Sub-Chandrasekhar mass carbon-oxygen white dwarfs with a surface helium shell have been proposed as progenitors of Type Ia supernovae (SNe Ia). If true, the resulting thermonuclear explosions should be able to account for at least some of the range of SNe Ia observables. To study this, we conducted a parameter study based on three-dimensional simulations of double detonations in carbon-oxygen white dwarfs with a helium shell, assuming different core and shell masses. An admixture of carbon to the shell and solar metallicity are included in the models. The hydrodynamic simulations were carried out using the AREPO code. This allowed us to follow the helium shell detonation with high numerical resolution, and this improves the reliability of predicted nucleosynthetic shell detonation yields. The addition of carbon to the shell leads to a lower production of 56Ni, while including solar metallicity increases the production of intermediate mass elements. The production of higher mass elements is further shifted to stable isotopes at solar metallicity. Moreover, we find different core detonation ignition mechanisms depending on the core and shell mass configuration. This has an influence on the ejecta structure. We present the bolometric light curves predicted from our explosion simulations using the Monte Carlo radiative transfer code ARTIS and make comparisons with bolometric SNe Ia data. The bolometric light curves of our models show a range of brightnesses, which is able to account for subluminous to normal brightness SNe Ia. We show the model bolometric width-luminosity relation compared to data for a range of model viewing angles. We find that, on average, our brighter models lie within the observed data. The ejecta asymmetries produce a wide distribution of observables, which might account for outliers in the data. However, the models overestimate the extent of this compared to data. We also find that the bolometric decline rate over 40 days, Δm40(bol), appears systematically faster than data.