Fig. 3: Camera calibration increases performance of an uncooled, industry-grade CMOS camera for SMLM and diffraction-limited fluorescence imaging.

Maps of dark current (a) and noise (b) at 50 ms single frame exposure time for an uncooled, industry-grade CMOS camera (characteristics shown in Fig. 1d, e). Simulations of astigmatism-based 3D PALM without explicit consideration of pixel-wise effects show a similar pattern in the localization bias, but of greater amplitude as compared to the cooled sCMOS camera (compare Fig. 2), both laterally (c) and axially (e). Explicit application of CMOS-specific fitting largely removes the bias for PALM (d, f) as well as STORM and DNA-PAINT (g) and restores the theoretically achievable root mean square error in the localizations (h). i, Experimental 3D PALM data of clathrin tagged with mEOS3.2 in a U2OS cell using the same camera and applying CMOS-specific fitting including thermal effects. j, Axial view of region indicated in i. k Experimental 3D STORM data of Nup107-SNAP, labeled with AF 647, in a U2OS cell using the same camera and applying CMOS-specific fitting including thermal effects. l Gallery showing lateral and axial views on individual nuclear pore complexes indicated in k. m The axial view on the region indicated in (k) shows two parallel lines from the nucleo- and cytoplasmic rings 57 nm apart. n First frame of experimental raw data from time-lapse TIRF imaging of AP-2 tagged with eGFP in live U373 cells recorded at 1000 ms single frame exposure time. o NCS corrected and (p) ACsN corrected frame. The entire time-lapse for (n–p) is shown in Supplementary Video 2. q The noise of a pixel of high dark current is strongly reduced via both approaches after appropriate characterization of the camera including thermal effects. The signal has been offset-corrected by the pixel value of the first frame from the time-lapse.