Soft Decode and Forward (SDF) cooperation techniques have been shown to improve the performance of Distributed Turbo Coding (DTC) schemes, in which the Source Node (SN) and Relay Node (RN) employ convolutional codes, while the Destination Node (DN) employs a turbo decoder. However, this technique has not be able to mitigate the typical error floors that are associated with DTC. In this paper, we mitigate these error floors by upgrading the SDF technique to a scheme in which turbo codes are employed at the SN, RN and DN. We demonstrate that this scheme is capable of offering significant gains over both Amplify and Forward (AF) and classic direct-transmission schemes, even if only partial decoding is performed in the RN. More specifically, in our Partial Soft Decode and Forward (P-SDF) scheme, there is no excessive penalty associated with terminating the RN's iterative decoding process before convergence is achieved, as a benefit of the soft nature of its modulation scheme.