AFAIK the fastest autofocus lenses for any 44x33mm sensor are f/1.7; the fastest autofocus lenses for FF (36x24mm) are f/1.2, i.e., one stop faster. Depending on the aspect ratio for final use, those are essentially even or slight advantage to FF. But with FF there are tons of autofocus f/1.4 lenses, and for 44x33mm there are far fewer f/2 options. So overall advantage to FF.
I agree in the principle point, that you get better low light performance with a current 35 mm format system than with a crop-medium-format-system. But the math of mixing sensor sizes and F-stops together does not work that rigorously for let's call it super fast lenses. F1.4 or F1.2 can be considered super fast lenses, but what does that means in practice?
First if you look at the bayonet of a Sony E-mount camera with full frame sensor, you notice that the edges of the sensor are shaded by the bayonet ring for perpendicular light. So there can be lenses made to be super fast, but only in the image centre. At the borders and the edges they are not F1.2, as there is simply no geometrical room for light to propagate to the sensor at F1.2. This is not a problem from the perspective of light sensitivity or vignetting (you can correct that digital), but from the perspective of bokeh. You get much "more"* bokeh in the centre and the bokeh get's less to the edges. *=larger confusion circles for point sources.
I think that misuses "bokeh", which I understand to mean the
quality of the out-of-focus areas. Do you mean to say that the depth of field changes? I think that's what you mean because you're discussing circles of confusion. So a couple of questions:
(1) Away from the center, are these lenses not f/1.2, or are they in fact f/1.2 but not T1.2? IOW, is there a difference in this regard between illumination and depth of field?
(2) Considering your "in practice" standard, are there any regular and current very fast lenses that don't vignette substantially wide open? You mention Sony f/1.2 lenses; are the Fuji GF f/1.7 lenses, the Hasselblad f/1.9 lenses, Nikon f/1.2 lenses, and/or Canon f/1.2 lenses substantially darker in the corners than in the center? In other words, in practice, to what extent is this a Sony issue, versus an issue with all current very fast lenses?
Another problem is that the common first-curtain-electronic shutter truncates your bokeh (and decreases light sensitivity).
Controversial opinion here, but I really see EFCS as nothing more or better than a workaround for certain shutter-shock situations. Otherwise, to me the choice is between full mechanical shutter and full electronic shutter. Yes, I realize there are a few cameras that don't have full mechanical shutters;
to me that's a substantial omission. Again, not asking others to agree, I think I'm reasonably aware of the pros and cons, and each of us is entitled to prefer different tradeoffs.
You must check, whether you can shoot a super fast lenses with mechanical shutter or not.
Other than flash X-sync limits versus 'high speed sync', and running out of mechanical shutter speed in full 'Sunny 16' conditions, where e.g. you might hit a wall e.g. at ISO 100, 1/8000 s, and f/1.8, is there any other problem?
And I tried, whether you have F0.8 or F1.0 is barely noticeable in the centre. At the edges it's the same.
How about we say that half a stop is rarely if ever more than a fairly subtle difference?
If we leave the light sensitivity and look at bokeh: If we talk about depth of field, going the fast lenses route is in my opinion a dead route. You can do that, but your'e constrained to persons facing the camera in order to get both eyes at high resolution.
Don't get me wrong: I think the current popularity of ultra-fast lenses is excessive to bordering on a bit silly.
For my taste they're misused more often than they're used well. For my uses and taste, a headshot is best at about the equivalent to FF 135mm and f/5.6 or so. I want both eyes very sharp and I don't want mushy ears. But for shooting full-length shots and more distance and printing / viewing them at moderate sizes, wider apertures can work well.
Also, to some extent output use / size matters. Depth of field that's too thin when printed 16x20" or even 8x10" might look good in a half-page magazine photo, for those who still remember magazines.
