Experimental Physics and
Industrial Control System

Hi Jens,

 

We have started using EtherCAT hardware this year. It’s all Beckhoff at the moment. We put everything on the (copper) EtherCAT bus and do not use Modbus TCP. We have an EtherLab EtherCAT master running on a RT_PREEMPT patched Linux PC. EPICS interfaces to the EtherLab master software using Diamond’s EtherCAT module. Connected to the master are several EK1100 EtherCAT bus couplers and one CX9020 running TwinCAT 2. We have not started using any safety terminals yet. I am looking into motion control right now but have not tried any solutions yet.

 

The EK1100 nodes are being used where we would have used VME in the past. The CX9020’s main task is pressure interlocking (this is done in TwinCAT 2) but we expose some of its I/O to EPICS for control via CSS and logging in the archiver. This has only been running for 2 – 3 months now on a new beam line that has not seen much use yet but we are happy with what we are seeing so far. We are planning to continue using Beckhoff embedded PCs in future interlock systems although I am trying to encourage the guys who do the PLC programming to use TwinCAT 3 in the future.

 

Hi Bryce,

 

Are you saying that the embedded PC / safety terminals cannot respond to a digital input quicker than 8 ms? This is a little surprising. In the system above we can get the EtherCAT 2 x cycle time worst case response time performance. I.e. outside world all the way up through EPICS and back out again. We are using a 1 ms cycle time so we respond within 2 ms. Like I say though we are not using any Beckhoff safety equipment yet and I did not try this through the CX9020 (which would be an interesting test that I might do soon).

 

This is how I tested the WCRT (and once again I’m talking about the EtherCAT bus here and not Modbus TCP):

Using the EPICS/DLS/EtherLab master system I hooked a waveform generator up to a digital input (EL1144) connected to one of the EK1100 couplers. This signal was also feed into an oscilloscope. In EPICS I connected the input’s binary input record to the binary output record that sits on top of a digital output (EL2024-0010) which was connected to the same EK1100. This digital output was connected to the oscilloscope. I then checked the time between the input rising edge and output rising edge by eye. I never saw it over 2 ms. I didn’t up the input frequency over 250 or 300 Hz (I can’t remember now). If we wanted to I am confident we could drop the cycle time to 500 us without much work and see the expected WCRT (i.e. 1 ms).

 

Damien