Getting an accurate estimate of site-specific ground thermal conductivity from thermal response tests (TRTs) is crucial to the efficient and sustainable use of ground source heat pump systems. In a conventional TRT, the ground thermal conductivity is estimated by perturbing the ground with a positive heat injection and the response is measured in time. Despite the simplicity, the conventional TRT does not take into account the ground thermal response to cooling pulses. This may lead to mode-biased estimation due to the presence of natural convective effect. To address the problem, a constant-temperature TRT that comprises simultaneous heat injection and extraction was proposed. A test data analysis method was developed based on the finite line-source model. On this basis, a very first field test was performed in Taiyuan, China, demonstrating the use of the proposed TRT concept with full-scale measurements. Results show that, in the case of this field test, the ground thermal conductivities derived separately from heat injection and heat extraction are 1.83 W/(m K) and 1.65 W/(m K), respectively. The significant difference (10.9%) indicates that when heat injection is involved, the natural convective effect has enhanced ground heat transfer to increase the conductivity estimate. Thus, as compared to the conventional approach, the use of the proposed TRT can improve the characterization of the ground thermal response and conductivity. Relevant results and the test protocol reported in this study will be useful in providing the background information for this technology to be adopted in the ground source heat pump industry.