Sorry for posting on such an old thread but I think it makes more sense to have all information in one place than to start a new one. I tried pemu01's approach and it worked. I reduced my convergence problem (green: 2 px left, red: 2px right) to about 1px. The picture is much clearer/sharper now and I'm enjoying my HS20 again (although I've set my eyes on the JVC X-35).
BUT: It wasn't as easy as it sounded. After doing the steps more than 20 or 30 times (you have to watch some movies on a tv while doing this or you go crazy) I became more and more careless. Each cycle took about 6-8 minutes: switch off, take apart, adjust, assemble, switch on, check, swear and switch off and wait for it to cool down. Due to my increasing carelessness, I started touching the panels and the prism. Because I didn't take enough care when cleaning it, I've got a tiny green glowing spot on the bottom in dark scenes. Not a big enough problem to make me open it again. So, it took me about 10 hours and it did cost some nerves but it was worth it.
One advice: For me it was not possible to adjust the panels by 2-5 pixels. Every time I re-tested, I overshot the goal significantly, making it sometimes really bad (50px off) and I almost gave up. When I modified the approach, I got better results:
1) Lift the panel up higher than necessary.
2) Thighten the 3 screws only a little - so that you can still move the panel by hand but it is fastened enough that when disassembling the parts again, the panel doesn't move.
3) Center the panel as good as possible, assemble everything and switch the projector on.
4) Now take one of the long casing screws and carefully (!) insert them through the grove for the lcd cable. Place it on one corner or the middle (depending on the offset) of the panel and gently tap e.g. with the rear of a screwdriver on the screw.
5) In case of a horizontal offset place the screw on the side of the panel and use the screw as a lever to move the panel (also by tapping on the screw). Important: when using the chip board as resistance for the screw, hold it down on a big chip so the screw doesn't lift it up out of the plug while the projector is running. Because the screw can not lay flat on the side of the panel, the horizontal movement will also result in some rotation which has to be corrected by tapping the panel into place from the top.
6) switch it off, disassemble it and tighten the screws of the panel - everything VERY careful so you don't move the panel by mistake.
Bare in mind: Doing this on a running projector is dangerous in two ways: You risk an electric shock if you touch a part of the power supply. And you might destroy your projector if you touch some electronic components with the screw or the screwdriver. But the advantage is, that with some practice, it's possible to move the panels almost pixel by pixel, you can eliminate rotation easily and with about 5 cycles per panel (only two are necessary) you're done.
I'm fascinated by this approach as everybody keeps telling you, that the HS20 convergence problems can not be corrected and one would have to buy an ridiculously expensive prism block. Well, considering the effort, it's almost impossible... but it was fun to be daring enough to take the projector apart.