Go Down

Topic: ATxmega (Read 5267 times) previous topic - next topic

I just totally bumped into this.  If a second gen arduino is in the works I would think of using these guys.
Higher resolution ADs and DAs on the chip sweet.  The higher pin counts might cause some board layout issues but the new functions are pretty cool.


http://www.atmel.com/products/AVR/default_xmega.asp
"Who left the fridge open?"
-Tugg Speedman
(Scorcher VI - Global Meltdown)

Cheater

Thats a seriously high end chip. I want one. ;D

4 DMA channels, 32mhz, onboard crypto.

Very nice. I dont think its designed for beginners though. ;)
Its more of a speed demon.

I had not read anything about them just ran across it yesterday.  
Yeah the on board crypto that is crazy but all the other stuff looks really neat.
From reading some of the press releases it does not seem like it is that complicated.

The ATMEGA64(100TQFP) is about $7.33 per 100
The Xmega64(100TQFP) is about 3.75 per 10k, I am guessing that the prices are almost the same, probably made in the same fab.

Only the large pin packages are available right now but it would be cool to take a look at one of the Xmega64 in the 44TQFP.

These chips have the same AVR core so  it will be interesting to see if these can be applied to the non-industrial community.
"Who left the fridge open?"
-Tugg Speedman
(Scorcher VI - Global Meltdown)

ckiick

Yeah, I got a little excited about this too, until I read some of the details.

The Xmegas are "low power" devices. Which means that they do not do TTL I/O.  So one of these could never be a drop in replacement for an arduino chip.  Sure, you could probably rig up a board that did voltage conversion between the Xmega and the exposed pins, but that would probably run your costs and complexity up to the point where it's no longer worth it.

I wish chip makers in general (and microcontroller makers in particular) would take a page from the FPGA world and separate the IO voltage from the chip supply voltage.  That would make them much more flexible.
Chris J. Kiick
Robot builder and all around geek.

AVRman

It is great to hear that they're continuing the AVR instruction set and have found a way to build it into a more powerful device. My only concern is when they will actually get these devices to silicon. I love playing with new stuff, but hate waiting for it  ;)

unsped

is there a concern they aren't going to? what alternatives would take over?

AVRman

I notice that they are quick to launch datasheets for chips that aren't ready for production yet. Take for example the AVR32 flash based devices. Very powerful, cool looking products, but you can't get em unless they're samples which is kind of a pain. 6 months after the datasheets were released, still nothing at the major suppliers.

Well, Atmel had kind of a rough year on the US stock market -everyone did. Not to say this has anything to do with it, but I remember it trading shares at $1USD

Cheater

Quote
I notice that they are quick to launch datasheets for chips that aren't ready for production yet. Take for example the AVR32 flash based devices. Very powerful, cool looking products, but you can't get em unless they're samples which is kind of a pain. 6 months after the datasheets were released, still nothing at the major suppliers.

Why would you start making big batches of chips before anyone has any use for them? :P
The samples are so products can be designed around the chips just in time for them to come in to production.

AVRman

I guess I'm just really impatient.  :)

torontogeek

Hi, I am new here, but have spent 30+ years designing hardware.

1.. ATxmega engineering samples are available to qualified users. So far it seems the only ones they have are ATxmega128A1, I have two due here today or tomorrow.

2.. Providing 5V I/O capability isn't even common on FPGA's! Some have 5V tolerant inputs, but not all. Most have maximum 3.3V on the IO cells. For example the very inexpensive and popular Altera Cyclone-II uses 1.2V for the core and in my case I use 3.3V on the IO. You can drive 74HCT (or equivalent type) logic from the 3.3V outputs. This is because the TTL threshold is within the range of the 3.3V outputs. In adapting some older designs to use FPGA's instead of processors I have connected directly to things like 75HCT573, 74HCT574, 74HCT138 devices.

Because most of my work involves EMI/EMC certification and we have proven designs based on 5V parts - such as MAX232's and the like we want to make as few changes as possible to reduce the risk. So in many instances I use SN74LVCC3245 dual voltage translators to go between 3.3V and 5V subsystems. Unfortunately they are not available in DIP packages (at all) but if an Arduino design was done with a 3.3V core then something like this could easily be hand soldered anyway.

3.. Look at something like the Diecimila board - it uses the FT232RL which uses a 3.3V logic core. Although you can choose to run the IO section at 5V if you need to.

4.. 5V logic became common when structures within the IC's were pretty large. As technology has improved and geometries have shrunk, the chips CANNOT be made to work on the higher voltages. If you want the low power and the performance advantages, then just suck it up, you will have to use translators.

The geek!

idover

they're competing with intel! :)..

westfw

I talked to Atmel about the xmega parts a bit at ESC.  According to the guy I talked to, their big feature is this "Event System" that enables the peripherals, in conjunction with the DMA controllers, to do a lot of "common" things without having to involve the CPU at all.  Aside from allowing higher performance if the CPU is awake, this also allows the CPU to spend more time in assorted SLEEP modes, consuming less power.

I dunno if the features are that relevant to an arduino-like product.  It's  a form of multiprocessing, which is a whole huge can of worms from a software development perspective.

torontogeek

#12
May 11, 2008, 01:14 pm Last Edit: May 11, 2008, 01:15 pm by torontogeek Reason: 1
Quote

I dunno if the features are that relevant to an arduino-like product.  It's  a form of multiprocessing, which is a whole huge can of worms from a software development perspective.


Two good issues.

The XMEGA gives an obvious upgrade path when an application is now too demanding for an ATmega but perhaps cannot justify the cost, complexity or power consumption of a larger processor. I have several applications where going to an XMEGA will save me a cmplete re-write, and lsave me from learning YANP - Yet Another New Processor.

If you were to ignore the new features, you get a MEGA-AVR with lots more I/O flexibility. But managing an event system and using it with DMA are not for the faint hearted.

But an XMEGA Arduino? Something using an XMEGA is going to be very un-Arduino like!

The Geek

torontogeek

Quote
they're competing with intel! :)..


Not really, because Intel has sadly left pretty much all of the traditional embedded markets now. In the days of the 8080. 8085, 8048, 8051, 8032 even 8088, 8086, 80188, 80186 Intel ruled the roost in many respects. But now even the 80386SX and other newer embedded devices are discontinued.

So people who traditionally used Intel now need to migrate. Many that I know are choosing soft-cores, such as the Altera NIOS-II, but things like the ARM-7TDMI, ARM-CORTEX-M3 are also very popular alternatives with the XMEGA-AVR set to join the ranks.

The geek!

westfw

Quote
Something using an XMEGA is going to be very un-Arduino like!

Violent agreement!  It's an interesting part as an upgrade to the regular AVRs, but probably not for arduino.  There's a growth path from Arduino-->other AVR-->ATxmega-->AVR32, or something like that...

Go Up