Why ND filters ruin dynamic range on modern high‑ISO sensors – stop using them

I've been shooting the same 24‑70mm f/2.8 on a full‑frame 45‑MP camera for years and finally decided to test the hype around ND filters. I set the ISO to 1600, used a 0.9‑stop ND, and compared the RAW files with and without the filter. The histogram shows a clear compression of the shadows, and the highlight roll‑off is noticeably harsher. Even though the filter claims neutral density, the extra glass introduces micro‑refractions that the sensor's high‑resolution Bayer array amplifies, effectively reducing the usable dynamic range by about 2‑3 EV.

What makes this more interesting is that the in‑camera noise reduction algorithms can't compensate for the extra scatter, so the final JPEG looks flat, and the RAW still shows increased noise in the shadow detail. I've tried brand‑name filters, variable ND sliders, and even the cheap screw‑on types – all of them produce the same loss, just at different magnitudes.

Bottom line: if you're shooting at high ISO for low‑light or street work, ditch the ND. Use a faster lens or adjust your aperture and shutter speed instead. Anyone still swearing by ND filters for creative exposure control should reconsider the trade‑off they're making with image quality.

Thoughts? Have you measured similar dynamic range drops, or do you think this is just anecdotal?