I know nothing about controls, but I can tell you about
ethernet.  If you have delays longer than 200 mS using
TCP, something is very wrong.  50 mS with UDP.  Usually,
on short runs, the propogation delay is less than the
resolution on whatever you will have to measure it with;
the only appreciable delay is after the interface circuitry
(i.e., in processing the packets).  Only on VERY heavily-
loaded networks are collisions any problem.
<als>

I've never used Ethernet/IP to control drives, but I think it would be fine if there's not a lot of devices on the network or your application doesn't not require determinism (meaning that it needs to trigger the motor at a repeatable rate...not too early...not too late).

I worked on high speed applications where if I triggered the motor too early, it wouldn't pick up the product.  If I triggered it too late, it would cause the product to pile up.  For my application, a steady PLC scan time and network determinism (SERCOS) was needed.

Also, have you considered ControlNet?

Hope this helps.

I would not use TCP for critical applcaition where delays would cause a problem,i.e. Steppers, High Speed counter, Life Saftey applictions. By it's defined protocol TCP will delay in certain events, Collisons, no response from request and most damaging the complete dropping of packets. TCP is considered a non-deterministic protocol. Even with fine tuning its parameters for retries and timeouts, it could delay or drop packets.

snagit88
an old colleague of mine has recently been writing a few technical articles about use of industrial ethernet within motion control applications. He's mainly focusing on Profinet and profibus based systems but this article should provide some food for thought:
http://www.manufacturingtalk.com/news/sie/sie162.html

EtherNet/IP uses UDP multicast for I/O connections, and in a switched environment this tends to get through plenty fast.   I've happily run 10 ms cyclic connections to 1794-AENT FLEX adapters without a hiccup.  Even when I am running TCP traffic to the same ControlLogix bridge module, the I/O updates only show a few milliseconds variance at the most.

I presume these are steppers that require just discrete I/O for commands ?  I wouldn't attempt a pulse train over any network, but as you described just commanding the steppers to go, it ought to be no problem at all.

Rockwell has a good performance and application publication for EtherNet/IP systems that has switch considerations and network bridge loading worksheets.  

http://literature.rockwellautomation.com/idc/groups/literature/documents/ap/enet-ap001_-en-p.pdf

snagit88
As you can see from the Rockwell info too, and to answer your original question, there should be no major concerns using Ethernet/IP. It is certainly the way for future.

Thanks for all of the info! We ended up just running a long cable to the machine. The customer was a little concerned with any delays. But, knowing that Ethernet/IP has this cabability in the future will make me more willing to go this route.

Thanks.

here is a good article on ethernet
http://www.mrplc.com/kb/index.php?page=index_v2&id=3&c=13