Seeing that the new XMegas devices had a EBI interface with SDRAM support, I quikly hacked a breakout board for a MT48LC16M8 chip (a 16MB * 8-bit SDRAM) which the datasheet claimed the EBI would support.
Yes, it would have supported it, IF the chip had had a port more.
8-bit SDRAM support is only supported in 4-port mode, which NONE of the existing Xmega devices have!
Only 4-bit devices are supported in 3-port mode.
The thing is that the XmegaA datasheet that covers all Xmega's, and the EBI application note happily specifies this 4-port mode.
The actual Xmega128A1 datasheet only briefly mentions EBI in a short paragraph!
But deeper investigation show that the Xmega128A1 are missing the PORTL that would be needed for 4-port mode.
So, that means you can only work with 4-bit devices (or just use only 4-bits of an 8-bit chip like I did).
What's the problem then?
The problem is that with 3-port mode you can ONLY use SDRAM, there's no support for SRAM.
Doesn't sound too bad, you say, as 16MB is plenty of memory.
Well, yes, but it also mean you can't connect anything else that uses a regular address/data bus, such as ethernet devices, LCD displays, CF cards or any other memory mapped device.
Also, the configurable chip selects doesn't work, as there's no port pins for it (despite having 5 unused 8-bit I/O ports!!), so you have to bit-bang the Chip Select.
The EBI implementation in the Xmega is crippled.
Crippled so bad it needs a wheelchair on crutches.
The ideas I had for this chip is simply not possible, I have to use the SAM7SE as planned before the Xmega showed up.