Seeing, mount jitter, the aperture's own diffraction, and the pixel each impose their own MTF. The real system is their product. Read off which term dominates, get the Gaussian-equivalent FWHM, set your sampling in ″/px (the number your gear gives you; FWHM/1.6 is the target), and watch what deconvolution does to the noise on the true system curve. Diffraction fixed at λ = 550 nm.
Same frequency axis (cycles per arcsec) so you can read them off against each other. The bold white curve is the MTF that actually exists — everything below the seeing/mount rolloff is gone before diffraction ever matters. Fit all shows the full band out to the diffraction cutoff; zoom focuses on the combined rolloff.
Amber is the gain deconvolution must apply to undo the real system (1/MTF) — the noise amplification at each frequency, up to your Nyquist.
A dim background at your target SNR, blurred by the real PSF and sampled, then inverse-filtered to Nyquist. Same demonstration as before, now on the true system.