The design, simulation, and fabrication of a flow sensor to be integrated into an implantable micropump is presented. The flow sensor operates by the method of thermal anemometry, in which heat is dissipated from a resistive element held in the flow of the fluid. The rate at which heat is carried away is dependent on the flow rate and is directly related to the thermal conductance. A control circuit utilizing the constant-temperature anemometry mode of operation is used to generate a change in voltage in response to change in thermal conductance, and subsequently, flow rate.

A mathematical expression describing the sensor sensitivity based on thermal effects is proposed, based on the thermal spreading resistance and basic heat transfer laws. The mathematical model is refined using finite-element analysis, and a complete formulation for the effect of sensor area, length-to-width ratio, and fluid velocity on thermal spreading resistance is determined. The refined thermal spreading conductance equation can be used to replace assumptions made in initial mathematical analysis.

An original fabrication process is presented and investigated, in which a p-doped polysilicon bridge is encapsulated in silicon oxide and silicon nitride using surface micromachining techniques. A sacrificial polysilicon layer and KOH etching are used to form half of the complete fluid channel in the bulk of the silicon wafer. When the fluid channel is sealed with a complementarily etched wafer, the sensor bridge is situated in the middle of the fluid channel, optimally placed for maximum sensitivity. The fabrication process yields functional sensor bridges, with even the most fragile sensor shape withstanding the process.

An analog constant-temperature control circuit is developed, based upon a Wheatstone bridge with a feedback op-amp. The circuit produces an output voltage dependent on thermal resistance, and will reduce the effect of fluid temperature fluctuations on circuit output. This relies on introducing a second sensor bridge into the fluid channel, and limiting its self-heating. A design based on replacing one element in the Wheatstone bridge with a “transimpedance virtual resistor”, and another design based on reducing the power to one leg of the bridge are simulated, and their performance compared. PSPICE simulations are used to optimize the circuits in order to maximize the ratio of sensitivity to fluid velocity to sensitivity to ambient temperature.