[Qucs-devel] Re: ASCO

[Stefan J. <st...@gr...>]

On Fr, 27.01.2006, 09:01, asco developer wrote:

Hello!

> ...
>> Attached the curves with original/first/second try from you.  It looks
>> really good and also shows the tradeoff thingy.  I can't verify that
>> any of these is an optimum or not.  This is hopefully ensured by your
>> software.  I assume so. :-)
>
> These type of algorithms never guarantee that the best point is ever
> found.  You can just be reassured that "most of the times" you are
> "close" to it. Which, at least for me is quite satisfactory.
>
> The idea when using ASCO (and similar approaches) is to use
> mathematics to get a justification why is 1 and not 10 or 50. Then,
> the software is used to see if a better trade off can be found,
> without using some of the approximations we have used to derive the
> equations.

I see.

>> A function like "at(S[2,1], 12e9)" (S21 at 12GHz) or "find(S[2,1], 1)"
>> (frequency where S[2,1] is 1) just went up on my TODO list. :-)
>
> I can propose something like
> 1) y, yval, yvalue... y(s(11), 1GHz)
> 2) x, xval, xvalue... x(s(11), 2)
>
> which can be generalized to an
>
> y(vdb(node1), 5GHz)          => magnitude in dB
> y(v(node1), 5GHz)            => magnitude
> y(vp(node1), 5GHz)           => phase at 5GHz
> MAX(vdb(node1), 5GHz, 10GHz) => maximum
> MIN(vdb(node1), 5GHz, 10GHz) => minimum
> AVG(vdb(node1), 5GHz, 10GHz) => average
>
> Its like finding some scalable and logical language on how to measure
> a given waveform which can be implemented gradually. Just some ideas!

I put the following things onto my TODO list:

  yvalue (S[1,1], 12e9)
  xvalue (S[1,1], 1)
  max (S[1,1], 2e9: 5e9)
  min (S[1,1], 2e9: 5e9)
  avg (S[1,1], 2e9: 5e9)

Whereas S[1,1] can be vectors, and matrix vectors with a single dependency
only (in this case it is "frequency").  The range specification is max(),
min() and avg() can be either:

  xval1:xval2  OR  (values from xval1 to xval2)
  xval1:       OR  (values from xval1 to last)
  :xval2       OR  (values from first to xval2)
  :                (all values)

The specified values are going to be inclusively.

Each given x value *must* be in the dataset, otherwise the functions return
an error.  When given a y value (as in e.g. xvalue()) then the nearest
x value will be returned.  Or should be an appropriate interpolation be
applied?

Any more suggestions, objectives, comments?

Thanks in advance,
  Stefan.