> I'm a big Atmel fan as well, but I've started using ARM's, and I probably
won't
> go back. For my needs, the ARM ended up being cheaper than the XMega
series.
> TI has parts with the Ethernet mac AND phy built in (all you have to do is
connected
> the pins to a RJ45 connector with built in magnetics, throw in some
firmware, and
> you're on the net!).
Yeah, the bang for buck you get is kinda crazy on the ARMs compared to the
AVRs, but for whatever reason (probably just past familiarity) I still find
it easier to bash out AVR stuff faster. (Of course the last two ARM designs
I did were an OMAP35xx and an i.MX53 and even Atmels new 200-400 page
datasheets are *trivial* compared to the 4000+ page datasheet and 4000+ page
user's manual on the bigger ARM SOCs!)
I went to the TI Stellaris training a while back (got the little robot thing
they gave away), but I haven't done any real designs with it. The built in
MAC/PHY combo is cool though. Finding something that lashed up to the
parallel PHY signals on the OMAP was a PITA.
> I only went with ARM because Atmel kind of screwed me on the ATXmega256A1.
> I had a datasheet, and an assurance from a factory rep that it was coming
soon. One
> year later it hadn't been released, with no date given as to when (or
even if) it was
> going to be released. I already had a board laid out and ended up having
to redo the
> whole CPU section. (Grrr!) That's when I looked into the ARM series.
I got burned on the ATXMega128A1 that supposedly supported x8 SDRAM. Except
for the little detail in that it doesn't. Then the one that is going to
have that capability has been on "next quarter" status for about a year and
a half. ;-) OTOH, I do like their really small parts (Tiny13A, Tiny24A,
Mega48, etc.) and kinda sprinkle them around on things probably more than I
should.
> ST Micro, Atmel, TI, Freescale, etc all make ARM variants, so it's easy to
change
> sources when different features are required (or a company screws you
over).
> Programming them is like programming a PC. No more bank switching code
(woo hoo!),
> and no more separate pointers for FLASH and RAM.
I've used the NXP LPC24xx and the STM32 recently for clients. I like the
STM32 quite a bit. Of course our present-generation emulated JAMMA setup is
Atmel (the old ARM7 AR91R40008 ;-) and most of our 'commercial' arcade
multigames were various ARMs too. (Atmel SAM9261 for the Sears Extreme
Arcade, various speeds of the PXA270 for the Golden Tee Clubhouse Edition,
Arcade Legends 2, Ultimate Arcade 2, etc.)
I had a real unpleasant experience with the i.MX31 (we tried to answer
"which operating system do you want to use-- Windows CE or Linux?" with
"none of the above" and had a six-month detour trying to get the right
information to use the 3D accelerator), then I had to *not* use the official
PMIC because it wasn't available in a ball pitch that we could manufacture
affordably so I had to DIY a power controller... and that was not much fun
to make work right. FYI, if the clocks come up before the right power rail
sequence on an i.MX31, you're calling 1-800-NEW-CPU. :-( And the DDR
controller was finicky. Not much fun at all.
To their credit, Freescale fixed a bunch now with the i.MX53 and that design
actually mostly fired up first try (again some PMIC weirdness, but nothing
fatal).
> The ZVG uses Atmel parts. Obviously a firmware update could turn the ZVG
into a
> real nice pattern generator, but if I were to redesign it, I would
definitely go ARM.
> They're just so damned easy to use, and with that kind of power, throw in
an emulator
> and you have a CCPU on chip! It'd probably end up being cheaper as well,
or at
> least very similarly priced.
Yup. I'll say that I have a STM32F103RFT6 on a design with a vector
generator right now. ;-) A half dozen years back we did an emulated vector
system that used a raster display and didn't see the light of day, so we're
bolting that on to a 'real' vector generator now instead. We still need to
write a CCPU core, but we already have our 6502 and Z80 cores in ARM
assembly which covers most of the rest of the vector titles we're interested
in.
> For your design, you could probably have run the vector generator almost
completely
> in the background, using the DMA controller and timers, and in the
foreground run an
> FFT on the ADC inputs and based your patterns on frequency, beat, as well
as amplitude.
> Same part count, similar cost.
I really just used the XMega because we have one on our ArcadeSD board, so
it was easy to hack the op-amps on and drive the laser from that without
needing to do a new PCB. (We were having a chiptune concert at the arcade
the next day and I wanted to bring something for visualizations, so it had
to be zero-leadtime hack on something else. ;-) I must say though that the
XMega's event system can really do a lot of interesting stuff too-- again,
virtually a complete VG without any CPU-- but as soon as you start dealing
with anything >8 bits the ARM really starts to shine, and certainly for
emulation there's no contest.
I still think it'd be fun to do a little Asteroids 'dongle' that just plugs
in to the ILDA connector on a laser and plays the game directly without
needing lasermame or a USB DAC or any of that stuff. Probably of very
limited interest, but 'neat' counts for something. :-) At least with lasers
you really only need to give it 'endpoints' and as the mirrors slew you get
your 'line' as a result. But there's a bit of black-magic associated with
getting the endpoints right that I never bothered to figure out. It worked
"good enough" for what it was needed for and I left it alone after that...
-Clay
---------------------------------------------------------------------------
** Unsubscribe, subscribe, or view the archives at http://www.vectorlist.org
** Please direct other questions, comments, or problems to chris@westnet.com
Received on Tue Nov 15 00:58:45 2011
This archive was generated by hypermail 2.1.8 : Tue Nov 15 2011 - 06:50:03 EST