Welcome to LinuxCNC.org

gera229 wrote:Hi Gera,

Hmm...that's stretching my memory - and it's only a couple of days . As far as I can recall, it was slightly less. I'll switch back inot that version of the system, crank it all up and rerun the sniffer.

Actually, I'm more looking forward to seeing you run your CNC machine rather than testing the latency once again haha.
12k and under is not bad at all especially if you use a controller card.

OK, this is an approximation (for "last interval" values) of the (live) sniffer after a couple of hours' running:

This is, of course, with "isolcpus=1" (per 07_rtai), hyperthreading off in the BIOS, and a number of pdfs (including a coule of the larger LinuxCNC manuals), Firefox (a few tabs), "Eye of Gnome 2.23" with the original (and only) screenshot loaded, Synaptic packet manager, Ayam (NURBS modeller), the system monitor, the sniffer of course, a folder window, glxgears, and three terminals (one of which glxgrears is running from as "glxgears &"). The machine is also connected to a VPN session on the iMac as well (it looks much nicer on the iMac's sharper screen than on the rather iffy touchscreen, and is about the same size).

Note that the "Max Jitter" values haven't changed in those two hours. Also note that the system monitor shows the single operating CPU as being max-ed out at 100 %, the other one is sitting idle at 0 % (of course).

This is on the first system that I installed; 12.4 % of memory (out of 4 GB) used, and o % of swap space used. The CPU heat sink feels "mildly warm", but it also has a 90 mm fan (connected to the D525's three-pin chassis fan socket) blowing across one end of it (the mess of cables on the desk has pushed it out of way of the bulk of the sink).

Interestingly, the network part of the monitor is showing sustained high ethernet activity, due to the VPN link with the iMac, no doubt.

gera229 wrote:
Heh, you and me both! I'm still fussing (some uncharitable but accurate persons might say "fooling") about with cables and enclosures, but it's getting there - slowly.

Another note: I have tested the mains swtitch for the drive power train - it didn't blow the socket's fuse into a squillion tiny particles and trip the ELCB this time . Following that I hooked up the toroid and got readings from the two combinations of 2ary wires, then hooked one red and the black centre-tap up to the PMDX power board, which gave an output of 80-ish V (I haven't tested the current yet).

I can't get the breakout board fired up until I've got the parallel port cable - whichever config I decide on - soldered up. Which, with my soldering "skills", is going to take quite some time.

AlexN wrote:ELCBs (or RCDs as we call them in the UK) can be rather troublesome. I ended up powering my VFD through a solid-state relay controlled by LinuxCNC. This is deliberately of the type that only turns on at the mains zero-crossing so the surge current into the capacitors is much less.