On Fri, 2007-04-06 at 23:23 -0700, Vasile Surducan wrote:

> True simulation of any serial programable device can be made only in
> the real world.
> Mostly because there are small timing differencies between various
> manufacturers products, so even a simulation works, a real connection
> may not work. So why spending time and nerves on simulation?

Vasile,

Having written my own PIC/System simulator I'm clearly biased... However I
second Gerhard's reply. A simulator is an excellent tool for development.
I also understand the premises behind your concern.

Here's my view. When confronted with a new project, I have two ways in
which to develop. I can either start writing the code and debug on real
hardware -or- I can spend time to create simulation behavioral models for
the crucial devices in my system and develop code on simulated hardware.
Both approaches are valid, and typically (in my case) both are applied.

I view the creation of a simulation model as an investment in a tool that
I can use over and over again. Sometimes the initial cost is prohibitive.
However, over time you can accumulate quite a number of useful simulation
models that you can then leverage off of for other tasks.

I'm not sure if you followed the LCD Font thread or not, but this is an
example where a simulation model is invaluable. First of all, I don't even
have real hardware! So it's impossible to even test the code out on a real
device. But I've created a simulation model of a graphics LCD device that
allows me to go ahead and begin development. I know for a fact that this
model has deficiencies, but over time I'll work these out.

The graphics LCD will be used in a proprietary product that consists of
many other components, most of which have simulation models. The models
have been refined to the point where I have a high degree of confidence
that if firmware works in the simulation that it will work in the real
device. If there's a discrepancy, I resolve it and fix the simulation
model. However, I *always* test firmware on a real device before releasing
it to a client.

In addition to development, there's also the issue of maintenance. For
example, suppose a bug is discovered in your firmware after it's been out
in the field for a year. If you fix this bug, what's the degree of
confidence you have that this fix doesn't have an impact elsewhere?
Simulation regression tests and unit tests are invaluable here!

Scott
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist