Hey Marc, Be carefull with this. This code is running in Userspace while etherlabs ethercat stack is running as Kernelmodule.
Ethernet Packets in Linux are transfered to the Userspace over a mechanism called "Soft IRQ". This will mess up any Realtime dependency especally on little ARMs like Raspberry etc if we want to do something, lets say, every millisecond. See this link and others on the blog for details about the Linux Network Stack. https://blog.packagecloud.io/illustrated-guide-monitoring-tuning-linux-networking-stack-receiving-data/ The network stack in Linux is realy not meant to be "realtime". There some other mechanisms like "packet collecting" or "pause" on heavy load because all was developed to defend "the server" against DOS attacks in the early days not for "realtime". Beckhof actually wrote a special driver for one of there PCs that is using a polling mechanism to work around all those DOS defend mechanisms. https://elixir.bootlin.com/linux/latest/source/drivers/net/ethernet/ec_bhf.c I also enjoy this conversation. Best regards Michael -------- Ursprüngliche Nachricht -------- Von: Marc Wang <marc.w...@uwaterloo.ca> Datum: 01.03.22 17:57 (GMT+01:00) An: EMC developers <emc-developers@lists.sourceforge.net> Betreff: Re: [Emc-developers] LinuxCNC is in Debian! Hi, FYI, I got interested in the conversation and ended up finding this repository on GitHub: https://github.com/OpenEtherCATsociety/SOEM The code for the master is released as a GNU General Public License version 2. Regards, Marc -----Original Message----- From: Rod Webster <r...@vehiclemods.net.au> Sent: March 1, 2022 7:42 AM To: EMC developers <emc-developers@lists.sourceforge.net> Subject: Re: [Emc-developers] LinuxCNC is in Debian! > From my understanding Beckhoff is very open regarding the master (who >even does not need a special hardware, when driven by an RTOS) and are >very restrictive when it comes to the slave (where a ASIC / FPGA is >needed in any case) This was also my understanding when I read the licensing requirements. Heck they said they would even share source code with you if you were building a master... I have posted on the ethercat section of the forum calling for technical guys to get this done https://forum.linuxcnc.org/ethercat/45239-lets-get-ethercat-into-the-linuxcnc-distribution-need-technical-people-to-help Please send me an email at ether...@vmn.com.au if you can help with the tech stuff. Rod Webster *1300 896 832* +61 435 765 611 Vehicle Modifications Network www.vehiclemods.net.au<http://www.vehiclemods.net.au> On Tue, 1 Mar 2022 at 21:12, Rainer Stelzer <r.stel...@sgs-robotics.de> wrote: > Hi Gene, > > So once you've bought the controller, there is no other restriction. > > And I assume that royalty is 2x per axis, once for each end of the cable. > > No, they don't charge twice. > Part of marketing a Fieldbus-System is to make it widely accepted in > the industry, so Beckhoff (and others like Siemens with Profinet) > has/have to find a balance between protection and easy/low cost > accessibility. > > From my understanding Beckhoff is very open regarding the master (who > even does not need a special hardware, when driven by an RTOS) and are > very restrictive when it comes to the slave (where a ASIC / FPGA is > needed in any case) > > Makes also sense from commercial point of view, because in a typical > system there are much more slaves needed that masters. > > > Or can one receiver handle more than 1 axis with acceptable latency? > > A slave in EtherCat's daisy chain topology has an input and an output > Interface. (the biggest difference to most other ethernet based > fieldbusses) > > The incoming stream is shifted out with 1 bit delay to the output. > > At 100Mbit/s -> 10ns delay added up to each slave in the chain. > > The difference between one slave 3axis servo pack and three single > axis servo amps is ~20ns. > But to synchronise the output, all slaves have to wait for the full > frame received. > > Even on a 10 axis system this does not even come close to the total > frame time. > Not to mention the jitter caused by a (soft)RTOS > > > cheers > > Rainer > > > > > _______________________________________________ > Emc-developers mailing list > Emc-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-developers > _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
