CANopen performance

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
We are having problems with failures when attempting relatively rapid
CANopen write requests. The initial estimate for the system assumed
that four 1 Mbit CANbuses would support 80 writes at 50 Hz. In theory
it looks easy, looking at bandwidth (assumed 50 bits per frame,
gives 50 kbit/s per CANbus).

The question is basic, but probably not easy to answer: is it reasonable
to expect to do 20 writes at 50 Hz = 1000 writes per second on
each CANbus using CANopen? The web-page http://www.canopen.us/faq/0153.htm
has made me wonder whether the answer is no (it mentions 2-3 ms CANopen
overhead per cycle),

Thanks,

Philip Taylor

Re: CANopen performance
On 8 Feb 2004 10:43:22 -0800, philip_b_ snipped-for-privacy@yahoo.co.uk (Philip)

Quoted text here. Click to load it

What do you mean by write requests ? A PDO or SDO ? A write from
master to slave or message transmission from slave to master ? If this
is a PDO, which mode are you using ?

In the simplest case, the PDO transfer does not add any run time
overhead compared to a simple CANbus broadcast.

What kind of controller chips (Philips vs. Intel) are used at the
CanOpen master resp. slave ?
 

Quoted text here. Click to load it

Those figures sounds like SDO transaction times on a low speed bus:-).

If this is really PDO overhead, the only explanation for this kind of
overhead would be some intelligent CanOpen cards with an own processor
that needs some handshaking with the main processor.

Paul


Re: CANopen performance
On 8 Feb 2004 10:43:22 -0800, philip_b_ snipped-for-privacy@yahoo.co.uk (Philip)

Quoted text here. Click to load it

That article refers to the time from signal input to the slave,
processing at the master and outputting an other signal on the same or
different slave. It also assumes 1-2 ms processing latency at the
master after a "frame received" interrupt.

While this is the worst case two way delay, in a typical environment,
multiple slaves transmit their measurements simultaneously (arbitrated
by the hardware), multiple signals are processed at once in the master
and multiple output frames are sent from the master immediately after
each other. Thus, there is quite a lot overlapping input-processing-
output transactions and thus, the bus can be highly utilised.

Paul
 

Site Timeline