Recently I've rather gone off the whole direct toner to PCB approach. Just doesn't seem to be robust enough. Having recently fabricated some PCBs in the Chemistry department I've seen some of the insane artwork they use. The artwork is (or was, before it was outsourced) produced in the university printing studio. The artwork is staggeringly good - incredibly black. High resolution too.
The process is as follows:
I really can't stress enough how black the artwork is! So I started thinking about modifying a laser printer to be an imagesetter for this film stuff.
At lunch Rob and I started talking about it, and came to the conclusion that we really want to just image straight onto the PCB photoresist. However, there is a serious problem here - PCB photoresist is sensitive to UV light, not IR. UV lasers are horribly expensive.
Ok, so what about an LED printer? I didn't really know how they worked, but it seemed likely that you could mod them to change the LED wavelength. Unfortunately this is not the case. They are actually a line of LEDs, one per pixel! So for a printer with a horizontal resolution of 600dpi there's a line of LEDs in the printer that's 600 LEDs/inch over 9 inches of paper width. Of course these can't just be SMT LEDs on a board! So something horribly complex is at work, and they cannot be replaced with UV ones. Also, 1200dpi ones don't really exist for reasonable amounts of money.
So we're stuck with modifying a laser printer to fit a UV laser. So I started looking for UV laser prices. They barely exist - certainly no UV laser pointers around yet. Then I remembered the DIY air (nitrogen) laser - the only laser you can build at home for a finite amount of money. Almost no money in fact, just an HV supply and a pile of aluminium and acetate. It emits at 337.1nm UV. Awesome.
But one problem - it's pulsed, and the pulsing is built into the laser physics. The guy says the laser pulse lasts <10ns, but doesn't state the recovery time between pulses. This would presumably be reduced by actually using a nitrogen atmosphere rather than air.
So is the pulsing speed sufficiently fast to not bugger up the raster scan of a laser printer? Lets say imaging an A4 page takes around 1s. That's 1200dpi * 11.7" = 14040 lines/page. So one line takes something like 1/14040=71.2us. If we assume a 50% duty cycle on our air laser then that gives us 71.2us/20ns=3560 laser pulses per line, or 3 laser pulses per pixel. So it's sketchy, but possibly doable.
We also don't know whether the amount of energy delivered by the beam is sufficient to expose the board, but I guess we can just pump more nitrogen/current into this laser to address that. Assuming we don't want an enormous reverse engineering excercise we're stuck with the a fixed scan/print speed.