[etherlab-users] Domains and datagram timeouts.

[Florian Pose]

Hi Aaron,

On Wed, Oct 07, 2009 at 08:10:23AM -0700, Aaron Edsinger wrote:
> Hello. We've just bumped the total number of ET1200 slaves on our bus
> from 24 to 35. We're polling at 1Khz. We're now seeing messages like:
> 
> WARNING: 246 datagrams TIMED OUT!
> WARNING: 24 datagrams UNMATCHED!
> EtherCAT: Domain 0: 2 working counter changes - now 66/102
> 
> Reducing the poll rate to 100Hz doesn't fix the problem so I don't think
>    it is rate related. Subsets of slaves work fine, so we also don't
> think it is necessarily a hardware issue.
> 
> However, we're using a single domain of size 4080 bytes. This means the
> data image must be broken into 3 Ethernet frames. With 24 slaves we only
> required 2 frames. We're considering splitting the data image into 2+
> domains, but I don't see that this should help. We'd like to keep comm.
> with all slaves at 1Khz if possible.

This may be a buffering problem of the used network adapter. Maybe it
can not store three frames before you fetch them from the hardware.
Splitting into more domains will not be helpful, if this is the case,
because the width of the process data keeps the same. Do you have the
possibility to try a different network adapter?

-- 
Best regards,
Florian Pose

http://etherlab.org