We consider coarse vector quantization for the fronthaul of distributed or "cell-free" massive MIMO, which is also related to Cloud Radio Access Network (C-RAN) architecture. We propose the use of lattice vector quantization (LVQ) based on Voronoi codes, and discuss how they can be adapted for efficient quantization of correlated signals, using a near-ellipsoidal Voronoi codebook. The Bussgang decomposition is extended to model these quantizers, and some numerical results obtained. A comparison of the quantization error with that of the Linde-Buzo-Gray (LBG) shows that for strongly correlated signals, while simple Voronoi LVQ has around 3 dB poorer performance, the near-ellipsoidal codebook is very close to LBG.