Zero-Crossing is closely related to the Gardner timing error detector. Instead of using only the upsampled signal, ZC also makes use of hard decided symbols.

Compared to Gardner (as used here), zero crossing combines the two times upsampled signal $x_{i}(n)$ with hard decisions $s_n$ on the input signal $x(n)$. $$ t_\mathrm{err} = \mathrm{Re}\left\{\sum_{n=0}^{N/2-1}\left[x_i^*(2n-1)(s_{n-1} - s_{n})\right]\right\}, $$ where $N$ is the averaging length.