Thank you very much for your thoughts. This is excellent advice.

Thanks!

-at

"Dick Pierce" wrote in message
om...
"at" wrote in message
...
How good results can you get by using computer programs to design
speakers?
There are many such progs available. How much trust would you place into
the
results? If you make a cool design by a program and build it, what can
you
expect in reality?

Can you recommend some good progs that would work? What are the most
useful?
And how can you get most use of them?

The accuracy of these programs is limited by a number of contraints:

1. Your understanding of the limits of the model. Any program based
on, for example, the Thiele-Small model will work well (given
other contraints listed here) in predicting the resulting response
with in the piston band of the system, but will not include
diffraction and baffle effects. If you expect the predicted
response of a 12" woofer at 800 Hz to be accurate, you WILL be
disappointed because you did not understand the limits of the
model.

2. As in all such excercises, garbage in give garbage out. And that
means if you plug in driver parameters from a catalog or a spec
sheet and expect the results to reflect reality you WILL be
disaapointed, because you failed to account for the fact that
there can (and often is) large variations in the operating
parameters of drivers AND differences between the specs and
purchased reality. Similarily, the performance and drivers
changes with conditions, and if by not accounting for these
changes you design a system that violates the constraints, you
WILL be disappointed in the results.

3. The sophistication of the model and the accuracy of the model
description. If you make assumptions about some of the detailed
parameters of the design, you WILL be disappointed. For example,
you design a system which assumes perfect absorbtion in the
enclosu if your design depends upon that assumption, you WILL
be disappointed in the result, because you made unrealistic
assumptions

Now, notice that all of the above comments have the operative word
"you" as the agent of failure. This is based on my experience extending
over 30 years of doing this wort of work, more importantrly, of watching
many others do it. My observations show clearly that the discrepancy
between the predicted design and implemented reality is due almost
entirely to the program user: that person's failure to understand
what the model does and what its limitations are, providing unrealistic
parameters to the program, and making unrealistic assumptions about
the design. I have seen a half dozen speaker design programs take the
same garbage data and result in the same garbage design, and those same
half-dozen programs take the same good data and operating under the
same reasonable contraints spit out a VERY competent and realizable
design.

Now, beyond that, there's another factor you didn't ask about. While
all of the programs, within their constraints, do a very reasonable
job, some of them a simply miserable to USE because the person who
wrote the program spent most of their time working on implementing
the math and NO time on dealing with simple usability issues. What
good is the most sophisticated, realistic modeling program if you
end up being so frustrated using it that you're ready to shoot the
computer followed by yourself? This problem, regrattbly, plagues more
programs than it should, and it's not just limited to speaker
software, to be sure.

An example of this are the older versions of LMS. It's a VERY good
mathematical model under the hood, but it has a LOUSY user interface.
I don't know if they've cleaned it up or not, but it was bad enough
that I found it difficult to recommend even considering the sophistication
of the model. It also was buggy in the sense that you could see gross
mathematical truncation or roundoff errors occuring in plots.