The differential code bias (DCB) in satellites of the Global Navigation Satellite Systems (GNSS) should be precisely corrected when designing certain applications, such as ionospheric remote sensing, precise point positioning, and time transfer. In the case of COMPASS system, the data used for estimating DCB are currently only available from a very limited number of global monitoring stations. However, the current GPS/GLONASS satellite DCB estimation methods generally require a large amount of geographically well-distributed data for modeling the global ionospheric vertical total electron content (TEC) and are not particularly suitable for current COMPASS use. Moreover, some satellites with unstable DCB (i.e., relatively large scatter) may affect other satellite DCB estimates through the zero-mean reference that is currently imposed on all satellites. In order to overcome the inadequacy of data sources and to reduce the impact of unstable DCB, a new approach, designated IGGDCB, is developed for COMPASS satellite DCB determination. IGG stands for the Institute of Geodesy and Geophysics, which is located in Wuhan, China. In IGGDCB, the ionospheric vertical TEC of each individual station is independently modeled by a generalized triangular series function, and the satellite DCB reference is selected using an iterative DCB elimination process. By comparing GPS satellite DCB estimates calculated by the IGGDCB approach based on only a handful (e.g., seven) of tracking stations against that calculated by the currently existing methods based on hundreds of tracking stations, we are able to demonstrate that the accuracies of the IGGDCB-based DCB estimates perform at the level of about 0.13 and 0.10 ns during periods of high (2001) and low (2009) solar activity, respectively. The iterative method for DCB reference selection is verified by statistical tests that take into account the day-to-day scatter and the duration that the satellites have spent in orbit. The results show that the impact of satellites with unstable DCB can be considerably reduced using the IGGDCB method. It is also confirmed that IGGDCB is not only specifically valid for COMPASS but also for all other GNSS.