[etherlab-dev] buffered mode and triple-buffer

[Jeroen Van den Keybus]

>
> I am NOT a twincat-professional,


Me neither, but I believe that by default the IgH master copies the SM
config stored in the SII to the slave. That includes the all-important
control word (the Data field in the Configurator dialog e.g. 0x00010026)
that also defines the buffer mode. So I'd think you must program the SII
EEPROM using the configurator. There is one catch: I see from the source
(master/sync.c) that de direction and watchdog bits get overwritten by what
you specify in the ec_sync_info_t struct, so be sure to set these right.
Also, I think that the physical address and offset are derived from the
PDOs you try to map into them.

Increase debugging using 'ethercat debug 1', try to run and look for the
SM: line in the dmesg log to tell you what control, address and length
values were written to the SM. It should match the ones you see in the
configurator.

but if I run this configured twincat-project (see screenshot_drive_tlgs.png
> & screenshot_master_tlgs.png).Depending on analysing the
> ethercat-configurations runs fine with just the expected working counter
> changes with twincat. It does some (IMHO) weird sync-management with
> triple-buffers (see screen_shot_drive_tripple_buffer.png in the
> attachments). Is there a way to copy this behaviour with our
> Ethercat-Master?
>
> When I have a look in the source-code of the master, in theory, I could
> change the fmmu-configuration to ask these logical-adresses as the
> twincat-project does, but this is not the exact point. I do not see how I
> could copy the behaviour having the slave ordered using "triple-buffers"
> (what ever this exactly means).
>
>
The FMMU config is assembled during PDO configuration. No need to enforce
virtual addresses. The only important thing to look for here is that the
same physical addresses are used. They have defaults in the SII EEPROM as
well, but you overwrite them with ec_pdo_info_t. I think passing NULL for
these makes the master use the defaults. Some doc is in include/ecrt.h.
Same as above: check in dmesg at FMMU:.


Btw: Ethercat itself seperates it's sync-management into buffer- and
> mailbox-mode right? Am I right, If I say, the default behaviour configuring
> a syncmanager for, -let's say -just exchanging pdo's, is buffer mode?
>
>
The default behaviour for PDOs should be 3-buffer mode (transparant,
coherent access to a memory area) if you can live with duplicated or
skipped frames.

Has anybody else experienced this problem or does anybody have a idea what
> I could do?
>
>
I'm still in the steep part of the learning curve, so I may have said wrong
things here. Since no one answered you for about a week, I thought to try
and help you out with what I learned so far.


J.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.etherlab.org/pipermail/etherlab-dev/attachments/20131127/66010c10/attachment.html>