Re: [CAD_CAM_EDM_DRO] Just How Accurate Is CNC?

Chuck Rice wrote:

>At 8:30 PM -0600 11/8/04, Jon Elson wrote:
>
>
>>softtechinit wrote:
>>
>>
>>
>>>What sort of repeatability and positioning accuracy should one
>>>expect from a CNC mill? Would you call 0.001" typical or
>>>unacceptably sloppy?
>>>
>>>
>>>
>>>
>>There are a whole raft of things that cause loss of accuracy. You can't
>>lump them all together. There are things like axis straightness and
>>orthogonality, which cause increasing errors as you move farther.
>>There are things like screw pitch error and variation, which can
>>be compensated for. There are things like backlash that are quite
>>repeatable, but can cause poor finish, out-of-round circles, etc.
>>.001" on a 1/4" part is pretty poor. .001" on a 12" part is getting
>>into the stratosphere, when you include stress release in the
>>part, temperature variation, etc. To hold .001" over 12", you need to
>>keep the part in a temperature controlled room. So, it depends
>>on what you are doing, and the size of the parts.
>>
>>I have a Bridgeport with high accuracy screws and .001" backlash,
>>and it is more than accurate enough for me to make some fairly
>>exotic parts at times.
>>
>>
>
>I am still new to this, but one thing that has been bothering me. If
>you have a full feed-back system, why is the feedback taken from the
>lead screws? Seems like you would eliminate the blacklash completely
>if you measured the table location directly instead of through the
>slop in the screws. The controller would just need to move the table
>till it was in the correct spot. seems to me that when you want to
>measure something, you measure as closely to the thing as you can,
>not through a lot of linkages. -Chuck-
>
>

The sole reason is COST! Price a 1000 cycle/rev shaft encoder.
Surplus, you
can get them under $100, new, maybe almost $300. With a 5 TPI screw,
1000 CPR gives 4000 counts/rev, * 5 TPI = 20,000 counts/inch or
.00005"/count.

Try pricing a linear encoder with the same resolution, and you are
talking about
$3000, minimum. Now, a standard DRO scale is .01 mm, or roughly .0005"
per count. This is really too low for proper CNC control even to .001"
positioning.

A second reason is that just because you can READ the machine position
with a linear scale on the table, doesn't mean you can CONTROL the position
to that accuracy. If you put the position feedback sensor on the table,
there
better be VERY little backlash in the screw and nut, or the servo will
rattle
back and forth across the backlash as fast as it can, possibly even burning
out the motor or servo amp.

Jon