Fast lenses, and High ISO

Started Jul 18, 2014 | Discussions thread
Austinian Veteran Member • Posts: 7,544
Re: Understanding ISO

mosswings wrote:

Austinian wrote:

Great Bustard wrote:

The most important factors about the sensor that manufacturers do not tell us are:

  • QE (Quantum Efficiency -- the proportion of light falling on the sensor that is recorded)
  • Read Noise (the additional electronic noise added by the sensor and supporting hardware)
  • CFA (Color Filter Array)
  • Microlens Efficiency

Which of these are likely to see significant improvements in the fairly near term (next few years)?

If by significant you mean something that might garner an extra stop or more of performance, it's hard to say. There has been recent work by Panasonic on a different type of CFA that eliminates the losses inherent in current implementations and that theoretically could yield a stop, perhaps more, in sensitivity:

It's been over a year since this announcement, but no updates. One of the problems of this technique may lie in tailoring the color splitting response to produce acceptable color rendition. Given all the complaints about current CFAs, it will need to be no worse.

From what little I've read about various difficulties in formulating appropriate dyes for CFAs, adding more constraints to the CFA dyes' requirements sounds like it might be difficult.

BSI (backside illumination) is the current go-to for increasing sensor QE. Sony calls it EXMOR-R and it is found on no larger than 1" sensors as it involves thinning the sensor wafer down to a few 10s of microns in order to expose the pixel wells from the backside. Doing this for a larger format sensor is still impractical as it dramatically weakens the chip.

Sony's trying to curve their sensor to eliminate corner vignetting and the need for tricky microlens tailoring. It also has the side effect of improving sensor response by straining the sensor lattice, but again it's an expensive technique.

Will this work for a wide variety of DSLR lenses, or will optical issues limit it to fixed-lens cameras, since existing lenses 'expect' a flat sensor?

Read noise can be attacked in several ways, but another way of dealing with the problem may be in redefining the entire imaging process. This is what Eric Fossum has been working on with his Quanta Imaging Sensor. It basic trades the charge-integrating approach of today for a photon-counting approach using a combination of extremely dense binary-response pixel arrays, high frame capture rates, and heavy postprocessing - 100+MP arrays, 1000 frames/sec capture rates using fairly conventional CMOS technology. In doing so one can trade off resolution for DR, tailor tonal response directly, possibly directly compensate for camera motion, etc. etc.. The downside is that it requires pixel read noise to be about 4 stops better, but since it's not trying to do a linear amplification, more options for doing so are open to the designer. Research chips are in development now, but we have a long road to go.

I've read some of Eric's posts in PS&T about this. My uninformed reaction is that such image processing sounds like it may need a pretty healthy PC/GPU setup (at least), but maybe dedicated silicon could do it in-camera.

Thank you, good info.

 Austinian's gear list:Austinian's gear list
Sony a7 II Sony a7R III Rokinon 14mm F2.8 IF ED MC Sony FE 28mm F2 Samyang 135mm F2.0 +3 more
Post (hide subjects) Posted by
(unknown member)
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow