[hpsdr] Call for Comments and Discussion - OzyII

[Bob Cowdery]

Hi all,

A few extracts and comments below. These make me think that the first 
step should be to establish what the objectives are because the 
possibilities are wide ranging.

Bob
G3UKB

"If we are going to go down the path that some are suggesting and design 
a Linux board with all the requisite resources, then why not add a DSP 
and call it Sasquatch? I think it would be a shame to invest all that 
energy (and money) and not go the extra step."

HPSDR is really a mix of plug-in single function boards and separately 
some multi-function boards. They both have a place but I would much 
prefer to keep single function as then the parts are replaceable and can 
evolve more easily.

"Linux even with real-time extensions is still only soft real-time, 
whereas what you need is something even more performant, small, simple 
and fast, preferably with guaranteed performance. To this end, with our 
project, along with using an ARM7 core from ATMEL, we used FreeRTOS 
<http://www.freertos.org> which fits the bill nicely when combined with 
the free TCP/IP package."

I can see the sense in this which gets an RT system, full TCP/IP stack 
and development capability. If it really is the case that Linux would 
burn CPU and not have predictable behavior then it's a  good  
compromise.  I would still prefer  Linux as it opens up more  
possibilities.

"I'd say UDP is the way to go. Not only is UDP almost trivial to 
implement into an FPGA once you have raw Ethernet TX/RX working, its 
functionality is easily sufficient for modern HDR systems. Datagram loss 
is not an issue when streaming over a (dedicated) local net, and a host 
that can't keep up with an UDP flow will certainly choke on the same 
flow over TCP."

This seems to be a pragmatic proven option but doesn't add huge value 
over OZY1 except cable length and multicast. These two alone may be 
sufficient justification.

 1248175760.0