Because of its simple principle and high adaptability to severe operational conditions, the capillary-tube viscometer has been widely used for viscosity measurement. However, difficulties in accurately correcting the end effect induced measurement deviation will result in great uncertainty for measurement results. In order to solve this problem, in this work, we studied factors affecting the end effect by conducting the high pressure nitrogen viscosity measurement at low flow velocity with an improved capillary-tube viscometer. The experimental results indicated that the influence of the end effect became less significant with the decrease in flow velocity (v) and tube inner diameter (d) and varied inversely with the length of tube (L). We defined the ratio of measured viscosity to standard viscosity obtained from the NIST database as the viscosity deviation coefficient (Ce). From the Ce vs v, Ce vs d, and Ce vs L curves, we have observed that there existed a threshold velocity (vthreshold), a threshold diameter (dthreshold), and a threshold length (Lthreshold) at which Ce got closer to 1.0. It suggested that under certain experimental conditions, the influence of the end effect on gas viscosity measurement became negligible. Based on that, we established end effect free capillary-tube viscometry and compared the nitrogen viscosity results measured by this method with the data provided by the NIST database. The results presented a good match with error within 1.2%. These insights will contribute to improving the accuracy of a capillary-tube viscometer especially under high pressure.