Retrofit - LinuxCNC + EtherCAT

Hi everyone,

I’ve read through a lot of forum threads and would like a sanity-check before I start converting a large manual lathe (≈ 7 m between centres, three controlled axes: X, Z and spindle) to CNC.

Outline: IPC running LinuxCNC  →  Beckhoff EK1100 EtherCAT coupler.

1) Spindle (big induction motor that must stay)
   • Ideal: EtherCAT-ready VFD.  
   • Fallback: keep the old VFD and drive it with ±10 V from an EL413x analogue-out.  
   • Right now I only need the spindle to reach the commanded RPM.  
     Do I really need extra feedback (index pulse or full encoder), or is open-loop ±10 V enough for a first-cut retrofit?
     I can add an EL5101 / EL5042 later if it saves headaches.

2) Linear axes (X, Z)
   • Replace the old drives with EtherCAT servo drives; current & velocity stay inside the drives, LinuxCNC closes the position loop.  
   • New linear scales will be fitted. Where should their signals go?
       – straight into the servo drives, or  
       – through a Beckhoff encoder terminal (EL50xx) so LinuxCNC sees raw position, or maybe both.  
   • I’m after whatever gives the most reliable closed loop without over-complicating wiring.

3) Limit switches
   • Where do people usually land these in an EtherCAT setup?  
       – dedicated digital-input terminal (EL1004 / EL1008),  
       – directly into the drive I/O and pass the state via PDOs,  
       – or something else?

4) Still unsure about
   • Pitfalls when driving an older VFD from a Beckhoff analogue-out terminal.  
   • Best practice for scale feedback on a long machine: drive only, EL50xx only, or split to both.  
   • EtherCAT servo brands that work well for retrofits (good low-speed torque, reasonable cost, easy integration).  
   • Cleanest way to wire limit switches in an EtherCAT chain.

Thanks a lot for reading. I’m grateful for any advice, success stories or warnings you can share!

Kind regards,
vformi

If the VFD currently works, I would leave it as-is until everything else works.

If it ain't broke don't fix it.

Why did you choose to go EtherCAT?

A spindle isn't really an axis.

Thanks for catching that: the machine is X + Z plus a spindle.

There’s no VFD at all—this is still a fully manual lathe with a star-delta spindle and gearbox.

I picked EtherCAT mainly because the bed is 7 m long and I don’t want voltage drop or noise on long analogue/encoder runs. One Cat-5e solves the length and loss issues.

Vast majority of servo / servo drives have "normal" cabling between the servo and the servo drive.

Don't see any difference due to ethercat just because one axis is 7m long?

(I'm a fan of ethercat and think it is the way forward. So neat just plugging servo drive into linuxcnc with single ethernet cable and getting.servo drives status / encoder position etc)

I’ve mainly worked on much smaller machines, so on a lathe this size, I’m not sure whether the long motor/encoder cables would become an issue. If it's okay to leave the drives in a central cabinet and just rely on good shielding and routing, I’m happy to do the same—this is more of a concern than a hard opinion.

EtherCAT still appeals because it lets me pull extra data from every drive (load torque, temperatures, alarms, second encoders), and keeps future expansions simple.

I think Beckhoff and Rexroth have a distributed Ethercat drive system where the drive is built into the servo and there is only one power and data cable.  I've never used them and I'm sure they are pricey.  If you use quality cabling I don't think the length will be a problem.   You could also mount an EK1100 on the carriage and/or the far end to locally terminate signals to Ethercat vs running them individually back.

I would suggest connecting the limit switches and linear scales directly to the servo drive...I think you will get the best closed-loop performance by letting the drive handle it all.  If you are going to all this trouble I would also be sure and get absolute encoders so you don't have to deal with homing on this large a machine.