Here we go again!! New sensor technology!!

[antiquark]

Interesting conversation. Speaking of quantum efficiency, does anyone know what ISO rating would correspond to an efficiency of 100%? I.e., what is the highest possible ISO? Would the Nikon D3s already be at that point?

 

[Benjamin Marks]

I am waiting for the 8-track version.

Ben Marks

p.s. pretty cool image.

 

[semilog]

Interesting conversation. Speaking of quantum efficiency, does anyone know what ISO rating would correspond to an efficiency of 100%? I.e., what is the highest possible ISO? Would the Nikon D3s already be at that point?

QE is really important but not the whole game. One can have a sensor with very high quantum efficiency but very low performance due to noise.

There are (more or less) three sources of noise:

1. Shot noise.
This is "counting noise" (think of tossing small numbers of marbles into a cup a few feet away -- there is a random element to how many marbles end up in the cup). This noise is equal to the reciprocal square root of the number of events detected. The only way to reduce shot noise is to count more events: higher QE, faster lenses.

2. Dark current.
Some sensors are better than others. Colder sensors are better than warm ones. Dark current accumulates as a function of exposure time. This is the main reason why astronomers and microscopists like to cool sensors to -40° or colder. Improvements continue to be made here. It is likely that one of the most under-rated reasons for improvement in high-ISO performance is improved thermal management. (It would be really interesting to see tests comparing how different cameras perform at lower and higher ambient temperatures.)

3. Read noise.
Note that until DA conversion, CCDs and CMOS sensors are analog devices! Read noise is the noise penalty incurred in shifting accumulated charges off of the CCD or CMOS sensor, amplification of the signals, and digital to analog conversion. Great strides have been made here and are continuing to be made. Thermal management also can make a difference.

My guess is that the D3/700 cameras have decent but not yet spectacular QE (they still use front-illuminated sensors, after all). Sony's microlens tech was industry-leading for many years, and Sony is supplying Nikon's high-end sensors. This helps a lot with QE.

I expect that the Nikons do have superbly low read noise (Sony has excelled at making chips with low read noise), good thermal management, and truly outstanding post-capture signal processing.

Yes, there is still room for significant improvement. Just not as much room as some people think.

The improvements that I find most exciting are in *price*! Look at the performance of the (Sony) sensor in the Pentax Kx. That's a relatively inexpensive DSLR, and its performance is for any practical purpose as good as any APS-C DSLR ever released.

 

[ernstk]

Tiny sensors that will fit into a mobile phone or similar sized point and shoot capable of delivering images on a par with full frame digital SLR's kind of seems inevitable to me!

We seem to enjoy mocking each stage of digital development as we no doubt mocked it's initial introduction ... this was the first digital camera:

Do you think we've come very far?

Do you think development is going to stop because we think it's come far enough?

Not bloody likely! 😀

Cool picture. I notice the Digital Equipment Corporation (DEC) logo on the cassette. I worked for these guys for 10 years in the 80's and 90's but don't recall ever seeing this...

Ernst

 

[antiquark]

There are (more or less) three sources of noise:

Thanks for the lucid explanation!

 

[percepts]

OMG.
It'll be at least 3 years before canon and nikon are using this technolgy and 5 before Leica get there arse into gear. Do you think I should sell my current kit now just to avoid redundancy.

 

[Pickett Wilson]

"There are (more or less) three sources of noise:"

Small children are the most annoying.

 

[semilog]

You folks have reminded me. There's a fourth major source of noise: John Cage.

 

[percepts]

I forgot we do have the "great green bush cricket". I discovered one of these in my garden about 10 years ago. Gave me a big shock cos I didn't even know such things existed in the UK. I thought it was some monster locust blown over from Europe till I caught it and looked up the identification for it.

 

[Finder]

You folks have reminded me. There's a fourth major source of noise: John Cage.

I have 4'33" on CD. That has very little noise.

 

[semilog]

I have 4'33" on CD. That has very little noise.

But not a lot of signal, either.

By the way, you haven't come across the sheet music for a four-hands transcription of the piece, have you?

 

[IgalSc]

OMG.
It'll be at least 3 years before canon and nikon are using this technolgy and 5 before Leica get there arse into gear. Do you think I should sell my current kit now just to avoid redundancy.

I think you should sell it now and put on deposit with highest interest available to have the Leica with this technology in, let's say, 5 years 🙂

"There are (more or less) three sources of noise:"

Small children are the most annoying.

disagree. maybe most noisemakers, but not annoying

 

[JPresley]

My first K-x test shot at iso 12,800 taken on St. Denis street in Montreal at night. It was underexposed by a stop judging by the histogram.

 

[JPresley]

I'm sure the D3 and the quantum film camera are better. But not bad for $500.

 

[semilog]

My first K-x test shot at iso 12,800 taken on St. Denis street in Montreal at night. It was underexposed by a stop judging by the histogram.

This raises another point. Most CCDs are have better intrinsic sensitivity in the blue-green than in the red. This effect is further accentuated by the Bayer mask which gives you twice as many green photosites as blue or red. Reddish illumination or subjects generally look worse at high ISO than blue or -- especially -- green. These effects are clearly visible in the example shown. Red sensitivity is a major area where current sensors can be improved.

As you say, this is just astonishingly good for a camera that will (almost certainly) be available used in ex+ condition for $300 within a year from now.

 

[JPresley]

Sensitivity might be good enough now (APS-C and bigger) that other improvements might make a bigger difference for amateur use.

Abrupt saturation/clipping in the highlights looks like another area where digital could improve. Imagine a chip that could not be overexposed. I guess the ultimate imaginary sensor would be one in which each pixel counted photons as they arrived, and didn't max out unless you pointed the camera straight at the sun.