Draw an inset for two unrelated plots

Question

Suppose I have the following voltage vs. time data set, recorded by a decent oscillosocpe:

As you can see, there is some very fast changes between t=0 and t=1 sec for this measurement.

So I recorded this physical phenomena again, but now on a shorter time window, so the noisy part can be seen better:

Good! I have two csv files that I can plot with pgfplots.

What I am asking is how to produce this figure:

Using tikz's spy library, you can zoom in on a portion of a TikZ picture.

This drawing above is very similar in its appearance to tikz's spy tool, but it is made of two "completely unrelated" plots. spy is good only for making an inset of one plot.

Any one have an idea?

Note: These plots were created with Desmos and Windows paint. If you want to post an answer then a circle spied by a triangle is good too. This wiggling unrealistic waveform is just for demonstration purpose, to clarify the idea.

Thanks!

Update:

I read now the tikz manual and came across the Coordinate transformations subsection (25.3).

I just ran the following code, to compile successfully:

\documentclass[border=2pt]{standalone}
\usepackage{tikz,pgfplots}
\pgfplotsset{compat=newest}
\begin{document}
\begin{tikzpicture}
\begin{axis}[
grid=major,
]
\addplot[samples=10,domain=0:2]{x^2-2};
\end{axis}
\begin{scope}[xshift=8em,yshift=-20em]
\begin{axis}[
grid=major,
]
\addplot[samples=10,domain=-3:2]{-x^2+2};
\end{axis}
\end{scope}
\draw[red,thick] (0,0) -- (2,1); % <--- A decoration
\end{tikzpicture}
\end{document}

So I got two different axis environments, separated one from the other by the {scope}[xshift , yshift]. Now I added a thick red line from (0,0) to (2,1) but those two coordinates are not sticking to the (first) plot coordinates system, as you can see from the axis ticks.

I changed that to

\draw[red,thick] (axis cs:0,0) -- (2,1);

and got an error

Missing \endcsname inserted. \pgfcrdmth@x l.18 \draw[red,thick] (axis cs:0,0) -- (2,1); ?

So I need to give this coordinate system a name, so TeX know whom I refer to? How do I do that?

Answer 1

Are you looking for one of those?

\documentclass[border=2pt]{standalone}
\usepackage{pgfplots}
\pgfplotsset{compat=1.17}
\begin{document}
\begin{tikzpicture}
\begin{axis}[name=axis A,
grid=major,
]
\addplot[samples=10,domain=0:2]{x^2-2};
\end{axis}
\begin{scope}[xshift=8em,yshift=-20em]
\begin{axis}[name=axis B,
grid=major,
]
\addplot[samples=10,domain=-3:2]{-x^2+2};
\end{axis}
\end{scope}
\draw[red,thick] (axis A.north east) -- (axis B.north east)
(axis A.south west) -- (axis B.south west);
\end{tikzpicture}
\begin{tikzpicture}
\begin{axis}[grid=major,
]
\addplot[samples=10,domain=0:2]{x^2-2};
\draw (0,-2) coordinate (bl) rectangle (1,2) coordinate (tr);
\end{axis}
\begin{scope}[xshift=8em,yshift=-20em]
\begin{axis}[name=axis B,
grid=major,
]
\addplot[samples=10,domain=-3:2]{-x^2+2};
\end{axis}
\end{scope}
\draw[red,thick] (tr) -- (axis B.north east)
(bl) -- (axis B.south west);
\end{tikzpicture}
\end{document}

Answer 2

In this case, I imported my underlying graphs as images, so you could focus on the syntax of the \stackinset, where I place argument #5 (the "spy" boxes) atop argument #6 (the underlying graphics), in this case, aligned at the bottom-right corner. However, you could replace each of the \includegraphics calls with (in this case 3cm x 3cm) tikz drawings, or you could create them in standalone mode and import them, as I do here.

The key is to lay out the two images in a size you like, and that will provide you the graph coordinates to use for your "spy" boxes.

\documentclass{article}
\usepackage{tikz,stackengine}
\begin{document}
\stackinset{r}{}{b}{}{%
\begin{tikzpicture}
\draw [thick] (1,0) -- (4,0) -- (4,3) -- (1,3) -- (1,0);
\draw [thick] (.25,4) -- (.75,4) -- (.75,6) -- (.25,6) -- (.25,4);
\draw [thick] (.25,4) -- (1,0);
\draw [thick] (.75,6) -- (4,3);
\end{tikzpicture}%
}{%
  \stackon[.5cm]
    {\kern1cm\includegraphics[width=3cm,height=3cm]{example-image-a}}%
    {\includegraphics[width=3cm,height=3cm]{example-image-b}\kern1cm}%
}
\end{document}

Using the OP's sample graph, I could adapt. However, I highly recommend setting your graph up in a way where you know exact size of your images. Further, while both graphs are done in a single tikz picture, it may prove simpler to stack them as two separate tikz constructions.

\documentclass{article}
\usepackage{stackengine}
\usepackage{tikz,pgfplots}
\pgfplotsset{compat=newest}
\begin{document}
\stackinset{r}{.2cm}{b}{.5cm}{%
\begin{tikzpicture}
\draw [thick] (3.3,0) -- (10,0) -- (10,5.5) -- (3.3,5.5) -- (3.3,0);
\draw [thick] (.5,7.0) -- (2,7.0) -- (2,8.5) -- (.5,8.5) -- (.5,7.0);
\draw [thick] (.5,7.0) -- (3.3,0);
\draw [thick] (2,8.5) -- (10,5.5);
\end{tikzpicture}%
}{%
\begin{tikzpicture}
\begin{axis}[
grid=major,
]
\addplot[samples=10,domain=0:2]{x^2-2};
\end{axis}
\begin{scope}[xshift=8em,yshift=-20em]
\begin{axis}[
grid=major,
]
\addplot[samples=10,domain=-3:2]{-x^2+2};
\end{axis}
\end{scope}
\draw[red,thick] (0,0) -- (2,1); % <--- A decoration
\end{tikzpicture}
}
\end{document}