Bipolar operation of DC circuits

Normally, a bridge rectifier chip is used to convert AC to a pulsing DC. However, it can also be used to make a DC circuit insensitive to input polarity. This is really handy because I have destroyed more circuits than I like to remember because I accidentally switched the polarity of the power supply.

The downside is that the input voltage has to be about 1.4V higher than the required supply voltage. The power dissipated in the bridge rectifier is 1.4 times the current. This may not be a good solution in low voltage circuits with less head room, but for most normal applications, it is a simple way to protect the circuit from inadvertent polarity changes.

Make High-Quality Double-Sided PCBs

Why make your own printed circuit boards when you can get them commercially made for low cost? The cheapest source I know is Oshpark. While they do provide a great service, it takes at least 2 weeks to receive the boards. For prototyping, this can be a major hurdle. Each design iteration will then take a month, and a project may need many months to get done. The DIYer can build the board and assemble everything in one evening. That advantage is really hard to beat.

Besides  time, there are other reasons to make your own board.  Commercial services charge by the board size, not by complexity. Larger boards will cost more even if they are completely blank. Once I had to make an over-sized PCB because the parts had to be spaced far apart for a specific reason. It was a very sparse board, but getting it made from even the cheapest commercial source would have been expensive.

Having said that, this is not really about saving time or money. If the phrase “building your own” does not excite you, then read no more.  Building a board may not take two whole weeks, but it will take many hours. Like in everything, there is a learning curve. It’s not going to work the first time. If the process itself is an unpleasant experience then it will not be worth any amount of time or money saved. Remember, we are talking about building a near-professional quality board, not something that is thrown together to check the functionality of a circuit. The reason people build instead of buy is because they enjoy learning and perfecting new skills. If you are detail-oriented, have lots of patience and take pride in craftsmanship, then this might be for you.

Now on to the details…

Laser printed toner transfer method

The laser printed toner transfer method is by far the easiest and the best quality for the home builder, even compared to UV exposure of pre-sensitized boards (see FAQ for explanation). This requires a special paper that is coated with a release agent. Apparently, some glossy paper can also work, but it is better to get the real thing from, which is also available through major outlets like Mouser and Digikey. They are about a $1 per sheet. You also need the laminator, which is about $90 from Amazon. Some people have used a clothes iron, but the temperature and pressure from the laminator will produce far more consistent results.

The key to successful toner transfer is to find a good printer. This is the tricky part because there seems to be a huge variation from one brand to the next regardless of the advertised dpi resolution. You need to make a test pattern of fine lines and try it on different printers and examine under a microscope. We are looking for a high density of toner coverage in the lines, clear white spaces and proper spacing of the lines. The expensive HP printer in my office gave surprisingly poor results under a microscope. The images below show 5mil lines with 5mil spaces from this printer. Notice the specks of toner in the white spaces and the two merged lines.

Printing services at office supply stores are good places to experiment because those printers tend to be well maintained. After some trials, I found a printer at Officemax that worked reasonably well. It is a Xerox Workcentre 5775. For 10 cents a page, I was able to get some decent results.

HP Laserjet 600 M601 (5 mil lines and spaces – microscope image). Notice the specks of toner in the white spaces and the two merged lines.
Xerox Workcentre 5775 (5 mil lines and spaces – microscope image). This is a much better print.

When printing, don’t bother cutting the toner transfer paper into smaller pieces to save paper. Just use the whole sheet. Duplicate the design to fill the entire sheet if necessary. Besides, Officemax said they will not put any sheets in their printer with pieces of loose paper stuck to it.

Toner transfer

The FR-4 dual copper clad board should be trimmed at least half an inch larger than the design outline. The extra room is useful for handling the board. Deburr the edges, clean the board and tape the printed sheet face up on the board. Drill through at least three holes along the edges. These are for alignment and should have been previously designed on the board layout. Smaller holes are better – I use #80 drill bits. Then remove the paper and scrub the board with a Scotchbrite pad until the copper is shiny and new, and wipe with isopropyl alcohol.  Avoid using water because the fine copper particles will immediately oxidize and stick to the board leaving an ugly residue. Finally, wipe with a lint-free cloth. The towelettes they sell for cleaning eyeglasses works well for this. Once the board is fully dry and clean, place one print face down on the board with small pins or wires to align them to the previously drilled holes. I use 10mil wires. Carefully tape the paper to the board along the edges. Kapton tape is better than Scotch tape because it will hold up to the lamination heat. Then flip the board over and do the same with the other side.

Aligned toner transfer paper. Notice the small wires for alignment.

It may be tempting to do the toner transfer one side at a time. But this can actually lead to more problems. The paper and board will expand by different amounts under heat, and if we do one side at a time, it is difficult to get the alignment right, especially with large boards.

Let the laminator run for about 15 minutes at full heat. Then feed the board from all four sides. We can’t assume that the top and bottom sides will be heated equally, so it is best to flip the board and feed another four times. This will be a total of eight feeds. Let it cool off a little, and spray some water on the paper. The release agent should become activated and the sheets should separate by themselves without any effort. Then carefully peel the tapes off and discard the paper. Hold the board under clean running water for a couple of minutes. This is important. We have to thoroughly remove the release agent that was on the toner transfer paper, or else it will prevent the sealant from sticking in the next step.

Feed the board eight time from each direction.
Successful toner transfer


The toner has a lot of pores and will produce pits in the copper traces if used without the sealant. The sealant goes into these pores and makes a better etch mask. The sealant comes in large sheets and look very similar to the old carbon copy foils, except these are green. They are available from the same places that sell the toner transfer sheets. Wrap the green foil around the board and tape it while holding it stretched without letting it fold or crease. Then feed it through the laminator several times as before. Then remove the tapes and peel the foil off.


Feed the board eight times again with the sealant foil
Sealant foil peeled off

At this stage, there will be excess sealant on the board. Lines may look merged or smeared, especially between very narrow traces. During my first few attempts, I had mistakenly assumed that the process had failed, and decided to start over again. In reality, this is a very normal effect, and the excess sealant can be easily cleaned off.

Get a roll of 3M clean-release masking tape and stick on the entire surface of the board on both sides. Apply a brayer roller or firm finger pressure to work it into the traces. Then peel the tape off. This should remove all of the excess sealant that is not on the original toner surface. This step should also convince you that the toner and sealant are very robust and won’t scrub off that easily.

Masking tape to lift off the excess sealant
Peel the tape off to remove the excess sealant
After masking tape lift-off

The microscope images below show the difference on a 5 mil test pattern with and without the tape lift-off. The effect is dramatic, so this is an important step not to miss (and I have not seen this step mentioned anywhere else).

This is the image after the toner transfer (5 mil lines – microscope image)
This is the image after applying the sealant foil (5 mil lines – microscope image). Notice the excess sealant in the gaps and merged lines.
After applying the masking tape all the excess sealant has been removed (5 mil lines – microscope image)


The images below show the effect of the sealant on the toner.

Just the toner contains a lot of microscopic voids (microscope image)
After applying the sealant all the voids are sealed (microscope image).



This is the time to carefully inspect the traces for any breaks and pin holes.  They can be easily fixed with a fine-point Sharpie marker pen. The ink seems to hold up well to the etchant.

Minor defects touched-up with a Sharpie



Now the board is ready to be etched. I use Ferric chloride, although there are less corrosive options as well. This can be done without any spills, mess or fumes by using a ziploc bag. Put the board in the bag, and then pour some etchant into it and seal the bag. I also double bag it in case of leaks. Lay it on a flat table, and use a brayer roller to spread the fluid around keeping all part of the board wet. Flip the bag about once a minute and use the roller on the other side. The etch time strongly depends on the temperature, so a hair dryer will significantly accelerate the etch rate. The board should be clearly visible through the bag, so there is no need to take it out until it is fully done. Once done, open the bag, retrieve the board (with gloves), place it on a stack of paper towels and wipe the chemical off. Then clean it more thoroughly in running water. With this method, the etching can be done without spilling a single drop of chemical anywhere.

The waste etchant should be poured into a bottle for proper disposal. It is extremely corrosive and should never be poured down the drain.

Etching in a ziploc bag
Etched board. Notice the Sharpie ink is still intact.


Strip Toner

Now the toner/sealant can be stripped with a solvent such as acetone. But don’t make the mistake of wiping the board with a wet acetone towel. This will smear the toner into the board irreversibly and it will look awful.  Instead dip the entire board in a bucket of acetone, wait for about 2-3 minutes for it to penetrate the toner,  and while it is still immersed in the acetone use a soft bristle brush to loosen the toner off the board.  Don’t take the board out of the acetone until all of the toner is completely separated from the board. Some toner may still get smeared into the board, but this seems unavoidable.

After stripping the toner
After stripping the toner (closeup)



This is the process of applying identifiers and text on the board, though it is not done with the conventional silkscreen printing method. The same toner transfer technique is used to apply the pattern to the board using the previously drilled alignment marks. Because the surface is not perfectly flat after the copper etch, some of the pattern that falls near the copper edges will refuse to transfer properly.  To minimize problems, the design should avoid crossing the boundaries of the etched areas as much as possible.

Silkscreen text applied

Solder mask

Solder mask is what adds the professional touch to a board. It also makes soldering fine-pitch surface mount components tremendously easier. There is a product called Dynamask 5000. It is basically a peel and stick epoxy film and it works extremely well. Unfortunately, it doesn’t seem to be available in the U.S. except through sellers overseas.

The film is green in color and comes in sheets. Cut the film about half inch larger than the design. Peel off the protective film on the textured side (use two pieces of Scotch tape to separate the films). Starting from one edge, use a brayer roller compress it on to the board. Do not let the film touch the board except where the roller comes in contact. The idea is to prevent air bubbles from being trapped. Then warm up the laminator to 200F and feed the board once. Repeated feeds or higher lamination temperatures can soften the film too much and cause them to reflow, and this seems to create blisters and dry spots. It is also a good idea to use finger pressure to work the film into the etched grooves. The film is really meant to be laminated under vacuum, so this method will inevitably leave some air bubbles trapped behind the film, especially inside narrow grooves. If there are large bubbles, the film can be easily peeled off and discarded.

Once everything looks good, it is time to expose the solder mask with UV. You will need a UV lamp. For small boards, a nail curing lamp works just fine, which you can get for $20. For larger boards, you may need a flatbed UV lamp, which cost about $75. Flatbed lamps are better because the cover has a latch to compress the transparency onto the board to improve the image resolution.

First, print the solder mask pattern on a transparency sheet such that the toner is on the side facing the board. This will require the top side image to be mirrored during printing.  Attach the transparency to the board with tapes, making sure it is in very close contact with the board. Any gaps will blur the image.

How long to expose really depends on the lamp intensity. The  specifications call for 100 to 500 mJ/cm2, but you would need a UV power meter to use that data.  It is easy to do a few test coupons to figure out the right exposure. Too much exposure can allow the light to bleed through the dark areas of the transparency. Too little will make the film dissolve in the developer. Due to the reflection, the film over the copper areas will get exposed faster than in the etched areas. So pay attention to how the film behaves on both areas. In my case, I was getting about 2mW/cm2 on the flatbed UV lamp, and 5 minute  exposures gave good patterns.

Flatbed UV exposure lamp

The film needs to sit in a dark place for about 30 minutes to allow the exposed epoxy to cure. Then remove the transparency and peel off the top protective layer from the film.

Developing is done by mixing some sodium carbonate powder in water to a ratio of 1% by weight. Do not use the store variety washing soda; it gave really bad results, probably because it contains other additives besides sodium carbonate. Get the pure sodium carbonate from chemical stores or from the same vendors who sell the Dynamask film.

Dip the board in the solution and very lightly scrub with a bristle brush. The specifications call for the solution to be heated to 90F, but that seems to make the develop run too fast. It was more controllable when the solution was colder. It took about 5 minutes for the develop to complete. Excessive develop times made  the film too soft with ragged edges, so it is important to monitor the process carefully.

Develop with a bristle brush

After the patterns are fully developed, rise the board in clean water and expose in UV again for several minutes. The purpose of this exposure is to fully cure the epoxy. The timing is not important but it should be significantly longer than the original exposure. Then run the board through the laminator again, this time using the maximum temperature. When applying the solder mask on the back side, it is a good idea to protect the front side with some sticky shelf liner paper. This way the developer will not soften or attack the solder mask on the front side. When both sides are done, I also bake the board in an oven at 100C for about 30 minutes to further harden the film. Make sure the board lays flat in the oven or else it will warp.

Successful solder mask
Solder mask (microscope image)


Component Holes

Now the component holes can be drilled. We need micro sized drill bits for this. The best kind of drill bits are the ones with the 1/8″ shank. The standard ones without the shank are hard to center on a drill chuck due to their small diameters. Mcmaster-Carr sells these in many different sizes, and it is good to have a bunch of them on hand.

During drilling is when you discover how good the front and back side alignment turned out to be. There is not much we can do at this point except hope it turned out ok.


Plated vias represent the last frontier for the homebuilder. There is no simple way to do through-hole plating at home, so people have always been exploring alternative techniques such as wire solder, conductive epoxy and rivets. There are problems with all of these approaches, but I found wire soldering to be the most reliable, followed by hollow rivets.

Wire soldered vias

Soldering a wire is the simplest way to connect a via, but it tends to leave large acorn-like protrusions even when cut with a flush wire cutter. These joints are not only unsightly, but they also prevent placement of components above these vias. However, this is by far the most foolproof way to make vias.

Hollow rivets

Hollow rivets are a tidier alternative. I have been using 0.6mm copper rivets (0.4mm inside diameter), which are available on ebay and various other sources. There is also a special tool for compressing these rivets, but it is horribly expensive and I can’t imagine anyone really buying such a tool. The rivets can be installed by hand. On the layout, I design 75mil pads with 25mil drill holes. Open up the holes with a #72 drill and insert the rivets with a tweezer.  Flip the board over and place it on a flat metal surface making sure the rivets are in contact with the metal plate. Under a magnifying glass, use a  thumb tack to slightly flare the rivet tails. Then use the flat end of a tool (such as a hex screwdriver) to compress the tail, making sure the screwdriver is held vertical to the board. The compression should produce a distinct snap when the rivet sets. It does not take much force since copper is soft.

Rivets inserted and the board flipped over
Flaring the rivet tail with a thumb tack
Flared tails (microscope image)
Compressing the tail with the flat end of a hex driver
Compressed rivet tail (microscope image)
Rivet flange on the other side (microscope image). The registration error is evident from this image.

But there are some serious problems with rivets as well. Many of my vias ended up being open circuits. I am pretty sure these were due to corrosion on the copper. Since the copper pads were all exposed to sodium carbonate during the solder mask step and then subsequently to high heat in the laminator and oven, it is not surprising that they would have oxidized. Furthermore, simply folding one metal onto another seems like a pretty bad way to make an electrical connection.  I ended up using a small dab of solder paste around the rivets and reflowed them. After this all the vias worked fine, most likely because the flux in the solder paste cleaned off the oxides.

So in the end rivets had to be soldered anyway, so they were not that different from soldering a wire. However, they were more flush with the board that the wired vias. The images below illustrate the difference in profile heights.

Wire soldered vias with 5 mil wires
0.6mm OD rivet vias with solder paste


Trim & admire

The final step is to trim the board and use a file to smooth the edges.

The end result is a board that looks professional and very close to what you may get from a commercial source. And the whole process can be done in about 4 hours.



  • How does toner transfer compare to UV exposure of photoresist?

All commercial PCB manufacturers use UV exposure of photoresist using a photoplotted (not laser printed) mask. The resolution of photoplots can exceed 50,000 dpi, whereas a laser printer tops out at about 2400 dpi. In order to exploit the advantages of photoplotted masks, you need a very uniform UV source and alignment capability, which is far beyond the average home shop. Most home builders who use UV use a laser printer to make the transparency and use that as a mask to expose a photoresist coated board. This doesn’t make much sense. Why? Laser toner is not fully opaque to UV since it obviously has a large number of pin holes (which is why we use the green sealant for the toner transfer). UV wavelength is only 365nm, so it is very difficult to seal these pinholes and make it opaque. Combined with optical diffraction and scatter in the film, it is not difficult to see that the resulting image will be worse than the original print. On the other hand, a direct fusion of the toner does not suffer from any of these effects. I can get 5 mil lines with the toner transfer, which is about as good as any commercial PCB service.

  • If toner transfer is better for the home builder, why are we using UV exposure for the solder mask?

Firstly, the solder mask epoxy requires UV to harden, so there is no way around it. Secondly, solder mask features are typically much larger than the copper traces so we can get away with some sloppiness. Unfortunately, this means we need the equipment for both processes – a laminator and a UV lamp. But luckily, these are pretty cheap.


  • Why not electroplate the vias?

It is difficult to get reliable results with electroplating outside an industrial setting. Although there are some home setups, it is messy, requires a lot of chemicals that have short shelf lives, and still the results are not that great. The first step is an electroless plating process to activate the holes, and the second step is to do the electroplating. MG chemicals has a youtube video on this.

  •  How good is the front and backside alignment?

This depends on how accurately you drill the alignment holes, and how carefully you place the sheets. Additional errors will creep from the thermal expansion of the board. Overall, the alignment should be better than 5 mils for small boards.


Amazon EC2 vs Intel NUC for a small server

NUC stands for Next Unit of Computing, a phrase used by Intel to describe their fanless computers. It is a complete computer that measures roughly the size of a book. This came at a time when I was considering abandoning my clunky 10 year old server at home and replacing it with Amazon’s cloud computer.

So, the question is this: which one is better for hosting a small website/file server? An Amazon EC2 instance, or a dedicated NUC that lives in your basement? Here are a few numbers that convinced me that NUC is the better choice.

First, it’s important to clarify that I am talking about a small website with only a few dozen visitors per day that will host wordpress blogs (this one), family photos and files. For this, the smallest EC2 instance (t2.micro) will be more than enough. It has 1GB RAM and a SSD or magnetic storage. SSD is faster than magnetic storage, but it costs more.

The NUC unit contains an Atom processor, 4GB RAM and 240GB SSD. Here is the cost and performance breakdown.

t2.micro + SSD t2.micro + magnetic Intel Atom NUC
Base hardware $75/year $75/year $140
RAM 1GB (included) 1GB (included) 4GB ($33)
Drive (240GB) $252/year $132/year $91
Energy included included $22/year1
Managed DNS Elastic IP Elastic IP $25/year2
Lifespan3 1 year 1 year 5 years
Bandwidth Included Included Included4
Avg cost $327/year $207/year $100/year
CPU benchmark5 11.1 seconds 11.1 seconds 32.8 seconds
Drive benchmark5 17.7 Mb/sec 2.5 Mb/sec 47.4 Mb/sec

1Average power consumption of 25 watts, with an electric cost of 10 cents/kWh.
2A managed DNS service such as
3The dedicated hardware is assumed to last for 5 years.
4EC2 includes network access; for the dedicated server, I am assuming everyone has high speed internet access that can be shared with the server without a significant penalty.
5Benchmark testing was done with sysbench.

One of the biggest advantages of running an EC2 server is the ability to remotely reboot and fix the computer if there is a problem. With a dedicated server, if the computer fails to boot or loses network connectivity, there is no way to fix the problem without physically getting to the computer. This could be a problem if you are far from the computer,  out of town or out of the country.



The NUC server is less than half the cost of running an EC2 server. This assumes the bandwidth is already included. Disk access times are significantly faster with the NUC, which makes it suitable as a local file server. The biggest penalty is the processing power – the Atom processor is about 3 times slower than the t2.micro. For a simple website and file server, this is probably not a major disadvantage. The biggest advantage of EC2 is the ability to repair and reboot the computer remotely.