My project idea ... midi sequencer ...

[bsmithyman]

I don't have any hardware yet, as it may take a while before I can actually start this project up. But in the meanwhile, I was thinking on using XC IDE to start up a few things. One, is the actually sequencer data, to see how many notes+events I can have with around 32k of memory. Another idea would be to use one chip-core just for the sequencer storage and playback, and another one for everything else. That way I have memory memory for the sequence. Or, in the last case, an external eeprom memory. (256k or 512k) I just need to check how fast I can read/write to those chips. I guess I could just use a small buffer to interface with those?

Wk

There are some ongoing discussions about adding memory to the XMOS boards/chips, the biggest issue being the large number of I/O lines needed to interface parallel memory chips. Serial options are available, but note that if you use an eeprom like you suggest, the write delays can be pretty long. A search of the forums may help with your planning, but the general consensus is that a memory manager thread is the way to go. The XDK has RAM on board, so if you're looking for an example implementation, that's one option.

[TonyD]

Have you considered Ferroelectric RAM. Ramtron make some large FRAM's with a 40MHz SPI interface.

Looking at their site they offer:

FM25V10 1Mb (128Kx8)
FM25V20 2Mb (256Kx8)

[williamk]

Thanks. In any event, the chip memory should be enough for a start. I still need to create my structures to check how many events I can get with 32k of memory. (leaving 32k for the program)

I'm thinking on doing 3 structures:

1) the SONG structure
2) the CLIPS structure
3) the EVENTS structure

The SONG will hold where the CLIPS (patterns) are played, repeated, etc ... EVENTS are the actual notes, bend, wheel, aftertouch, midi-cc, etc ... So a clip structure just points where the clip start in the EVENTS structure, and the size/len. The SONGS structure tells where CLIPS are played.

Wk