Re: EPICS and large CAN systems

[Ralph Lange <Ralph.Lange@gmx.de>]

On 16.03.2011 17:57 Rolf Keitel wrote:

Hello Elliot,

just to elaborate a bit more on the previous messages:

I'm adding my 2 cents between the lines here.

• we run about 1000 power supplies via can-bus
  

BESSY runs about 500 power supplies via CAN.

• when we started this 15 years ago, we deliberately did not use a higher-level protocol like can-open to keep things simple and to get more devices on one loop. Over the years, we adapted Andrew's driver to also work with PMC modules, but I agree with Andrew that this is not the solution you are looking for.
  

When we started 15 years ago, we deliberately chose to implement a subset of the CAL standard (that later evolved into CANopen), so we are able to talk to CANopen devices using our protocol. (And we were lucky to have bet on the right horse.)
We use VME cards by esd that drive four CAN buses, each.

• in order to avoid the problems Burkhard's problems, we developed "hub"-boards, which daisy-chain the CAN-bus and individual CAN controllers are fed from them by short stub-cables, effectively introducing a local star-topology at the hub.
• the advantage of this is that we can take individual controllers off the bus without breaking the daisy-chain
• the disadvantage: this is not a clean solution, but it works because we run the CAN-bus at a relatively low speed of 100 kHz.

To solve that problem, we are daisy-chaining power supplies by simply putting the plugs into the middle of the CAN bus line (very much like christmas lights). That way, pulling a plug from the PS does not interrupt the bus line at all, which stays physically intact all the time. As the "short leg" of those T connections is very short (few mm), we are running this at full speed (1MBit) up to the full possible segment length and >30 power supplies.

~Ralph