Jerky Small Straight Line Moves

[dertsap]

Those are pretty straight forward parts to machine , sounds like they aren't using arc fit in bobcad . I use bobcad as well and it should put out the arc moves for simple stuff like that

[underthetire]

Exactly. 35 years old. Anyone around here still running a 35 year old PC?


Chuckle. How fast a CNC goes does not depend on whether it is running Mach or a Fanuc controller. It depends on the design of the hardware and, above all else, the power of the MOTORS.
Mach3 can simultaneously drive up the 6 fully synchronised axes. An ESS can output step pulses at speeds up to 1 MHz for those 6 axes. Let's do the maths for my Mach3 machine:
Single step resolution: 0.8 microns.
Assuming 1 MHz step pulses, we have 1e6 * 0.8*e-6 = 0.8 metres/second.
I respectfully suggest that this would be fast enough.
Note two things here: Mach3/ESS would bring that 0.8 m/s to a halt at a point within 1 micron of the nominal. The second point? You would need a very large block of concrete to anchor this beast down!


Fair enough.
I am running Mach3 version 042.063. Prior to this we had previous versions of Mach3: the upgrade each time was FREE. Try getting a free upgrade from Fanuc. Prior to that we had Mach2, and before that Mach1. And Mach4 is coming along nicely. Adding functionality to the ESS is also just a SW upload at program start time.

This is where modern computer-controlled systems can differ so markedly from older hardware-based systems: an upgrade may require nothing more than a restart. You don't have to replace a box full of electronics. Found a bug somewhere? Next version. Need to support another g-code? Update the SW - usually for FREE. Want to add several custom M-codes? Write them, put them in the Macro directory, restart.

Why didn't Fanuc go the PC route? It just might be because 35 years ago a 'PC' was barely able to support a word processor, let along do heavy maths. (I exaggerate, but you get the idea.) Now they are locked in, and are not game to restart.

Cheers
Roger

Yes, but your talking almost micron "steps' which is a rather large amount for industrial controls, and your completely ignoring the feedback loop, which is more important than feed forward loop. Fanuc is proprietary, same with the motors and amplifiers, no weird breakout boards to slow down communication. Speed is one thing, speed with resolution and profile accuracy is a whole different animal. Let me know when you can run a ballbar at 300ipm and hold .0001 circular interpolation with just step and direction.

Sent from my A3-A20FHD using Tapatalk

[RCaffin]

If Mach still has issues with small straight line moves at high feed rates

It does not have any problems at all with those.

your talking almost micron "steps' which is a rather large amount for industrial controls

Are you saying that 0.8 micrometre steps (metric!) are too big? They seem to give me very nice results.

your completely ignoring the feedback loop, which is more important than feed forward loop.

Sorry - I don't understand this bit. I never mentioned a feed-forward loop.
With digital controls the feedback loop from the encoder to the driver is ... well, everything.

no weird breakout boards to slow down communication.

Um ... you do realise that the chips in a Break out Board are as fast or faster than the chips in a Fanuc control? It's no problem to put a 1 MHz pulse train through a BoB.

Let me know when you can run a ballbar at 300ipm and hold .0001 circular interpolation with just step and direction

I think that will be more a function of the mechanical engineering than the SW. That said, let's look at your figures.
300 ipm: 5ips. The limit here would be the motor and gearing - on ANY machine. Yer pays yer money and ...
0.0001" circular interpolation: forgive me if I translate that into metric. That's 2.5 um or 3 Step pulses on my machine. Yeah, no problem at all. Mind you, I question how many spindles have a TIR small enough for that to be meaningful.

The thing about modern PC-based CNC is that the parts are not proprietary : there is competition among the vendors all the way through the chain. Quality and performance goes up and price comes down. We benefit. For example: one of the driver mfrs has a standing offer about one of their latest motor drivers: if it blows up, they will repair or replace it for free (unless you have covered the PCB with swarf!).

Cheers
Roger

[dertsap]

what quality production machines are being sold that are running mach .
Hass is pretty much the bottom of the barrel when it comes to decent quality production machinery (not to offend anyone) and I can't picture them ever using mach above their own proprietary controls . Mach was developed for the hobbyist and was a means for the hobbyist to have a cnc control on the cheap .

[samco]

we are in the process of converting an early 80's matsuura. (yasnak/yaskawa control)

The spindle and servo drives are being re-used. They are matched to the servos and are working well.





This is all done with linuxcnc and mesa interface hardware. Closed loop control (not step/dir) about 50% done I would say. The basic stuff is working but some hardware issues and making it more user friendly (Z has some backlash)

Linuxcnc has a new trajectory planner that works very well - and as always has great path following. (you can specify how close to the programmed path you want linuxcnc to follow)

The first video above is a program made up of very short line segments.

sam

[RCaffin]

This is all done with linuxcnc and mesa interface hardware. Closed loop control (not step/dir) about 50% done I would say.

Um - I may be wrong here, I often am, but I seem to remember that LinuxCNC and Mesa also use Step/Dir with DC servo motors. Closed loop of course, but still emulating a stepper. They may of course be sub-micrometre steps: that is normal.

Yes, the TP is very important. But that is getting to be 'known technology' today.

Cheers
Roger

[samco]

Um - I may be wrong here, I often am, but I seem to remember that LinuxCNC and Mesa also use Step/Dir with DC servo motors. Closed loop of course, but still emulating a stepper. They may of course be sub-micrometre steps: that is normal.

Yes, the TP is very important. But that is getting to be 'known technology' today.

Cheers
Roger

You are wrong..

Mesa has both solutions (as does pico systems and others). Step/dir - analog with encoder feedback. Or you can mix and match.

In my situation - I am using the mesa 6 axis servo interface board - http://electronicsam.com/images/mats...423_155545.jpg

you can see the encoders going cables hooked in - and the analog/enables going to the drives.

Internally in linuxcnc there is a a motion controller running in realtime that sends out velocity levels to amps while reading back position. This is run through a pid loop for process control. (you tune using following error or pid error to get the best motion)

sam

[samco]

In the first video you see in the background of the computer screen halscope - an osciliscope within linuxcnc. This is what I was using to view the following error of the axis.

[barchdesigns]

That's exactly what I was hoping to achieve with my retrofit, but I hadn't seen it done on a machine of your size yet. Originally I was thinking of replacing all the spindle and servo drives but it seems more logical to leave them since they are indeed matched to the motors. The Fanuc drives seem pretty easy to repair/maintain as I found over the last year. I have a thread of how I replaced the Fanuc Spindle drive in my machine with a Delta VFD, but then ultimately went back to the original Fanuc drive. http://www.cnczone.com/forums/spindl...25824-cnc.html

Replacing just the NC portion of the machine with something that is faster and cheap to replace these days seems the way to go once the Fanuc 6M NC starts having problems. Ideally I would like to figure it all out before having to tear the machine apart. I was even thinking of making a Fanuc plug interface breakout board so that I could easily go between the new "Test NC" and the original Fanuc NC just by unplugging the Honda connectors and switching the cables. I haven't seen anyone actually do something like this though and it leaves me wondering if people just buy retrofit professional packages and have a company come in and do it.

[RCaffin]

Mesa has both solutions (as does pico systems and others). Step/dir - analog with encoder feedback. Or you can mix and match.

Most interesting. Mesa and Gecko go rather different routes when it cmes to details, although both use encoder feedback.

My Gecko 320X drivers take Step/Dir from Mach and encoder F/B from motor into a single axis servo driver. The onboard FPGA does all the control and PID work. It seems the Mesa approach is to have a separate board for the actual loop control, followed by a power amplifier board, with the enhancement that they get several channels on each board. The link between the control and power boards is done in analog.

The difference is that the Gecko driver conceals the link from the loop control to the power amplifier, which has some advantages when it comes to speed of response. Mind you, in most cases the speed of response is unlikely to be a large problem in either case.

Both approaches work, but I am fairly sure that the Mesa loop control board takes Step/Dir signals from LinuxCNC or EMC2. Mach and LinuxCNC/EMC2 are not all that different underneath.

Cheers
Roger

[samco]

Yikes.

Linuxcnc did not support step/dir initially. One of the original conversions that nist did was at GM on a K&T machining center. Stepper support was added after nist put EMC into public domain. Internally a stepper setup in linuxcnc is setup like a servo - closed loop feedback to motion..

There is no step/dir signals going between linuxcnc and mesa hardware. Period.


Lets take a look at your gecko setup.

SW4 “ON” and SW5 “ON” = +/- 256 count following error limit (default)

Do you mean you can have up to 256 encoder counts of error before you know there is an issue? Wow - on the matsuura that has micron resolution - that would be .010"/.26mm. What is your counts/inch or mm?

Also - mach 3/4 (mach 4 doesn't have any path following control - yet) doesn't have an easy way to set path following. it has 'CV Dist Tolerance_____Units' and 'Stop CV on angles > _____Degrees'

So how close to programmed path is your machine running? Can you tell me? I could tell you - using linuxcnc...



https://groups.yahoo.com/neo/groups/...essages/148027

^ and that isn't even taking into account the following error of your gecko drives. (which I could profile too - if I hooked encoders / linear scales up to your machine)

I can tell you that currently the matsuura has a peak following error of .001" at 400ipm So - if I run a program with G64P.001 - the worse case ever path deviation would be .002" And - if the following error gets above my set point - linuxcnc tells me.

sam

Most interesting. Mesa and Gecko go rather different routes when it cmes to details, although both use encoder feedback.

My Gecko 320X drivers take Step/Dir from Mach and encoder F/B from motor into a single axis servo driver. The onboard FPGA does all the control and PID work. It seems the Mesa approach is to have a separate board for the actual loop control, followed by a power amplifier board, with the enhancement that they get several channels on each board. The link between the control and power boards is done in analog.

The difference is that the Gecko driver conceals the link from the loop control to the power amplifier, which has some advantages when it comes to speed of response. Mind you, in most cases the speed of response is unlikely to be a large problem in either case.

Both approaches work, but I am fairly sure that the Mesa loop control board takes Step/Dir signals from LinuxCNC or EMC2. Mach and LinuxCNC/EMC2 are not all that different underneath.

Cheers
Roger

[RCaffin]

Hi Samco

Fortunately the specs you are quoting for the Gecko driver are not the full story - they are not even a small part of the story really.

Yes, you can program the Gecko over what error is required before it reports a Fault. Typically this means that the axis has run into and end stop or something rather robust and things have gone really bad. One tries to avoid that ever happening

In addition to this hard fault state, the driver has 3 LEDs to indicate Follower performance. They have these meanings:
Red: Fault. You hit something. Bad.
Yellow: Warning, meaning that you are about to go into Fault because you are >128 Steps out of sync. In my experience, this LED is of little practical use as the transition to Fault usually follows quickly. It can be of some use when debugging a drive train by hand, but that's all.
Green: In-Position, meaning the axis is within 2 encoder pulses or Steps of where it should be.

My experience is that the Green In-Position LED stays firmly lit for all of my machining. Since my machine has a single step resolution of 0.8 micrometres, that should mean the performance is within 1.6 um. Let's be cautious and allow for the possibility that I can't see a 1% or 2% flicker on the In-Position LED, so that there could be errors of up to 5 um, or 0.0002" (0.2 thou). Over short distances the error from the ball screw is probably within this limit as well. Over long distnaces the ball screw error will probably dominate, but I have not measured that.

I have been monitoring the parts produced on the machine and they stay within that 5 um tolerance, or perhaps a bit better. Touch probing is definitely better than that, down near 1 um.

I will leave the LinuxCNC Step/Dir argument to someone else with greater experience of LinuxCNC, as it has no value for me.

Cheers
Roger

[RCaffin]

There is no step/dir signals going between linuxcnc and mesa hardware. Period.

How very odd. Especially in light of the EMC2 page on hardware, which says

any Stepper-motor or servo-motor drive that accepts step/direction signals could be used with EMC2. These are just some examples of inexpensive stepper/servo drives.

In fact, that page does not mention any other form of interface as far as I can see.

Ah well, I am not fussed. Obviously some Linux CNC SW does drive Step/Dir.

Cheers
Roger