Warranty Void if Removed

Much has happened since my last blog post since, though published only minutes ago, it was written a few days back. This lag between life and blog is largely the fault of flickrfs.

The main change has been rebuilding the printer into new printer's chasis. The wooden arrangement was good to find out what needed doing, but I chopped off much more plastic than I needed to. The first printer was, it turns out, a 2100, and there were a few differences between it and the new 2200dn. The dn was electronically knackered, and wouldn't produce an image, so I used all the parts of the 2100. With the new chassis the printer is more rigid and easier to dis/reassemble as necessary.

As the results of the previous post show, there is much work to do to improve image quality. In my experience laser printers have difficulty printing onto shiny things. Prints onto the acetate we use rarely come out well. Small tracks are largely ok, but big areas come out patchy. As we all know, a PCB is either perfect, or an ornate coaster.  Two solutions spring to mind:

1. Print twice - the voids are likely to get filled by the next print
2. Make PCB rough - eg. by pre-etching briefly

Now both of these options clearly suck, but that's what we're stuck with. I prefer option 1, as we're gonna have to get enough repeatability to do double sided boards anyway. Option 2 is horrible because it adds another chemical stage and the board has an enormous surface area, so it oxidises at an incredible speed.

I started working on mechanical improvements to make option 1 viable.

I wanted to add a pair of high-friction driven rollers to keep the board square. The registration unit contains a driven rubber roller and a set of nylon spring-loaded rollers pushing down onto it. Combining the registration units from both printers gave me a pair of sprung rubber rollers, the lower one being driven. I mounted it at the front of the printer. I also added a platform and guide rail to allow boards to fix the lefty-righty alignment.

After this addition, the board fed nice and square each time, but as the trailing edge left the rollers the image screwed up in two ways - there was a little smear where the rollers pinged together kicking the board a little. Also, the image was much lighter for the rest of the print as the back of the board tilted up under the force of the imaging rollers, reducing the pressure at the imaging point. So I added some low friction nylon rollers after the imaging stage to prevent the board from tilting. This got rid of the dodgy greyness, leaving just the smear to fix. 

To fix the smearing issue, I plan to set up a mechanism to lift the sprung rubber roller as the trailing edge approaches. Bit of a faff, but it should be quite nice afterwards.

At some stage I mounted a second guide rail using g-clamps. Whilst it seemed like a good idea at the time I have the tiniest inkling that it may now be causing problems.

I experimented a great deal, but still alignment in the driven axis is a problem. I thought some uncertainty was creeping in as the drive rollers pushed the board into the imaging rollers. I started thinking of ways to start imaging after the board had already passed into the imaging section. Whilst this would waste a couple of inches of board at the top, it may eliminate a problem. Tom had the cunning idea of mounting the leading edge sensor so that it looked at the trailing edge instead. The signal therefore needed inverting, but fortunately it is a photointerruptor that is usually triggered by the absence of a black plastic lever that gets pushed out the way by the board. Thus it could be inverted simply by passing the board through the photointerruptor's gap, removing any uncertainty due to lever wobble.

With this sensor modification repeatability is now good, but not perfect. However, I have an inkling that the second guide fence is the problem. One could imagine that as the driven rollers grab the board, the slightest amount of skew could result in the board binding in the fences and skidding here and there. This wouldn't be a problem with just one fence.

This is an interesting project because so far the trickiest problems have been identified in little thought experiments, rather than by observation. Hopefully the trend will continue.

Images are a little sparce in this post. Did I mention flickrs sucks?