Dear community, we have been experiencing patterns in the FSC curves where it drops sharply in medium- to high-resolution range (see attached) during benchmark sessions with apoferritin. Global CTF refinement indicated large beam tilt ~1 mrad, but the FSC curves become normal-looking afterwards. We were careful with the alignment and did all the CTF-based calibrations under SerialEM. TFS rechecked and calibrated beam tilt but the same pattern persists. Any idea what may have led to such? Many thanks!
How far off is your pixel calibration? Beam tilt and incorrect pixel size can both manifest in the same oscillations; check your reconstruction against something like EMD-11638. 
1 Like
It seems like you were using a wrong pixel size. Please double check the pixel size.
Thank you so much for your replies. unfortunately, it’s not something simple as a wrong pixel size. we calibrated the pixel size and it was pretty much spot on. we did 3 back-to-back benchmarks, and the beam tilt got progressively worse, with the latest one exceeding 2-3 mrad.
Acknowledged, thanks for responding. What was the timeframe for the benchmarking? Separate runs? Do you have any heavy equipment nearby? Any building works? No unshielded or field-swept magnets anywhere nearby?
Thanks again. Below are the beam tilt values for the 3 benchmark data collections lately. The first two with the same sample grid using apoferritin from ThermoFisher. In each case, 300-400k particles were used in the final reconstruction:
-
On 2/23/2026, Tilt (mrad): [-0.46252727 0.5635262]
-
On 3/6/2026, Tilt (mrad): [-0.82926129 1.36199]
-
On 3/10/2026, Tilt (mrad): [-1.88156933 3.55318185]
FSC curves before the CTF refinement from these benchmarks attached. The reported beam tilt values from the Global CTF refinement got progressively worse over the 3 week period. In all 3 cases, CTF refinement was able to “correct” the abnormal beam tilts, and the final reconstruction reached about the same reported resolution of 1.8-1.9 angstroms with smooth/normal-looking FSC curves. The pixel size used for processing was correct based on calibration.
Our site was established back in 2017. There has been some constructions nearby about 500 meters away (large new building being built) but that has been going on for well over a year and this happened rather lately.
