The rat's nest of wires taunts me as I poke the soldering iron deep into the mess, trying to add a connection. It teases me as I try to squeeze it into the project box with loose connections that I then have to dig through and find, painstakingly. It confuses me with bad solder joints.
But I am strong. I learn how to use the Eagle PCB (printed circuit board) editor, with its symbols, devices, packages, net-lists and other arcana. I untangle the rat's nest line by line, learning to love the 'ping' noise when I make a connection. I have a piece of artwork on the screen by the end. Ready to go. Now, how do I make the PCB real?
But I am strong. I learn how to use the Eagle PCB (printed circuit board) editor, with its symbols, devices, packages, net-lists and other arcana. I untangle the rat's nest line by line, learning to love the 'ping' noise when I make a connection. I have a piece of artwork on the screen by the end. Ready to go. Now, how do I make the PCB real?
*DRAFT*
I need speed and low cost but I don't have volume to offer a PCB vendor. Even if I were willing to pay the typical few hundred dollars in setup fee, I would have to wait several days for shipping. The time penalty for design errors would force me to 'get it out the door' - stressful and sometimes counterproductive.
So, despite my reservations, I've started developing an in-house circuit board fabrication technique for very early prototyping. It is not meant to replace a PCB house (of which there are many very good quality), but rather be used in very early stages - ideally to replace the 'rats nest' stage. It is also not a hobby-style process which tries to minimize overall cost. Rather, I am looking for a fast, reliable process that I can use on a regular basis. No exotic, hazardous or difficult-to-obtain materials or steps requiring a lot of labor.
I need speed and low cost but I don't have volume to offer a PCB vendor. Even if I were willing to pay the typical few hundred dollars in setup fee, I would have to wait several days for shipping. The time penalty for design errors would force me to 'get it out the door' - stressful and sometimes counterproductive.
So, despite my reservations, I've started developing an in-house circuit board fabrication technique for very early prototyping. It is not meant to replace a PCB house (of which there are many very good quality), but rather be used in very early stages - ideally to replace the 'rats nest' stage. It is also not a hobby-style process which tries to minimize overall cost. Rather, I am looking for a fast, reliable process that I can use on a regular basis. No exotic, hazardous or difficult-to-obtain materials or steps requiring a lot of labor.
The process I have developed takes an afternoon for board of moderate complexity (two layers, 30-40 vias, 30-40 discrete parts mostly SMT). Here is a quick overview:
Overall, the process takes an afternoon if nothing goes wrong - but can take a few days if you get stuck on a problem (like I did LINK). The most unwieldy part is soldering and testing the vias. A mistake here can cause mind-bending headaches later, and the wires sticking out of the board is one of the biggest differences between this board and a production board. Other steps, like mounting and testing components are part of the PCB development itself - you find your errors during these steps. Some people will try to make extremely fine details - but I feel that this is better left for a later step, when you are working with your PCB vendor and finalizing the design.
Following are some more details and illustration of a few of the steps. It's taken a while to work out the details, but they do make the process faster and more efficient. And of course, if anyone has comments/suggestions/improvements, please do share!
1) Design with in-house production in mind - because vias and holes are so costly in terms of labor, use surface mount components, try to use as few vias as possible, and space them widely. Use thick traces and look for trouble areas near the edge of the board, or crowded areas where faults could occur. Use via pads that are large enough to tolerate error in the drilling process. I also set the ground pour to have a generous isolation spacing, as shorting to ground is one of the main sources of faults. SparkFun has a good PCB design guide. These design principles will also help with production boards.
I also have an Eagle script that exports all of the layers to PNG files that can then be loaded by a graphical editor. I use Adobe InDesign (multiple pages) and 'place' the PNGs on a couple of 8.5x11 pages. The links can then be automatically refreshed if you change the design. There are a few things I put on the pages:
2) Cut big and trim later - Cutting the PCB with about a quarter-inch margin around the edges helps greatly when I tape the transfer paper in place. In a later step I trim the margins back precisely.
3) Light the drilling area - I use a small LED lamp shining directly on the drill area, which creates a shadow telling me precisely where the bit will land. I also don't worry about the precision too much - if you get the hole within the via pad it is usually sufficient. The bigger offenders are large holes (>1mm) where the bit can wander.
I use complete sheets of 8.5x11 stickers so I can run it directly through the printer, and remove the sticker with acetone. The Avery is not much more expensive than the Staples brand. I've also tried the transparent vinyl shelf liner but I had to cut it out (a time consuming step), and it did not stick as firmly.
4) Find the right transfer medium - after trying several types of papers:
5) Minimize the prep - for a while I was trying to use tarnish remover and scrubbing the board a lot. But after some testing, it seems that a short scrub and a wipe with acetone is enough. Wear gloves to keep the board clean.
6) Use a lightbox for alignment - for a while I was using a flashlight, but it pays to be able to lay the pieces down flat.
I use a Sharpie to fill in the pads around the vias, and any missing spots.
7) Etch in a ziploc bag. There are many choices of etching chemistry - but for now I stick with small quantities of ferric chloride in a ziplock bag. This minimizes the amount of chemistry I need, as well as keeping the mess down. I simply put the ziplock bag into a container bag when I am done, and re-use it. I agitate using a roller - it seems to take about 10 minutes. I've added a small amount of acetic acid to my etch chemstry, but it doesn't seem to matter for such small amounts.
8) Placement marks and annotations are not critical - they can be aligned by eye, and as long as you can make them out, it is ok.
9) Check for faults before soldering vias! I visually inspect the traces, and anywhere traces run close by I use my multimeter in continuity test mode to check for shorts or gaps. It is worth the extra time to get a good transfer, as a fault can take minutes to find and correct.
10) Solder the vias using copper wire - this step is most time consuming, and requires some care to avoid shorts and gaps
11) Print a bill of materials, and keep track of which parts you've already soldered
12) For ICs I follow a procedure to replicate a reflow oven profile
Before plugging the board into power, do a final continuity test on power and ground - there have been many times when I've gotten impatient and skipped this step, only to damage some critical component. If possible use a lab power supply with a current limit for early testing, and set at a modest current (100mA) - a telltale 'click' will tell you something is shorted.
- Design and lay out the board using a computer-aided design program like Eagle. CAD programs help you create the schematic, then lay out the board using an editor which tells you which pins to connect to which, and will also check for clearances, etc. I use the Freemium version, which mainly limits the size of the board and the number of layers (two). When I'm ready for a four or more layer board, I will probably upgrade for the $1500 or so. I have a couple of tricks to lay the board out in Illustrator to make it easy for later steps. Time: hours to days.
- Cut the raw board out. I use an inexpensive paper cutter from the local office supply store. It is probably also possible to use a knife and straightedge, but this is fast and less hazardous. I cut it larger than I need - this is important for attaching the transfer design to the board securely in a later step. Time: 5 minutes
- Drill the holes for the vias which connect the top layer to the bottom layer. I print the pattern on a piece of label paper. I use a Dremel in a small drill press and cheap high speed steel bits from an online source, and sand the board down afterward with a small piece of fine grit sandpaper. It may be possible to get cleaner holes if you spend on tungsten carbide bits but I have not seen any need for this, nor any problems with breakage. For about 30-40 vias, this takes me 20-30 mins
- Print the design using a laser printer on a transfer medium. People use a number of transfer mediums, including plain paper, glossy paper, parchment paper, special water release papers. I've found my special paper which has very high resolution, comes off cleanly, and I can buy at Home Depot for a few bucks a roll. This is where a little bit of optimization can help speed things up a lot. Time: 5 minutes
- Prep the board. I scrub both sides of the board with a Scotch-brite sponge and soapy water, and wipe down with some acetone on cotton pads to remove any oils. I handle the board with gloves from this point on to reduce fingerprints. Time: 5 minutes
- Transfer the design to the board. In this case, I use a laminator which can create the high temperatures and pressures for a good transfer - this is a critical step. It takes longest to get alignment right, but a few tools here like a light-box can speed things up. Time: 15 minutes (but early on you may have to repeat a few times to get it right)
- Etch. Lots of options here but I use a simple etch chemistry I bought online and a simple 'in the ziploc bag' etch. Time: 10 minutes.
- Transfer labels/annotations to the board - part names like 'R1', 'C1', placement information - all of these will help when mounting components. I use the direct transfer process again here. Time: 20 minutes.
- Check for faults! There is a method to this madness - and this can be time consuming as you begin to really learn your design for the first time. Time: 10-60 minutes or more
- Solder the vias with wires. This is probably the most time consuming and labor intensive step, and I am looking to improve this significantly. A good chunk of this is checking for faults in the via solder joints. For a board with 30-40 vias, it takes me about an hour
- Mount passive components. I smear solder paste compound on the pads, then place components using tweezers. I re-testing for faults once the passives are in place. For a board with 30-40 components, it takes me about an hour
- Mount and test ICs - I handle these more carefully as it is harder to debug once they are in place. If I am feeling especially paranoid I will use a thermocouple to monitor the temperature.
Overall, the process takes an afternoon if nothing goes wrong - but can take a few days if you get stuck on a problem (like I did LINK). The most unwieldy part is soldering and testing the vias. A mistake here can cause mind-bending headaches later, and the wires sticking out of the board is one of the biggest differences between this board and a production board. Other steps, like mounting and testing components are part of the PCB development itself - you find your errors during these steps. Some people will try to make extremely fine details - but I feel that this is better left for a later step, when you are working with your PCB vendor and finalizing the design.
Following are some more details and illustration of a few of the steps. It's taken a while to work out the details, but they do make the process faster and more efficient. And of course, if anyone has comments/suggestions/improvements, please do share!
1) Design with in-house production in mind - because vias and holes are so costly in terms of labor, use surface mount components, try to use as few vias as possible, and space them widely. Use thick traces and look for trouble areas near the edge of the board, or crowded areas where faults could occur. Use via pads that are large enough to tolerate error in the drilling process. I also set the ground pour to have a generous isolation spacing, as shorting to ground is one of the main sources of faults. SparkFun has a good PCB design guide. These design principles will also help with production boards.
I also have an Eagle script that exports all of the layers to PNG files that can then be loaded by a graphical editor. I use Adobe InDesign (multiple pages) and 'place' the PNGs on a couple of 8.5x11 pages. The links can then be automatically refreshed if you change the design. There are a few things I put on the pages:
- Top and bottom layer copper - the top and bottom are printed in a row symmetric across the centerline of the sheet, so that any distortions in the printer are mirrored. Don't forget to mirror the top layer image.
- Top and bottom layer silkscreen, in the same arrangement
- Drill pattern and legend - right side up so it can be printed on a label and also for reference.
- Top and bottom layer copper - unmirrored so you can check the traces visually
- Top and bottom layer silkscreen - un-mirrored so you can see where everything is placed
2) Cut big and trim later - Cutting the PCB with about a quarter-inch margin around the edges helps greatly when I tape the transfer paper in place. In a later step I trim the margins back precisely.
3) Light the drilling area - I use a small LED lamp shining directly on the drill area, which creates a shadow telling me precisely where the bit will land. I also don't worry about the precision too much - if you get the hole within the via pad it is usually sufficient. The bigger offenders are large holes (>1mm) where the bit can wander.
I use complete sheets of 8.5x11 stickers so I can run it directly through the printer, and remove the sticker with acetone. The Avery is not much more expensive than the Staples brand. I've also tried the transparent vinyl shelf liner but I had to cut it out (a time consuming step), and it did not stick as firmly.
4) Find the right transfer medium - after trying several types of papers:
- Vinyl shelving paper backing. This is my go-to choice. It is transparent, smooth, and does not stick to the toner once it is released. Disadvantage - it has to cut down to size, peeled from the vinyl, and taped in place on a piece of 8.5x11 printer paper to feed through the printer. It is also very inexpensive - so I don't waste time trying to conserve paper, and feel comfortable throwing out mistakes.
- Glossy 8.5x11 paper - I wanted to be able to just put a sheet of paper in the printer. Very inexpensive, but it is not as transparent so more difficult to align, did not transfer as cleanly, and left fibers behind, which dissolve in the acidic solution but also caused the toner to come off in certain areas.
- Wax paper - transfers well but the backing falls apart in water
- Parchment paper - holds up better than wax paper, but the rough surface makes for poorer, less dense transfers
- Dextrin-coated paper - absolutely great for releasing, but is thicker and expensive - and I end up cutting down the 8.5x11 sheets to conserve paper anyways. It is only available from one source as far as I can see.
5) Minimize the prep - for a while I was trying to use tarnish remover and scrubbing the board a lot. But after some testing, it seems that a short scrub and a wipe with acetone is enough. Wear gloves to keep the board clean.
6) Use a lightbox for alignment - for a while I was using a flashlight, but it pays to be able to lay the pieces down flat.
- Being careful to keep the board clean, I lay the board atop the pattern, either top-side or bottom-side. I look for the pattern through the drill holes.
- I align the majority of the via holes. They will not be perfect because they are hand-drilled, but as long as they are within the pads it is fine.
- I tape the board to the pattern with a piece of Scotch tape. I've also used high-temperature Kapton tape, which will leave less of a residue, but is more expensive and harder to handle with gloves on. I tape within the outer margins of the board and check that the transfer paper is held firmly against the board.
- Fold the transfer paper so that the other side of the transfer is over the board. Lining up the holes, tape the transfer pattern in place. I trim the transfer paper so that the tape has a solid area to hold onto. Again, the important part is that the transfer paper is held firmly against the board.
I use a Sharpie to fill in the pads around the vias, and any missing spots.
7) Etch in a ziploc bag. There are many choices of etching chemistry - but for now I stick with small quantities of ferric chloride in a ziplock bag. This minimizes the amount of chemistry I need, as well as keeping the mess down. I simply put the ziplock bag into a container bag when I am done, and re-use it. I agitate using a roller - it seems to take about 10 minutes. I've added a small amount of acetic acid to my etch chemstry, but it doesn't seem to matter for such small amounts.
8) Placement marks and annotations are not critical - they can be aligned by eye, and as long as you can make them out, it is ok.
9) Check for faults before soldering vias! I visually inspect the traces, and anywhere traces run close by I use my multimeter in continuity test mode to check for shorts or gaps. It is worth the extra time to get a good transfer, as a fault can take minutes to find and correct.
10) Solder the vias using copper wire - this step is most time consuming, and requires some care to avoid shorts and gaps
- I use wires clipped from bundled wire, about 1 inch long, and match the via size so that they fit snugly
- Dip the end of each wire in solder paste
- Put the wire through the via and bend out of the way, this will also help it stay in place while soldering
- Flip the board over and using a small pin apply solder paste to the other sides
- Add extra solder paste top ground vias (without a small pad)
- Reflow with a hot air gun - watch that solder flows up the side of the wires
- Check for continuity, being especially careful with ground vias, where the solder can form a doughnut shape around the wire without actually creating contact
- Clip the wires short with a wire cutter
11) Print a bill of materials, and keep track of which parts you've already soldered
12) For ICs I follow a procedure to replicate a reflow oven profile
- Check the pads for continuity to power and ground, shorts between pads, and that the passive values are correct.
- Melt a thin layer of solder paste over the pads, removing excess with a solder wick (too much solder will cause bridging)
- Apply a dab of flux to clean the chip and help it stick
- Get the chip from the packaging - I keep mine in ziplock bags with desiccant to help with moisture
- Place the chip, checking the orientation
- Use a 'helping hand' to suspend a thermocouple right above, barely touching the chip. If it pushes down too much the chip may slip.
- Start a timer. Hold the hot air gun (set at 260C) about 3 inches away and wave over the chip area so that it reaches about 100C; hold for 1 minute
- Bring the gun closer, about 2 inches away and concentrate the flow over the chip. The temperature should rise quickly to ~150C. Watch as the flux boils, then the chip should shift slightly as the solder melts, probably within thirty seconds; the goal is to keep temperatures below 215C and time below 1 minute
- If necessary, give a gentle push with tweezers to ensure centering while holding the hot air over the chip.
- Remove the hot air, let the chip cool
- Test power and ground connections for shorts and the correct impedance. Inspect the pads for solder bridges. If you find any, run a soldering iron (at 300-350C) quickly over the pins.
Before plugging the board into power, do a final continuity test on power and ground - there have been many times when I've gotten impatient and skipped this step, only to damage some critical component. If possible use a lab power supply with a current limit for early testing, and set at a modest current (100mA) - a telltale 'click' will tell you something is shorted.
RSS Feed