(abbreviated) We consider how tight binaries consisting of a super-massive black hole of mass $M=10^{3}-10^{4}M_{\odot}$ and a white dwarf can be formed in a globular cluster. We point out that a major fraction of white dwarfs tidally captured by the black hole may be destroyed by tidal inflation during ongoing circularisation, and the formation of tight binaries is inhibited. However, some stars may survive being spun up to high rotation rates. Then the energy loss through gravitational wave emission induced by tidally excited pulsation modes and dissipation through non linear effects may compete with the increase of pulsation energy due to dynamic tides. The semi-major axes of these stars can be decreased below a 'critical' value where dynamic tides are not effective because pulsation modes retain phase coherence between successive pericentre passages. The rate of formation of such circularising stars is estimated assuming that they can be modelled as $n=1.5$ polytropes and that results of the tidal theory for slow rotators can be extrapolated to fast rotators. We estimate the total capture rate as $\sim \dot N\sim 2.5\cdot 10^{-8}M_{4}^{1.3}r_{0.1}^{-2.1}yr^{-1}$, where $M_{4}=M/10^4M_{\odot}$ and $r_{0.1}$ is the radius of influence of the black hole in units $0.1pc$. We find that the formation rate of tight pairs is approximately 10 times smaller than the total capture rate. It is used to estimate the probability of detection of gravitational waves coming from such tight binaries by LISA. We conclude that LISA may detect such binaries provided that the fraction of globular clusters with black holes in the mass range of interest is substantial and that the dispersion velocity of the cluster stars near the radius of influence of the black hole exceeds $\sim 20km/s$.