Ok, I'll
try to keep it brief. I attended the first two sessions devoted to AVR
at the ATMEL corporate headquarters in San Jose yesterday. The room
was large, and packed with people. They had pastries, bagels, and
drinks in the morning and provided a table of sandwich fixings for
lunch. I was also surprised to be handed an 'AVR Butterfly', which is
an ATmega196 dev board w/ LCD, speaker, and a cool joystick-button. I
guess they're planning on selling them for $20 at some point, but their
fairly cool..
The fist person gave an overview of their AVR line, their new products,
and an overview of what was going to be released. Though there were
some technical discussions, the bulk of the first session was more of a
sales pitch. But they do have some new interesting products. For
example, they're coming out with a secure version of the AVR, which has
built-in cryptographic features, a platform to do Ethernet development
(you must first buy the dev kit, then sign an agreement that you're not
going to port their source to another microprocessor platform, but then
can use it royalty-free). They have one that has a built-in RF
transmitter. They don’t have any with built in receivers, so I'm not
seeing a huge benefit here. They also are going to start putting LCD
drivers directly inside the AVRs (ATmega196 supports it now). That's
what they were using to drive the LCD on the AVR Butterfly, and it
seems like either I have a bad LCD, or their technology sucks. It
flickers a lot and some segments just don’t seem to work right.
They're also retooling a lot of their older products by adding
features, making them fully backwards compatible adding more flash, and
dropping the price. One of the handouts had a picture of what was
being phased out and what was replacing it (i can try to scan it if
people can't find the info elsewhere). They talked a bit about the
compilers, said that the AVR was designed in conjunction with IAR to
work well with C, talked briefly about compilers. They mentioned
avr-gcc, but it was almost a footnote, and the big negative associated
it was lack of support -- though I personally find support great on
this list.
The second speaker (Tony, I think), went on about designing
hardware/software. I'll just go over the bits I found interesting...
They suggested using 2 caps and an inductor rather than a single
decoupling cap. Using the internal calibrated oscillator is sufficient
to drive the UART, and if you use a 32k xtal for RTC purposes, then you
can use that to recalibrate the internal oscillator on the fly to
compensate for temperature. For power saving mode, it's advantageous
to run at a high clock speed and sleep or use a low clock speed an not
sleep (though I'm sure there are a lot of variables involved). They
stressed using local variables, and if you were to use global
variables, to pack them in a struct (for a smaller code base). They
demo'ed the IAR compiler and it was able to produce a 20% reduction in
code size after it had it's optimizations turned on. The code was
using to recode the AVR Butterfly to be a thermostat, reprogramming it
via the JTAG interface. Oh, and they're also going to phase out the
JTAG and replace it with a device that supports 1 wire debugging to
save on pins. The newer products will support the one wire debug
feature, and I'm guessing the JTAG replacement will also work with the
older JTAG interfaced products. It's not out yet, but they said it
will be around the same cost.
That's just what I recalled off the top of my head. If you have
questions, please feel free to email me. Otherwise, I think it was
worthwhile event, though way too early for my blood. And I would have
liked to stay for the 8051 and ARM sessions, but their chair didn't
agree with my behind, and I wanted to avoid traffic going back into San
Francisco.
Reza
|