Hi @fxminato,
Thanks for the question. The reason for this discrepancy is that GPU-accelerated Simulate Data job (available only in v4.4 and later) and the Legacy Simulate Data job have different ways of calculating noise from signal power and SNR. The Legacy Simulate Data job defines the SNR as the ratio of the standard deviations of the signal (particle slices) to the gaussian white noise. The GPU-accelerated Simulate Data job instead defines SNR as the ratio of the variances of these two quantities. Since the latter definition is more common, the new GPU accelerated job was chosen to have this definition. The Legacy job’s SNR definition was kept unchanged so as to not change the behaviour of the job as it may be used in current workflows.
Note that the tile images only display the lowpass filtered images, which also would cause images to appear higher SNR than they actually are, since the signal power for most cryo-EM volumes is stronger at low frequencies.
I hope this clarifies the discrepancy!
Michael