CAN bus

We prefer isolated CAN -- one less ground loop to debug.  We've used the ISO1050 for years, but it's not FD-compliant and doesn't regulate the voltage on the isolated side.  Also, the ISO1050 is out of stock everywhere, natch.  But isolating your CAN bus is something to consider.

On Sat, 14 Aug 2021 07:49:04 -0700, jlarkin@highlandsniptechnology.com
wrote:

>On Sat, 14 Aug 2021 10:03:08 +0100, Michael Kellett <mk@mkesc.co.uk>
>wrote:
>
>>On 13/08/2021 16:25, jlarkin@highlandsniptechnology.com wrote:
>>> A customer wants us to build a box (basically a cinder block bristling
>>> with connectors) that includes a CAN bus for control. We'd include
>>> ethernet too, for more general use.
>>> 
>>> Can someone recommend a good chip or chip set to do slow CAN? A good
>>> eval kit or equivalent? A good way to make a PC talk CAN?
>>> 
>>> We'd have an internal uP, one of the ST Arm things.
>>> 
>>> 
>>> 
>>There are several possible CAN pitfalls, mainly to do with physical 
>>interfaces and protocols.
>>There is low speed fault tolerant CAN and the more common high speed not 
>>fault tolerant CAN. There are ISO specs.
>>Then there are protocols, like CANOpen - but there are zillions of them.
>>
>>The CAN peripheral on the ST chip will probably be OK for anything 
>>hardware except CAN FD. I've known people get fussy about clock speeds 
>>and want "exactly" the odd speed they are using which may not fit well 
>>with the processor clock.
>>
>>If you need to write code to support some of the complex protocols it's 
>>a lot of work.
>>
>>On the other hand if your customer has a sensible and documented home 
>>brew protocol it can be quite simple and problem free.
>>
>>MK
>
>I've been told that ST has software drivers for their CAN interfaces,
>and like their ethernet stuff, it's pretty bad.
>
>We are replacing some old boxes and the customer probably wants them
>to be drop-on replacements for their old home-made stuff. That sounds
>awful.


We are doing CANBUS on our newer stuff.  Using STM32s, NXP and TI
processors.  We have a couple guys working on most of the CAN
software. 

Expect it to take a while to get it right.  A single processor is
probably a bit easier though than multiples as long as your
requirements aren't too high.

At least the address filtering and such is done in hardware

boB