gnuplotgnuplottex
Here is my MWE. Why doesn't it reaches the x-axis?
\documentclass{standalone}
\usepackage[miktex]{gnuplottex}
\begin{document}
\begin{gnuplot}
set terminal epslatex color
set xzeroaxis lt -1
set yzeroaxis lt -1
set grid xtics lc rgb '#555555' lw 1 lt 0
set grid ytics lc rgb '#555555' lw 1 lt 0
set yrange [-2:2]
set xrange [-2*pi:2*pi]
plot abs(sin(x))
\end{gnuplot}
\end{document}
I would use the pgfplots package for this. You can generate the data with gnuplot by using \addplot gnuplot {<expression>};, and then read the generated data using \pgfplotstableread{\jobname.pgf-plot.table}\table. After sorting this table, you just access the first element, which now contains the local maximum.
\documentclass{article}
\usepackage{pgfplots}
\usepackage{pgfplotstable}
\begin{document}
\begin{tikzpicture}
\begin{axis}[ymin=0,ymax=0.05,xmin=0,xmax=2000,grid=both]
\addplot [domain={0:2000},samples=1000]
gnuplot {(20*x)/(100000 + 50*x + x**2)};
\pgfplotstableread{\jobname.pgf-plot.table}\table
\pgfplotstablesort[sort cmp={float >},sort key={[index] 1}]\sorted{\table}
\pgfplotstablegetelem{0}{[index] 1}\of{\sorted}
\let\maxy=\pgfplotsretval
\pgfplotstablegetelem{0}{[index] 0}\of{\sorted}
\let\maxx=\pgfplotsretval
\node at (axis cs:\maxx,\maxy)
[circle, fill, red,inner sep=1.5pt,
pin={
[fill=white]40:{(\pgfmathprintnumber{\maxx}, \pgfmathprintnumber{\maxy}})
}
] {};
\draw [red] (axis cs:\maxx,0) -- (axis cs:\maxx,1);
\end{axis}
\end{tikzpicture}
\end{document}
There are several tools for EPS to PDF conversion: at least ps2pdf, epspdf and epstopdf. It can be verified that the first one of these (which uses GhostScript) creates a PDF of page size (say, A4 or US Letter) and not content size, while the other two work as expected. Some internet digging shows that GhostScript accepts the -dEPSCrop option, which clips the picture to its bounding box.
So, use either of epspdf, epstopdf or ps2pdf -dEPSCrop to properly convert the picture.
Best Answer
What happens is that gnuplot is sampling the x-axis. It means that it does not compute the function for all the possible points in the interval
[-2*pi:2*pi]but only a finite number of points. The corresponding y-values are then linked together with a line to complete the plot. If the number of samples is too low you obtain a behaviour similar to yours. The best course of action is then to increase the number of samples considered by adding something likeset sample 1000(tested with gnuplot but not with gnuplottex).In gnuplot with
set sample 100:In gnuplot with
set sample 1000: