(I would like to thank Kim Constantikes who computed these Allan curves and shared them with me. I sincerely appreciate it.)

Here is the Allan standard deviation curves for the gyroscopes in Nexus S.

Figure 1: Allan standard deviation of the gyroscopes in the Nexus S. The unit of vertical axis is "rad/sec".

One should be careful while using the gyroscopes in the Nexus-S. As shown in the following figure, there is a significant (and systematic) change in the sampling time even when the sampling mode is set to "SENSOR_DELAY_FASTEST".

Figure 2: Sample number vs Sampling time (nanosecond)
Notes:
  1. According to the Nexus-S's kernel drivers both the accelerometer and gyroscope are produced by ST. The sensors were labeled as k3g and kr3dm in the drivers. However, I was not able to find any such model in the ST website. I believe ST just re-brands one of its sensors for the nexus-s.
  2. The comparison of the above AV curve and the AV for curve for the iPod reveals that the gyroscope in the Nexus-s is approximately 4 times worse (in the power sense) than the gyroscopes in the iPod, which are also produced by ST.