[En-Nut-Discussion] Any Work-in-progress or else for the XMEGA

[Bernard Fouché]

Le 14/03/2011 16:18, Uwe Bonnes a écrit :
> The only advantage of the XMEGA I still see is that XMEGA in QFP64 has 0.8
> mm pitch versus 0.5 mm for the STM32 series. 0.8 mm pitch is a lot more
> friendly to prototyping. Let's see what problems arise when I try the STM...
>
> Thanks
>
Here we are starting designing prototypes with the 256A3 (for a target 
not running NutOS BTW) since it appears that it is one of the rare XMEGA 
available for real. (being now nearly 'fabless' I wonder if Atmel does 
not risk to be 'product-less' and then 'customer-less'...).

Our main concern is power consumption in MCU's 'idle' mode, which can be 
much lower with an XMEGA than an ATMEGA.

The DMA subsystem is very handy, I've made tests using DMA with SPI 
(using SPI master mode on a UART) and this way one can clock the SPI bus 
most of the time at half the crystal speed without calling ISR code for 
each byte but only when a full transaction completes. The result is much 
faster/shorter transactions (so the MCU can spend more time idling...). 
One can use DMA for USART Tx and ISR for USART Rx, for instance for a 
datalogger that output a lot of data to a PC but rarely gets data from 
the PC.

Other things I found handy compared to ATMEGA:

- being able to toggle bits in some I/O ports without having to do 
read-modify-write sequences.
- PDI programming (not interfering with other features on the 2 needed pins)
- lot of serial ports
- coherent design of I/O ports
- cleaner memory map for all embedded devices like USART, etc. well 
designed for the DMA & event systems.

I did not try much the event system yet.

Problems encountered during my tests:

- nearly impossible to find Atmel's 'XPLAIN board' on the market.
- AVR ONE! only runs on AVR STUDIO 4 on Windows while AS4 can use it on 
Linux for AVR32. (did not try much AS5 yet)
- a lot of errata on the available MCU stepping and no new release yet 
(AFAIK).
- Atmel code from application notes sometimes bugged or not up to date 
with current MCU stepping.