A 1.55-microm continuous-wave heterodyne Doppler lidar (HDL) with three receiver-detector units is used to validate experimentally the findings presented in the Guérit et al. companion paper [Appl. Opt. 41, 2232 (2002)] on the effectiveness of independent realizations to improve HDL performance (velocity or power estimates or both) at a low carrier-to-noise ratio (CNR). In fact, noise has a detrimental effect on the accumulation techniques, so in the Guérit et al. companion paper, the chances of getting "heavy" speckles in HDL signal from many receiver-detector units on a single- or several-shot basis are investigated theoretically and numerically with the Zrnic HDL model. The experimental results enable us to conclude there is a very good agreement (better than 95%) between the performance computed from actual HDL data and from the theoretical prediction.