Art:

Thanks for explaining a little more about what goes on within the program, and the particular problems you are working at solving. I really appreciate, and value your feedback.

My first CNC machine goes back to 1980; teletype machines, and paper tape!

Like a complex machined part, the solution to the threading problem is breaking the job down to the individual pieces of the puzzle, and solving them one at a time.

Thank you, and those involved with you, for your work to make Mach a better tool for us all. 

John

Hood:

1.) 20,000 per second? Check your math.

2.) My threading comment was a generalization, and I did refer to higher RPM's for smaller diameters.

3.) Are you spinning 2" shafts 1000 RPM when cutting a thread?

I think we all are interested in seeing this particular problem within Mach code, solved in the best possible way. Also, I appreciate this forum because of the vast pool of skilled people willing to help other people. My shop is 4-axis milling (2) machines, no CNC turning yet, and (2) engine lathes.

Still learning.

 

   



   

 

A 500 line encoder will be seen by Mach in quadrature so thats 2,000 pulses as far as Mach is concerned. 2,000 x 600 RPM IS 1,200,000 pulses per minute so divide by 60 seconds and you get 20,000 pulses per second.


Depends on material but I very rarely do as big a thread as 2inch but if I did then if mild steel I would be over 1000rpm, stainless probably about 600.



Hood

ART,
Here's some more testing data for you. I have attached pictures of the two test pieces along with

data for each.
The tests were done using my new pc.
Two pitches, namely .025 ( 40 tpi ) and .1 ( 10 tpi ) were done. The leads increase for both at
.0075"/inch.Now the prior tests with a borrowed pc were for .050 ( 20tpi ) and the lead error on those
was basicaly 0.009" / inch.I will do another test with mine at .050 but i think that you could say the
following: The lead error seems to be a constant 0.008" / inch increase.
 
These threads basicaly all had a single scribed lines.Test #2 ( the .1" pitch ) had two scibed lines
and my best quess is that it happened around the tenth pass or so based on the depth. The graphs
show per pass the led error for each thread root. So you can actualy see the changes between
groups of threads or revolutions. Additionaly there is a list of noted rpm variations from the threading
diagnostics and test#2 data it shows specifics during that threading pass #.

Look the info over and if you have any questions just post.

By the way, for days i have been checking my spindle rpm and it always tested 115 rpm, today 120
rpm ( and read with three different indicators ) . Not even going to mention why i think that is so,
and no, i didn't check the house voltage today. 

RICH

ART,
OOPS!
I did something realy stupid and caught it today. All of my recent post's which say that the lead is increasing is wrong.  The LEAD IS .......DECREASING. The amount of lead decrease is 0.008" / inch.
I found and used Chip's Screen set,  so now have kernal speed information.
I'll go back and modify my recent posts. The data is good, just again, it's decreasing.
RICH

Some time ago I read the thread in it's entirety. I have since followed it closely. It have been very interesting reading and I,m very impressed on the effort laid down by everybody in trying to get the threading to be as good as possible. I have thought a few times on saying something but have decided to keep shut as I probably not have had something useful to say.

The last post by RICH where he says the lead is decreasing is however particularly interesting. I have been thinking since some time ago on the fact that Mach can't speed up if that was necessary in order to catch up on an increasing spindle speed. The thought I have had is what if the spindle speed measurement done just prior to each pass is slightly slower than the actual spindle speed when threading occur. Then the pitch will be to large. I understand the spindle speed is measured in only one revolution prior to the threading on each pass. It places a huge demand on the exact measurement of the time between those two index pulses on that one revolution. When threading at say 500rpm, that is 3,000 degrees/second or 333 µs/degree. I have been thinking on what is the repeatability of the index trigger. An error of a couple hundred µs of the exact time of the trigger signal and the wrong way may cause a noticeably deviation of the pitch on long threads, especially at high rpm's.

I hope I remember the previous posts by ART correctly as what I have said now else definitely would be incorrect. It may very well be incorrect even if I remember ART's previous explanations correctly. In that case disregard what I have said here.

Steinar

Steiner,
The best i can do "easily" is measure visualy when the sensor turns on and off and that is repeatable within 5 minutes of an angle. So that's one ball park useless number.  A 0.003" error between two actual scribed lines was around 10 degrees of a rotation. Now doing 50 moves using a G32 Z-2.0" F.050 from Z=0 showed all 50 moves ending within a 0.001"distance which tells me that the trigger is very repeatable and the axis travel and acceleration together was consistent. Then i even broke that down by .1" moves from Z=0 to 1,1.1,1.2, etc and all was good. The start Z=0 and ending point Z=-2.0 was confirmed using a calibrated allignment telescope and scale. The scope is 40X at 12" and has an optical micrometer ( measures to 0.0001" ). So can't do too much electronicaly, but, the end physical result is what is important.

ART,
When i run the driver test PPS is 25455. The kernal speed viewed on Chips lathe screen that we used before for testing, shows a pulse frequency of 25452, but if you go out and come back into Mach it will change to 23654. It will continue to show the 23654 for however many times you exit and come back into Mach. Now reboot and the the above numbers will repeat as described again.

I did a third test piece at 0.1" pitch, the lead decreased by .008" /  inch and the pulse frequency on the screen was 23654.

Like that show that was on TV " You Asked For It". Well they are the pulse frequencies. 
RICH

Hi Rich:

  Seems pretty stable.. Almost like its correctable by adding a correction factor for a person to enter. I suppose one could make the feedrate command .8% faster.. but Ill come up with some tests to prove what the heck it is if I can.. Still dwelling a bit while I finish up my workshop..

Art

Winters coming and you need to get the shop done.
A user enter correction factor would be nice.  I certainly would want others to confirm what i found. So if it's user correctable and they do testing, than at least It comes from a know base point of reference.

RICH

Rich:

 Well, at least youve proved the kernal speed we end up with is not a factor. I suspect we can eliminate that, there is correction in the code for that, and it appears to work as all kernal speeds appear the same. Im thinking we simply are a few rpm off in spindle speed measurement.. ( .8% perhaps?) or somethign similar..

Art