[Arm-netbook] KiCAD based A10 break out / prototyping PCB finished

[Bari Ari]

On 02/06/2013 07:02 AM, jm wrote:

>
> I agree Ajith there is a lot of R&D and knowledge out there
> about reflections and PCB design. Please do continue to tell me I'm
> wrong and why - because all these links enriches us all. My only defence
> is that it doesn't agree with my experiences. Its a chance to check up
> on all the theory and if I'm wrong, I'll admit to it, explain what went
> wrong and redo the PCBs. I have budget for several iterations, so I'm
> not desperate to get it right first time, second time etc.

There are critical timing issues with DDR3. You have multiple lines of 
data and control signals. The data lines have to be stable before they 
are latched or the stored data won't be correct. Its the same for reads, 
the CPU DDR3 controller needs the data lines to be setup and stable 
before they are latched. So data needs to arrive slightly first, then 
control and clocks. There are lots of lines and they need to be matched 
in length as well as impedance or the control states or data will be 
corrupt. There are also other issues here that I'm not going into with 
high speed PCB design like reflections, dynamic impedance, skin effect, 
via design, series resistance, etc etc.

Is it possible to do this with limited PCB tools like Kicad. Yes it is 
but it depends on some luck and lots of wasted manual time spent pushing 
and shoving things around. Sometimes people get lucky and their PCB 
layout works. There is a reason why experienced high speed designers use 
tools like Allegro or Altium for PCB layout since they can get solid 
boards designed in a fraction of the time.

A software analogy might be: Can you use punch-cards or BASIC to write 
an application for the A10? Sure but do you really want to?

-Bari