The Allan Variance technique is a very powerful mathematical method which allows a user to assess the real performance of any gyro in relation to the dynamics of his application, taking into account the effects of bias, noise, drift and long term sensor instability. It's beyond the scope of this glossary to fully describe the technique (contact us or browse the web for more information) but here is a summary of the key points: Bias is always measured by averaging successive data samples - the challenge is always to select a suitable time over which to average sampled data. Averages taken over short time intervals will be dominated by noise, those over a longer period by longer term drift. The technique involves selecting a range of time intervals (eg from 0.01s to 500s) over which to average data. The variation (standard deviation) from one averaged time period to the next is calculated and plotted against the averaging interval in log-log form. The resulting graph has a characteristic 'bathtub' shape. By examination of this graph, it is possible to compute the key defining characteristics of the gyro, namely

- Angular random walk

- Bias instability**-** Rate random walk