There is a way to do this. *IF* you really want quantitative answers, you
can go through a laborious process (but kinda fun, too) of characterizing
each step of your system's block diagram and the impact you find along the
signal path.
You have to start "microscopic" then get to the "macroscopic"
Essentially you characterize the noise source and its impact to the S/N
ratio and convert that into Bit Error Rate (BER) After you have BER, you
can then quantify the impact to your data stream (the retransmission
requirements, and likelihood of a second transmission requirement) and then
calculate the efficiency of your transmission system.
I've done this for simple receivers and it is not a trivial process. Try
contacting Prof. T. Kailath or Prof. Robert Gray at Stanford. They taught
the courses where we learned to do all this and they both have some books on
the subject.
Characterizing the noise source is not easy. You must assign an energy
level and statistical distribution to it. There are several noise sources
that are easy to characterize, like Johnson noise in resistors, etc. More
difficult to characterize are the industrially generated types. However,
Poisson distribution (characterizing spike noise which arrives at random
times) convolved with Gaussian energy envelope (to represent the fluctuation
in power) is fairly decent. And spike noise is the most disruptive to
digital systems if their presence coincides with a transition edge and
"confuses" the logic.
Armed with noise characterization you can then determine the BER for the
receiver (Don't forget to include the modulating effect to the noise
distribution caused by the radiation-reception mechanism). If the receiver
takes action based upon a spoiled bit, you have to include that sequence in
your statistical analysis too. Then there's the statistical possibility of
errors that can occur and "hide" not being detected by the CRC, which you
have to include too. Most of this analysis is steeped in statistical theory
and statistical detection theory and is extremely tedious.
But at least once it's done, it's done, then you can apply all types of
noise sources to your model and see their effects. You'll discover that the
system is more sensitive to certain types of noise. (Like that wasn't known
to begin with.)
The results will be a plot of efficiency versus noise levels. My guess is
1/4 a man year of effort. Or, you could, just find some "threshold" points
using very gross assumptions. Through the use of these threshold points,
you'll have a good argument for justifying avoiding the long full-blown
analysis.
For example, assume some BER and find the effect on efficiency. Then with
some "hand waving" you can probably develop "not to exceed" limit for BER
caused by the radiation. and simply find a few points (more like areas).
Good luck.
Let us all know what you find.
- Robert -
-----Original Message-----
From: Jose Miguel Rio <[email protected]>
Cc: [email protected] <[email protected]>
Date: Tuesday, December 28, 1999 11:39 AM
Subject: RV: Network efficiency
----- Mensaje original -----
De: Jose Miguel Rio
Para: [email protected]
Enviado: martes, 28 de diciembre de 1999 18:48
Asunto: Network efficiency
Hi folks:
I have had an enquiry of a customer and I need some help.
How can electromagnetic radiation decrease the efficiency (speed) of a
computer network. I have heard that the transmissions errors increases, so
the retransmissions are more frequent and the network is slowed.
Do anybody know how can I quantificate this effect?
Thank you
Jose Rio
Advanced Shielding Technologies
---------
This message is coming from the emc-pstc discussion list.
To cancel your subscription, send mail to [email protected]
with the single line: "unsubscribe emc-pstc" (without the
quotes). For help, send mail to [email protected],
[email protected], [email protected], or
[email protected] (the list administrators).
