Machsupport Forum

Mach Discussion => General Mach Discussion => Topic started by: JohnHaine on November 14, 2018, 04:16:16 AM

Title: Probing speed
Post by: JohnHaine on November 14, 2018, 04:16:16 AM
I am involved in developing a new probing technology for CNC machines for tool setting.  It will work with both lathe and mill, and for example in the lathe can be used to calibrate tool offsets very quickly, and routinely to set the Z reference to the end of a bar in the chuck.  At the moment it is limited to rather slow axis movements but we are working on an improvement that should make it considerably faster.

So to my question for those of you who use probes: what axis speed to you feed at when moving the probe towards the material, whether using touch-trigger style probes or other sorts?

Thanks in advance for your input!

John.
Title: Re: Probing speed
Post by: RICH on November 14, 2018, 07:16:28 AM
John,
I probe at 1in/min in Mach3 when setting lathe tools and only one touch off  is used.
Yes, that is slow, BUT, accuracy is +-0.0002", repeatable. Overall time to populate
a tool table for 20 different type lathe tools takes under/about 20 minutes.

Probe feedrates need to address the users configuration, system accuracy, which include software reaction time, and axis accel/ decel  time.
Probing should not cause damage to part or tool. Must also say that any probing will only be as accurate as the users machine "system".

Probe feedrate should be an a user defined rate depending the what the probed values will be used for. If  I recall correctly min slow probe distance for Mach3 must be greater than 0.010".
 
I also probe to acquired points to automaticaly make a CAD drawing of a part. The rate varies depending the routine that will be used.

Variables associated with probing should be user adjustable / defined if the probing is not dedicated to a specific system.

FWIW,
 RICH
Title: Re: Probing speed
Post by: JohnHaine on November 14, 2018, 04:27:49 PM
Thanks Rich.  At the moment I probe at ~5 mm/min so considerably slower than you, hopefully we can increase that considerably.
Title: Re: Probing speed
Post by: joeaverage on November 14, 2018, 05:22:08 PM
Hi,
if  I understand correctly the moment a probe event is detected the advancing axis is decelerated to a stop. If it were crash stopped
there is the potential for to lose steps and therefore the probing would be inaccurate and the machine would have lost reference.

For a given probing speed Vp and a de-acceleration (given by the motor tuning), a, then the distance traveled, d, from the probe event (t=0)
until stopped at ts:

d=1/2 * a* ts2
where ts is solved from Vp=a*t s
or:
ts= Vp/a

(Gallieos equations for constant acceleration)

So you can calculate the overrun and effect on accuracy, or even make allowance for the overrun in the calculation of the probe point.

I probe PCB material for the purpose of applying a leveling correction. It accommodates any warp, twist or bow in the material. I normally probe at 50mm/min.
My steppers are tuned to an acceleration of 375mm/s2. I have tried probing at 10mm/min and noticed no difference in the probe file that results.
I conclude therefore that the axis acceleration is such that the overrun is so small that it is masked by the lost motion (flexure + backlash) which is less than 4 um.

There is one circumstance where a significant probe inaccuracy can occur. When a probe makes contact the motion controller, be it a parallel port or an external
controller, how quickly does the controller apply the de-acceleration? In the case of my ESS the best I have been able to measure (using my storage scope)
is 15us between the probe event and the step pulse stream changing to max de-acceleration. Note that probing is supported at hardware level by the ESS.
If I waited for the probe event to be signaled to Mach4 before the de-acceleration was applied then the delay would be about 20 ms, about 10ms each way comm
delay. It would be worse, much worse, in Mach3 given the longer comm delays.

My contention is that: IF the motion controller supports probing at the hardware level resulting in near instantaneous application of de-acceleration AND the maximum
acceleration of the de-accelerating axis is reasonable THEN excepting extreme probe approach velocity the probe results will be accurate to within the flexure/backlash
of the machine.

IF however probing is programmed as opposed to realtime implementation THEN even moderate approach velocities produce significant overrun and therefore
inaccuracy.

Craig
Title: Re: Probing speed
Post by: JohnHaine on November 15, 2018, 04:33:00 AM
Thanks Craig, good to have measurement results on the response time.  For Mach3 my understanding is that the processor is in a very tight loop while probing, of course Mach3 displaces Windows and even MS-DOS and effectively runs bare-metal, so I guess the probe input may raise a hardware interrupt so response time could be pretty quick.  Your measurements suggest a way I could evaluate this.
Title: Re: Probing speed
Post by: joeaverage on November 15, 2018, 12:11:32 PM
Hi,
prior to using the SmoothStepper I used the same probing routine with a parallel port. I never measured the response.
I never felt it was necessary.

Why don't you try some different approach speeds to see if there is any measurable difference?

With one touch plate in the machine try approaching at your current 5mm/min and then try at 100mm/min.

With my current setup (Mach4/ESS) I would not expect to see any measureable difference until I get to 200mm/min approach speed.

Craig
Title: Re: Probing speed
Post by: JohnHaine on November 15, 2018, 12:21:42 PM
To clarify, I would like to know what speeds people expect to use with their current probing methods so I can set a performance objective for the new approach.
Title: Re: Probing speed
Post by: joeaverage on November 15, 2018, 01:01:31 PM
Hi,
kool.

I use a software accessory called Autoleveller for probing PCB blanks. It generates a probe file which is then used to modify (slightly) the
Z height Gcodes of a PCB isolation routing file. Getting the depth of cut correct to 0.01mm (or better if you can!) is critical to the success
of making PCBs, especially surface mount boards with fine features.

I have one board 0f 135mm X 150mm and it is probed at approx. 10mm intervals with a total of 210 probe points. The Autoleveller program
actually produces the Gcode to probe the board....a bit like a Wizard....very convenient. I program it with 1000mm/min rapids between probe locations,
the initial height above the blank from which to start probing is 2mm and the probe approach speed is 50mm/min.

To probe all 210 locations takes 7 minutes, tedious but not unduly so. The routing file takes nearly 10 hours.....and requires as many as eight 0.5mm
endmills......that's pretty tedious!

I have tried increasing the probe approach speed to 100mm/min but noticed on a very small reduction in the overall time it takes to probe the board.
While I would expect 100mm/min approach to give me accurate probe results I expect more certain accuracy at 50mm/min! Given the extra time taken to
probe at the slower approach speed is only of the order of a minute I did not persevere with approaches of 100mm/min.

Where you can save some time is to start with the initial height of the probe above the board at 1mm rather than the 2mm I commonly use.
The PCB blank when double sided taped to the spoil board is flat and true to within 0.1mm or thereabouts. Thus I could probably get away
with withdrawing the probe to only 0.5mm and still be confident that it would clear the PCB and rapid to the next probe location without
contact. That would save some cycle time.

All up I tend to probe at 50mm/min......it has proven to give me accurate results to within less than the accuracy of my machine and yet fast enough
to be useful.

Craig
Title: Re: Probing speed
Post by: ZASto on November 15, 2018, 01:34:30 PM
WOW 10 hours :D
I usually have ethhed board, without drilling in much, much less time. Up to 200x300mm in max 2 hours, from the moment when I send a file to my frends that have photoplotter :).
Title: Re: Probing speed
Post by: joeaverage on November 15, 2018, 02:36:59 PM
Hi,
the PCB material for this board has a very heavy copper layer, 12oz/yard sq. or 420um! Normal 1 oz board has 35um
copper layer.

There are a number of consequences when using a PCB like that. First is that you cant really etch it, there is to much undercutting
to allow fine features. The second is that you can take only small step-over cuts to prevent the endmill from overloading and breaking.
My board has up to 600VDC applied and so I want 5mm clearance between adjacent tracks and at 0.2mm step over requires a large number
of repeatitions.......hence the long cycle time.

Craig
Title: Re: Probing speed
Post by: joeaverage on November 15, 2018, 03:46:59 PM
Hi,
just to show how the calculation works I will demonstrate using the parameter of my mill, namely acceleration of 375mm/s2.

At 50mm/min approach:
V=50/60
  =0.8333 mm/s
ts=V/a
   =0.8333/375
   =2.22ms
d=0.5 * 375 * (0.002222)
  =0.925um
ie the overrun is les than 1 um, plenty good enough accuracy.

At 100 mm/min:
V=100/60
  =1.666mm/s
ts=1.6666 / 375
    =4.444ms
d=0.5 * 375* (0.004442)
  =3.7 um

ie at 100mm/min the overrun is about the same as the accuracy of my machine, ie its acceptable.

At 200mm/min:
V=200/60
  =3.33mm/s
ts=3.333/375
   =8.88 ms
d=0.5 * 375* (0.008332)
  =14.8 um

ie at 200 mm/min approach the overrun is nearly 15um and about three times the accuracy of my machine. If I am probing normal 1 oz board,
35um copper, then the probing overrun amounts to 43% of the thickness of the copper layer which is unacceptable.

Craig
Title: Re: Probing speed
Post by: RICH on November 15, 2018, 10:21:37 PM
See attached FWIW.
RICH