default precision

[deadlykitten]

Hello, today i runned a rough face milling program :

___________ipw > duration
0.003 .. 0.010 > 1:30
___________0.9 > 1:23
_____________0 > 1:18

it can be used not only for rapids, but also when cutting ; however, caution is required in this last case, so to modify ipw according to operation

about HiCutPro, does somebody know what it does ? or how to test it ? kindly !

[deadlykitten]

i just had a revelation let's begin : is a cnc going where it should ?

this means : G01 axis = position, will move "axis" to "position" ? final position always is "position±error" (relation 1)

if error = 0, than this means that the cnc will deliver "real positioning", and such a thing does not exist yet

is obvious that cnc's producers struggle to reduce error

axis from my cnc's deliver 0.001 mm acuracy; does this mean that "error = 0.001" ? answer is no

let's consider this example :

N1 G01 axis = position - a bit
N2 G01 axis = position + something
N3 G01 axis = position
N4 G01 go somewhere

at line N3 machine is moving on "axis" from a "greater value=position + something" to a "lower value=position", so let's consider image 1 , where " movement direction " is shown from right to left

in the moment when cnc "reached" line 3 shown into attached image, N4 (next block) is executed; what means "reached" ?
it means that "control_position"="position"+"ipw for mills / droop for lathes"

however, after reaching line 3 ( shown into attached image ), axis movement continues until it reaches line 2 ( shown into attached image ); this means that "control_position"="position"+"eps 1"; here it stops

"control_position" is the position comanded by the cnc
"control_position" is the value displayed on the monitor
"control_position" is always ahead of the "real_cnc_position", because cnc's are designed to work this way thus, in this case "control_position"="real_cnc_position"-"folowing error", because machine is moving to left
how there is a "folowing error", also a "stop error" must exist, to avoid letting the "real_cnc_position" going in front of the "control_position"; thus "control_position"="position"+"eps 1=stop error" (relation 2)

"real_cnc_position" is recorded from a "device" like "absolute encoder", or the "AbsoScale", with some precision : "real_cnc_position"="device position"+"device error" ( relation 3 )

backlash exists; more ore less in above example, after line N1, machine moves to right for "something + a bit"( N2 ) and after it moves to left for "something"( N3 )

is "a bit" < , = , > than "backlash" ? this depends on how much is "a bit ( from the program / code )", and "backlash" on axis arround "position" ; so an error appears = eps 2, shown into attached image; eps 2 depends on "backlash - a bit" ( relation 4 )

if "a bit" >= "backlash" then eps 2 = 0
if "a bit" < "backlash" then eps 2 = 0 .. 100% backlash, depending on | backlash | - | a bit |

( . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . )

so, let's reload relation 1 : final position always is "position±error", where "error" depends on :
- a "stoping error", related with the "following error", as described into relation 2
- a "device error", from the "absolute encoder", or the "AbsoScale", as described into relation 3
- "eps 2" = 0 .. 100% backlash, as described into relation 4

[deadlykitten]

how to minimize positioning errors ? ( errors described in previous post )

1) "stopping error" : G09, G61 ( = G09 modal, canceled by G64 )

2) "device error" :
... the "absolute encoder" is enough for most cnc operations
... "AbsoScale" is better than "absolute encoder"
... and the cherry on top : 0.1um control

3) "backlash errors" :
... far as i know, Okuma verifies and revises each axis backlash at each 30mm
... avoid low zig-zag movements
... G60

[deadlykitten]

so far was local behaviour : positioning another question : is a cnc going how it should ?

this means how close is a cnc movement to a real(=theoretical) trajectory ?

some available movements and functions :
...1) segments ( G01 ), arches ( G02 / 03 ) > most used
...2) interpolation ( G101 / 102 / 103 ) > less used, but still
...3) regression ( nurbs )

i will begin with 2) interpolation: milling on a spindle frontal with CX is not so precise as milling :
... on a spindle frontal with real Y axis, or
... on a mill

thus, axis interpolation ( CX ) :
... delivers good results if shape is near spindle center ( X close to 0 )
... losses precision when X increases
... delivers real trajectory when X does not move during cutting, but this is a particular case, when an arch has it's center on spindle axis

1a) segments and arches - let's check out the segments in attached image is a grid with squares equal with cnc's displayed accuracy : 0.001

let's consider that machine should move among a segment between those 2 yellow points; those yellow circles are sugesting the "positioning error", because machine is never on a middle of a square, as shown in previous posts; however, let's consider "positioning error=0"

question : because machine displayed accuracy is 0.001, will the machine move among the blue squares ( 0.001 x 0.001 ), or among the yellow line ? well, yellow line is theoretical, will never be reached, so, will the machine move with 0.001 acuracy, or better ?

G01 triggers an internal function, that will comand X & Y motors to move syncro, with a sincronization factor = depending on neccesary X and Y axis travel ( syncro = dy / dx or something like that )

a movement is generated like this : acceleration - straight palier ( constant speed ) - deceleration
actually, like this : acceleration - straight palier ( constant speed ) - deceleration - moving slow to ipw - stop

motors will accelerate and break differently, with light differences, but there will be a duration when bought motors will be on the straight palier, with their speed syncronized : in this moment, motors are running continuously, and this generates a path pretty close to a real/ theoretical line ; this movement is slowed down by the "following error", but still, results are more better than the grid

1b) segments and arches - let's check out the arches - same thing happens with arches, only that syncronziation changes , because dy/dx changes ; motors are internally comanded to move in this manner, with best available electronics accuracy; if an arch is coded not by G02 / 03, but with a row of G01, than the result is not the same, because machine will accelerate & break at each segment ; arches accuracy is not as good as segments, but is still ok

3) regression ( nurbs ) :
... machine will try to maintain a constant speed as long as possible, without breaking and accelerating at each segment
... machine will try to keep up the syncronization as long as possible, and reduce differences between syncronizations (actual and next), modifing the syncro when needed, thus getting away a bit from the trajectory, within a given tolerance
... is the equivalent of G01 or G02, but for a complex shape; thus, is reading the trajectory, and after it walks on it as best as possible

[deadlykitten]

momentum among a path

once a movement begins with a given feed, machine will try to keep it :
... among positioning / leaving is all ok
... when cutting, will require more power, so to avoid losing speed
... if too much power is required, than "torque limiter" drops in

obvious, more power means a lower accuracy, yet enough for roughing

best accuracy is achieved when machine is moving gentle

[deadlykitten]

so, about this precision, i think caution is required when exporting complex shapes for milling, with small segments, thus the machine can not reach and maintain the desired feed

if i knew all this, i would not have posted this thread

[deadlykitten]

It's all great fun.
Cheers
Roger

Hy RCaffin, please, i am a bit in doubts about senzors used to locate parts, and i remembered your posts in this thread

attached images are from a nikken catalog

those senzors start to flash , bip when they are in contact ... and i suppose that with extra senzors they can be programmed to work like a renishaw

about the ones that need conductive materials :
... for example on a lathe : does the electric circuit goes turret-senzor-part-chuck-chasis-turret, thus is it a closed circuit or opened ?
... for a mill : does the electric circuit goes spindle-senzor-part-vice-table-chasis-spindle, thus is it a closed circuit or opened ?

and how are working those senzors for non-conductive parts ?

... once i was ready to send a senzor back, because i thought it was not working, but it proved to work on the lathe being in it, while in my hand it did not work kindly !

[deadlykitten]

The tip is floating while the body is connected to the frame of the machine. Touch something metallic which is electrically connected to the machine and you complete the circuit. Beep! This won't work on plastics

hy Roger ... this means that i can not detect a metalic part inside a plastic vice ?

thank you for all the infos, i only needed a light idea ...i can not comprehend such stuff

you are the god of probes all the best !

[deadlykitten]

If you can hook up a wire from some metal part of the machine to the metal object in the plastic vice, then it should still work. It does not need to be heavy wire: very light wire would do

hy Roger ... actually i thought of something else

is not like i will craft probes, but wait a bit to do a drawing so that circuit is closed even on plastic

... this is as far as i can go with electrics ; kindly !

[deadlykitten]

hello, about the main idea of this thread : it may be possible that the machine checks diff to be in droop not only at corners, but always; obviously, diff is always checked, but i dont know what is the comparison factor

some heavy cutting tests + diff charts proved that Okumas are reliable, diff being lower than few microns

to be sure of such things, special equipment is needed

diff deviations are higher when cutting specs increase, and using the cncs at this state will make it behave more roughly when it is used at lower specs

in other words, accuracy of the machine can be and it is tested, but tools used are not easy available ...

having them running in background may deliver numbers, giving you trust of what a cnc is capable of kindly !

[deadlykitten]

I have had success in decreasing cycle time by "changing accuracy" of the machine. for example if you are on a horizontal mill and you rapid from left side to right side, you will see significant delays at the end of every rapid move "back, over, in" while the machine confirms that the DIFF is small enough to proceed with the next block.

hy mr wizard, if i may, would you use this trick with same confidence, for a " slow rpm face milling ", and also for a " high speed tapping " operation ? i am looking for a method to quantify the 'confidence' / kindly

OFAST
  NOEX LV01 = 0
  GOTO NHERE
OSLOW
  NOEX LV01 = 0.0001
  NHERE NOEX VINPX=LV01 VINPY=LV01 VINPZ=LV01 ( VINPA=LV01 )
RTS