I'm trying to find $f_Z(z)$ with $Z=2X-Y$, for $X$ and $Y$ with joint density function $f_{XY}(x,y)$: $$ \begin{cases} x/8 & 1 \le x \le 3 \land -1 \le y \le 1 \\ 0 & \text{elsewhere} \end{cases} $$
and I want to do it without using the convolution theorem.
My strategy was first finding the distribution of Z $F_Z(z)$ and then finding its derivative, which is the density function of Z $f_Z(z)$.
So, what I'm trying to find first is: $$F_Z(z)=P\{Z \le z\}=P\{2X+Y \le z\}=P\{(x,y) \in D_Z\} =\iint_{D_Z} f_{XY}(x,y) \,dx \, dy$$ with $D_Z$ being the half plane $2x-y \le z$.
I'm not sure how to find the limits of the integrals. I was trying something like this:
If $z \lt 1$, the joint density of $XY$ is zero, so the distribution of $Z$ is zero (I think I'm doing something wrong here). The plot is (sorry about the quality of the plots, I didn't knew how to make them look clearer): plot1
If $1 \le z \le 3$ the joint density of XY equals $x/8$. The plot is: plot2 In this case, $D_Z$ is the shaded area inside the rectangle. So, the distribution of Z for $1 \le z \le 3$, is: $$F_Z(z)=\int_{y=-1}^{y=1} \int_{x=1}^{x=\frac {z+y}{2}} \frac {x}{8} \, dx \, dy = \frac {z^2}{32}-\frac{5}{96}$$ Then, the density function of Z is $$f_Z(z)=\frac {z}{16}$$
But this result does not match the one that I obtained doing the same thing using the convolution theorem ($X$ and $Y$ are independent).
I also found the rest of the limits on $z$, but didn't do the integrals because I think that the only one that I did is wrong. They are: $$3 \lt z \le 5$$ $$5 \lt z \le 7$$ $$z \gt 7$$ All of these were found by finding the intersections of $z=2x-y$ with the corners of the rectangle (I can elaborate if necessary).
Thanks in advance to anyone that can be of help.