Help with Mesa card seletion

Brief history. Recently converted a VMC to Mach. Gutted the entire machine, kept DC brushed servo motors, encoders, 160vdc power supply and limit switches. Machine had a geneva type tool changer with 2 dc motors for in/out and fwd/rev.

Got everything working and realized that I was very uncomfortable with the way Stop and Feed/hold work in Mach. I was using CSLabs controller.

So now I am looking at removing windows PC and CsLabs hardware and redoing with LinuxCNC. The mesa boards seem to be the way to go. Trying to read over their site but there are just so many different boards and I really just can not pin down what I need.

So, I have 3 servos that run from three new DG4S-16035 (CncDrive). The encoders are wired directly to these drives. 5 volt differential with no index. The step and direction signal is single ended. I currently have a differential to single ended converter for step and direction. They have basic 5 volt fault input and output available.

I currenty have about 16 inputs and 13 outputs for various limits, switches, coolant....
Currently, all my limit switches and hard switches are wired with 24vdc. This could be changed to 5 volt pretty easily.

I have a new vfd for the spindle. Simple 0-10vdc control with 24v inputs for start, fwd and rev. This could be changed to sourcing inputs. Currently no encoder feedback but would like to have a setup that this could be added later.

And finally, I have a MPG from CS Labs attached to their MPG expansion module. I know this isn't going to work but would need more I/O in addition to whats listed above to handle a manual jog wheel.

Hoping someone would be kind enough to shed some light on what boards I should be looking at to do this. Not even sure what interface I should be doing. PCI, ethernet....

One more question on the encoder feedback for the three axis. Since the drive must have these signals, I am assuming that the encoder signals would need to be split and go to the PC boards as well.

Thanks. I want to keep the step and direction as I already installed the new drives and I'm comfortable with them.

In Mach, I have have found that Feed/Hold has a major issue. There is some buffering going on since the cslabs is an external motion controller via ethernet. My machine has max speeds of 300 ipm which is 5 inches per second. After I got everything up and going, I quickly realized that if you press feed/hold during a G0 move, the machine will continue motion for many inches (sometimes 5 inches or more). Has to do with running out the buffer I'm told. Then I found out that if the current motion was an Arc, there were additional bugs relating to when it came to a stop. Meaning it may or may not finish the current arc. When you press start, it may or may not finish the arc before making the next move. This randomly and frequently results in the tool plunging into the material and destroying the tool and part. It is very unpredictable which in my opinion makes the feed/hold button unusable inside of a G code program.

Then I played with the Stop button. It will come to a stop very quickly. What happens next is pretty much the same thing. It may or may not complete the last move that was Stopped midway before executing the next line of code.

Its funny, I had Mach3 running via parallel port on a knee mill for 3 years and never really noticed any issues. Low and behold, I started testing these same senarios on it and indeed, there are issues. Just never really used the kneemill in the same way plus it is way slower than the VMC so not as noticable.

Bottom line, I just don't feel that Mach3 is safe on the larger machine with faster moves, and larger tools at stake. I choose it because I was somewhat familiar with it from the kneemill conversion. Searching the net, I don't see anyone complaining about basic issues like this with Linux CNC. I don't think swapping out the hardware to Linux CNC is going to be that bad. What I'm not looking forward to is the learning curve to make it all work

Thanks for that. So if I did the math right, my accel/decel is set at .254 G's= 2.49m/s^2
Max velocity =300ipm = .127m/s

So the time to come to a stop is .050 seconds (plus the 1 ms of course). So that calculates to .118 inches of travel. Maybe I messed the math up. Been awhile. At 1 G, I calculate .051 inches. So maybe I'm doing it wrong because you came up with 1/8" at 1 G.

Could you clarify. I thought velocity/acc=Time
Then X=V*t+1/2a*t^2

In any event, an 1/8 or even 1/4 is perfectly acceptable. In mach, its multilple inches and loose control of program path.

Well, you might also look at the Pico Systems Universal Stepper Controller.
(Full disclosure, I make those.)

Step & direction output, up to 4 axes, can accept encoder signals, 16 digital inputs, positions for 8 SSRs for outputs, and optional DAC for spindle speed control.

See pico-systems.com/osc2.5/catalog/product_...ath=3&products_id=19

and

pico-systems.com/osc2.5/catalog/product_...ath=3&products_id=28

Jon

So, could someone chime in and clarify about how I could expect the Stop and Feed Hold buttons to work.

Example Gcode

G0 X0 Y0
G1 X1 Y0 F50
G3 X-1 Y0 R1
G1 X0 Y0

So in that program, we rapid to 0,0. Then we G1 to the right. Then we go counter clockwise 180 degrees to x-1y0, then we move back to center.
In this simple arc, what happens if I hit STOP somewhere about half way through the arc?
If I hit FeedHold (Pause) in same place?


What I am used to happening on the OEM software is that a STOP (Reset on Haas) would stop all motion and the spindle. If start were pressed again, program would start from the beginning. If Feed/Hold (Pause in LinuxCNC) is pressed, it would decelerate all axis quickly to a stop, spindle would be left turning. Pressing Cycle/Start again would just continue on with no lost position or tool paths.

My only real concern here is to not run into exactly what happened using Mach and CSLabs. Reading the Mesa 7176 overview, I see the phrase "The 7I76 is a step/dir oriented breakout with 5 axis of buffered step/dir outputs". I think this "buffered step/dir outputs" is a big problem with the Mach setup I just installed. Since the breakout board is using buffered data, what happens when a Pause is seen. Does it need to "run out the buffer" and then pause?

Thanks again. Gonna dive in and take a whack at it.

You certainly will not have those lost incomplete move problems you had with Mach. The machine will stop as fast as it's configuration will allow, the tool will remain on the programed path, and resume motion exactly where it left off completing what ever move was in process (line, arc, helix, spline curve...) .
You will still obviously get a tool mark where you paused, because of machine deflection, and the tool spinning in place when stopped, but that is unavoidable.