[time-nuts] Updated Shera controller

Hal Murray hmurray at megapathdsl.net
Thu Jul 29 01:48:55 EDT 2010


stanley_reynolds at yahoo.com said:
> Full circle back to the software, the number of units sold, the cost per
> hour and time to complete project would determine the software cost. Would
> not  surprise me if the software would be the biggest expense till you break
> the 1000  unit mark unless the cost per hour was very low. As a hardware guy
> at heart it  is hard for me to assign a cost/value to software ;-) 

That line of thinking is probably appropriate for a commercial project.

For a hobby/volunteer project, software can be free.  Consider Lady Heather 
as an example.

In this context, there are two types of software.  There is the software you 
run on the board you build.  There is also the software you use to develop 
the software you run.

I'm assuming a volunteer would write the software just like volunteers have 
designed boards.



I'm not familiar with windows.  I think PIC and AVR come with free software 
for windows which works well with their low cost development platforms.  The 
compiler may be crippled to get you to buy the real version from somebody, 
but I'm pretty sure it's good enough to get well off the ground.

I'm not familiar with what's available from the vendors for ARM.


gcc has good support for PIC, AVR, and ARM.  There may be better, but it's 
well past good enough.  (It runs on windows if you use cygwin.)

If you aren't using the vendor packages, you also need a utility to download 
the compiled bits.  I'm pretty sure I could find one, and/or write one from 
scratch.


There might be a third type of software, a library that you use in the 
software you write, say a FFT package that's optimized for the CPU you are 
using.  In the context of low volume hobby projects, it's probably 
simpler/cheaper to use a bigger/faster CPU.

Or perhaps you need an OS.  If you depend on a commercial OS, somebody would 
have to buy a license.  Linux is free and runs on ARM.  NetBSD runs on ARM.  
I'm not sure about the other BSD variants.  That's 1/2 :)  I expect most of 
the code we would be interested in would be low level, just collect the data 
and pass it off to a PC to do the number crunching, display, and archiving.  
As such it doesn't need an OS.



-- 
These are my opinions, not necessarily my employer's.  I hate spam.







More information about the time-nuts mailing list