Lark Board RoadTest

How would you  do with this Lark boards if you got one?

Clem Martins wrote:

 

How would you do with this Lark boards if you got one?

Lark is designed (and priced) for high end image and other signal processing, which isn't something I do these days.  I'm most interested in inexpensive FPGA boards with Spartan-3A/3E or Spartan-6 that students or hobbyists can afford so they can use my Flavia: the Free Logic Array.

@Clem

The problem with the Lark board, and the reason I didn't apply for the road test although at least 50% of my work is FPGA related, is that to use this kind of board other than just as a script kiddy running  a pre-built example or two, requires very considerable effort.

To make any sensible use of it you need to really exploit the combination of ARM9 and FPGA on one chip  - which means that you are going to need to be pretty good at ARM9 applications processor work and FPGA. And then you'll need to put in a LOT of effort (to really get  a feel for how useable it is I would estimate at least 4 weeks if you are already an Altera FPGA user and 6-8 if you are not.). Much more than this for a serious demo application.

In this context a free board is a non-issue.

To answer your question more generally - I'm very unconvinced by mixing application processors and FPGA on one chip - it only makes sense as a cost reduction over using separate parts or if the very tight coupling is useful. We are yet to see a  killer application for the Xylinx Zynq which has been around for ages.

You could do something like an pocket ultrasound scanner using the FPGA for the scanner processing and the ARM for user interface, graphics etc. but I have no idea if the FPGA has enough welly to do this.

I'm much more interested in the FPGA and micro-controller combinations like Micro-Semi's smartFusion2 but the pricing isn't competitive. But one very often teams up an Arm Cortex 3 or 4 with an FPGA and having them in one chip might well be very handy - but only if it costs about the same for the parts as buying the FPGA and micro separately.

MK

Michael Kellett wrote:

 

To answer your question more generally - I'm very unconvinced by mixing application processors and FPGA on one chip - it only makes sense as a cost reduction over using separate parts or if the very tight coupling is useful. We are yet to see a killer application for the Xylinx Zynq which has been around for ages.

We looked into using a Xilinx Virtex-II Pro over 10 years ago.  That chip included one or more PowerPC cores.  It made sense technically, since our latest products were all based on PowerPC, specifically Freescale/Motorola MPC860.  However, its pricing was impossible, so instead we used an IBM/AMCC PPC405EP processor plus an external Spartan-IIE, connected using PCI.  That ended up being a great combination, and cost about 1/3 of the integrated chip.  Xilinx Spartan chips are so much cheaper than Virtex (or whatever they call them nowadays) that you're better off being extremely clever so you can fit your design into a Spartan.

I haven't seen the Zynq pricing come down to something acceptable to me.  The original Xcell article quoted "starting at $15", which would have made them extremely attractive.  However, all the Zynq boards I've seen (except Parallella) are substantially more than $100 and don't appear to be coming down.  Plus you need to use Xilinx Vivado software, and I have no idea how much computer and human time resources that requires.

JMO/YMMV