Spatial zero point variation relative to SDSS (Eddie Schlafly)

I promised I'd show some examples of this during DRAVG. Here are figures from ten arbitrary fields from the July MD release. Each pixel represents one cell of one chip. The median number of stars used per cell for computing the zero point via matching to SDSS is about four. An overall mean per-exposure zero point has been removed.

There are obviously good and bad images here, but chip-to-chip variations in zero point and within-chip variations in zero point are generic.