[Ovmsdev] Incentivising / Rewarding
s2000 at sounds.wa.com
Tue Sep 3 11:35:05 HKT 2013
On Sep 2, 2013, at 18:28, Mark Webb-Johnson wrote:
>> Custom class USB allows you to use the features of USB that
>> separate commands from data and make the size of the transfers
>> more definitive.
> I really would prefer driver-less, if at all possible. A simple
> serial control VCP. With a powerful-enough processor (and the
> difference between an 8bit PIC and a 32bit PIC is US$2), we have a
> lot of opportunity to offload work to the USB-CAN adaptor itself
> (for example high-level filters), and I don't think we will be baud-
> rate limited. Perhaps if there is an advantage with custom class
> USB, we can offer both interfaces?
Offloading is possible no matter what class the device is. Baud-rate
is not really the issue. The one thing I've noticed is that the
serial protocol can get really messy, and that's basically about
separating commands from data, or otherwise detecting the boundaries
of packets of information. That's what the USB protocol is good for,
and it can really simplify the protocol. Also, I've seen some designs
where the device gets stuck in data mode, requiring the device to be
restarted to get back into command mode, and that wouldn't be an
issue with a custom class.
Multiple interfaces would be a great way to gain the advantages of
both approaches, time permitting.
>> Any reason to avoid the 12V vehicle power? I suppose that
>> simplifies the hardware.
> Yeah, just to keep it simple (and cheap). We would still need 5V-
> >3.3V. Unattended logging would be cool, but only 128KB RAM in the
> device, and adding an SD-CARD would complicate things.
128KB can be enough if the logging is both specific to a certain
subset of messages, and compressed or compacted. In any event, I
imagine that a simply USB charger could keep the board powered for
> I'll try to rough up a schematic, to see how many components this
> can be brought down to. For China manufacturers, saving even 1
> component brings the cost of both materials and assembly down. I
> remember working with the factory to remove 1 single chip costing US
> $1.10 down to a bunch of passive components costing US$0.10 -
> resulting price difference was US$3 (I think they really didn't
> want to use that chip, not just for the cost but also for the
> hassle of obtaining it).
So, passives do not count as a whole component, not compared to a chip?
> I really want to get this to DB9 - MCP2551 - PIC32 - USB, with
> minimal external components.
I don't see why that wouldn't be possible. A lot easier than the
OVMS, which has lots of cool features and subboards.
More information about the OvmsDev