Fig. 6.

Homogenisation for the capture of source/sink mismatch. (a) The nozzle structures used to evaluate the performance of homogenisation. Activation delay (AD) is measured at the location indicated, and calculated by comparing activation times to those seen when simulating the same domain but with no obstruction present. (b) AD and conduction block as predicted by homogenised models using linear boundary conditions for closure subproblems (9), for a range of exit widths $r$ ($l=\text{200}\text{µm}$). Failure to activate the measure point (conduction block) is marked by an $\mathit{\times }$ placed at the top of the vertical axis. Good agreement with the finescale model is achieved for homogenised models with grid spacing $\Delta x\le \text{100}\text{µm}$. (c) AD (dot colour) and conduction block ($\mathit{\times }$) in the finescale model, and the best-performing homogenised model using linear boundary conditions ($\Delta x=\text{100}\text{µm}$), for multiple combinations of $l$ and $r$. Delay is well estimated by the homogenised model, and block is perfectly predicted in all trialled scenarios. (d) Maps of activation time (AT) for two finescale problems (top row) and the equivalent homogenised models (bottom row). Despite changes to the effective channel width in the homogenised models, the timing of activation is well recovered. Noticeable delay in activation is seen where the homogenised model’s larger grid size results in a sudden change in channel width (light blue circles)..