[Ovmsdev] OVMS CAN-USB Call for Assistance

[HONDA S-2000]

I like the MCP2551, but I don't think it's guaranteed to work without  
a 5V boost regulator. The USB specifications are clear that the 4.5V  
minimum needed is not always guaranteed to be present.

As for the crystal, it's all about the accuracy of the internal  
oscillator. USB calls out a precision in the specification, and as  
long as you meet that you're fine. Older chips had an internal  
oscillator that was not precise enough. This newer PIC32MX796 has a  
0.9% oscillator, meaning that it's plenty accurate enough for USB. I  
trust Microchip, in that they don't claim their internal oscillator  
will work for chips that can't do it, but this one can.

That said, there's nothing wrong with putting the traces on the board  
(other than the space they take up). I'm partial to smaller SMD parts  
like the ECS ECS-120-20-30B-DU, but those are more expensive than the  
large canned crystals.

Does the CAN spec (ISO-11898) say what the required precision is for  
the clock?

The SN65HVD235 has a bus listen-only feature that seems like a really  
safe option for a project like this. I imagine that the device will  
mostly be listening to CAN messages and not attempting to inject  
potentially dangerous messages into an unknown vehicle. Supposedly,  
the listen feature also helps with autobaud.

Brian


On Sep 4, 2013, at 19:22, Mark Webb-Johnson wrote:
> Looking at the wholesale pricing, and given that we have 5V already  
> and lots of experience with MCP2551, I think I'll stick with that  
> for the initial design and then see what the factory says.
>
> At the moment, I'm trying to make a rough design using  
> PIC32MX795F512H+MCP2551 - DB9 at one end, USB at the other. No  
> final decision on microcontroller yet, but I just want to get  
> something to the factory to firm up estimates on what this is going  
> to cost. I'll tell the factory that the SN65HVD233 is an option for  
> them, if it is easier/cheaper.
>
> The PIC32 is completely overkill for this, and will most likely be  
> 50% of the BOM cost. But, I really want lots of RAM on this thing.  
> If we drop RAM requirement to 64KB, we can save a few bucks, but it  
> really doesn't seem worthwhile.
>
> Regarding the oscillator, every design I've ever done uses an  
> external oscillator. I know the PICs (and others now) have internal  
> FSCs, but am concerned about timing accuracies for CAN and USB. The  
> data sheet says the internal FSC can be used for USB (and  
> incidentally is always used for 48MHz USB when in suspend mode).  
> Anyone have any experience with this? My gut feeling is to wire the  
> board for an 8MHz external crystal (and two caps), and then see if  
> we can get it working reliably with that disabled. If we end up not  
> needing it, we can leave it unpopulated at manufacture time (saving  
> a buck).
>
> Regards, Mark.
>
> On 5 Sep, 2013, at 4:01 AM, HONDA S-2000 <s2000 at sounds.wa.com> wrote:
>> The MCP2551 requires a minimum Vdd of 4.5V to ensure ISO-11898  
>> specifications, but USB Vbus is only guaranteed to be 4.01V under  
>> certain conditions like at the end of a bus-powered hub.
>