[etherlab-users] VoE message latency

[Gavin Lambert]

Hi,

 

Given a standard cyclic process running (PDO exchanges occurring but no
other queued SDOs etc), what would be the expected latency of an SDO or VoE
transfer (separately both out to a slave and back, ignoring the processing
time within the slave itself)?

 

I'm in the process of developing a custom slave (for a custom master app)
and am trying to decide whether a particular kind of data transfer
(asynchronous and acyclic, but still with a performance deadline, albeit one
about 20x more relaxed than for PDO transfers) is achievable with SDO/VoE
transfers or whether I would have to reserve PDO space for it (even though
most of it would go unused most of the time).

 

I'd prefer SDO, given that it's more standardised, but I don't think it'd
work without support for complete access for both read and write, and it
looks like for some reason Etherlab does not support complete access reads
at present.

 

Or to rephrase the question a bit:

1.       If an SDO/VoE request is queued by the realtime cyclic thread, how
many cycles will (typically and worst case, ignoring slave failure) run
before the request is completed (received and acked but not replied to by
the slave), and conversely how many cycles would it take for a message
posted by the slave to get into the master and acknowledged?

2.       Does this change if the request is queued by a different thread?
(And is this even allowed?  From what I can tell it looks like most such
operations should be cross-thread-safe, but it'd be good to get
confirmation.)

 

Regards,

Gavin Lambert

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.etherlab.org/pipermail/etherlab-users/attachments/20130429/d2303568/attachment-0002.htm>