> > Given a significant number of parameters and the need to assess circuit
> > performance over multiple values for each parameter, the
> resulting number
> > of cases can be large. Varying each of 50 parameters through a 20% range
> > in 1% increments yields 1000 cases. And while some parameters can easily
> > be varied while measuring, others cannot; how does one step a
> tube through
> > its tolerance range for contaminants, or through its tolerance
> ranges for
> > inter-element spacings?
>
> You are making mountains out of what turn out to be mole-hills. It's
> pretty easy to find the extremes in phase shift and loss at several
> points, and determine if oscillation is possible.
If they were indeed molehills, the questions surrounding VHF oscillation and
suppressor effectiveness would have been resolved long ago.
> I think maybe the mental picture you have of the system is
> incorrect, perhaps you are assuming wild gyrations on multiple
> frequencies. Multiple complex resonances and feedback paths are
> typical for "unclean" layouts, like a breadboard wooden chassis
> PA with point to point wiring, but not in modern PA's with decent
> construction.
Words like "wild", "unclean" and "decent" are subjective, and thus not
conducive to proving anything. Modeling eliminates the need for imprecise
formulations in natural language. One describes a system with equations,
establishes the domain of each parameter via measurement or specification,
and then runs those equations over those domains to produce results that can
be repeated and objectively assessed by everyone, not just one person poking
a scope probe around one instance of a circuit.
> My point is the model would take longer to construct than the
> measurements would take, and almost certainly not be able to
> include anywhere near everything.
The measurement-and-argue approach to resolving the issues at hand has yet
to succeed. Creating and validating a model is certainly more time-consuming
than taking measurements. But after years of non-convergence, one should
consider optimizing for results rather than time.
> > > IMO, if modeling programs can't handle a simple inductor
> > > properly...they need to be carefully watched with other components.
> > Are you asserting that no modeling program properly models an inductor?
> > The fact that some do not hardly means that all do not. There are plenty
> > of low-quality scopes about, but that fact doesn't invalidate the
> > measurements made with high-quality ones.
> Please tell me the name of one that models a solenoid inductor
> properly at radio frequencies. I really would like to purchase a copy.
As previously stated, I am not an RF engineer. The last analog simulator I
used was written in APL and ran on Multics, a predecessor of Unix. I am
assuming that competent models exist for devices as physically
straightforward as a solenoid inductor. While the answer to my question "Are
you asserting that no modeling program properly models an inductor?" was not
direct, the response implies "Yes", at least in the case of a solenoid
inductor. That is surprising and, if true, dissapointing. Perhaps the
modeling software companies themselves will be interested in responding to
this allegation. I will canvass them, and hopefully identify one or more for
you to evaluate; IEEE Spectrum carries simulator advertisements, and Alta
Vista should turn up the rest. To save time, please name the modeling tools
that you've already tried and rejected, and the versions (or approximate
dates) involved.
73,
Dave, AA6YQ
--
FAQ on WWW: http://www.contesting.com/FAQ/amps
Submissions: amps@contesting.com
Administrative requests: amps-REQUEST@contesting.com
Problems: owner-amps@contesting.com
|