What are the benefits of Ethercat and what are the deciding factors that make people choose sercos over Ethercat

Ethercat is similar in performance to Ethernet Powerlink, both are open and well supported in the marketplace for all industrial networking purposes, including motion. SERCOS was originally intended for motion and typically is run over optical fiber but is in transition. The new SERCOS III will go over Ethernet, therefore will have broader application. However, Ethernet Powerlink and Ethercat are supported and stable today, whereas SERCOS II will go away and SERCOS III products are just appearing.

John may be optimistic in stating that B&R's Ethernet Powerlink is similar to Beckhoff's EtherCAT. I considered both before deciding on EtherCAT. Ethernet Powerlink is slower, though much faster than legacy field busses and sufficient for many applications. It also requires special ethernet hubs. National Instruments recently selected EtherCAT for remote I/O in their CompactRIO. I haven't used Sercos III, but Beckhoff uses it in their motor drives because it has features that make it preferable to EtherCAT there, particularly in configuration control.

Hi William, thanks for your sharing of your evaluation. I'd love to hear what convinced you that Ethernet Powerlink would be slower than EtherCAT since that is not what we are seeing on any practical applications. EPL also does not need special hubs as far as I can tell whereas EtherCAT requires specific ASICs on its nodes. If you are interested, I'd love to hear more of your thoughts on EPL's performance.

A disclaimer. I am an automation user and don't represent any manufacturer or vendor.

A B&R rep displayed the X20 system in 2006, which uses EPL. I recall he estimated response times from analog input to digital output ~2 ms. That is our main concern, to quickly kill a test if an analog reading goes out of limits. We prefer <1 ms in some tests with tiny rocket engines. In our EtherCAT system, I measure a 0.4 ms response. That is not the full-up system since still awaiting parts, but with 16 chans (2 chan analog modules), at 0.1 ms cycle time. Performance was not the only concern with B&R. At that time, they used 3rd-party reps and didn't appear fully committed to the U.S. market. We had also started with EtherCAT in 2004, but wanted to keep considering alternatives (still do). Indeed, some in-house engineers kept asking "why not Allen-Bradley?" (still do).

Other impressions have come from plots in Beckhoff or National Instruments seminars and magazine articles. I have never seen measurements that show EPL faster than EtherCAT, and most show it many factors slower. I recall a statement at a seminar (Beckhoff or NI, don't recall) that EPL requires old-fashion hubs that are uncommon today, since based on an earlier ethernet standard. I don't have hard-copies to quote, only memory, and am no ethernet expert. Perhaps a Beckhoff tech guy will chime in (after proper legal guidance).

EtherCAT does need special chips at the sensor/module end, but Beckhoff makes them affordable. A few transducer vendors have jumped on-board, but I am sure they would like more. They also offer an IP core, which is easy if the device has an FPGA, which I understand is the way National Instruments implements EtherCAT in CompactRIO. The PC/controller end originally required certain Intel NIC ethernet chipsets to run the drivers, but I think they eliminated those restrictions. I use Beckhoff PC's, so no issue.

Anyway, the original post asked for comparisons between EtherCAT and SERCOS. You B&R guys jumped in with EPL. I think Beckhoff agrees that SERCOS III has advantages in motor control since they use it there, but better that they chime in.

The performance of EtherCAT, SERCOSIII and Ethernet Powerlink is defined by the transmission performance of the 100Mb/s Ethernet media. As long as the packet size is small and the number of stations are also small, there are no remarkable speed differences. The ring topologies of of SERCOSIII and EtherCAT has speed advantages if the number of slaves are huge.
But this ring topology has a lot of disadvantages ... but this is an other issue.