I'm trying ethercat for a new machine. It makes sense to go with ethercat because it originally had absolute servos for all axis and for the tool changer magazine. If I'm running a mesa setup i need to get the absolute encoder reading at the start via modbus.

I've been running Linuxcnc for a ling time for 3 of my machines. Also like to play with ethercat as it's more interesting.

I've seen a video of ethercat running at 8kz servo thread but also found out that some Ethernet ports do not support that due to power saving Ethernet stuff.

Running a faster servi thread might solve most of these issues i guess?

Ok es wird Birnen mit Äpfel verglichen das eine sind 1,4 das andere sind 2,2 bzw 3,5 Kw was die für Drehmoment haben sagt keiner und
was an Leistung am Fräser an kommt auch nicht .
Des weiteren ist alles was aus C kommt nicht berauschend.
Wenn man was gutes will muss man Industrie Produkte kaufen oder erheblich Abstriche machen.

Gruß
Juergen

I am getting out of ideas.

It's a new computer so if you haven't done so already, you can check overall latency with and make sure the min and max are below +- 20 microseconds or so.

If that doesn't give anything I think you should go back to the working setup you had
with a laptop. If I understood right that setup worked, with the Realtek dkms driver.
Follow your own notes in this thread.

Ethercat master is always distributed as a package separate from linuxcnc due to licensing constraints.
So there is no better or worse linuxcnc version, they are all the same what concerns ethercat and locally compiled master.

Rod you should check out the new V8 Sheetcam. Increased license price, no linux version,
marginal support, company sold and lead developer left the ship (if I understood right).

Anyway, I do most prototypes, one-offs of everything. Sometimes 2 or maybe even 10 but not more. And never again.
I find the Conversational tools to work well, but could be expanded upon.
I already have G-code when it enters QtPlasmaC.
Nesting of several G-code files would be very helpful. Positioning, rotation. Some tools exists in Conversational.
Interactive positioning would be nice. Moving lead-in positions would be nice, but I see the problem.
Very rarely, if ever, I go via dxf and Sheetcam.

Ok Peter, so that is resetting my original understanding. But I have been unable to get the index pulse, which is clearly present in halscope to trigger the position reset. The sensors I am using for both the toothed wheel (a pair in quadrature) and a separaste index tooth, is ATS616LSG by Allegro. Datasheet
Just wondering if there is some magic in the devices that makes them unsuitable here. ( a straw to clutch at).

Re the speed variation, then I don't think there is anything in the config that is altering the analogout.
I have these:- But I think they are only used for the pyvcp speed display bar and at speed led. Can the act of processing them anyway do something?  I guess the thing to try is a minimum basic clean build and see what goes from there.

setp lowpass.spindle.gain 1.000000
net spindle-vel-fb-rpm => abs.spindle.in
net spindle-vel-fb-rpm-abs abs.spindle.out => lowpass.spindle.in
net spindle-fb-rpm-abs-filtered lowpass.spindle.out



# ---Setup spindle at speed signals---

#sets spindle-at-speed true
net spindle-vel-cmd-rps => near.0.in1
net spindle-vel-fb-rps => near.0.in2
net spindle-at-speed  <= near.0.out
#net spindle-at-speed spindle.0.at-speed <= near.0.out
setp near.0.scale 1.000000
setp near.0.difference 1.666667
 

Thank you so much, Mr Hakan.
I set udev rule, and then I don't need to type "chmod 666 /dev/EtherCAT0" anymore to exec LinuxCNC.

I'm using  PC,  HP Elite800 G4 which have Intel NIC , And I clean installed for 3 times. (In BIOS I disabled SecureBoot )

But all results was same.
Ethercat is unstable. (Alarm A12 flicking,  I can't do anything.  But I could reset if A12 happend before with EoE disabled master, and I could Jog, Run GCode with Yaskawa servo with DELL OLD PC)

I have attached dmesg file.

By the way, which LinuxCNC version (iso?) are good for EtherCAT( Especially DisablingEOE) ?


 

Interesting, a lot of output there.
I was just about the write that it must be the connection to linuxcnc, hal that's the problem.
Then I noticed that it disconnects when I turn the wheel. Almost always, when idle.
You have that too?

If your current Mesa setup already delivers stable rigid tapping, and you’re not confident you can guarantee the same margin of performance and robustness with EtherCAT, then from an engineering and risk standpoint there’s no strong reason to change.

EtherCAT can certainly work and can reach high performance, but it requires more integration effort, tuning, and understanding of the drives and control cycle to achieve the same level of determinism you already have with Mesa.

If the goal is simply to get “the same performance” you already have, the change becomes technical risk without a clear benefit. In that case, the most conservative and rational decision is to stay with Mesa.

Moving to EtherCAT only really makes sense if there is a clear advantage (architecture, cabling, drive availability, industrial standardization, etc.) and you’re willing to invest the time needed for commissioning and tuning to extract the same level of performance.

It seems remora is working ok. There must be issue with GUI. Try using axis gui first. UART just shows startup msgs on remora.

sorry I thought I answered. Because the QTP post is shipped by Sheetcam its set up for it to be master (to be polite to Les). But there are some variables at the top of the post to make QTP to be the master. In this case, it just sends the tool number to QTP and it retrieves that numbered tool from it's list. QTP has a tool that takes an exported sheetcam tool library  and converts it to  QTP format. read the docs.

Finally Sheetcam has both widows and Linux versions of their software and a single license allows you to use both.

You should also look at monokrom as recent advances by Joco with the internal cam is pretty amazing! Plus it also does its own hole processing by replacing any hole gcode with optimized leadins.

This interesting,I have thiked on this in the past, and certanly there are a lot of tools all of them are separated eforts, I think the features that software  should include but not limited to:
-Open DXF natively and generate the G-Code,
-Includethe database aproach used in monokrome to optimize holes.
-Ease of use, from final user stand point.

Is there a real delay with rigid tapping when we're using ethercat?

I'm running my old okuma mill with a mesa setup and it's working just fine. Although max I've tapped is 600rpm.

Don't really know how fast the Spindle encoder is read by the mesa. But i guess it's definitely updating every servo thread cycle.

Can't we expect the same performance for rigid tapping with ethercat?

I think there is a small but important distinction here between a process/sequence watchdog and a true safety watchdog.

A true safety watchdog is about detecting potentially dangerous or uncontrolled physical motion and forcing the system into a safe state (typically STO / safety stop), for example:

If |actual_velocity| > Vmax_safe when the axis should not be moving → trigger STO

If |pos_cmd - pos_fb| grows beyond a safe limit → trigger STO

If encoder feedback is lost while moving → trigger STO

If an axis moves while it is not enabled → trigger STO

If the real-time servo thread watchdog trips → trigger STO

If a safety PLC reports an invalid condition → trigger STO

These kinds of checks usually live in the safety layer (drive safety functions, safety PLC, safety relays, or core safety logic), not inside a process component.

What the current code is doing is more a sequence/process watchdog: it detects that homing did not start in time, did not finish in time, or the drive reported an error. That’s important, but it’s not the same as detecting unsafe or uncontrolled motion.

So I agree with Hakan’s point: the component should probably just signal a fault condition via an output pin, and the machine/safety layer can decide whether that becomes a controlled stop, an axis fault, or in some cases a real E-Stop.

I would just edit the ini file but that means you cant use pncconf or stepconf again. be sure to read ini file docs and homing configuration docs