An experimental test set for magnetic bubble devices has been constructed in which the spacing between the garnet film and the Permalloy overlay is variable. The experimental uncertainty in spacing is approximately m, and spacings as small as m have been attained. Bias margin data are presented which were taken at 1 Hz on a 20 micron period chevron circuit as a function of spacing. The collapse and strip-out fields begin to be affected when the spacing is comparable to the garnet film thickness, increasing as the spacing decreases. At larger spacing the high-bias failure mode changes from collapse to uncorrelated bubble motion. A theoretical model which accounts for some aspects of the spacing dependence of the strip-out and collapse fields is described. This model approximates the circuit by a continuous Permalloy sheet. At the low spacing required for efficient use of the rotating field, the model indicates that ±10% nonuniformity in a 2 micron spacing over the device area results in a degradation of the bias field operating range by about 12%.