Re: Linux board with 10 CANs

[Tom Evans <tom_usenet@optusnet.com.au>]

On 15/05/2015 5:56 PM, François Beaulier wrote:
> Hi,
>
> I'm starting the design of a CPU board, based on a computer on module
> with a very common SOC like i.mx6 or am335x.

There are SOCs with 8 CAN ports built in, like the Freescale MPC574X 
chips. That should be a lot easier to work with, and should be supported 
without any extra drivers.

The following Selector Guide tells me the Vybrid SVFxxxR has *FIFTY* CAN 
ports. That may have multiple typos, as Freescale's web site doesn't 
admit to that part.

http://cache.freescale.com/files/microcontrollers/doc/roadmap/BRAUTOPRDCTMAP.pdf

> CAN bitrate needed is quite low (50kbps)

So at about 110 bits for a minimum sized Standard ID packet, that gives 
you 450 messages/second max. That gives you a maximum service latency of 
2.2ms for as many buffers as the chip has. If your data has 8 
bytes/packet then that drops to 3.5ms/buffer (7ms for a chip with 2).

That wouldn't be impossible with MCP2515s, just inadvisable. Especially 
with Linux as it isn't good at any service guarantees.

But I've successfully run three MCP2515's on a Coldfire part. The QSPI 
helps a lot. There are a lot of parts with multiple QSPI ports, and that 
makes it less onerous (or DSPI or ECSPI). The i.MX6D has 5 on it. So two 
"real" CAN ports and two MCP2515's on each of the SPI ports, with a 
spare SPI port. Or three per port.

> I don't want to use MCP2515, i had trouble with it on a previous
 > design because of the lack of buffer in the chip.
> I can't imagine putting 8 MCP2515 and not missing any frame, but may be
> i'm wrong ?

Did you turn the BUKT bit on? If you spread them across lots of SPI 
interfaces it should get easier as they should all be accessed in parallel.

Tom