Too good to be true?

EtherCAT

http://www.automation.com/pdf_articles/EtherCAT_Introduction.pdf

It looks good...


Anyone used it yet ?

Yes we are working on a project that uses this system. So far everything is coming together real nice. Would have been better to have taken their training class, but with a little online tech support all seems real nice and straight forward. We are about 6 weeks into the project, I will post some up dates as the project goes on. This is a very large project so time will only tell.

It looks like a fancy way to convert Etheret to some other media and protocol.

What makes it special?

The idea is to reduce the number of small packets. Many small packets make Ethernet very inefficient. If the master can send one big packet that ALL slave devices can read then all the slaves will get their data at the same time. The trick is that each slave must know where its data is in the big packet. This is similar to other buses like CC-link and interbus s. I/we think this is the best way for Ethernet to work for industrial purposes but there are special hardware requiresments.

The key phrase is the document is in the EtherCat Operating Principle where is says each frame passes THROUGH the node. This means there are no hub or switches as shown in Figure 5. The data in shifted in and then shifted out like sercos, interbus s and cc link.

Our problem with this is that someone owns the special chips that allow the Ethernet packets to "pass through" the node. This is not a standard Ethernet chip and this will raise the cost. Buses like EtherCat and Isochronous ProfiNet require special hardware that will require the redesign of products if they want to be compatible with these buses. I doubt the extra hardware for Isochronous Profibus and EtherCat will even be compatible with each other. Is the market big enough in the specialty high speed applications where this extra cost can be justified?

EthernetIP is not really isochronous. There is too much jitter for some applications, but it works well enough for most. EthernetIP does start to get very busy when there are many devices producing packets. To reduce the packet flood one often needs to use expensive routers to retrict the flow of packets to just between the producers and consumers. EtherCat would not require the routers as it isn't using the producer - consomer model.

I would like to see an open protocol where the master puts all the data for each slave in one packet just like EtherCat, but it wouldn't require the special hardware. There are many applications where the frame update rate only needs to be 30 frames per second. That can easily be achieve with normal Ethernet hardware. Guess what applications run at 30 frames per second.

Peter Nachtwey said:

Guess what applications run at 30 frames per second.

Click to expand...

Hmmmmmm......Video???

hehe

Bob