[daip] one more suggestion...

[Gustaaf van Moorsel]

Right after we discovered the aliasing, I suggested something along
those lines.  But it was never implemented at the time since the
polynomial fit appeared to work well in all circumstances.  Now
this seems no longer to be the case it may be time to reconsider.
It would seem to me that a two parameter fit (amplitude and slope)
should suffice.

Gustaaf

Elias Brinks wrote:
> HI Eric,
>
> One more thought. Michael Rupen, in one of his e-mails to me, stated:
>
> * According to Barry:
>   > The theoretical law, and I would be astounded if it were not closely
>   > followed, is that the complex amplitude in channel n is
>   >
>   >   1+a*exp(2*i*pi*(2*n-1)*W*w)
>   >
>   > where W is the channel width, in, say, Mhz
>   > w is the usual w expressed in, say, microseconds
>   > a is a channel dependent constant that I'm sure Mike Revnell could 
> supply,
>   >   being the ratio of the transmission at f0-(n-0.5)*W to that at
>   >   f0+(n-0.5)*W.
>
> Now, apparently this behaviour is sort of seen in the spectra, but not
> exactly. But, it DOES suggest some functional behaviour and it
> would seem that it might be an option to fit something resembling
> the above function in UVLSF rather than a polynomial. Presumably,
> such a function is less likely to "go wild". Have you ever considered 
> this?
> Would it be an option?
>
> Cheers,
>
> Elias