plc.myforum.ro

mayo · Administrator Joined: 14 Jun 2009 Posts: 700

A working demo at the recent Hannover Fair in Germany showed PROFINET operating at cycle times of 31.25 microseconds.
This high speed option has come through the adoption of the PROFINET Version 2.3 specification which as well as introducing other important features also contains three key technical innovations aimed at fast IO applications.
These are backwards compatible with existing specification devices.
Achieving this dramatic performance improvement relies on increasing the efficiency of the Ethernet data handling.

Here's how it's been done:

Fast Forwarding:
Switches play a key role in all Ethernet networks but they introduce delays since the ID of an incoming frame must be determined before a decision taken as to its destination.
Typically, delays of 5 microseconds are experienced.
In Fast Forwarding, the ID information is placed earlier in the Ethernet telegram, enabling faster identification and immediate forwarding.
Delays are reduced to 1.25 microseconds per switch, a significant improvement especially for linear networks where a switch is required at each node.

Dynamic Frame Packing:
Data packets in very fast IO applications like drives are typically small - maybe just a few bits - making poor use of Ethernet bandwidth in linear networks because the messaging overhead for each packet is large by comparison.
Dynamic Frame Packing places several data packets in each Ethernet telegram.
At each node the relevant data is dropped but no process data is retrieved.
The telegram gets smaller the longer the line, speeding average response times too.

Fragmentation:
Transmitting a standard Ethernet frame containing TCP/IP data can take up to 125 microseconds - that's defined in the Ethernet standard.
It places a lower limit on cycle times since guaranteed TCP/IP transmission is an integral part of the PROFINET specification.
All Industrial Ethernets face this limit.
To overcome it, TCP/IP telegrams are split into smaller units at the first network node.
The final node reassembles the TCP/IP message. "Fragmenting" the TCP/IP messages allows more PROFINET data to be transmitted in the same time period.

Using these innovations results in scalable PROFINET cycle times down to 31.25 microseconds!
At least three chips are emerging to support this advance - Hilscher's netX, Siemens' ERTEC and the TPS1 "Tiger" from Phoenix Contact.

There's a video of the Hannover Fair demo here.
http://youtu.be/NTbmLthgsaQ