Need help with unexpected/random stops on basic Parport/5 axis BOB setup

OK, so if I try to summarize things, it seems to me that :
- the CNC cabinet GND _is_ connected to the building GND
- I may have un-expectedly high levels of EMI, some of these find their way to the PC, perturbating the PC screen and the reading of the inputs on the parport
- powering the PC from outside the CNC cabinet seems to alleviate the problem : no more ripples on the PC screen, and no more false readings of the e-stop
- EMI seem to come from the axis Delta vars : when I power only the spindle VFD, I see no ripples on the PC screen

Does this seem correct ?

When the PC is powered outside the cabinet, its GND connection is distinct (?) from the CNC GND. Is that why it seems to fix things ? If that were the case, this would mean that the perturbations come not from faulty isolation of the Bob, but rather from the common GND ?

@Todd : can you give some more details on how I could add an EMI before the vars ?

Thank you all for your help and time
Laurent

Most of the servo drives I've installed have there own built in power supplies and are ran from mains power. The manuals for these drives specify an AC filter to be wired in the power supply line (preferably) as close to the drive as possible. For example the manual for one of the drive's I've used (a Teknic SSt-3100) specifies a "Shaffner model 350-12-29 filter". (Please don't ask me how these filters work or how to choose the right one for your drives, that is well beyond my knowledge.)

Here is a little reading.
support.automationdirect.com/docs/emi_mitigation.pdf

Thank you for the reading material.

I'll also check my vars doc (Delta ASD A2) and see what they tell me.
Meanwhile, I'll probably just run the CNC with the PC powered outside from outside the CNC cabinet, assuming it is safe enough ?

Section 2.6, page 2-8

Two things I would suggest.

1) Fit a filter in the spindle drive power supply cable. Look on eBay for "RASMI" (one of the manufacturers) and pick one with the right number of phases and current capacity. (but be aware that these will often trigger a GFCI or RCD, so you need to look at their leakage current spec too)

2) Not a fix, but pragmatic, put a debounce HAL component between the e-stop input and LinuxCNC. 10mS extra delay before LinuxCNC knows about an e-stop won't matter, assuming that e-stop is also handled in hardware?

Hello,

Sorry, I've been so busy on an urgent (aka running late) project, that I didn't update this conversation.

I've tried several things with little success, and in the end, found that simply powering the PC _screen_ outside the CNC cabinet is enough to solve this problem...

So, I now have the PC unit powered from the CNC cabinet, the PC screen powered from an independent wall socket, and all is well in the world. Go figure !

I think I'll stop at this for the time being, as I still plan to replace the current office PC/parport/basic bob setup with a fanless PC mounted inside the cabinet, Mesa Ethernet card, and remote touch screen control.

Thank you all for your help and support, LinuxCNC world is the best