Light Machine Corp. Benchman XTr (retrofit)

WOW! Thank you dynamyte24! Good information to have! I will check mine tomorrow, if it is there, and see how they match!
Mac.... you brought up something I had not really considered before... PID tuning... I am rethinking the Mesa controller and think I might go with a Kflop/Kanalog system as they seem to have more support when it comes to PID testing. Might make it more attractive in a re-sale situation. Not really thinking of resale but hey, 20 year old Bridgeport mills still sell because there are no barriers to use them. This thread rocks!

steve_a wrote: I am rethinking the Mesa controller and think I might go with a Kflop/Kanalog system as they seem to have more support when it comes to PID testing

Are you sure about that? You might be looking in the wrong place.
The Mesa cards don't do the PID loop, that happens in LinuxCNC. This has the advantage that you can "watch" the controller and see exactly what it is doing. It has the potential disadvantage that it isn't a super-fast PID loop, running typically at 1 or 2 kHz.
(There will often be a hardware loop in the drives controlling motor current, I am only talking here about the position loop)

If your drives take velocity input from the motors then there is also a hardware velocity loop. In this situation a 1kHz LinuxCNC position loop is more than adequate.

Are you aware that Kflop is an alternative to LinuxCNC, ie, not something that you can use _with_ LinuxCNC?

The other mill I helped convert uses LinuxCNC in open loop and I would have thought that would be all the Kflop/Kanalog would need to work. I see a lot of Mach3 stuff for this system but you do have me re-thinking again.
I actually don't like the idea of giving the PID task to the computer. I think a better system is to have the closed loop handled by dedicated electronics in the machine with the computer just sending g-code commands for the system to handle. I know the history of closed loop with LinuxCNC is great but as a hardware person, "philosophically" I prefer self contained systems.
The original system tasked the motion controller with running the PID for closed loop and as far as the software was concerned, it was open looped. The real negative with the Kflop is USB communication but because it is just interpreting g-code, that seems pretty safe.
I have yet to successfully tune a PID. It seems " more art then science" , to quote Galaxy Quest. The Ziegler-Nichols Method seems to be the most straight forward but it has you run the controller into an oscillation before you can do any calculations. Can be very scarey! Anyway, the point is, I want to be sure I can actually tune the darned thing and that is driving my concerns with the Mesa. This will be easier once I find out if the PID information is still in the computer and if it closely matches the numbers dynamyte24 posted.

So the only values from Rex's PID numbers I found usable are the P and the following error. Everything else was not so great. Unfortunately, now my x-axis is not functioning, so it will be some time before I am back up and running on that axis. First I thought it was the motor, so swapped it out with a backup and it worked for about 1 minute before going out again. Then I swapped the amplifier over to the a-axis amp and it still does not work... Time to check the wiring agan because the output is working, but the motor is not.

I am seriously considering getting new amplifiers, ones that have a better track record with LinuxCNC. Does anyone have a recommendation?

Well, Good news bad news. I can confirm the information from Dynamyte24. I found the PID information for this machine.

The X-Y-Z

Kp 40
Ki 2
Kd 80 < this is the only difference. Dynamyte24's was 70.
Kv 80
Kvff 80
Follow Error .1
SPU 50,000
Int Range 40
Accel 20
Decel 20


A axis

Kp 40
Ki 2
Kd 80
Kv 80
Kvff 80
Follow Error .1
SPU 40,000
Int Range 40
Accel 20
Decel 20



Spindle (My spindle is 5,000 rpm but the numbers are the same:

Kp 10
Ki 10
Kd 40
Kv 0
Kvff 0
Follow Error .1
SPU 8000
Int Range .1
Accel 50
Decel 50
So the good news is that these are pretty much consistant numbers. the bad news is that it doesn't sound like this helps you Mac. That makes me wonder what the difference might be.

Steve, check the machine serial# shown on the software with the actual serial# of your machine.

The software # for mine is: 110-55AE-0176
Machine # is 021-55AE-0184

The configuration files were shipped on 3.5 floppy while the actual CNC program was on a CD.


Rex