[Azalin]

It's nice reading a cool project every day here.

Very nice machine canuck.

[the_canuck]

I've been reading up on dual loop control and magnetic linear encoders. I'm going to give it a shot on the x and y to see if i can further remove any chance of backlash creeping into my cuts.

I plan on adding magnet linear encoders to the x and y and then hooking them up to Linuxcnc and changing the step generator to a velocity mode, and running a PID with the encoders as feedback in the loop. In the current state my machine is run from linuxcnc like a stepper setup with no control loop through linuxcnc. The servo drivers I'm running have a control loop but they don't go through Linuxcnc.

The goal is to eliminate the possibility of backlash and hopefully improve some problems I've been experiencing which I believe are related to backlash in the rack and pinions.

The z-axis is a C1 ballscrew and I believe I don't currently have any issues there.

To make the upgrade happen I'm also going to upgrade the size of the x-axis cable chain as it's really full. I'll also need to add another breakout board from Mesa so I can read the encoder signals.

Programming everything to work is going to be the hardest part. I've seen some Linuxcnc examples that are pretty similar to what I want to do. I think it will be doable.

My est. for parts cost of the upgrade to be $750 using used encoders found on ebay. The rest I will have to buy new.

Andrew

[pippin88]

Why magnetic rather than glass?

[the_canuck]

I need an 80" encoder for the x-axis. Glass seems to be much more expensive and fragile.

Andrew

[Arbo]

How is pricing on the renshape board? And would it be good as a negative to make a positive mold from?

[the_canuck]

How is pricing on the renshape board? And would it be good as a negative to make a positive mold from?

Adhesives Distributor, Dow Corning Distributor of the Year. The adhesive and sealant resource for engineers. in Denver has some in stock. Shipping would be expensive as the stuff is heavy. Prices depend on dimensions and density. 460 1" x 20" x 60" is in the order of $280. The thicker stuff is more expensive.

460 machines better than 440 and has a better surface finish. You could use it direct as a mold if you are careful. It sands and polishes pretty easily.


Andrew

[the_canuck]

Parts are in for my linear encoders. I was waiting on the Mesa breakout board. Now to find time to get it done.

Probably not this week unfortunately.

Andrew

[the_canuck]

Update:

Finished out replacing the x-axis cable chain to a 2x larger one and ran the new wiring for the linear encoders a couple weeks ago.
Then last weekend, I spent the whole time cleaning the shop and putting up more shelves and finally unpacking some boxes from my move nearly 3 years ago.
Today I added the connectors and Mesa 7I85 in a separate small box and got everything wired and box mounted to the machine.

What's left, is to redo my HAL files, flash the new custom bit file to the Mesa 6I25, install the magnetic strip encoders, and install the magnetic readers on x and y axis.

Andrew

[the_canuck]

As of the weekend, I have all the new bits installed and working.

The RLS magnetic encoder system is pretty awesome. I highly recommend it. I currently am getting 25400 steps per inch out of the encoders.

6I26 with the 7I85 works seamlessly. I was able to get LinuxCNC up without errors pretty quickly after the updates to the machine. I then copied over all the HAL and INI settings for the encoders. My setup is basically 'like' a closed loop stepper configuration with the DMM servos and DYN2 drivers and rotory encoder controlled by LinuxCNC with step/dir setup. The magnetic encoders are feeling a LinuxCNC PID loop that is a velocity-PID output that feeds the step/dir commands to the servo drives.

I'm now in the process or redoing my control loop tuning. Now that I can see the loop responses on HAL Scope it makes it so much easier to setup the servos. Just for this fact alone I recommend adding the linear encoders.

My plan of attack for the tuning is to tune the servos with rotory encoders first. My old tune looks to be too aggressive and was causing issues on first try with the new loop working. I was getting too much overshoot and steady state oscillation. Round 2 I'll work on the LinuxCNC loop that uses the new encoders. After one session of tuning I'm down to a following error of =/- 0.006" on reasonably quick movements and direction changes on the y-axis. I would like to get that down some more but there's also some backlash to take case of in there as well.

The main issue I currently have with tuning is a periodic oscillation during constant velocity moves. My current theory is I have the servo tuning still too aggressive which reacts poorly to the rack and pinion transfer velocity 'noise'. Tonight I'll try lower the gains on the DYN2 drive and see if it improves overall following error.

I will be stoked if I can get the following error to +/- 0.002" and still have reasonably quick loop responses. I now have lots more 'nobs' to turn and being able to use the HAL scope is an order of magnitude easier to dial in the loops.

Once I get the y-axis done I'll start in on the x-axis which is a bit harder to do as I dual drive that axis so I have 2 servos to tune plus the linuxcnc PID loop.

I'll post some screen shots and my HAL and INI files once I have the tuning dialed.

Andrew

[Arbo]

Made any aluminum molds with the beast yet?

[the_canuck]

Not yet. Planning on converting to ball screws before doing more molds. I have a design completed and Aluminum plate already to go once the machine finishes it's upgrade.

[the_canuck]

Here's what I've been up to for the last few months. I've been busy, just not on the machine.

Shop upgrades and building some RC models.

[Arbo]

Not yet. Planning on converting to ball screws before doing more molds. I have a design completed and Aluminum plate already to go once the machine finishes it's upgrade.

Cool.

I am about to get a sheet of solid surface and start cutting molds out of it ... dug out a design dale did a while back, gonna start with that.

[the_canuck]

Some minor progress...

Replaced the C-clamp that been holding my z-axis cable chain in line since the first movements. Band sawed out a Corian bracket and installed it a few days ago. Works great and looks a bit better than the orange C-clamp. I'd take a picture but my phone battery is dead.

I've been talking with HIWIN on some of their ball screws but still haven't completely made up my mind. Their sales rep may stop by the shop this week if he gets time. He's in town from Chicago which is pretty cool of him to reach out.

Andrew

[Arbo]

All forward progress is positive progress. Look forward to seeing some aluminum molds.. one of us has to pull it off.

[moparx12]

Here's what I've been up to for the last few months. I've been busy, just not on the machine.

Shop upgrades and building some RC models.

Ball screws on x and y? What size/lead were you thinking of going with?

[the_canuck]

Ball screws are on order from Hiwin in IL. Ordered 2020 size (20mm lead, 20mm diameter) 2 start C7 rolled with R1 rotating nut assembly for both x and y axis. I'm not too worried about the C7 rating as I already have linear encoders on the x and y which I should be able to use to compensate for the mediocre accuracy of C7 screws. Ground screws were out of my budget and are probably not needed anyway.

I also picked up some AL plate to make all the servo motor mounts, etc. I found some basic 20mm shaft mounting brackets from China I hope to use to hard mount the screws to the machine. Rest of the brackets and spacers I will have to make. For the pulleys that attach to the nut, I've yet to find some that will fit correctly and will probably have to modify whatever I pick up. I'm leaning towards a 2:1 belt reduction setup with a 20 tooth pinion on the servo and a 40 tooth on the rotating nut assembly. I'll lose a bit of speed over the rack and pinion but fingers crossed I fix my accuracy and tuning issues.

I'm in waiting mode until the screws arrive.

Andrew

[the_canuck]

Ball screws arrived Hiwin this afternoon. They are 4 start higher rated version and they look pretty nice. They should be a great improvement over the rack and pinion setup. Now the fun part of the project begins. Making the brackets with my current setup before I tear it down. I have a friend who is a full time machinist that has volunteered to help on the machining of the pulleys which will speed things up.

I checked the Chinese mounting brackets and they are not going to work, so will have to make up some as well. I'll be able to make them exact size so they will work better than the off the shelf ones anyway.

Picture is from Hiwin's website. Mine are slightly different but I'm sure you get the idea. I'll post pictures of mine later.

Andrew

[Arbo]

Ball screws arrived Hiwin this afternoon. They are 4 start higher rated version and they look pretty nice. They should be a great improvement over the rack and pinion setup. Now the fun part of the project begins. Making the brackets with my current setup before I tear it down. I have a friend who is a full time machinist that has volunteered to help on the machining of the pulleys which will speed things up.

I checked the Chinese mounting brackets and they are not going to work, so will have to make up some as well. I'll be able to make them exact size so they will work better than the off the shelf ones anyway.

Picture is from Hiwin's website. Mine are slightly different but I'm sure you get the idea. I'll post pictures of mine later.

Andrew

I was wondering how they would be better than R&P, because such things are slower... but then I remembered you are using massive servo's that can make those things fly.

[mactec54]

I was wondering how they would be better than R&P, because such things are slower... but then I remembered you are using massive servo's that can make those things fly.

Depending on the pitch of the Ballscrew, being a 4 start Ballscrew , it could easily be as fast as a R&P machine being much smoother and way more accurate