On Fri, 11 May 2007 14:58:20 +1200, Russell McMahon wrote:
>> This is pretty funny, I got a chuckle when I read Russel's post
>> initially.
>>
>
> Nice to know I am amusing :-).
>
>> During the Q&A I asked them "Isn't the AVR32 pretty much the same
>> peripheral mix, memory space and CPU horsepower as the ARM7/ARM9 core
>> chips?
>>
>
> What made me give i 'the time of day' ws the USB2 OTG interface. All the
> rest was pretty good too, but having hardware 12 Mbps OTG buolt in is
> rare.
>
> Are there any ARMs with OTG capability>

I haven't looked in about a year or so, that's when I last did any work
with ARM and USB. At the time there were quite a few "higher" ARM chips
with OTG, for example the Intel PXA270 family (which we were using). We
ultimately dropped USB OTG from the product, initially our thinking was one
USB port to do both host and client interface. We ended up with both a host
and client port instead. The main reason was lack of driver support and the
complexity of the OTG interface (hardware AND software).

It turns out it's actually easier and better supported on the software side
to do a full host interface than an OTG interface. There is a pretty
convoluted negotiation that has to take place for the "host" to decide it's
a host and set everything up correctly. It also depends on the connected
device to respond properly and that doesn't always happen either.

There's also a lot more host stacks available for OSs like WinCE and Linux.
OTG is not very well supported on the device side either. All the client
device manufacturers assume you'll be running a PC (Windows, Linux, etc.),
or an embedded device running WinCE or Linux and have a real host port so
that's all they write drivers for. Running without a "real OS" and you are
in the land of writing your own low level drivers and the device
manufacturers don't want to even talk to you or if you come bearing loads
of $$$ they might talk to you.

I'm not aware of any ARM7 or ARM9 class CPUs that have OTG but there are
definitely some with host ports or both host and client ports.

I'm no USB expert but the software guys (who were very good and had some
significant USB experience) concluded that OTG was a good idea on paper but
that it was a fairly daunting task to implement. Their conclusion was that
it was just easier all around to use a chip with a real host port and
implement that. I know the hardware interface is more complicated for OTG
as it has to be switchable between host and client and all that entails.

Back in the day when the only chips that had host ports were PC South
Bridge chips or PCI USB controllers, OTG probably looked like a good idea
for embedded systems. But with the host controllers built into a lot of MCU
chips these days it doesn't make as much sense (to me). Maybe things have
gotten better on the OTG front but I still haven't seen many devices that
work with it.

Matt Pobursky
Maximum Performance Systems

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist