I got this ellipse yesterday with center on x-axis going through points (0, p) and (0, -p) and touching unit circle twice (double contact) at x = t
$$ x^2+y^2-1+\dfrac{1-p^2}{t^2}(x-t)^2=0 $$
Here is the context when I asked for the solution Perspective view of longitudinal great circles - ellipses inside a circle
I changed the variable names to $p$ for pole and $t$ for touching point to clarify that this ellipse equation does not contain the semis $a$ and $b$ or the center directly.
Now how to determine center and semi-axis? What I finally did after tons of mistakes is multiply out etc. and put it into general equation, keeping the variables/parameters:
$$ (1 + \dfrac{1-p^2}{t^2}) x^2+y^2 - \dfrac{2(1-p^2)}{t}x - p^2 = 0 $$
I have no $xy$ ("B") or $y$ ("E") in this case, so the long converting equations on wikipedia/ellipse shrink a bit, e.g.
$$ x_0 = \dfrac{2CD}{-4AC} $$
I this overkill? The semi-axis have a much longer equation.
Because the person who offered the solution guided me to "completing the square". Maybe he was thinking of a specific value pair for $p$ and $t$. But even so this method seems cumbersome, because I don't have nice coefficients, and also I want to find several ellipses with different $p$ and $t$. I must also add/admit that I never heard of this method.
To me it seems these general-form-coefficients to canonical-form-parameters equations on wikipedia are made exactly for a case like this. Now I can go: with given $p$ and $t$ I get such and such A, C, D and F, and thus my $x_0$, $a$ and $b$ are this.