Hi Paul,
The data here represents the output of a Fourier transformed time series. Since
the data isn't infinitely long, there is FFT leakage from the peak bin into
adjacent bins. Window functions can in some cases reduce FFT leakage, but
leakage is impossible to remove completely. With that in mind, if bins adjacent
to the peak bin receive unsymmetrical amounts of leaked spectral power, then the
actual peak is not perfectly centered in one bin. Curve fitting isn't perfect,
but it will allow the calculation of a reasonable approximation of the true
location of the peak to a resolution less than the bin width.
I agree that in this dataset there is a dearth of bins adjacent to the bin
containing the local maximum. Nevertheless, a curve can be fitted to even this
small group of bins.
--
Dave
On 12/10/20 21:34, Paul Oxley wrote:
> Dave
>
> Unfortunately, curve fitting will not help resolve points between bins. You need
> more points to develop a curve. The data has a peak with a large slope on both
> sides. Data outside the peak and two bins on either side is not material to the
> position of the peak.
>
> A guess might be obtained by averaging the three points.
>
> Paul
>
> Paul
>
> On Thursday, December 10, 2020, 09:25:08 PM EST, Dave Typinski
> <
dav...@typnet.net> wrote:
>
>
> Hi Paul,
>
> In a perfect world, yes. We can, however, work with imperfect data and obtain a
> reasonable sub-bin-width approximation of the peak's location using curve fitting.
> --
> Dave
>
>
> On 12/10/20 21:15, Paul Oxley wrote:
> > Dave
> >
> > If you want to know where the peak occurs, you need more sample points.
> >
> > Paul
> >
> > On Thursday, December 10, 2020, 08:43:17 PM EST, Dave Typinski
> > <
dav...@typnet.net <mailto:
dav...@typnet.net>> wrote:
> >
> >
> > Hi Michiel,
> >
> > Here's a plot of the data showing one peak with some prominent side lobes. The
> > idea is to fit a Gaussian / Lorentzian / whatever curve to the main peak to
> > determine its location along the horizontal axis with sub-sample precision
> > (i.e., interpolation between two samples).
> >
> > I expect the true peak is somewhere between bin numbers 897 and 898.
> >
> >
> > What I'm looking for is a Python package that does this sort of peak curve
> > fitting. No sense reinventing the wheel if it already exists.
> > --
> > Dave
> >
> >
> > On 12/10/20 13:50, Michiel Klaassen wrote:
> >> Hi Dave,
> >> There seems to be no answers to your question, and I myself think it is also
> >> difficult to answer.
> >> Can you give an example graph with (paint) indication what you want.
> >> regards,
> >> Michiel
> >>
parac.eu <
http://parac.eu>
> >>
> >> Op do 10 dec. 2020 om 04:46 schreef Dave Typinski <
dav...@typnet.net
> <mailto:
dav...@typnet.net>
> >> <mailto:
dav...@typnet.net <mailto:
dav...@typnet.net>>>:
> >> <mailto:
sara...@googlegroups.com <mailto:
sara...@googlegroups.com>>
> >> <mailto:
sara-list-...@googlegroups.com
> >> email to
sara-list+...@googlegroups.com
> <mailto:
unsub...@googlegroups.com>
> >> <mailto:
sara-list%2Bunsu...@googlegroups.com
> <mailto:
2Bunsu...@googlegroups.com>>.
> >> <mailto:
sara...@googlegroups.com <mailto:
sara...@googlegroups.com>>
> >> <mailto:
sara-list-...@googlegroups.com
> >> email to
sara-list+...@googlegroups.com
> <mailto:
unsub...@googlegroups.com>
> >> <mailto:
sara-list+...@googlegroups.com
> <mailto:
unsub...@googlegroups.com>>.
> > to
sara-list+...@googlegroups.com <mailto:
unsub...@googlegroups.com>
> > <mailto:
sara-list+...@googlegroups.com
> <mailto:
unsub...@googlegroups.com>>.
> > to
sara-list+...@googlegroups.com <mailto:
unsub...@googlegroups.com>
> > <mailto:
sara-list+...@googlegroups.com
> <mailto:
unsub...@googlegroups.com>>.
> to
sara-list+...@googlegroups.com. <mailto:
unsub...@googlegroups.com.>
> --
> --
> You received this message because you are subscribed to the Google
> Groups "Society of Amateur Radio Astronomers" group.
> To post to this group, send email to
sara...@googlegroups.com
> To unsubscribe from this group, send email to
>
sara-list-...@googlegroups.com
> For more options, visit this group at
>
http://groups.google.com/group/sara-list?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "Society of Amateur Radio Astronomers" group.
> To unsubscribe from this group and stop receiving emails from it, send an email
> to
sara-list+...@googlegroups.com
> <mailto:
sara-list+...@googlegroups.com>.
> To view this discussion on the web visit
>
https://groups.google.com/d/msgid/sara-list/1750963326.6867069.1607654087876%40mail.yahoo.com
> <
https://groups.google.com/d/msgid/sara-list/1750963326.6867069.1607654087876%40mail.yahoo.com?utm_medium=email&utm_source=footer>.