Plasma Questions

Andy spoke about THC pid control on this thread
forum.linuxcnc.org/27-driver-boards/3176...ement?start=20#82390

I googled for some C based PID code examples and its pretty much as he states it.
Its pretty easy to add pid control to a THC component.

I was going to look at this eventually. One thing I was not clear on was how the PID control should work and given the THC component plays with velocity I am even more confused.....

My thinking was that given the THC was called 1000 times per second, it probably should wait until the commanded Z axis move was completed before applying another adjustment to prevent see sawing. Most PID examples are based around a heating element or something that is much more variable than a discrete movement though space and time. What do you think?

The only thing I don't quirte understand

rodw

I was looking a using the pid component because it is very tested and works well with lcnc. The output of a pid calculation is usually a velocity command to the actuator to reduce the position error. Oh and by the way the heater examples can be misleading. Most heating servos are bang bang meaning the actuator (heating element, think your water heater or furnace) is driven at either off (zero velocity) or on (maximum velocity) with no in between states allowed. In the plasma case we are using the arc voltage as a position command/feedback. For a given set of cutting condition ( cut current, cut speed, material type, material thickness, consumable type and age, air pressure/flow) the arc voltage is proportional to torch height above the material. According to the Hypertherm gurus the torch height needs to be held to better than +/- .005" ( .127mm) of ideal for good cut quality. I can tell you from my experience +/-5.7 volts change in arc voltage (+/- .005" cut height with my torch and power supply) produces a noticeable change in cut quality. For us lowly small shops with air plasma systems this means using the manufactures cut tables as a starting point and using test cuts (fixed height and speed) to refine the arc voltages we need for a given set of cutting conditions with our equipment. I know this sounds like huge tolerance but remember a piece of sheet laying on slats is really flopping around when you start to think of +/-.005" tolerance with all the thermal movement, imperfections on the slats and thickness variance of the sheet, a mill table and a good mechanical connection this isn't. I know this sounds pretty picky but the first time you have a customer waiting to pick up a few hundred parts that all need secondary operations to clean up the dross you will see the light about cutting process control and why it matters. I primarily produce art and decorative pieces and my first wholesale order gave me a case of tennis elbow (wire cup brushing dross with an angle grinder) that took 6 months to recover from.

John

John,

Thanks for your time to describe some of this. I did realise a heating element is only on or off. The last thread I linked to was the one that eluded me in an earlier post.

Now I've regathered my thoughts. Looking at the THC code, it looks like the output from the code is a position command
that is calculated and set in this line
So I guess my question was really that if we are looping through this 1000 times a second on the servo thread, the correction is likely to take longer than 0.001 of a second to complete. So if we move to a more sophisticated PID based correction model, should the correction we apply be allowed to complete before the next correction is made? My gut tells me we should.

I know nothing about this. I started my build as a ghetto home project as it seemed a cool thing to do. Since then, I have taken equity in a manufacturer who spends an obscene amount on outsourced profile cutting, 99% of which could be done on a plasma cutter. I've substantially increased the specs to be industrial quality and so far that has been achieved. I thought this was well sorted in LinuxCNC and I see its not quite there yet. I do have an external THC to fall back on if the Mesa THC card does not do its job. There are not many plasma tables in our country so there could be a commercial product emerge if I solve the same problems you are talking about..

I've got a lot of CNC stuff ordered to finish it off but it will be a few weeks away and I have to design the table yet. I will try and at least get some switches in place to simulate the THC in the next week or so but It will be a while before I actually get far enough to turn the plasma on! Oh! and I won't be the one with the grinding wheel!

should the correction we apply be allowed to complete before the next correction is made?

This has come up before. I believe it should. In the experimental THC you can set a tolerance for the Z axis THC correction location. That is Z has to be at or close to the commanded corrected location before the next voltage check is made. There is also a programmable delay that can be added between changes.
After Christmas I will have more time and will write up more information, and post the experimental THC in a new thread.

+/-5.7 volts change in arc voltage (+/- .005" cut height with my torch and power supply) produces a noticeable change in cut quality.

Are you saying on your system a change of .005" cut hight results in a 5.7 volt change?

Rick G

Rick thanks, seeing there are a couple of guys ready to run with this, are you willing to share your experimental thc (even if privately)?

Can anyone shed any light on the lines from JT's plasma HAL example I posted previously

I had a look at this, the touchprobe is outlined here wiki.linuxcnc.org/cgi-bin/wiki.pl?Touch_Probe

And the digital motion one is outlined here linuxcnc.org/docs/devel/html/config/core-components.html
motion.digital-in-00 - (bit, in) These pins (00, 01, 02, 03 or more if configured) are controlled by M62-65.

So I'll leave it to you to describe the Gcode part.

I have looked over the work you posted in the Plasma section of the O codes thread and found it very informative
[/quote

Can you post a link to this thread?

Hello

Thank you for the replies.

tommylight

Yes, I really am trying to push the complication on this. I have a THCad board installed and actually plan on using JT's THC component for some early testing. I am aware the the update rate of the THC comp using a THCad board is nearly an order of magnitude faster than the commercial units and much faster than the mechanics of the Z axis stage. I am really looking for input from the group from people that actually use this setup. I fully expect to hear that I am wasting my time and that this approach will work so much better than my old Torchmate AVHC there is no need to change anything. I cut about 1,000lbs (454kg) of 14ga (1.9mm) HR steel sheet per month, not much by your job shop standards buy plenty for one semi retired guy.

Rick G

Yes. I have a Duramax torch and use the Finecut consumables so my typical cut parameters for 14ga HR steel are:

45A Cut current
220 in/min cut speed
.1g (about 37in/s/s) acceleration on X&Y (really too slow but all the better my cheap table will do reliably)
.09" pierce height
.07" initial cut height

While this is a slight departure from the Hypertherm cut charts it is where I get my best cut quality with my equipment YMMV.

The measured arc voltage into a long straight cut @ .07" height is about 80V on my old Torchmate AVHC and confirmed using a DMM. So 80V / .07inch = 1142.857 Volts/inch *.005 inches = 5.7V voltage change for a .005" change in cutting height if all other parameters are fixed.

I cut mostly artsy and decorative panels and often don't have enough space for longish tangential lead ins and lead outs on the interior cuts. Using the correct length lead in is a great way to compensate for torch acceleration and arc stabilization for rectilinear engineering pieces.

rodw

I know this will start a real fire. If you plan on using your system for even semi commercial work be sure to use a real machine torch and power supply from a US or Western European supplier (Hyperthrem, ESAB, TD are the ones that come to mind). If you don't you'll wish you did! Build your table as stiff as you can afford and make sure you can get at least .1 g acceleration at speed without shaking the torch. Thin materials (< 3mm) need high speed and acceleration for good cuts.

I am looking for decoding of the operators or notation of the two specific lines of HAL code I posted earlier. I have found no other examples of this notion and don't understand it. I get the whole general HAL nets and pin thing for connecting signals.

forum.linuxcnc.org/plasma-subroutines-library

Again thank you everyone for your comments.

John

So what doesmean?

Well, I'd say somebody has cut and pasted some HTML code somewhere becasue it is HTML forNow it will make sense to you. See www.w3schools.com/html/html_entities.asp

And yes, I know long term Hypertherm is the way to go. I've got a Traffimet S75 machine torch. The table will be very stiff, probably with 100mm x 100mm SHS legs and frame

rodw

Thank you. The last time I did any real programming HTML was still inside CERN.

One other thing you should consider while still building is what you are going to do for fume control. Plasma cutting is a smokey dirty process. Water table or down draft both have their ups and downs. Water table is simpler but only gets about 95% of the crud. Down draft can cost a lot of of heat or AC if you don't filter and recycle. The filters get big and expensive for the size you need.

I have put in the scale, ddt and pid components and connected them up to the THCad encoder so I can see how they react with a real arc. Initially I will just watch them with Halmeters.

John