It is possible to get equivalent pictures with FX and DX cameras:
- the two cameras must be shot from the same point (same perspective)
- the field of view of the lenses must be identical, for example 35mm on DX and 50mm on FX
- the entrance pupil of the two lenses must be identical, for example shoot the 35mm at f/2.8 and 50mm at f/4 (35/2.8 = 50/4 = 12.5mm)
In other words when comparing DX and FX, the DX lens must have a focal length one "stop" shorter, and the aperture must be one stop faster to compensate. This is only an approximation, one stop is sqrt(2) or about 1.4, while the difference between DX and FX is about 1.5x linear, but it's close enough for most practical purposes.
This will give two images with the same framing and same relationship between the focus plane and out of focus foreground/background (DOF, bokeh etc) This is because the two cameras are viewing the world through the same "window" - the same entrance pupil from the same position. The only difference is how the view from the window is projected onto the sensor - in one case it is projected further away over a larger area (longer focal length on bigger FX sensor), in the other case projected closer over a smaller area (shorter focal length on a DX sensor). Notice also that when the image is projected over a smaller area, the light intensity is higher, in other words the f/stop is faster.
This assumes the two lenses have similar rendering, depending on how the lens is corrected the DOF and bokeh may differ greatly even for otherwise similar focal lengths and apertures (see the thread comparing the ADS 85/1.8 with the Tamron 85/1.
. This comparison also ignores differences at pixel level. If both sensors have the same pixel pitch the smaller sensor will obviously have fewer pixels, it's basically a crop of the larger sensor, which puts greater limits on how much the image can be enlarged. If both sensors have the same number of pixels, the smaller sensor obviously has smaller pixels, which means each pixel has less light-gathering capacity, may have less dynamic range etc