[hpsdr] Call for Comments and Discussion - OzyII

[Henry Vredegoor]

Hi Chris,

I think the least problem is the hardware (complexity and component count
and such) given the number of example hardware implementations being around;
one trades hardware complexity for software complexity up to some point?
Laying out a high quality CPU board at reasonable production costs might be
more of a problem. (avoiding 4+ layers?)

Reading Phil's proposal it seems we would start (almost) from scratch.

Even if we could copy many existing software parts, I think it would be a
hell of job to get it all to work at the level of perfection and performance
we are used to under Linux. (Not that I find it not extremely fascinating to
see this being developed in an FPGA!)

Development steps I think one would have to roughly go trough:

1. Get the basic hardware right
2. Get YOUR implementation of the soft-core CPU/virtual machine on YOUR
hardware right
3. Get YOUR implementation of a basic OS on the virtual machine right
4. Get YOUR TCP/IP network functionality of the OS with YOUR hardware right
5. Get the applications (compilers!) working on the networked OS right
6. Get the performance of the OS/networking/applications right

IMHO not exactly a trivial job.

TCP-IP is the key to a whole new world of applications for SDR we are just
starting to explore.
Do not forget the importance of the support for really (inter-)networked
applications in the OS (FULL TCP/IP stack, Routing, DNS, etc.)

Henry.




> -----Original Message-----
> From: Chris Albertson [mailto:albertson.chris at gmail.com] 
> Sent: maandag 20 juli 2009 20:17
> To: Henry Vredegoor
> Cc: Phil Harman; hpsdr at openhpsdr.org
> Subject: Re: [hpsdr] Call for Comments and Discussion - OzyII
> 
> 
> On Mon, Jul 20, 2009 at 8:47 AM, Henry
> Vredegoor<henry.vredegoor at gmail.com> wrote:
> > ***** High Performance Software Defined Radio Discussion List *****
> >
> > Hi Phil,
> >
> > Last weekend I studied your proposal for a future, new Ozy, 
> OzyII and gave
> > it some thought.
> 
> > I think the implementation of the full Ethernet/TCP-IP 
> functionality in an
> > FPGA is a difficult and complex task and will take a lot of 
> development
> > time. (witch could be better spent for "real" HPSDR work 
> ..... :-)  )
> >
> > I think a real microprocessor based system, maybe an SOC 
> (System On a Chip),
> > running a (micro-) version of Linux, maybe assisted by some 
> FPGA hardware
> > and an added high speed LVDS interface to the ATLAS-bus 
> interface would be a better approach.
> 
> If the FPGA is big and fast enough you can implement a CPU core in the
> FPGA.  If you do this then you actually save chip count and complexity
> by not having the CPU chip.
> 
> I agree that you want a CPU powerful enough to run a general purpose
> OS.  I think that there are free ARM and free SPARC cores available.
> SPARC may have the best and most mature OS support.   The other thing,
> is that if a full OS kernel is running then you will need a bit of RAM
> and that has to be planned for now.
> 
> I think using Ethernet in place of USB will open up a lot of new uses.
>  The big advantage of Ethernet is the maximum length of the cable.  A
> long data cable means you can place the HPSDR box very close to the
> antenna.  Maybe even outdoors in a die cast aluminum box then run a
> fiber cable back to the computer.
> 
> No antenna feed and no antenna feed losses either.
> 
> 
> 
> -- 
> =====
> Chris Albertson
> Redondo Beach, California


 1248131593.0