Router Forums banner

Nebula Controller Failure :(

8K views 47 replies 10 participants last post by  4DThinker 
#1 ·
After they sat idle for a few months and survived being moved from one shop to another I finally got around to checking out the condition of our college owned Meteor and Nebula from Probotix.

A student was able to use the Nebula with no complaints last week. Wednesday this week I wanted to show the students how to use the rotary axis so we booted up the Nebula again. Something must have happened inside the unity controller. The right Y motor would not respond to any jogging or g-code command. As such I couldn't Home the machine, and any Y movement would rack the gantry. Attempts to jog the X or Z axis would quickly generate a flurry of limit errors with the occasional E-Stop pressed warning (it never was).

I've boxed up the controller to ship back to Probotix on Len's advice. Hopefully they can diagnose and repair it.

Sad am I. That Nebula has caused me a few fits over the 3 years we've had it, but always kept on running.

4D
 
#3 ·
Bad time for the controller to go down right at the beginning of the semester.

Could just be dirty contacts in the E-Stop button from sitting for a long time and then reactivating the machine might have caused the error codes. The E-Stop might have been triggered during the move and when reset might not be making good contact electrically.

Hopefully Len can get you back up and running quickly.
 
#4 ·
As I said in my first post, the Nebula ran perfectly the week before it died. We even homed it then jogged it all around the bed to lubricate all the guide rails/bearings and feed rods. The symptoms seemed like a cascading failure of internal components.

The e-stop errors only popped up once or twice when I was trying to jog the CNC a little in each X, Y, and Z direction. Ever more odd was that when the controller was turned off I'd see either left or right Y motors spin for a few seconds.

In any case that poor controller is now on the way to Florida. Our Nebula was one of their first extra wide Meteors. Although it does have Nebula stickers on the sides the name of the LinuxCNC configuration was MeteorXtraWide.ini. The controller was an early Unity model. I think even before they were officially being called the Unity Controller. It had no VFD input (output?) like my new personal controller does.

4D
 
#5 ·
At this point we've determined that the controller wasn't damaged. The PC running it apparently suffered a power spike that damaged both of the parallel ports. We've had electricians working over the weekends in our new fabrication lab wiring up tools, and my guess is that they caused the surge accidently. The PC had been left on over the weekend so it was the only victim.

I sent the controller back to probotix, but they couldn't find anything wrong with it. When it came back it caused the same errors on our nebula. Using the configuration app I moved the control path from the first parallel port to the second parallel port. The limit errors disappeared, but the right Y motor would only run in one direction. Swapped the Y motor cables and the flaw moved to the other side. I brought to controller home to try with my personal CNC and it ran perfectly. The only thing left to blame is the PC's parallel ports. We are now on the hunt for a new PC.

4D
 
#6 ·
LinuxCNC is an very stable real time control system, but it’s insistance on doing the real time step generation internally and using parallel port(s) is becoming somewhat of a liability as computers with parallel ports become harder to find. There are lots of old computers that are available cheaply, but they are old and many can be unreliable.

I wish there was a version of LinuxCNC that could interface to any of the dedicated hardware step controllers (i.e. the Ethernet Smoothstepper). The only such choices seem to be the Mesa cards. There are numerous small, low power (fanless) computers that would be ideal as CNC control computers in a dusty woodshop , but have no parallel port or room for internal cards. I use MACH4 on my machine, with a long Ethernet cord to allow some space between my computer and the CNC.
 
#7 ·
@BalloonEngineer When you say small, low power computers, I think about the Raspberry PI (not sure if you meant that small). At a woodworking show last year, a guy had built a small CNC router (think dremel tool) and it was ran off a Raspberry PI, using a GRBL controller. As I dug further, I found that the GRBL wouldn't handle the higher power requirements of the steppers that a larger CNC requires.

There have been a few guys who have successfully come up with hacks for a R. PI, but nothing I would be willing to use long term. The R. PI, and other small computers like that seem like they would be a great solution for running a CNC, but right now options seem limited.
 
#8 ·
You can run a CNC using a tinyG card and a 12 volt power supply, and any PC with a USB port mapped as a com port. It is limited to small nema23 steppers though. I have done this using salvaged guide rails and steppers from many old printers I stripped before throwing them away.

Right now my college is budget strapped, so they are looking for an old PC they can add parallel port cards to which can run LinuxCNC. We may get lucky and have a useful system for $20 in ports. Or we may have to try several old PCs before we find one that will work. A new PC from probotix is $400ish.
 
#11 ·
I appreciate the suggestion, @MikeMa. We have plenty of old PCs laying around the building, it is just that not all PCs will run Linux/LinuxCNC well enough. So far 2 have been identified as having room for 2 parallel port cards inside. We won't know if LinuxCNC will run properly on either until we try it though. Apparently any with nVidia video on board tend to stop linuxCNC in its tracks.

4D
 
#13 ·
By small computer I meant something like this:
https://www.amazon.com/Fanless-industrial-computer-Qotom-Q190N-S01-celeron/dp/B015H4L200/ref=sr_1_10?s=pc&ie=UTF8&qid=1508468077&sr=1-10&keywords=fanless+pc+-atom+-barebones
(No experience with this one, just first example I found). No moving parts, no fans - no issues with dust.

I am a believer in having a computer dedicated to machine control, with no other applications, no network access. Once you have a reliable, working system, I see no need for any OS or other updates. If I want to check something online or listen to music, I usually have an iPad with me in shop.

The machines I have seen based on raspberry pi or such are too small, slow and flexible for things I am interested in cutting (wood, non-ferrous metals, plastics). I’m sure that small machines like the shapeoko appeal to some, but I would not be happy with one.
 
#14 ·
I've seen PCs on an HDMI stick (mini PC stick/compute stick/etc.) Which potentially could run a CNC if they had a nice little dock to provide parallel/ethernet/whatever outputs to a CNC controller box. At least one has an ethernet jack on it. My old eyes like a big screen though, and heavy duty controllers don't come tiny, so you're going to have a cabinet of reasonable size anyway. The PC used is probably ideally at the mini-ITX size like Probotix uses mainly to be just big enough for all the parallel/USB/video outputs needed to connect it all together.

4D
 
#15 ·
The saga continues....

The college had several old PCs we could use for the Nebula. The first one tried seemed to work fine once it was stripped of its drive and add-on video card. A second parallel port card was added. The motherboard had a parallel port on it. We took the solid state drive from the Nebula's PC and put it into this old PC box. It booted up fine. LinuxCNC ran fine. I had to use the configurator app to get it to see the controller, but once done I had control of the CNC...

All except that right Y motor. Just as it had been doing before all this swapping about, the right Y motor would move back but not forward toward the front of the CNC.

At this point we have a controller verified to be working fine. We have a CNC with all steppers working fine (The right Y motor works fine when plugged into the left Y jack). We have new parallel and USB cables. The only thing left to replace is the solid state drive from the original PC.

There is a good chance this CNC was left ON with LinuxCNC running over a weekend when electricians were working in our new fabrication lab. I'm guessing they cut the power at least once, which could have corrupted any open/running files including the parallel port drivers unique to LinuxCNC.

On monday I plan to format and reinstall Linux/LinuxCNC on the solid state drive. Probotix provides all the needed files/info on their wiki page.

Finger crossed.
4D
 
#16 ·
Conclusion: As I typed up my "I give up" report to Probotix yesterday morning I started to list all the things I'd done to narrow down what might be causing the problem. The format and re-install of Linux/LinuxCNC didn't make any difference. We had swapped out parallel and USB cables for new ones. We had confirmed that the controller and all the steppers DID work, except not when all connected together. We abandoned the PC for fear the parallel ports had been damaged, and the only thing moved into the new PC was the solid state drive. Today proved it wasn't the problem. We had replaced the X and Z limits switches, and...

Not the Y switches.

Realizing this I walked over to the Nebula to have a glance at the Y limit switches mounted under the side rails. On the right side (the side giving me the problem) one wire was not plugged in correctly to the outer lead on the 2 back-to-back switches. I moved the wire, turned the system back on, and found all the problems (limit errors, flawed stepper response, and e-stop errors) had disappeared.

I'm guessing a worker (networking or electrician or painter) had been in the CNC room over the weekend when this problem started. They would have moved the Nebula to gain access to the wall or wall jacks behind and beside it. I suspect in pushing it aside the worker could tell he/she might have unplugged the one wire from that limit switch under the side rail. Not knowing exactly which lead it needed to be attached to he/she guessed wrong.

So another mystery is finally solved. These CNCs are in what is essentially a public area. Access to the room they are in is not controlled. This leads to people using or even just moving them without really understanding the consequences of that action. It speaks to an area where the CNCs designed and built by Probotix could be a bit more people-proof. The X and Y limit switches in particular hang exposed where accidental contact might find them. I don't have a solution (yet), but some sort of box over these switches might insulate them from both dust AND those accidental contacts.

4D
 
#17 ·
4D,

Didn't you get a limit switch code of some kind on the monitor? That's the first thing I check when something goes wrong. That usually takes care of it. My controls are set in one spot and don't get moved, and they seem to be made to withstand everything I've thrown at them so far. Except the monitor that got knocked to the cement floor 3 times before it broke. Got a bigger one now.

Glad the mystery problem it solved. But yes, there could be a better set up for those switches.
 
  • Like
Reactions: MEBCWD and artman60
#18 ·
We did get several limit switch errors when this problem showed up. As the right Y axis would only move backward, we started testing the X and Z and of course they also created limit errors as soon as we moved them. I'd had problems with the X and Z axes at the end of last semester so those were replaced with new one. The new switches didn't make any difference (we still got errors from them) we moved on to trying other things (cables and ports and ...).

At one point I reconfigured linuxCNC to use only soft limits, so all the limit errors stopped. The right Y axis still wouldn't move forward though. The Y limit switches only popped back into my mind as I was typing an "I give up" email to probotix. I had tried literally every things else BUT those Y limit switches.

4D
 
#20 ·
Knowing how it is supposed to look is what let me spot the problem.... once I bothered to kneel down and peek under the side rails for the Y switches. In the history of the 2 Probotix CNCs in our shop those Y limit switches had never been a problem before. I had no reason to imagine that one might have been re-wired wrongly. Yet that is exactly what had happened... over a weekend when no one should have been using the CNCs.

I'll know better next time, but hopefully all new-building work is done now and there won't be a next time. I suspect no other Probotix CNC has suffered the odd problems mine have. Mine live in a college where for the last 2 years they've been moved 4 times. They've gone from a classroom shop under my oversight to a larger all-college shop where potentially any student/faculty might use them without going through me.

4D
 
#24 ·
One thing this episode reminds me is that over the years I've been using the Nebula it has always suffered from limit switch errors/warnings when we've tried aggressive cuts in dense hardwoods with it. Vibration bounces open the switches. Probotix has come up with a "soft limits" option in their configurator that should stop the switches from stopping a cut. That doesn't eliminate the vibration, but they are also working on that problem. They have recently been showing (on facebook) photos of a new 60mm x 60mm gantry beam that is thicker walled and considerable stiffer than the current 30mm x 60mm beam they use. I don't know if/when this will become the standard beam on all new Nebulas or Asteroids (or all their CNCs) but they imply you can get the taller beam by request on any new CNC you order from them.

4D
 
#27 ·
One thing this episode reminds me is that over the years I've been using the Nebula it has always suffered from limit switch errors/warnings when we've tried aggressive cuts in dense hardwoods with it. Vibration bounces open the switches. Probotix has come up with a "soft limits" option in their configurator that should stop the switches from stopping a cut. That doesn't eliminate the vibration, but they are also working on that problem.
I have read about increasing the voltage levels for limit switches to eliminate false openings.

Steve.
 
#28 ·
The spring/flex in the switch parts seems to fatigue over time, meaning they become easier and easier to "bounce" open the older they are. I can't see any way more voltage would change the metal to stiffen it/slow down this fatigue. Of course more heat and some solder could close them permanently, but that would defeat their purpose. :( I suppose some increased voltage might arc over a small gap though.

There are other types of proximity switches, but I'll guess they would be too expensive in the numbers needed for each CNC axis.
Their soft limits option uses the switches only to find "home" where it looks up the axis extents to keep the head contained from there on. The axis extents can be found and edited in the nebula.ini file (or whatever name is used for your configuration). I change mine to reflect the actual distances my head can move in X and Y directions. It is slightly more than the default setup permits.

4D
 
#29 ·
Each axis limit switch has a debounce variable in the HAL file. This is mainly used, I think, for signal interference. Could possibly increase the setting a little and see if it helps issues caused from vibration.

# debounce the y-axis switches and connect them to signals
net switches-y1-raw <= parport.0.pin-11-in
net switches-y1-raw => debounce.0.0.in
net switches-y1 <= debounce.0.0.out
net switches-y2-raw <= parport.0.pin-12-in
net switches-y2-raw => debounce.0.1.in
net switches-y2 <= debounce.0.1.out
setp debounce.0.delay 10

Another option possibly would be to remount one limit switch on each end of the bed and put the switch target on the bottom of the gantry (where switches are now). This would require some additional wiring, though.

Without much current running through those switches, it doesn't take much to open the contacts.

Just some thoughts...

Dave
 
#30 ·
Your idea is appreciated, Dave. At the moment the "soft limits" option in their configurator has solved the switch problem for now. We had devolved to using a jumper to bypass the X switches after homing the machine at one time in its history. Soft limits does essentially the same thing.

The Z switch is the one I've had to replace the most often though. My students tended to bring the router up until smashing it before jogging the head around. I'd find the switch's internal bar/lever bent after a while.

4D
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top