Machsupport Forum

Mach Discussion => General Mach Discussion => Topic started by: Tarak on May 11, 2008, 07:06:07 AM

Title: Accuracy of DRO's
Post by: Tarak on May 11, 2008, 07:06:07 AM
What exactly decides the accuracy of the DRO's, I've noticed when I'm making different specialised screens that because I have them on a PC (no machine interface) I have the imaginery motors cranked up really high just for testing.
I think (not real sure though) that this is why I have the following issue:
When commanded to go to X22.00 Z12.00
The actual DRO's register the following,

Is there a way for me to have this to register correct and still have it moving fast?
Title: Re: Accuracy of DRO's
Post by: Hood on May 11, 2008, 07:13:03 AM
I have never seen any DROs off by that amount, most I have seen them off is for example I want X12.000, I have seen X11.999 and this is with mm rather than inches.  Can you give us some motor  figures you put in so that we can try and replicate it?
Title: Re: Accuracy of DRO's
Post by: jimpinder on May 11, 2008, 03:18:28 PM
DRO's are simply counters, they have nothing to do with accuracy, or speed or anything else.

The thing that decides on accuracy is the number of "steps" or pulses per inch or mm.
If your machine has 1000 steps per inch, then the best accuracy you can get is 1 thou. Any less than this and clearly accuracy suffers accordingley. Speed is set by the inches per minute setting on your motor tuning.

If you are moving two axis together then the relationship between those two axis is tangential. Mach 3 will move the furthest moving axis at the speed set in motor tuning. The other axis will move more slowly, but in time with the first axis, in the ratio of the tangent of the angle between the start and finish points. It sounds a bit more complicated than it is.

We know one or two of the angles from our schooldays - a 3,4,5 triangle for example, one axis would move 4 units whilst the other moves 3 (5 being the hypotenuse) and a 45 degreen triangle when the two sides are equal.

There may be others, but the vast majority are not simple numbers, and the short moving axis will have to be truncated. Even more complicated in a circular movement. It then might have to be truncated again, dependent on your "steps per unit figure"

Machs maths are done to 12 decimal places and then rounded up or down, and therefore the actual error is less than the error you see on the DRO.

In the end Mach 3 decides how many pulses to put out to your "stepper" - and the DRO merely counts the number of pulses and then
 translates this into a figure you can understand.

I must agree with Hood though. On a properly set up machine with realistic figures for step per inch, I have never seen "errors" more than 1 or maybe 2 tenths of a thou.

Title: Re: Accuracy of DRO's
Post by: jimpinder on May 13, 2008, 03:22:02 AM
Whilst I cannot see why you say your steps per inch are give and take 100 (it should be an exact figure all the time - unless some allowance is made for backlash) - well the answer is - stepper motor manufacturers will guarantee positional accuracy of their motors to within + or - whatever of each step. The step is correct, but the motor might stop a little to the right or left of a central position. The beauty about this, however is that in-accuracy does not accrue, since movement to the next position - whether one step or 10,000 - the final resting place will still be within + or - of the central position of that particular step. Steps are normally now 1.8 degrees so the error is very small.

My steps per inch ar 60,000. It follows then that there are 60 pulses per 1 thou of movement. In theory therefore I should be accurate to 1/2 of 1/10 of a thou. Unfortunately my Gecko drives deal in 10 micro steps. These are not full steps of the motors and, therefore their position cannot be accurately guaranteed like a full step - although one must say that if there are 10 of them, they must be more accurate than one full step. I can therefore certainly guarantee 6 steps per thou which gives me 1/2 a thou accuracy and if Gecko can add the other 1/10 then so be it.

At 253,360 then you have 253 (or 254) per thou, and even allowing for micro steps you (as you say) should easily split the thou into 25 - not 20. The distance you travel does not matter (providied your pulses per inch are correct).

To me it is the repeatability that is the amazing thing about CNC. I don't mind turning one bit by hand, but if I have another 19 to do - over to the machine - and they all come out the same (hopefully)
Title: Re: Accuracy of DRO's
Post by: Hood on May 13, 2008, 06:46:01 PM
It is amazing that those little pulses, and a lot of them, can consistantly be controlled to repeat themselves. So much so that they can make the same thing wrong over and over again just like instructed :)! 

 How true that is LOL

Title: Re: Accuracy of DRO's
Post by: jimpinder on May 14, 2008, 04:11:32 AM
Despite running it through Mach 3 on my office computer, running it through Mach3 on my workshop computer, watching it doing a "dry run" first, crossing my fingers etc etc -

I still wait with baited breath to see the first "cut" come out of the machine. If it's alright - well - GGGGRRRREEEEAAAATTTT :D ;D
Title: Re: Accuracy of DRO's
Post by: vmax549 on May 20, 2008, 02:33:37 PM
Darc what you are probably seeing is that you have the steps per in set very coursely and that is as close as Mach can come to the commanded position without going over the position. With high resolution steps it still occurs but is WAY back in the number string where it occurs and you will never see it.

OR you have set the dros to NOT round up to the displayed number of digits. That will cause the same effect.

(;-) TP
Title: Re: Accuracy of DRO's
Post by: Tarak on May 20, 2008, 05:39:54 PM
Thanks guys, I do only really notice it when I set the steps to an extremelly low number so the test is quick to sit through.