gta02-core component drawings (was Re: GTA02-core)

Werner Almesberger werner at
Tue May 12 02:15:24 CEST 2009

Dave Ball wrote:
> What's the benefit of  
> then merging them back into a library of everything?

One I could think of is that you don't have to select all the
individual libraries. But perhaps that doesn't make much of a
difference - I never quite got why KiCad needed libraries in
the first place :-) If the LoE turns out to be useless, we can
drop it in a while.

> Fab.  I was in general trying to keep the components similar to the  
> existing gta02 models, to minimise transposition mistakes when we're  
> re-creating the schematics.

Yup, good idea. And I see that you cleaned them up a little as well. 

> Does it make more sense to have a 'part' for each of those 19, group  
> them as I have (mem, IO, peripherals, power), or do something different?  

Hmm, good question. It sounds a litte radical, but perhaps it works
out nicely in the end. A few references:

- Samsung's SMDK2440:
  - Address, Data, DMA, Chip Select, Clock, Timer, ADC/TSP
  - SDRAM, IIC, IIS, SPI, USB, EINT+, LCD Data, LCD Control, JTAG,
    UART, SDIO, NAND Controller
  - External Interrupt, Camera interface, Power

- Samsung's SMDK2443:
  - EBI Bus (NOR, NAND, OneNAND, CF), EBI Bus Control, DRAM Bus,
    DRAM Control, System Control
    PWM/Timer, UART, SPI
  - Power

- Openmoko GTA03 (6410):
  - Power
  - POP Memory
  - all the rest

So the common number of parts seems to be three. But there's
certainly some appeal in being able to put a small fragment of
the CPU just where it's needed, without running lots of signals
across sub-sheets.

> Is there value in re-creating the part exactly as it's drawn in the 
> current GTA02 schematics to minimise silly copying errors?

I think treating the CPU as one monolithic block makes it just
too unwieldy. With regard to copying errors, I'd count on
massively parallel review to catch them ;-)

- Werner

More information about the Gta03 mailing list