Insects play an irreplaceable role in biodiversity survey, while light trapping is one of the most efficient methods in the insect collection and it is widely used for investigation of nocturnal insect communities. The traditional light trapping method, however, has the problem that affects the reliability of the survey results. Firstly, its lighting range is not controlled, which the trapping area cannot be defined. Secondly, while the standardization and stability of data acquisition is essential in insect ecology researches, the manual collection is not as standard as automatic methods. Therefore, we devised an improved device called "spotlight box trap" (SBT), and used the moth collecting tent (MCT) as the traditional method. The insect collection comparison study of the two light trapping devices was carried out in the sample plots and involved two forest stands (the pure plantation of Populus×beijingensis and Pinus tabuliformis), with moths (night active Lepidoptera) as the representative insect groups. The two light trapping methods were evaluated by comparing the differences and error levels of the number of individuals and species, species composition and dominance, species diversity index and morphological parameters of collected moths. A total of 6728 moths' specimens of 24 families, 164 species were collected; 1164 of 19 families, 93 species by SBT and 2573 of 22 families, 125 species were collected by MCT in Populus×beijingensis forest; 1044 of 21 families, 87 species by SBT and 1947 of 22 families, 117 species were collected by MCT in Pinus tabulaeformis forest. The results showed that:(1) the number of individuals and species in MCT was higher than that in SBT, and overall SBT had lower level of data error across sites. (2) Within the same month, the dominant moth community structure collected by the two devices was significantly different (P<0.05), and the proportion of shared species was lower in MCT than that in SBT, that is, the former device collected more endemic species. (3) In different plots, the data error levels of richness, diversity and evenness index of SBT were higher than those of MCT. (4) The moth species collected by MCT had overall larger body size (1.20 times that of SBT) and stronger flight ability (the relevant parameters are about 1.20 to 1.36 times than those of SBT). In conclusion, we consider that in the investigation of forest insect diversity, although the traditional light trap is commonly used, the data stability of the method is not reliable, and its representation of the forest is insufficient. This makes it difficult to conduct insect communities based on forest stands. The improved device, "spotlight box trap", is more stable due to its controllability of light domain and standardization of collection process, and has better applicability and application prospects in forest insect diversity investigation and monitoring.