lands.

Board 4-- (No picture.) The PCB artwork was shot down to half-size from a laser printer onto the UV paper and had some processing difficulties at .006-inches, but the whole process was a little sloppier than my usual efforts. All this suggests that anything below about .006-inch requires some extra care.