Using linear rails in Z axis--how many?

I made my cnc router with 80/20 linear rails--shaped plastic blocks that fit in the 80/20 rail. It seems to work fine on X and Y. I have had a bit of slop in lettering, so a small entry divot gets carved when making a cut like the crossbar on a capital A. I went through the bearings and tightened them up, greatly improving the divot. Now the Z axis sometimes loses registration, probably the stepper missing steps. It seems to be in the up direction (which makes sense), causing the bit to suddenly go deeper than planned in the work on the next plunge (or into the spoil board, >1" off). I'm using a Bosch router to power the bit. All the drive is done with 1/2" lead screws, 2tpi X and Y, 10tpi Z. Direct drive with 380 oz-in Nema 23 steppers/Gecko G540.

Things I could do (comments please!):
1. Add a spring to take up the router weight.
2. Replace the Z bearings with twin round linear bearings.
3. Replace the Z bearings with rectangular linear bearings.
For Z do I need a single rectangular linear bearing, or two in parallel? Two opposing? This is my central question.

Writing this I realize that I am assuming the problem is the bearings. It could also be my delrin nut on the lead screw, though the bearings seem more likely.

Thanks,
Steve

The consensus here is that those 80-20 linear bearings just aren't up to the job: http://www.cnczone.com/forums/80-20-...-bearings.html Replace them with something better - if not profile rails and trucks, then at least some supported round Thompson rails and trucks. Use two rails per axis, mounted parallel, and 4 trucks for best results.

The problem with Z losing steps is another issue, though. You can try reducing the acceleration of the Z axis, and if that doesn't work, adding a gas strut, which works better than springs for counterbalancing the weight of the spindle.

Thanks Andrew. Poking at it this afternoon I was coming to the conclusion that the weight had a lot to do with it. I ran two tests (stepconf in Linuxcnc) +/-0.5", 10ips screeched after a couple of dozen back and forth. The leadscrew was notably warm, makes me wonder if thermal expansion of the delrin nut is locking up the threads. The second test was at 1.2ips, no problem in the time I ran it.

Thanks

Are you using a way oil on those slides and screws, or running them dry?

I'm using LPS2, which is supposed to leave behind a waxy film. Way oil had not occurred to me, it will probably do a better job.

Both gas struts and metal springs have the same force/distance character, which doesn't mean the struts aren't better. It does seem to me that a counterweight would work best--keeping the force constant. Besides, I can try that with stuff on hand.

Steve

Give the oil a shot. Yes, a counterweight could work, at least enough to tell you if the spindle weight is really causing your problem. But gas struts work better than springs, because most springs aren't linear in their response, and constant-force springs tend to break after multiple cycles, especially if they're used on the Z axis for 3D work.