Highly accurate flow measurements with thermal flow sensors using the alternating direction method

We present a novel method to eliminate additive drift in silicon thermal flow sensors, the Alternating Direction method(ADM). In this sensor design the flow signal is contained in the gradient of a two-dimensional temperature distribution T(x,y) on the chip. This temperature gradient is induced by the convective heat transfer of the flowing medium, which is overall dependent on (i) the physical aspects of the flowing medium, (ii) the physical characteristics of the sensor and (iii) the thermal coupling of the sensor to its holder and environment. Ideally, the gradient in T(x,y) is a homogeneous function of (i). In real-world applications, however, the non-ideal aspects (uncertainty in the functional relationship of T(x,y) on the flow) are generally due to asymmetries in the silicon chip and its mounting, resulting in a further additive dependency of the gradient T(x,y) on (ii)-(iii). With thermal flow measurement of very low airflow velocities (0-30 cm/s), the factors (ii) and (iii) can have dramatic influence on the thermal gradient, for instance by inducing flow disturbances by the thermally induced convection. These additive dependencies contribute to unacceptable measurement errors in the low flow regime. ADM applies to vector sensors, having anisotropic sensitivity for the measurand S, and isotropic sensitivity for all other possible input signals, such as those induced by the above mentioned influences (ii) and (iii). ADM eliminates virtually all drift, providing the desired performance enhancement especially for the purpose of long term volume measurements. In this paper, the theory and possible applications of ADM are presented