[Voluntary_Exchg]

Hi All:

I am building a new mill from scratch, and will be converting a PM25MV when I can get one in.

I will be using high precision ball screws (hard to source reasonably off ebay), and I already have an adequate supply of SHS25 rails and blocks.

I will be using Parker servos series BE with J type encoders (2000 ppr). I will be running them using Parker Gemini GV U6E and GV U12E controllers.

Ultimately I might end up with as many as 8 axis on the scratch build, and maybe even eventually 8 axis on the PM25MV.

First off I must decide between the Klop/Kmotion setup or I am also considering a Galil DMC-4080 using CamSoft. Any other high performance suggestions? The second option is much more expensive. Naturally I expect this forum area to be partial to the Kflop, but any of you have much experience with Galil to compare?




Some general Gemini specs:

servo updates: 50 micro seconds
Velocity and torque modes: 12 bit +/- 10 V

Enable, Reset: 5-24VDC Selectable, 1/3 1/3 1/3 voltage switching threshold
Neg/Pos Limits 5-24VDC Selectable, 1/3 1/3 1/3 voltage switching threshold
User Fault 5-24VDC Selectable, 1/3 1/3 1/3 voltage switching threshold
Encoder 8 MHz Post Quadrature



Regarding wiring for the Gemini controllers for step & direction inputs, here is more info:

Gemini Step & Direction Inputs


The step & direction circuits on the Gemini are designed as RS422 differential. Because the inputs are not optically coupled, they cannot accept 24V, except with the GC-SDA adapter kit (described below). As with any single-ended signal, it is much more susceptible to noise when it is NOT differential. Parker recommends using a differential driver circuit.

To use with a 5V single-ended indexer/controller, do not connect anything to the Step- input. Connect the 0V or GND side of the indexer to Digital Ground. Connect the Step signal of your indexer to the Step+ of the Gemini. Use the same procedure for the Direction signals.

If you measure across the Step+ and Step- with nothing connected, you will measure a 2.5V difference. This is because the plus input has a 681 ohm pull-up resistor to +5V, and the minus input has both a pull-up and pull-down resistor. This allows it to accept either differential or single-ended signals. When wiring it with a single-ended input, a 5V signal on Step+ will be 2.5V above Step-, thereby registering as "High;" likewise, a 0V signal on Step+ will then be 2.5V below Step-, and will be seen as "Low."

It is possible to drive the Gemini with 24V Step-and-Direction inputs, using the GC-SDA I/O adaptor. This is a small, external PCB with user-selectable resistors that attaches to the Gemini's 50-pin I/O connector. It adapts the Gemini to +24vdc, +12vdc or any other voltage level. This is listed on the website as an accessory option for the Gemini. http://www.parkermotion.com/manuals/...-SDA_Guide.pdf

I was trying to decide between step-direction/cw-ccw modes or using torque mode? I think step-direct would be easier to establish, but I am thinking torque mode might perform better?

Here's this on Gemini update rates in a Torque Loop:

The Geminis (-U3E, -U6E, -U12E, -H20E) have a 62.5 micro-second update rate for the Torque Loop.

All the GV(6) drives have a 250 micro-second update rate for the Velocity and Position Loops.

The GV, GV6, GT, and GT6 all have I/O update rates of 1 ms.


Eventually the spindle and lathe axis might be run by velocity and absolute position servo motors.

So what dynomotion would will I need?

I am thinking that If I use step-direction then I might get by with just a kflop, but might still need another board for all the bells and whistles. But if I want to use the torque mode then I will need maybe a Kanalog or maybe a Konnect? Do you have any wiring suggestions?

Please comments/suggestions?

[691175002]

Since you can always fall back to Step/Dir mode, virtually all motion controllers will be sufficient. If you want to close the servo loop on the controller side you will need an analog interface. I would be surprised if a hobbiyest could measure a difference in the performance of those two systems so your decision should be based on other factors.

Over time I've come to realize that an important contributor to this decision is the availability of community support and the number of existing users. Mach3 is objectively awful, but so many people use it that any problem can be solved in a few minutes with google. People have implemented and documented ATC, probing, conversational, and more.

KFlop has a very bare bones interface and many basic features must be implemented in C. I have no experience with CamSoft but their webpage makes me want to vomit and I have a suspicion there won't be much community support.

I've recently been looking at switching to LinuxCNC. Its community is active and helpful, and the software is very powerful. LinuxCNC has roughly 800 pages of well written PDF documentation plus a wiki, forum, and chat room. With Mesa cards you can add any number of analog servo interfaces, encoder inputs, IO pins and more. Furthermore, Tormach just switched to LinuxCNC so we can expect it to grow even faster going forward.

[Voluntary_Exchg]

Will check out Mesa cards. And will take a closer look at LinuxCNC. Haven't used Linux in ages.

I am now thinking that velocity mode is the way to go. It seems that is preferred over torque mode for cnc? Do you folks agree? Even though this is starting out as a hobby, I want to do the job right so that it could turn into pro work later. So I would want the best performance I can get at a moderate price.

[Voluntary_Exchg]

If it is true that Linuxcnc wants a real-time connection to the motion control board, then I don't want this approach. I want to let the host send info in a non-time-critical manner by buffering high level commands to a multi-axis control board that is working under real-time constraints. I think that means kflop, or some other brand, or Galil.

[Voluntary_Exchg]

Now it looks like I will go with galil 4080 and maybe Mach4 when the post processor for Galil is ready . No advocacy for kflop on the Dynomotion forum!?

What I would like is a controller with the performace of a galil 4080 or better, combined with full built in g-code support. The Eding does not appear to be up to this level of performance, and I need to run 8 external servo drives and I don't think Eding can do that.

Aerotech is claiming that I can use their Nservo Servo Controller unit to control my legacy servo drives/motors. They have g-code compatibility. I guess you use their software controller (automation 3200) front-end to link to that. I think it is more expensive then a used Galil + Mach. Has anyone used Aerotech with Nservo?

[bhurts]

I love my k-flop but I don't know if it's right for you or not. Realistically I don't see how anyone can give you a good recommendation on any control without knowing exactly what your trying to do. Kmotion cnc is only good for 6 axis simultaneously as far as I know. You would need to say what each axis you have planned is going to be to even make a guess. At that point what software are you going to use to program the machine. This sounds really ambitious but without knowing more I'm not sure how practical it is. What is the goal of this project? Do you have a specific use in mind? No matter what you use to control the machine your getting into the realm where you will get little support from the manufacturer no matter who you purchase from. The after sales support would be very costly on any manufacturer unless you are highly qualified to do most of the work yourself. Probably many times the cost of the control on even the higher end choices. I'm not knocking you or assuming you can't do it. I just feel like there's not enough info to recommend a solution.

Ben

[Voluntary_Exchg]

x,y,z - 4 axis (slaving either on x or y), lathing/rotation =1 axis, spindle (servo controlled) =1 axis , 2nd rotational axis =1 axis. 1 extra axis for second spindle or maybe for tool changing or maybe for other purpose. If rigidness adequate, used for complex milling, other wise for less accuracy jobs.

[Eddard_Stark]

If it is true that Linuxcnc wants a real-time connection to the motion control board, then I don't want this approach. I want to let the host send info in a non-time-critical manner by buffering high level commands to a multi-axis control board that is working under real-time constraints. I think that means kflop, or some other brand, or Galil.

This was one of the reasons I went with KFLOP.

While Dynomotion doesn't have the community that LinuxCNC does, there is reasonably active community between here and the Dynomotion Yahoo group. What sets KFLOP apart is the active role Tom Kerekes takes in Q and A on the forums. AFAIK he designed the KFLOP and is extremely diligent about answering questions online. Extremely.

I am just getting my machine going (as in I have only run G-Code test my motors so far), but I am pretty sure the KFLOP can do what you described. I am using Channels 0 - 3 for X, Y, Z, Xslave and am about to setup channel 4 for my spindle. I have 3 open channels left on my config screen for a total of 8 axes. Post your question to the Yahoo Group and someone there, probably Tom, for sure can tell you if it is capable of doing what you want.

BTW, will you be documenting your build? I would love to see it. As I mention I am finishing a 5' x 10' stepper powered CNC router, but I already have my sights sets on more ambitious projects including a scratch built VMC. And why are you converting the PM25 also? I considered converting my BF20 clone, but I would rather put my effort into something more robust. At least that is what I am thinking today!

[Voluntary_Exchg]

Ya will probably try to get some photos once I get past the hardware buy stage. Then I can combine another hobby I love, photography. So far all my used servos and drivers work . With the Galil you have to be meticulous to get the proper connection options that work with your encoders and drivers. Most of the used stuff has the unwanted internal amps. Just want a straight no-amp 4080 that can handle standard quadrature encoders.

[bhurts]

Knowing what your axis setup will be roughly helps a lot toward making recommendations. I believe k-flop will do what you are describing. My reservations were due to how many axis would be controlled simultaneously in a g-code program. Kmotion cnc only supports 6 axis in g-code. Slave axis are not part of the equation they just follow the master. Spindle axis even if they are used for threading and tapping will also not be part of the coordinated axis system. From that point of view I believe you can make it work with k-flop and possibly kanalog depending on how you choose to control your motors. Hopefully this link works. Check out some of the other videos of his machine as well. It is run with k-flop and kmotion cnc.

My Home built 6-axis lathe , Checking A axis tracking - YouTube

Ben

[Voluntary_Exchg]

Can a g-code file be loaded directly into kmotioncnc and executed? (no need for mach then)

[bhurts]

Sorry that video sucks. It doesn't really show the machine. Try this.

My Home built 6-axis Mill, Cutting Gear Teeth with slitting saw - YouTube

Ben

[Voluntary_Exchg]

Well I checked the webpage and that is what kmotioncnc is for!

[bhurts]

Here's a link to his channel.

C Mac - YouTube

Ben

[bhurts]

Yes you are correct. Kmotion cnc is like mach3 but free. It is in my opinion superior in many ways.

Ben

[Voluntary_Exchg]

For servos, does Kflop do better with velocity, or torque? I presume feed-forward is implemented. Are complex trajectories like cubic spline a matter of using custom c code? If I wanted a command for parabolic trajectories in 3d, I would have to write a c program or maybe find one in a library somewhere? Has anyone seen a KFlop in action doing 5 axis simultaneous milling on a complex part like a turbine wheel or a 3d sculpture?

[TomKerekes]

Hi Voluntary_Exchg,

For servos, does Kflop do better with velocity, or torque?

I don't think one can say one or the other without additional information. If the Amplifier implements a good velocity loop then they should be similar. Both are supported.

I presume feed-forward is implemented.

Yes. See:

Step Response Screen

and

Attachment 291036

Are complex trajectories like cubic spline a matter of using custom c code? If I wanted a command for parabolic trajectories in 3d, I would have to write a c program or maybe find one in a library somewhere?

I'm not sure exactly what you have in mind but the KMotion Libraries don't specifically have spline or parabola commands. You would need to create those yourself. KMotion/KFLOP is optimized for short line segments to be handled at high rates. This allows any arbitrary multi-axis path to be easily defined (ie from any formula such as a Spline or Parabola if desired). We feel our Trajectory Planner is one of the most powerful at lookahead, filtering, tolerance, smoothing, rounding, etc, while optimizing all portions of the path without violating the axes constraints.

Has anyone seen a KFlop in action doing 5 axis simultaneous milling on a complex part like a turbine wheel or a 3d sculpture?

I'm not aware of any. Here is my old favorite albeit slow and only a prototype:



HTH
Regards