Fast sensors making fast lenses obsolete?

Started Jul 29, 2013 | Discussions thread
EEmu
Forum MemberPosts: 69
Like?
Not particularly
In reply to amalric, Jul 29, 2013

The 16Mpx sensor most certainly did not improve sensitivity by 2 stops, that was dynamic range. It got slightly more sensitive (53% vs 41%) and increased saturation levels a decent amount (40%) but what it really improved was read noise. (Note that DR is basically just saturation vs read noise, while high ISO performance is more complicated because it involves the read noise and shot noise.)

The E-M5 improved in high ISOs because it cut the E-P3's particularly bad read noise by a factor of 4. This directly improved the dynamic range by two stops, but the effect on noise was a lot more subtle because you need to account for the shot noise.

If we were to list out the SNR vs photons acquired for some cameras at ISO1600 starting from 2000 photons (near saturation) by 2 stop reductions (so 2000, 500, 125, 31, 8, 2):

E-P3: 43, 20, 7.9, 2.5, 0.7, 0.2
E-M5: 45, 22, 11, 5.0, 2.1, 0.7
Ideal: 45, 22, 11, 5.6, 2.8, 1.4

You'll note that the E-M5 out performs the E-P3 noticeably around 125 photons captured (-4 stops) and is about a good -6 stops as the E-P3 is somewhere around -5 stops. That's where the ~1 stop of high ISO performance comes from.  On the other hand, the Ideal (0 read noise) camera is only a stop better somewhere in the single digits of photons where the signal is worthless regardless of the noise level. The DR of the Ideal sensor would be fantastic though ;-).

In short, the E-M5 brought sensor noise levels down to 'close enough'. Even perfection isn't going to add another stop.

So from here, it's increases in sensitivity:

Quantum efficiency: about 50%, so 1 stop improvement is possible
Bayer filters: about 30-70%, so 1-2 stops improvement depending on color
Micro lenses: about 90%, so ~1/4 stop possible

Maybe I overlooked something, but without a fancy quantum mechanical breakthrough there's not much more than 3 stops of room for improvement. In practical terms, probably more like 2. Interestingly enough, between Panasonic's light splitters and the organic detectors we might actually see those 2 within the next year or so. Exciting.

Still, I am not seeing the room for improvement to eliminate fast lenses for low light shooting. Is there enough that an f/2.8 will perform as a f/1.4 now? Yes. However, that's only good enough if f/1.4 is good enough now. Personally, I've been in plenty of situations when it's not.  Anyways, even if you can do adequate shutter speeds at acceptable ISO, wouldn't you rather have good shutter speed and/or good ISO? (Provided you don't need deeper DoF, or course.)

Further, this 'good enough' bit is rather the antithesis of an ILC camera.  Why shoot a prime? A fixed super zoom will provide good enough quality and enough range.  Why shoot m43?  If it's good enough today, then a smaller sensor in a couple years will be too.  Perhaps the camera of the future is an improved FZ200...

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Huh.New
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow