There are many variations on this; but, you could drive @ slow speed toward a limit switch that is fixed in position near one end and when you hit it stop, reserve at much lower speed until you come off limit switch. This will take the lash error off the home position. There are many refinements I could suggest such as two limit switches for fail safety and time-constraints on seeking switches; but, this is this basic routine. When you come off the switch, load a zero or clear your postition value and you will be doing great.

  Hope this help a bit.

      Randall May
      www.principaldynamics.com

I was hoping there was something else to this.

It looks like I will have to tell the machine which direction the home switch is..

Thanks!

I don't understand your statement, I don't guess; but, yes, I can't think of a home routine scenario that doesn't begin with an assumption about which direction to assert to obtain home. Given a particular axis length, say 3', there is a big difference between being 7" away from one end and 7" away from other end. I might suggest; however, that if you chose to have 2 different home switches, you could htne drive to either and, depending on which switch you hit, load a zero or the full scale of axis length into position value. It would be fine to do this, I've just never done it myself. The initial calibration would require you to "know" the interval betwixt the 2 though so that you could genrate the correct preload!!

   Hope this helps more!!

     Randall       www.principaldynamics.com

Your second reply solved my problem!

I am obviously a newbie at this.  Your statements have helped me to see that I was approaching the problem wrong.

During normal run, homing the machine is easy because it knows where its at, and knows where the home switch is. If I store the value on shutdown, it will remember it the next day if homing is needed.  My biggest concern was dealing with what happens during a power loss when the position value does not get stored.
It seems that if this should happen, or the value doesnt get stored for whatever reason, some type of human input is going to be required.  
I have tried to limit the number of limit switches for various reasons.

Thanks for your help.  This forum is very educational.

The most accurate homing sequence, given that you have a servo motor and I am assuming standard encoder with index pulse is as follows:

Upon power up if home switch is not tripped, drive toward it until it is.

Upon tripping the home switch, drive away until the switch untrips.

Now continue to drive slowly until you detect the index pulse from the encoder.

This method ensures the unit will home to precisely the same place every time, and is in fact the precise way that all motion control systems that I have designed work.

What about an absolute encoder? I also use a laser range finder apparatus on an automated anodizing plant. It simply bounces a laser beam off of a reflector installed on the car.  This replaced a previously installed barcode system that used a very very long barcode.  This did not work reliably due to issues reading the code. All of these could possibly work for you.