Uzebox Xmega Project!

[uze6666]

Hi guys, I'll post on this thread the progress made on the "next-gen" Uzebox -- code named Uzebox xmega.

A couple weeks ago, I ordered a nice little breakout board from Sparkfun, sporting a nice ATxmega128A1. It runs up to 32Mhz, has an external memory bus interface, DMA and a gazillions other peripherals and timers. Since it comes with no pins, I had to make it "wirewrappable" in order to start prototyping. I simply used 40-pins wire wrap sockets split in 2. The result...something I can work with!




I'm still missing some required parts (like crystals, jacks and so on) and will report back in a couple days when I'll receive them.

Cheers and happy new year!

-Uze

[paul]

That looks like a lot of fun

[Firedawg]

Awesome! Keep us posted to your progress.

The Dawg

[mapes]

I see on other topics the discussion of the UzeBox successor. I noticed that there has been discussion regarding the 1284P and the XMega chip.

Will the XMega chip require rewriting any of the games or will it port natively.

I am semi-egar to know when it will be out and see what new stuff we can throw at it. I would really like to see what it can do.

[uze6666]

Initially I wanted to use a pure YC (luminance-chrominance) video interface to avoid using an AD723 (the 3.3v version of the AD725). But then I realized the quality would be somewhat lower and moreover all games graphics would have to be redone to go from RGB to YC color space. Moreover, it would have not been possible to make a JAMMA version. I'll have to stick to the AD723. Soooo, to answer the question, yes, existing games should simply have to be recompiled against the xmega kernel. That is unless some directly used ports and peripherals. But so far, I don't think anybody did that. It just sad that not everybody released the sources to their game. The Uzebox xmega won't have a game library as rich as his predecessor.

I'm still working on it, but switched to used the xmega256 instead. The chip is a tqpf64 with a pin pitch that will be easier to solder and get manufactured at ITead. It has 256K flash and 16K RAM...4X upgrade! Plus I've heard it can be overclocked to over 40Mhz...

Oh, and we will need an emulator if we expect this thing to catch on. Volunteers??

-Uze

[hpglow]

Glad to see the project is still moving along. I never actually finished my box. Well or my first project. But I stop in here now and then to check out what is happening.

It would be real nice if we could get like a full 1 mb of ram to work with on the next box.

[uze6666]

Yeah, it's moving along, but only in thoughts unfortunately. I'm still deeply divided on the subject. The xmega is cool, but it's not *that* an amazing upgrade compared to the '644. Lately I even started to seriously consider an ARM Cortex M3. But with either chips I can't seem to strike a winning combination of features vs. simplicity that I got with the '644. I'll express my though as pro and cons, I'd be glad to hear what the community thinks.

Atmega644
PROS

• DIP footprint, easy to solder by anyone
• Simple 8 bit part, small number of peripheral, easy to learn
• Open source tool chain, AVR-GCC
• Easy and cheap to program with a 35$ ISP programmer
• Lot's of community support for AVRs
• Bootloader support, integrated EEPROM
• We have a fully working emulator

CONS

• Overclocked
• Relatively slow speed
• No external memory bus
• 4K of SRAM

AtXmega
PROS

• Faster speed than '644, runs at 32Mhz, probably overclockable to ~45Mhz
• DMA, event system and other new peripherals
• External memory bus
• Up to 256K flash
• 16K internal SRAM
• Same toolchain and programmer as for the '644
• High backward compatibility with the '644
• Kernel would only need minor modifications

CONS

• SMT footprint, 64-pins TQFP not friendly to most hobbyist, worst with 100-pins
• External bus only available on 100-pins TQFP packages
• 32Mhz still relatively slow by today's standards
• 8-Bit architecture
• uzem needs a serious update

ARM Cortex M3
PROS

• Fast, 70Mhz+
• USB embedded connectivity
• DMA and other peripherals
• External memory bus
• Models with up to 1MB flash
• Models with up to 132K internal SRAM

CONS

• SMT footprint, not hobbyist friendly
• External bus only available on 100-pins+ TQFP packages
• Toolchain is different
• Complete kernel rewrite
• new programmer required (JTAG or other)
• Steeper learning curve
• no emulator

-Uze

[Jeroen]

What is the difference in price?

How much output pins are available on the ARM processor?

[uze6666]

ARMs are a bit cheaper than the Xmegas. Both are in the range 8-15$ per unit. 100 pin chips have 80 I/O pins.

-Uze

[Jeroen]

AtXmega got 78 I/O pins, so there is no real difference.

Personally I think I would choose the ARM because of the higher speed and much bigger flash/SRAM.

I don't have any experience yet with programming for the uzebox, but will programs that just use standard C and kernel calls compile?
The only thing I can think of what will break is the PROGMEM for placing data in code memory. If you define PROGMEM to something that doesn't do a thing, will it work? The ARM does have more memory, so if the game fits in the old 64kB it will definitly fit in the more-than-64kB of the ARM.

A thing that I would personally like very much would be easy access to the unused I/O ports.
Only thing needed would be lines from the I/O pins to holes in the PCB where I can solder headers. I haven't done any PCB design myself, but I guess that wouldn't add any costs (or not very much) to it but would make it not only a usebox II but can also be used to do other stuff.
Maybe connect an IR receiver and make a remote control or whatever someone can think of