> ... EtherCAT itself poses some of its own issues, with
> expensive slave devices and intellectual property issues with
> the protocol and custom hardware to implement the protocol.

Expensive slave device is a problem with Ethercat. In quantity it is possible 
to get a good micro controller from around one dollar and upwards.

> LinuxCNC solves a great many real-world problems with low
> cost, relatively generic hardware, so most users have not felt
> a great need for distributed control ...

There are at least some more or less distributed protocols. What make 
distributed control interesting are these cheap micro controllers with special 
purpose peripherals for closed loop servo control. These micro controllers 
however usually only have support for one or two servo motors so distributed 
control is needed.

For outputs it is possible to get similar perfomance as in ethercat by 
multicasting messages to all nodes at once on a dedicated communication path. 
For inputs however there are collisions or a need for more communication 
channels but if need for high speed feedback is removed lower perfomance in 
this direction could be accepted.


For short distance SPI for output with only SCK/MOSI signals with a rate of 
around 15Mbit/s and for inputs UART connected in ring could do the trick very 
cheap. If message is echoed around the ring there would be no collisions for 
the price of some delay for many nodes.


> > > I have been thinking about a new protocol. Ethercat are a perfect real 
> > > time network there data is inserted while packet is pass thru, actually 
> > > it is very similar to cascaded shift registers although CRC is also 
> > > added. I have not studied protocol yet but this kind of network it should 
> > > be possible to put node sourcing/sending data at any node or vary number 
> > > of nodes with exactly the same message circulating the network. It would 
> > > however be possible to achieve something similar for data sent from 
> > > linuxcnc if multicast is used although only for outputs.
> > >
> > > I would propose to put axis position control at motor controller, 
> > > advantages:
> > >   1. Motion control would be reduced towards youtube style real time 
> > > demand since positions could be sent beforehand and buffered.
> > >   2. Position control could be sent to all nodes with one multicast 
> > > message.
> > >   3. If interpolation is used it would be possible to reduce number of 
> > > messages.
> > >   4. Divide outputs between several nodes would only affect real time 
> > > performance for hopefully slow feedback.
> > >
> > > Problem would still be the same with inputs although I think most of them 
> > > have less demand on response time.
> > >
> > >
> > > Nicklas Karlsson

Nicklas Karlsson

------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://pubads.g.doubleclick.net/gampad/clk?id=278785111&iu=/4140
_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to