Hello Guest it is April 14, 2021, 03:34:00 PM

Author Topic: Really need help  (Read 27848 times)

0 Members and 1 Guest are viewing this topic.

Offline simpson36

*
  •  1,369 1,369
    • View Profile
Re: Really need help
« Reply #20 on: August 13, 2009, 10:20:55 AM »

To my mind the driver test is there to weed out the really bad computers that have no chance of running Mach in their present state rather than saying it will definitely run Mach.

Hood

You may view it as a pass/fail, but it also 'grades the papers' so to speak, and provides what appears to be a benchmark useable as a guide to setting us Mach. The logical assumption, and the one that I myself made, is that a passing grade at XX khz is a green light to set up for that speed. This is where the problem lies. Another area where the 'sin of ommission' is committed by the Mach documentation. I discovered that the driver test is not neccessarily indicative of how fast Mach can run on a given computer only after much wasted time and frustration  :'(

Info worth sharing, methinks.

Offline simpson36

*
  •  1,369 1,369
    • View Profile
Re: Really need help
« Reply #21 on: August 13, 2009, 10:36:27 AM »
. . . . I was monitoring the following error in the drives software . . . . .
I have been jealose of this ever since you first mentioned it some months ago   :-[

Soon I will have my hands on two new servo drives that have this capability. The tuning software for the CNCdrives products reportedly can both graph and provide textural data for real-time monitoring of the following error.  ;D Should be fun to play with as well as very useful.

FYI, Rutex tuning software has real time textrual following error, but the tuning software cannot run while Mach is active. Leadshine tuning software graphs following error (as well as a host of other parameters) in real time also. I do not know if their software can run concurrently with Mach.

Offline Hood

*
  •  25,838 25,838
  • Carnoustie, Scotland
    • View Profile
Re: Really need help
« Reply #22 on: August 13, 2009, 04:11:01 PM »
The big benefit of the SS is the pulserate, especially whe you are talking 2000 plus line encoders which is common on the industrial orientated servos, heck the spindle motors I have on the mill and lathe have intelligent encoders which can be set to 2 million counts per rev, even the SS wouldnt be much use there ;D

The pulse from the parallel port will vary from computer to computer, some will be very nice others not so nice.  I have scoped most of the computers ports I have ever used with Mach but the SS is in a class of its own even compared to the best PP pulse. Thats not to say the PP is no use, it usually does a very good job.

Hood

Offline Hood

*
  •  25,838 25,838
  • Carnoustie, Scotland
    • View Profile
Re: Really need help
« Reply #23 on: August 13, 2009, 04:23:34 PM »
. . . . I was monitoring the following error in the drives software . . . . .
I have been jealose of this ever since you first mentioned it some months ago   :-[

Soon I will have my hands on two new servo drives that have this capability. The tuning software for the CNCdrives products reportedly can both graph and provide textural data for real-time monitoring of the following error.  ;D Should be fun to play with as well as very useful.

FYI, Rutex tuning software has real time textrual following error, but the tuning software cannot run while Mach is active. Leadshine tuning software graphs following error (as well as a host of other parameters) in real time also. I do not know if their software can run concurrently with Mach.


The thing i really like is not only can I see graphically in the 4 channel scope window but also text of the realtime errors and  it will store the max error both plus and minus. Heres a couple of screenshots of the software, no drives connected so no info there ;D

Hood
Re: Really need help
« Reply #24 on: August 13, 2009, 04:31:38 PM »
Hey Guys...   I hate to say this, but I fear we are straying from my original question.

I think we might be digressing from the original subject a bit...
I went to the shop today and I stripped Windows XP of all of its extranious bits, as outlined in a Mach-3 document I have.  

I tried running my profile, and for abiout three runs, the X-Hand-wheel returned to its starting point and the Z-Handwheel fell four steps short.  I thouht things were getting better. Actually, for what I am turning, I could live with jogging "Z" back to zero a few steps and correcting the Z-DRO., but on the fourth try, "Z" fell short by 25 steps...  Not acceptable!  I tried slowing the motor's max speed settings down from 250 to 200. I don't like the way G0 sounds at 200 mm/Min and the end position never got better.

So, perhaps trimming Windows down is necessary and a good thing, but it hasn't helped getting my tool post to return to its original position.

Then I tried my experimental g-code to run "Z" back & forth several times, then run X back and forth several times. The program moves "Z" from 0 to 128, then back to zero three times at full speed using G0. Then it runs "X" fore & aft from 0 to 128 and back to zero, three times .  I tried setting motor speeds at 200, 250, 300, & 350. With each increase in speed the motors sound smoother and get faster and every end run.  Every time the script finishes both hand-wheels end back up at zero... every time!  Then, I noticed something strange...

As the saddle moves from zero to 128 in either axis, I hear a "glitch" in the motorthat's operating... like it just skipped a step or two. Then on the return from 128 at about two-thirds of the way back (same as 1/3, the way out) , I hear the glitch again.  This occurs on both the X-Axis and the "Z"-Axis... wierd!  I put a mark beside the rails on each axis. It's like clockwork., When the saddle reaches either mark, from either direction, there is this little "bzzt". The faster the motors are running, the harder the "glitch" is to hear, but nonetheless, it is there.

I wonder if both these motors are skipping pulses every time and the lost pulses in one direction compensate for the same number of lost pulses in the opposite direction. That way, in this situation with one motor running at a time can always return the hand-wheel to zero??

Tell me I'm nuts!

Anybody know of a simple kit to build a TTL pulse-counter with about six digits of display?

Offline Jeff_Birt

*
  •  1,107 1,107
    • View Profile
    • Soigeneris
Re: Really need help
« Reply #25 on: August 13, 2009, 04:44:45 PM »
Problems missing steps can be from a variety of things. It can be related to tuning, the PC not putting out a smooth pulse stream, the drivers expecting a long(er) pulse width, etc. A slight 'tick' sound when the motors are running is a good indication of the pulse stream not being smooth enough. The only way to tell is with an o-scope. Mach's Driver Test will test the interrupt latency of the PC but it cannot test the actual PC's hardware's capability of pulsing a parallel port that fast. Even new motherboards and add-on parallel port cards can fail to work properly.

A hardware based pulse generator like the SmoothStepper moves all the timing critical stuff off the PC and onto hardware that can very precisely generate a smooth and stable pulse stream faster than any of us would probably have a need for. I recommend a SS with all the systems I sell and I have yet to have sold a SmoothStepper and have an unhappy customer.

When you get a chance to look at the pulse stream with a scope it will give you a good idea of what the PC is able to do. Also look at data sheet for your drivers and see what the minimum pulse width is that they expect to see.
Happy machining , Jeff Birt
 
Re: Really need help
« Reply #26 on: August 13, 2009, 05:01:34 PM »
Thanks Jeff,

This skipping steps issue is the only thing keeping us from doing production.
I will tell my employer and suggest he consider springing for a smooth stepper.

I have an inexpensive digital scope on order. I hope to use it to see if I can see irregulatities in my pulse stream...
I'm also tempted to buy a digtal up/down counter kit and see if I can count the pulses and compare with the number of steps required to move a given distance.

Offline Jeff_Birt

*
  •  1,107 1,107
    • View Profile
    • Soigeneris
Re: Really need help
« Reply #27 on: August 13, 2009, 05:14:26 PM »
The scope is what you need so you can measure the pulse width regularity. Compare the pulse width with what your drives expect to see. Feel free to contact me (follow link in signature) if you have any other questions, or just post them here :)
Happy machining , Jeff Birt
 

Offline simpson36

*
  •  1,369 1,369
    • View Profile
Re: Really need help
« Reply #28 on: August 13, 2009, 05:59:40 PM »
The big benefit of the SS is the pulserate, especially whe you are talking 2000 plus line encoders which is common on the industrial orientated servos, heck the spindle motors I have on the mill and lathe have intelligent encoders which can be set to 2 million counts per rev, even the SS wouldnt be much use there ;D

The drives I am looking at all have pulse multipliers, so pulse rate not an issue unless you don't have enough resolution for the accuracy you need.

What is important for the hi-res encoders is the bandwidth of the drives on the encoder side. The drives I am looking into at the moment can read 1mhz from the encoder. That's 6,000RPM with a 2500line encoider . . . plenty plenty for what I need.  The step/dir pulse input is far less than that, but the multiplier saves the day. One order of magnitude is plenty for what I'm doing.

My only hesitation on the SS is the lack of swap axis,  . . . deal breaker for me. I've been talking to some people about a completely different approach to that issue. May need programming assist from some of the big-dogs on that, but we shall see.

Offline simpson36

*
  •  1,369 1,369
    • View Profile
Re: Really need help
« Reply #29 on: August 13, 2009, 06:22:35 PM »
Hey Guys...   I hate to say this, but I fear we are straying from my original question.
I think we might be digressing from the original subject a bit...

Didn't realize that when you leave the room, eveybody has to remain seated quietly until you return  :-X

You have a host of suggestions, but you haven't done anything but try the same stuff over again that you already did, so there is really not much new to talk about on your issue.

If I remeber it correctly, the smooth stepper is less than $200 or there about. Your salary is a cost to your employer, yes?  It will probably cost your employer more for you to dink around with scopes and building pulse counters than to buy the solution the gurus here have suggested  . . meanwhile zero production.

Here is the real question: is this a critical path issue or not? If it is, and I were your boss, I'd be asking why you haven't already had Jeff-Birt overnight you a Smooth stepper. If it's not critical, by all means build yourself some new toys and play.