148 SI-Ethernet User Guide
Issue: 3
7.3.3 Checking for bottlenecks
There are three main reasons why a bottleneck occurs:
1. A drive is receiving more Ethernet messages than it can handle (8000 frames per second).
2. A drive is being asked to access more parameters than it can handle (6000 parameters per
second for M600 and M700, 500 parameters per second for M200, M300 and M400).
3. A segment of the network has reached it’s bandwidth limit.
For a full duplex 100Mbit/sec Ethernet network, assuming all Ethernet messages are the
maximum 1500 bytes in length, the bandwidth is 8000 frames/sec in each direction.
7.3.4 RTMoE Message synchronization
Cyclic messages can be synchronized or non-synchronized.
Only one synchronized cyclic link in each direction (one transmit and one receive) is possible so
these should only be used for high precision applications where the motion of multiple drives must
be closely coupled (e.g. printing applications). All other messages should be sent using a non-
synchronized cyclic link.
By default, all cyclic links are non-synchronous, if a synchronous link is required then the relevant
link number profile for both the transmit link and receive link must be set to “Sync”.
Synchronized cyclic data links utilise the IEEE1588 clock time distributed across the network. The
IEEE1588 clock can synchronize the drive's control loops to within a 1 µs accuracy, Pr 0.11.002
Option Synchronisation Active displays the active option slot providing synchronization. With
synchronized control loops the Ethernet interface can be used to transfer drive parameters
containing motion information, including those from the AMC.
Synchronous links work by including the time of when the data should be used along with the data
values. This time allows enough time for the cyclic link to reach all destination devices, the time
allowed for can be set in Easy Mode Maximum Network Delay (S.11.030). The receiving interface
will wait for its current time to match the timestamp in the cyclic link before processing the
message.
With normal Ethernet there are a number of variables that can impact upon the performance of the
network. These include:
• Delays through switches - Ethernet is a switched network and messages are typically copied
completely into a switch before being forwarded on. This is fundamental to modern Ethernet
and cannot be influenced by system design.
• Message length – the longer a message, the longer it will take to transmit and copy into a
switch before forwarding it on. For a synchronous cyclic link frame this delay is 12 µs, for a full
Ethernet frame it is 120 µs.
Message length can be controlled, but to maximise compatibility with other Ethernet traffic, it is
sensible to allow for full frame Ethernet messages where possible.
• Length of daisy chain - A daisy chain with Ethernet is really a chain of three port switches.
This can be controlled through physical network design e.g. using a tree structure to limit the
length of daisy chains.
The above delays (Latency) are managed in software, in conjunction with PTP, to ensure that all
device use synchronous data at the same time, but the length of the daisy chain must be
controlled by careful network design.
• It is unrealistic to assume that all messages will be full frames
• In reality the maximum number of frames/sec will be higher
• A more detailed frame analysis may be performed if necessary but the values
stated can be used to quickly determine whether bottlenecks could be a problem
To Next Page
Loading...
