[etherlab-users] ethercat + xenomai on arm926 cpu system

[Andreas Stewering-Bone]

Hello,

I also agrree. The arm926 is known to have a bad realtime performance.
Please have a look at www.osadl.de https://www.osadl.org/Latency-plot-of-system-in-rack-2-slot.qa-latencyplot-r2s1.0.html

Please check your realtime performace or try a cylic task with 100Hz.

I you have the choice please use preempt rt because this will be the future for realtime systems based on linux.

Best regards

Andreas

-------- Original-Nachricht --------
> Datum: Thu, 30 Aug 2012 09:35:28 +1000
> Von: Ben Lewis <egretengineering at gmail.com>
> An: etherlab-users at etherlab.org
> Betreff: Re: [etherlab-users] ethercat + xenomai on arm926 cpu system

> Hi Luca,
> 
> First I must say that I have just joined this list and I have not used
> Etherlab.
> 
> I've recently completed an EtherCAT application on an ARM based processor
> (IXP420, 533MHz).
> 
> Hardware specifications here:
> http://www.beckhoff.com/english.asp?industrial_pc/cp66xx.htm
> 
> My I/O requirements were small:
> 
>   * DI: 4
>   * DO: 4
>   * AI:5
>   * AO:2
>   * PWM: 2
> 
> I used TwinCAT for EtherCAT master.
> 
> I was not able to achieve a 1 ms cycle time.
> 
> A 2 ms cycle time was possible but with high real time usage and regular
> cycle time exceeds in low 
> priority tasks.
> 
> In the end I had to settle on a 3 ms cycle time. This gave me enough
> overhead to run lower priority 
> tasks (such as updating the visualisation) reliably.
> 
> On the face of it, it looks like your processor runs at about half the
> frequency as the one I was 
> using, therefore it would not surprise me if you are not able to achieve a
> cycle time below 4 ms.
> 
> I'm not sure if this is of any relevance to you, I'm just providing it as
> another point of reference.
> 
> Regards, Ben
> 
> 
> On 30/08/2012 3:20 AM, paluan.luca at tiscali.it wrote:
> > Hello everyone,
> > I'm trying to develop a real time server on the arm
> > (arm926) cpu and on a real time linux with xenomai.
> > Unfortunatly even
> > the basic examples seem not to work on such a system: the real time
> > task is set to work with a 2ms period like the 2 servos connected to
> > the ethercat net (even though 1ms would definetly be the best).
> > With a
> > bare example I immediatly receive warning from ethercat master of
> > skipped frames and if the real task has to do something more the server
> > crashes...
> > I tried various compile time optimizations and  I tried to
> > give up with real time task using hrtimer without results.
> > My system is
> > using the generic ethernet driver too, but could anyone give me some
> > advice about whether the cpu power is enough to manage the ethercat
> > driver in a real time environment?
> > Thanks in advance, Best Regards Luca
> >
> >
> >
> > Invita i tuoi amici e Tiscali ti premia! Il consiglio di un amico vale
> più di uno spot in TV. Per ogni nuovo abbonato 30 € di premio per te e
> per lui! Un amico al mese e parli e navighi sempre gratis:
> http://freelosophy.tiscali.it/
> > _______________________________________________
> > etherlab-users mailing list
> > etherlab-users at etherlab.org
> > http://lists.etherlab.org/mailman/listinfo/etherlab-users
>