Does anybody know what's next for sensor technology?

  • Thread starter Thread starter Guest
  • Start date Start date
G

Guest

Guest
What's in the works for sensor technology?

Dynamic Range
Current sensors are getting towards 14stops Dynamic Range. Cool. Getting closer.
Currently employed tricks to add DR:
-split beam (two lenses, two cameras, special rig)
-second scan (extra scan between each frame, but can cause motion artifacts)
-3 sensor w/ prism. (one lens, one camera, 3 sensors)

I have no idea what the Holy Grail of DR is, but I can tell you I'd rather have 21 than 12. 18 is fine. 12 is a bit limiting at times.

My question about the new tech, is it cheaper to keep pushing the sensor technology, or would it be faster and cheaper to design a sensor with dual imaging built in.
Like filling in the gaps around light wells with smaller light wells for capturing highlight detail:

HDRsensorpattern01 by onesielaff, on Flickr
or layered sensors:

HDRlayeredsensor by onesielaff, on Flickr
 
Color reproduction:

I wish for strides in this area, but display technology still hasn't caught up. So I do not even know where to begin.

Also, much of this happens outside the sensor in the processing of the captured data, but I think one area I would like to see strides is color.


It would be interesting, since many cameras out perform their displayed medium (tv, monitor, projector, web), and many output mediums are highly compressed, to have a camera which store more information in regard to skin tone, which could then be used in post to make the best skin tones. Also, a render engine or codec which gave more bandwidth to that part of the color spectrum. So, even if a video is highly compressed, skin tones have an extra portion of the bit rate, and never look as bad as they do now. It would be nice to have something which gave emphasis to skin tone.

Skintone is so important, because we are most distinguishing of the human form.
 
Last edited:
SENSITIVITY:

With cameras like the FS100, F3 and C300, low light shooting has become more beautiful than ever.

While I would love to be able to shoot clean and accurate colors at 128,000 iso, I think about the trade offs.

Sensitive sensors need to be filtered under daylight shooting. But too much ND, and there is image degradation.

The ideal with current technology, is to have 2 cameras. One for daylight, which shoots at crazy low ISO, maybe even iso20, and another for low light.

Even the DVX's had to be ND'ed outdoors. So it would be interesting to see a camera which excelled under bright conditions.

Is it possible for a sensor to have two fields of images on the one chip, like the image I posted above? That way, it could switch photo sites depending on the situation.
 
Last edited:
We already know that Sony has made a sensor with the photosites extremely close to each other.


And we've seen Arri and Red combine two images realtime to make higher DR images, so theoretically a sensor like this...

...could be made to do higher spans of light capture, as well as double as a daylight/lowlight combo sensor.
 
Looks like you've got this ground covered, good summaries of what's to come. I'm beginning to wonder what this means as we begin to surpass to film's DR, as far as lighting is concerned. Would silhouette techniques be harder to pull off due to the ambient light spillage that these sensors could pick up? I hear a lot about low-light capabilities making people lazy - which I believe is true - but could it also make things harder to keep hidden?
 
I suppose it could. It would require more flagging and negative fill, and failing that, a much brighter backlight.
 
Looks like the sensor I described was already invented and implemented in a point and shoot Fuji camera. I hardly felt like no one ever thought of it. It is a basic concept.

http://www.dpreview.com/news/2003/1/21/fujisuperccdsr

Thanks to Vladimir at sensor world blog.

I do not know if this is a good thing, or a whether it didn't help and they jus scrapped it.
 
David Jimerson said:
I suppose it could. It would require more flagging and negative fill, and failing that, a much brighter backlight.
Click to expand...

A highdynamic range camera would require a much richer capture format so that post manipulation of the footage is better. One advantage of having two sensors in one, is that you would have the option of using a single image with lower native DR. WHy would that matter? Well if the data coming off the chip is less than 12 bit native (see Canon explanation of C300 8bit process), then trying to add a lot of contrast could degrade the image.
 
Reading the pros and cons about the old ps camera:

http://www.dpreview.com/reviews/FujifilmF700/page15.asp

"SuperCCD SR doesn't live up to its promise (the improvement is not significant)" - DPReview

and more about the results:
http://www.dpreview.com/reviews/FujifilmF700/page10.asp


This is all consumer technology in a very cheap point and shoot consumer camera, so I wouldn't expect much from the performance, but it also makes me raise the question, are the smaller photodiodes that much less sensitive, and would the signal be clean enough if they were? Surely that could be designed using much of existing technology. Especially if a ps camera in 2003 has it (albeit implemented poorly).
 
Last edited:
Here is another HDR split pixel design used for the automotive industry. 10Db dynamic range.

 
A full frame 3.5k bayer pattern chip would have extremely large pixels. Twice the size, maybe even 4x the size of the current F3, and deliver a crispy 1080p. You would get something like 0dB, being equivalent to 3200 iso, with 18db being like 25,000ISO but with noise performance like the F3 and c300 at 12dB. In other words, insane performance, even beyond what we have now.

On the other hand , why? What would you need that for. I would prefer a 4k camera with 14stops DR, @ ISO100
 
James0b57 said:
A highdynamic range camera would require a much richer capture format so that post manipulation of the footage is better. One advantage of having two sensors in one, is that you would have the option of using a single image with lower native DR. WHy would that matter? Well if the data coming off the chip is less than 12 bit native (see Canon explanation of C300 8bit process), then trying to add a lot of contrast could degrade the image.
Click to expand...

Sure; just entertaining the thought as stated. A lot of this doesn't matter if the recording format can't show it.
 
TimurCivan said:
A full frame 3.5k bayer pattern chip would have extremely large pixels. Twice the size, maybe even 4x the size of the current F3, and deliver a crispy 1080p. You would get something like 0dB, being equivalent to 3200 iso, with 18db being like 25,000ISO but with noise performance like the F3 and c300 at 12dB. In other words, insane performance, even beyond what we have now.
Click to expand...
I like the sound of that!
 
For a momnet there I thought James was talking to himself :) but yeah I want more DR. I dont care much about color .......I can cc and grade that.
 
You must log in or register to reply here.
Back
Top