CAN bus over fibre - is it possible?
mikep
Does anyone here know if it's possible - or has it been done?
I want to use the available CAN protocol support ICs but run in a very noisy
environment. Is it possible to use cheap fibre gear, say, as follows.
Single NODE
+---------------+
| CAN IC |
+-------+-------+
|
+-------+-------+
| Single Wire |
| <-> CAN cct |
+-------+-------+
|
+---------+ | +---------+
FIBRE | | | | | FIBRE
LOOP======+ FibreTx +-----+------+ FibreRx |=====LOOP
| | | |
+---------+ ^^^^ +---------+
Electrical
Connection
Network.
+--------+ +--------+ +--------+
+------| NODE 1 |------| NODE 2 |--- ... ---| NODE n |-------+
| +--------+ +--------+ +--------+ |
| |
+------------------FIBRE LOOP---------------------------------+
I understand that there is a considerable latency from TX to the same unit
receiving the echo of its transmission. However, if the number of nodes
was small enough, would it be possible to use CAN bus ICs to communicate
through this medium?
Regards,
Michael Portmann
Lars Wictorsson
There has been discussions in the CANlist about this and you can
find the archives here:
http://www.vector-informatik.de/v1_99/english/canlist/canlist.html
Search for the word "fibre" in the latest text file dated from
17/10/99 and you will find some ways of doing it. It will work,
but there are cheap (but complicated ways of 2 fibres) or expensive
(but easy way of 1 fibre). the hardest things is to find a
tranceiver that is capable of both sending and receiving. I have
no experience in this though, I just remember that we had a discussion
about it late last year or early this year.
/Lars
----------------------------------------------------------
LAWICEL / SWEDEN Phone : +46 (0)451 - 598 77
Lars Wictorsson Fax : +46 (0)451 - 598 78
E-mail: la...@lawicel.com WWW : http://www.lawicel.com
Embedded hardware/software together with 8051/C16x/AVR and
smart distributed I/O with CAN (Controller Area Network).
See CANDIP at http://www.lawicel.com/candip/ AVR+SJA1000
----------------------------------------------------------
Robert Scott
>
> Does anyone here know if it's possible - or has it been done?
>
>I want to use the available CAN protocol support ICs but run in a very noisy
>environment. Is it possible to use cheap fibre gear, say, as follows
[..snip..]
An essential feature of CAN is its ability to perform non-destructive
arbitration. This ability is predicated on the existance of two
states - "dominant" and "recessive". If two transmitters are
attempting to start transmitting at the same time, the one that
is transmitting the "recessive" bit will recognize that the bus
is in the "dominant" state and abort its transmission. The dominant
and recessive states could be implemented with passive pull-ups and
active sinks, but I don't think there is the fibre equivalent of
dominant and recessive.
Robert Scott
Ann Arbor, Michigan (rscott {at sign} wwnet {period} net is my real email address.)
(My automatic return address is intentionally invalid to foil spammers.)
Brian Geary
>and recessive states could be implemented with passive pull-ups and
>active sinks, but I don't think there is the fibre equivalent of
>dominant and recessive.
>
Brian.
Greg Neff
mikep <mi...@NOSPAM.newsguy.com> wrote:
> Hi.
>
> Does anyone here know if it's possible - or has it been done?
>
> I want to use the available CAN protocol support ICs but run in a
very noisy
CAN over fiber is not trivial. CAN requires tight timing (less than
one bit difference) between all nodes on the net. As soon as you start
building repeaters, you introduce delays. Every time you get a signal
from an optical receiver, you have to clean it up to remove pulse width
distortion and jitter. These circuits add too much delay. you can use
optical splitters and combiners to eliminate repeater delays, but these
are very expensive.
I would recommend a different protocol that is tolerant of propagation
delays on the net, such as LonTalk (EIA-709.1-A). For off-the-shelf
LonWorks fiber optic components, including fault-tolerant loops, see:
http://www.control-by-light.com
--
Greg Neff
VP Engineering
*Microsym* Computers Inc.
gr...@guesswhichwordgoeshere.com
Sent via Deja.com http://www.deja.com/
Before you buy.
ne...@my-deja.com
Hi,
we use for some measurements a FO-Repeater from a company called IXXAT
Automation GmbH. It costs about 280 Euro. You need 2 modules: one module
transfers CAN-Signals to fibre, the other fibre to CAN.
I hope it helped you.
REGARDS
Christian
David and Tammy Purdy
you are talking about and it is pretty easy to do, that is, you are
essentially extending the optical isolator. It does require an active hub
(I have seen some talk about passive fiber hubs but have not used one).
I built a four port active hub and all of my boards can transmit either
fiber optic or copper twisted pair (not both at the same time). The hub can
translate back to copper twisted pair using the 82C251 or wire AND all of
the received signals and stay essentially all fiber.
This system is very reliable and allows cascading or stacking of fiber hubs.
Going from a CAN bus to fiber to CAN bus is more difficult and I have tried
it and read a very good article on it in the CAN in Automation proceedings
(was even at the presentation) and almost had it working but noise on
either bus seemed to be a problem. I have seen commercial products but can
not vouch for their performance either bad or good.
In summation, CAN over fiber is easy, and has some big benefits. Dare I say
it is simpler than copper? Connectors are a problem. Our solution has been
to use feedthroughs. We have had some cable made (for a moderate cost) that
includes a pair of plastic optical fibers and a pair of twisted pair copper
wires (for power) that we can run through a Olflex Skintop feedthrough. Why
feedthroughs? For us, we need to be water tight.
You might check out the following
http://www.intec.rug.ac.be/research/groups/hfhsdesign/fortress.htm as this
system also uses fiber in exactly the way I have described.
Hope this helps, feel free to email me if this is to confusing,
David Purdy
<ne...@my-deja.com> wrote in message news:8aakb3$6hn$1...@nnrp1.deja.com...