Mike's 6-Axis Articulated Robot

[macardoso]

Well, first bit of bad news. From an earlier post I mentioned the J6 motor was binding up and was extremely hard to rotate by hand. These should be buttery smooth. Anyways, I just learned that the motor is fragged. No hope of saving it.

I used a bearing puller to gently remove the stuck pulley.


There was some dark grit right at the rubber shaft seal, but nothing concerning. The seal surface was in good condition. Some grease behind the seal likely improves water protection.


Pulling the front casting revels the stator windings. The front bearing spins smoothly.


Removing the rest of the covers allowed me to see what was going on inside. At this point I suspected a fully seized rear bearing.


But alas, removing the stator coils revealed a shattered rare earth permanent magnet on the rotor. I have no way to repair this. There is no sign of impact to the motor or a crash of the robot. I can only guess that this was a defective motor or the motor ran away at a high overspeed and destroyed the rotor with centrifugal forces.


So the J6 motor is trash. My options are to replace with an identical part, or to replace it with a comparable servo motor like an Allen Bradley TLY series motor with -B feedback (Tamagawa 17-bit serial).

If I want to run this on the original control, I think I will need to find a 1:1 replacement. Sucks...

 

[macardoso]

Well double bummer. This motor is not easy to find. Found one ebay listing for $730 from China (I put in a super lowball offer), and a few from some sketchy online sites. Messaged them all but I don't expect to be able to get the exact one.

Another option would be to find the corresponding Panasonic motor and rework the connector. If I found an electrically identical unit, then I could fix the mismatched connectors or mechanical interfaces easy enough. The Panasonic MSMA5AZS1A looks to be the identical motor with different connectors on the wires, different wire/housing interfaces, and a straight shaft vs 2 flats. This would likely be my best bet. I currently see 12 under $200 from China on eBay.

Finally I can look for what motor options I have floating around in my box of goodies. Without home switches, it will need to be an absolute encoder. Also, unless the motor I have has the exact same electrical properties and encoder, I'd never be able to use the original control box. Ugh.

 

[strantor]

For replacement of the motors and drives, you might want to consider something a little more agnostic. I am not very familiar with Panasonic servos but if I'm going to assume it's one of those proprietary affairs where you use the Panasonic software for the Panasonic drive, and in the software you have to select a Panasonic motor part number in the list which tells the drive everything it needs to know about inertia, LCR, brake, etc. and there is no provision for entering these values manually.

From your previous posts I take it you're an Allen Bradley/Rockwell guy so I'm sure you're familiar with the Ultra 3000 and Ultra 100 drives. These are dime a dozen on eBay and great drives you can use with just about any servo motor.

Do you have any reason for sticking with the Minas drives?

 

[macardoso]

For replacement of the motors and drives, you might want to consider something a little more agnostic. I am not very familiar with Panasonic servos but if I'm going to assume it's one of those proprietary affairs where you use the Panasonic software for the Panasonic drive, and in the software you have to select a Panasonic motor part number in the list which tells the drive everything it needs to know about inertia, LCR, brake, etc. and there is no provision for entering these values manually.

From your previous posts I take it you're an Allen Bradley/Rockwell guy so I'm sure you're familiar with the Ultra 3000 and Ultra 100 drives. These are dime a dozen on eBay and great drives you can use with just about any servo motor.

Do you have any reason for sticking with the Minas drives?

Well you seem to have a great view into my head

I won't comment about whether or not I work at Rockwell since I am often posting about hacking stuff to make it work in ways they may or may not want people doing. I'll leave that to your imagination.

This time around, I actually have the matching original robot control for this robot. It uses completely custom boards and drives and does NOT contain any panasonic drives that I can see. Everything is DENSO branded and designed I am assuming. The servo amplifiers are "dumb" and receive (from my initial look into this) a torque command and produce AC output. These amplifiers do not have any closed loop control circuitry that I can identify. Based on this, I believe the DENSO motherboard running i386 DX, i387 DX, and i960 processors handles all the serial communications to the encoders, all the closed loop motor control, and all the motor model parameters. I can almost guarantee that stuff is not exposed to the user. If I want to have the option to use the original control, then I need to replace the motor with an identical or electrically identical motor.

I have a handful of Ultra 3000 and Kinetix 2000 drives from work that I can play with. The Ultra 3000, when used standalone, allows you to create custom data for any motor (provided you can get it) and run a wide variety of feedback types. From the manual:


The last robot I had ended up with Incremental w/ Halls after my feedback converter board. The "Intelligent" encoders referenced in the Manual are Sick Stegmann Hiperface encoders. Unfortunately, these drives do not support Tamagawa Absolute Serial encoders. Additionally, if you try to run the drive over SERCOS fiber-optic (the easiest way to get the drive to get commands from the PLC in "integrated" mode) you lose the ability to custom define motors without Rockwell supplied Custom Motor Files (.cmf). I figured out how to generate my own, but there is a lot of firmware level nonsense that makes this really tricky to get working, hence why Rockwell prefers to supply known good CMF files for their partner companies only.

The Kinetix 2000 suffers from the same CMF issue, although I have somewhat worked around this in the past. These rack based drives would be perfect for running these motors and even natively support Tamagawa 17-bit serial encoders, although the jury is still out if the Rockwell supported Tamagawa encoders match, or are compatible with, the Tamagawa encoders in my motors. I also see that the Tamagawa encoders on Rockwell motors are programmed with the motor data, so I don't know if the firmware will allow me to use unprogrammed motors, or if I can program them myself with the proper meta-data to make this all work.



So the last comment was replacing them with some agnostic motors and yes, this is an option, but it means the robot will never run on the original control again, and it would be difficult (and expensive) to find the right sized motors with compatible mechanical dimensions and electrical performance. I'd have to find matching servo drives for these as true agnostic motors really do not exist (more money) and I'd have to find a way to get these controlled by the Rockwell PLC (needed for robot kinematics) which pretty much limits me to analog interface modules (1756-M02AE).

Rockwell has new drives, the Kinetix 5100 and Kinetix 5300, which support Tamagawa 17-bit serial feedback and are on the more "open" Ethernet/IP platform which allows users to avoid the CMF files and enter data by hand, however I don't have any of these, and since they are newer, I doubt I'll get any to screw around with. I would probably have the same meta-data issue as well.



Finally, reusing the existing motors only works if Panasonic is really nice and provides me the (sometimes proprietary) motor electrical characteristics. I might be able to measure some, but I've never done that before.

Interfacing to the servos is the hardest part of this robotics nonsense in my opinion.

-Mike

 

[strantor]

Well you seem to have a great view into my head

I won't comment about whether or not I work at Rockwell since I am often posting about hacking stuff to make it work in ways they may or may not want people doing. I'll leave that to your imagination.

Well my imagination didn't have to work very hard to fill in the blanks there. Maybe I'm just an imaginative guy...

This time around, I actually have the matching original robot control for this robot. It uses completely custom boards and drives and does NOT contain any panasonic drives that I can see.

My bad. I thought I saw something about minas drives earlier. I was skimming, so maybe you were commenting on the lack of them. Or my imagination filled in the blanks again.

Everything is DENSO branded and designed I am assuming. The servo amplifiers are "dumb" and receive (from my initial look into this) a torque command and produce AC output. These amplifiers do not have any closed loop control circuitry that I can identify. Based on this, I believe the DENSO motherboard running i386 DX, i387 DX, and i960 processors handles all the serial communications to the encoders, all the closed loop motor control, and all the motor model parameters. I can almost guarantee that stuff is not exposed to the user.

Gross.

If I want to have the option to use the original control,

Preserving the original control will be at the bottom of my list of priorities when I get a robot unless it's delivered as a complete working system. Different strokes for different folks. I find much more gratification in designing my own stuff, then trying to patch up someone else's. Especially when they've gone so far out of their way to prevent you doing so.

then I need to replace the motor with an identical or electrically identical motor.

I have a handful of Ultra 3000 and Kinetix 2000 drives from work that I can play with. The Ultra 3000, when used standalone, allows you to create custom data for any motor (provided you can get it) and run a wide variety of feedback types.
View attachment 394815

So the last comment was replacing them with some agnostic motors and yes, this is an option, but it means the robot will never run on the original control again, and it would be difficult (and expensive) to find the right sized motors with compatible mechanical dimensions and electrical performance. I'd have to find matching servo drives for these as true agnostic motors really do not exist (more money) and I'd have to find a way to get these controlled by the Rockwell PLC (needed for robot kinematics) which pretty much limits me to analog interface modules (1756-M02AE).

Rockwell has new drives, the Kinetix 5100 and Kinetix 5300, which support Tamagawa 17-bit serial feedback and are on the more "open" Ethernet/IP platform which allows users to avoid the CMF files and enter data by hand, however I don't have any of these, and since they are newer, I doubt I'll get any to screw around with. I would probably have the same meta-data issue as well.

[...]

Finally, reusing the existing motors only works if Panasonic is really nice and provides me the (sometimes proprietary) motor electrical characteristics. I might be able to measure some, but I've never done that before.

Inertia can be measured. LCR can be measured. Back emf can be measured. You can get an LCR meter on amazon/ebay for cheap. I've got one, not sure how accurate it is, but it's better than nothing.

An idea that just occurred to me (disclaimer: I've never done this): when you auto tune an induction motor with a VFD in sensorless vector mode, it measures all these electrical characteristics of the motor for you. If you autotune in sensored vector it also measures the inertia. A lot of higher end VFDs support permanent magnet motors. I'd bet $20... no, make that $5, that if you connect your servo motor with a 4096ppr incremental encoder on its business-end shaft to just about any modern full-featured VFD and run an auto tune, it would spit out most if not all the data you would need.

P.s. check out SEW servo drives. They allow you to put in the numbers for non-SEW motors and there's no *.xyz files involved.

 

[macardoso]

I did have the thought that this catastrophic failure is likely the reason this robot was removed from service. Batteries were last changed in 2011 which puts the date of failure more recently than that. Support and spare parts ended for this robot 15 years ago so the previous owners probably couldn't easily repair the unit. Wish nothing was wrong, but given that I found such a big failure I doubt there are any other major issues lurking. Hope I don't jinx it...

 

[strantor]

I did have the thought that this catastrophic failure is likely the reason this robot was removed from service. Batteries were last changed in 2011 which puts the date of failure more recently than that. Support and spare parts ended for this robot 15 years ago so the previous owners probably couldn't easily repair the unit. Wish nothing was wrong, but given that I found such a big failure I doubt there are any other major issues lurking. Hope I don't jinx it...

Where I work there are 2 robots and the encoder batteries get changed annually. Our PM schedule is rather... well, like a list of suggestions rather than decrees. Compressor oil might get changed at 2k hours like the manual says or it might get pushed back until someone has time to do it; nobody makes time for that. Air filters, bearing replacements, chain stretch measurements, gearbox oil changes, belt replacements, etc., all of it. But the encoder batteries different; they get changed on the day they're supposed to, with religious fervor. I suspect this behavior is driven by past experience and anywhere else that has robots and maintenance guys that have been around a while, would probably operate the same way. So, I would estimate time of death closer to 2011 than to 2022.

 

[Cadillac STS]

Can you use a motor with a double shaft front and back and put the original encoder back end on it? Get a motor same specs and just replace that part leaving all the feedback servo encoders in place?

 

[macardoso]

Preserving the original control will be at the bottom of my list of priorities when I get a robot unless it's delivered as a complete working system. Different strokes for different folks. I find much more gratification in designing my own stuff, then trying to patch up someone else's. Especially when they've gone so far out of their way to prevent you doing so.

So my thought is this. The original control is more or less ready to run (especially since I own the teach pendant). I'm missing these components (items in bold are things I'll need regardless):

• Controller Power Cable
• Robot Motor Cable
• Teach pendant Cable
• Robot Batteries
• Control Batteries

So I'm pretty close to a minimum viable system, although the full system needs lots of other optional components. Missing cables could be homemade if needed. Having the control working as-is would mean I could teach the robot a program and move it around. I'd be a bit limited in complexity of programming and without the floppy drive backup system, I could only have 1 program and no parameter backup, but still functional.

My last robot took me about 18 months to get running on an AB PLC. I think I could go faster this time, but there are a lot of unknowns on the path, and having it working with the old control would be a big plus to me - even just for fun. I think I'd still want to get it running on the AB PLC due to the advanced capabilities to take advantage of.

Inertia can be measured. LCR can be measured. Back emf can be measured. You can get an LCR meter on amazon/ebay for cheap. I've got one, not sure how accurate it is, but it's better than nothing.

An idea that just occurred to me (disclaimer: I've never done this): when you auto tune an induction motor with a VFD in sensorless vector mode, measures all these electrical characteristics of the motor for you. If you autotune in sensored vector it also measures the inertia. A lot of higher end VFDs support permanent magnet motors. I'd bet $20... no, make that $5, that if you connect your servo motor with a 4096ppr incremental encoder on its business-end shaft to just about any modern full-featured VFD and run an auto tune, it would spit out most if not all the data you would need.

P.s. check out SEW servo drives. They allow you to put in the numbers for non-SEW motors and there's no *.xyz files involved.

Measuring that data could be pretty reasonable. I have access to lots of drives to borrow from work. doing that might be possible. There are some touchy parameters, like the voltage and inductive time constants that I'd much rather get published data for, but if I can't, there is a possibility to measure them.

I'll check out the SEW drives for sure. One of the biggest limitations is sticking to an architecture that incorporates the robot kinematics. The full path planning and kinematics pipeline is too complex for a hobbyist to reasonable code from scratch so, at least as I know right now, I'm stuck with:

• Original robot control
• Rockwell PLC (one of only a few companies I know that do "open" robotics integration) with a small number of compatible servo drives. Analog interfaces won't permit use of the existing encoders.
• Laptop running LinuxCNC and special hardware interfaces
• Some application of ROS (Robot Operating System) which I know nothing about.

 

[macardoso]

Where I work there are 2 robots and the encoder batteries get changed annually. Our PM schedule is rather... well, like a list of suggestions rather than decrees. Compressor oil might get changed at 2k hours like the manual says or it might get pushed back until someone has time to do it; nobody makes time for that. Air filters, bearing replacements, chain stretch measurements, gearbox oil changes, belt replacements, etc., all of it. But the encoder batteries different; they get changed on the day they're supposed to, with religious fervor. I suspect this behavior is driven by past experience and anywhere else that has robots and maintenance guys that have been around a while, would probably operate the same way. So, I would estimate time of death closer to 2011 than to 2022.

Yeah that's what I've learned. If the batteries go dead and you shut down the power to the robot, you're SOL. Have to do a full recalibration (which might mean hiring someone to come in). If the control goes dead, then you better have that floppy disk with the parameter restore or it is a dumb box.

Recalibration with this robot seems trivial, which is good news, but the missing controller parameter files is a real risk to getting this working with the original box.