The set of $\Phi$ badly approximable matrices has full Hausdorff dimension

Recently Koivusalo, Levesley, Ward and Zhang introduced the set of simultaneously $\Phi$-badly approximable real vectors of $\mathbb{R}^m$ with respect to an approximation function $\Phi$, and determined its Hausdorff dimension for the special class of power functions $\Phi(t)=t^{-\tau}$. We refine this by naturally extending the formula to arbitrary decreasing functions, in terms of the lower order of $1/\Phi$ at infinity. We also provide an alternative, rather mild condition on $\Phi$ for this conclusion. Moreover, our results apply in the general matrix setting, and we establish an according formula for packing dimension as well. Thereby we also complement a recent refinement by Bandi and de Saxc\'e on the smaller set of exact approximation with respect to $\Phi$. Our basic tool is (a uniform variant of) the variational principle by Das, Fishman, Simmons, Urba\'nski. We also prove some new lower estimates regarding the set of exact approximation order in the matrix setting, which are sharp in special instances. For this we combine the result by Bandi and de Saxc\'e with a method developed by Moshchevitin.