Thanks John I missed the attachment.
I sorted out the air supply and reassembled my table with the torch fitted yesterday and today, I should get the interconnect cable wired in so will be ready to cut something! The new 100slv torch is much heavier than my earlier one so I think I need to rethink my magnetic breakaway. I think I'll just buy a commercial one.
snapncut.com/
image missing https://i210.photobucket.com/albums/bb270/rodweb/Mobile%20Uploads/2019-01/IMG_20190124_160557_zpsmaz8okqk.jpg
Its kind going to motivate me now I have it all looking shipshape again.
Also put the inverter on a little trolley I had floating around so access to the control panel and the air regulator is easy to get to.
image missing http://i210.photobucket.com/albums/bb270/rodweb/Mobile%20Uploads/2019-01/IMG_20190124_160537_zpsekbj3gw6.jpg
PCW wrote: I believe "skunkworks" found this out also...
I suspect that the external offsets data path (from offset to commanded position) has one or more servo thread pipeline delays so reduces the maximum stable Z axis loop control bandwidth.
Thanks Peter, I've probably tried harder than most to get the EO THC control to work and it oscillates quite badly. There is one more thing I want to test and that is to move the component onto another slower thread. If that shows no improvement, I'll be following the path of Skunkworks, islander261 et. al. or trying out PhillC's system.
The reason I want to slow down the THC thread is that Jim Colt from Hypertherm has said trying to control the Z too tightly or adjusting too quickly results in oscillations. Also, a paper I found on Industrial PID control said that motors should not be controlled too tightly and need to be given time to make a correction or you will oscillate around the set point. I might be wrong, but I think that trying to control a motor driven axis based on a rapidly fluctuating voltage is quite different to the normal PID based servo control of LinuxCNC and the current EO THC paradigm. I know some THC's get down as slow as 64 Hz.
From some of the data collection I did, I think about 200 Hz might be a good starting point which will let most offset corrections complete before changing it again (on my hardware velocity and acceleration) . I also wrote some code that calculated a moving average so perhaps a 5 sample moving average at 1000 Hz could be paired with the THC thread that is running 5x slower to match the two sampling methods.
So now I have a third reason to try this, becasue PCW suspects that the maximum stable Z axis loop control bandwidth is reduced by the EO.
