Re: [avrdude-dev] xmega status

[Zoltan Laday]

Paul,

On 10/6/09 7:20 PM, Paul Thomas wrote:
> Hello,
>
> I'm new to the list, but I wanted to ask a few questions while my
> hardware design is still on paper.
>
> First, it looks like from both the mailing list and the source tree
> that programming xmega parts using the jtagmkII via the JTAG interface
> is supported. Is that right? Are there any problems/limitations? I saw
> one post that said the fuse programming didn't work?
>
>    
Yes, xmega parts are supported using JTAGICEmkII via the JTAG interface.
The xmega support is pretty usable, but not perfect.
I added most of the xmega support. Read functions were tested quite 
thoroughly and found to be
working. Memory write functions are also ok. Fuse read is ok, but there 
can be problems
writing them. One other known limitation is that the target is not reset 
after programming
finishes (so, eg. it starts to execute the newly downloaded code only 
after a manual reset).
(This works in AVRStudio in Windows.) PDI is not supported.
> Second, are any other programmers supported? It would be nice to use
> something like the USBtinyISP or the AVRISP2 (low cost). I saw this
> post (http://blog.frankvh.com/2009/09/22/avr-xmega-and-avrisp-mk2/)
> that says the AVRISP2 is supported by the latest avr studio. From the
> description it looks like the old ISP was 3-wire SPI whereas the xmega
> is only 2-wire (shared data). In the source tree I saw just a few
> references to the XMEGA PDI interface, but nothing to suggest it was
> supported. Also on the mailing list a couple of people asked about
> using the PDI interface, and it looked like it was undocumented from
> Atmel and thus unsupported by avrdude? Do I have this right?
>
>    
AFAIK, there is no support for the xmega parts on the AVRISPmkII.
As far as the physical PDI interface is concerned, it is not available 
on the AVRISPmkII only, but
it is on the JTAGICEmkII, too.
PDI interface is documented in Atmel literature, so the method of 
communication between the
programming tool and the target MCU is known.
As far as avrdude PDI support is concerned, however, this is not relevant.
The real question is if the protocol between the host machine (Mac or 
PC, serial or USB) and the programming
tool is supported by the avrdude code. This protocol is different on the 
JTAGICEmkII and the AVRISPmkII.
These protocols are documented in Atmel literature, too.
The catch, however, is that this documentation is outdated. I do not 
know too much about the AVRISPmkII
documentation, but I had to analyze the protocol thoroughly for the 
JTAGICEmkII when I was
adding xmega support. The situation is that the JTAGICEmkII uses 
undocumented extensions of the
protocol when dealing with new MCU families like the xmega and the AVR32.
I could add xmega support only by reverse engineering the protocol 
extensions.
Reverse engineering means that the serial line or the USB needs to be 
sniffed while programming
is performed with the factory supported AVRStudio tool in Windows and 
you should make
your best guesses what the undocumented messages do. This situation makes
it difficult to eg. find the culprit of such issues that the target is 
not reset after programming.
In the case of the xmega, about 20% of the messages are undocumented.
In the case of the AVR32, it is the other way around, about 80% of the
messages are not documented. This makes AVR32 support very hard or even 
impossible.
As far as PDI support for xmega parts on JTAGICEmkII is concerned, is 
undocumented,
but it does not seem to be particularly difficult to add support for it. 
It will, however,
have a command line interface implication, as the physical interface 
selector (JTAG or PDI)
should be added there.

Regards,

Zoltan
> Finally, what support is there for using a bootloader with avrdude on
> the xmega parts? Looking at this site
> (http://pramode.net/2008/01/23/avr-atmega16-bootloader-experiments/)
> it looks like the butterfly programming type is just talking to a
> bootloader. Has anyone tried something similar with the XMEGAs? Is the
> butterfly "type" what I should be investigating?
>
> thanks,
> Paul
>
>
> _______________________________________________
> avrdude-dev mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/avrdude-dev
>