[StevenB]

Before you go looking for another gyro, make sure you first understand the one you've got.

First, all gyros have drift. That's ok as long as the drift is constant, because you can calibrate for it. Let the gyro sit still and average the reading over a few seconds. Then subtract this value from all the future readings. Note that the drift will vary with temperature, so a lot of sensors will actually include a separate temperature output you can read and use to compensate.

In practice, I've found that 1 degree per minute would be really good, and even 5 deg/min isn't so bad. You might want to zero the gyro before you begin climbing, so that anything that happens beforehand doesn't matter. Depending on the gyro's sensitivity, hard slamming or slow movements will cause small errors, and you'll inevitably end up with some accumulated error.

Sensitivity is a more complex topic than drift. Sensitivity is (by my definition) the smallest motion that you can accurately measure. Since the ADC on the robot controller has a set number of bits, then the sensitivity is inversely proportional to the range. That is, for a 10-bit ADC (which gives 1024 steps), a 300 deg/sec gyro has a precision of 300/1024 = 0.3 deg/step. A 100 deg/sec gyro would give 0.1 deg/step precision. With an analog ADC, sensitivity and range are directly in competition, and you need to think carefully about what you actually need.

Of course, that isn't the full story to sensitivity. If your sensor is very noisy, then the precision with which you can measure is reduced. For example, if there is +/- 100mV, then with a 10-bit ADC, you'll get 1024 steps / 5V * 0.1 V = +/- 20 steps. In this case, using a higher-sensitivity gyro will let you get better measurements for small motions that you wouldn't otherwise get.