JohnSwanda Loc: San Francisco

Except the high ISO performance of the latest digital cameras is way better than film, probably the area where digital most surpasses film.

amfoto1 Loc: San Jose, Calif. USA

Digital cameras are electronic... and that means electricity flows through their circuitry, including the image sensor.

Image "noise" essentially occurs when the wrong color is recorded by one or more individual pixels on the sensor.

Two things are the primary reasons causing a pixel to operate incorrectly: heat and "cross talk" from adjacent pixels.

Individual pixel size, density of pixels and spacing between pixels on an image sensor all effect how much heat is generated and the degree of cross talk that occurs. The larger the sensor and less crowded it is both can reduce the amount of noise that will be seen in images. For example, a 24MP camera's sensor has 24 million individual pixels. If it's an APS-C (DX in Nikon-speak) camera, the sensor's area will be around 350 square millimeters.... so there will be roughly 68,500 individual pixels per square millimeter! In comparison, the same 24MP in a so-called "full frame" (FX) camera is being captured using a much larger sensor with an area of about 860 square millimeters... or around 28,000 individual pixels per square millimeter. Because it's far less crowded, dissipates heat better and has more space between individual pixels to reduce cross-talk, full frame cameras generally produce less image noise than APS-C. Or, another way of looking at it, the FF camera will be able to take images with the same or less noise at higher ISOs than the smaller format.

All digital cameras have a single, base ISO... 100 in most modern cameras. All the higher ISOs are produced by amplification of that base. The "native range" of the camera are produced by hardware amplification... basically by increasing the voltages traveling across the sensor at the time of exposure. Thus, the higher the ISO, the more heat and chance of cross-talk. Many cameras also have an "expanded range", which are actually shot at a lower ISO and then "pushed" by the camera's firmware... essentially under-exposed images with their brightness adjusted. Doing that further amplifies the appearance of noise.

Many things have already been done to help reduce noise in image and today's digital cameras are MUCH better than those of only 10 or 15 years ago. For example, most earlier digital cameras used CCD sensors (charge-coupled devices). Early on Canon opted to instead use CMOS (complimentary metal oxide semiconductor) sensors. They only used CCD in the several very early "DCS" models produced in collaboration with Kodak (approx. 1995 to 2000) and their very first DSLR built primarily in-house (the 4MP EOS-1D intro'd in 2001). All Canon models from then onward used CMOS, which Canon produced themselves since they were unable to out-source the sensors they wanted.

CCD were widely available to buy "off the shelf" from several sources (Kodak, Panasonic, Sony and others), so all the other camera manufacturers continued to use them for many years. But it was easy to see the difference.... CMOS made for much less image noise and so were usable at higher ISOs. In approx 200-2008 I was shooting a lot of events with a friend who used a pair of Nikon D200.... one of the last Nikon to use CCD. Due to image noise she was reluctant to use ISO 800, let alone anything higher. I was using considerably less expensive (and less "pro-oriented") Canon 30Ds with CMOS and had no problem using ISO 800 or even 1600. When Nikon D300 came available and were using CMOS, my friend was one of the first in line to upgrade! Now, I'm not knocking them because in other ways the D200 was a game-changer... Except for it's high ISO limitations, it set a new and higher standard for "pro-quality APS-C" cameras. In fact with it Nikon sort of invented that category of camera (thankfully forcing Canon and others to come out with similarly more pro-oriented APS-C).

By about 2008-2009 all DSLR manufacturers had switched to CMOS... following Canon's lead and success. All of today's full frame and APS-C are using that type of sensor. In fact, many point-n-shoots using smaller sensors have also shifted from CCD to CMOS for the same reasons (some still use CCD, though... as do many or most camera phones). Among the longest CCD hold-outs were so-called medium format digital. I would image that considerably smaller market has limited sources for sensors. And most MF users are pretty advanced pros who were more likely to be shooting off tripods and/or using lighting gear that make higher ISOs less necessary. Plus the much larger sensor size and less magnification needed to make prints from images both also helped offset noise issues. But in the last couple years pretty much all MF digital have also switched over to CMOS.

In addition, in-camera tweaks to both hardware and firmware have helped rein in noise. Each generation of cameras improved on it a little or a lot. When I was shooting with 6MP 10D around 2004, I tried to keep to 800. With the two generation later 30Ds, I'd use 1600 without too much concern. 50Ds that were my next upgrade weren't higher ISO capable, but that was still good considering they managed that while having nearly double the resolution (15MP versus 8MP). I bought a full frame 5D Mark II to use alongside the 50Ds, in large part for it's higher ISO capabilities, thanks to it's much bigger sensor... up to 6400 (even 12800 occasionally). Next were a pair of 7D that I used to 3200. Now I'm using a pair of 7D Mark II that can easily do ISO 6400, sometimes even 8000, 12800 and 16000!

Post-processing software has come a long, long way handling it, too. Years ago I wasn't happy with how Adobe software handled noise from the highest ISOs in my cmaeras (Photoshop only initially, then Lightroom too when it came available). I used to use Canon's own DPP for especially high ISO images. Adobe finally got it right in Lightroom 3 and Photoshop CS5.... so I was able to use them instead. Now I use an Imagenomic Noiseware Photoshop plug-in with my highest ISO images. I've experimented a little with Nik DFine, too... but still prefer Noiseware. It works well and offers a lot of user control over how heavily noise reduction is applied.

Will there ever be entirely noise free images? I doubt it! At the same time that it's been largely eliminated at a certain ISO, the cameras offer even higher ISOs. Some now have ISOs in the hundred thousands and even millions... though they may not be very usable... yet.

All the above is very simplified.... but hopefully answers your questions. Digital today is way, way beyond what was ever possible back in the days of film! Keep in mind that film had grain, different than digital noise, but still in its own way degrading images at higher ISO (actually ASA... tho it's the same thing). When I shot slides, the fastest film I'd use was ISO/ASA 200. Much of what I used was only 50 or 100. If I needed any higher color images, I'd switch to 400 speed print film. I tried to avoid ever using any higher than that. Black & white film was similar. Years ago I used ASA 25, then ASA 32 when I wanted to make big enlargements. ASA 400 (Tri-X and similar) was usable... and could even be pushed to 800... only because the grain doesn't look bad in B&W. Being able to shoot usable digital images at ISO 3200 and higher has allowed me to do things that were never possible with film!

burkphoto Loc: High Point, NC

Given our current limited understanding of physics, I can see your point. But time and again, throughout history, we have learned new things that took us in new directions. By quantum leap, I was referring to the discovery of a new principle, or the development of a completely new paradigm. It's happened several times in the history of digital photography. The engineers are pretty clever. And you can never discount the effects of folks outside the current photographic equipment community.

Thankfully, most people don't need clean ISOs above 25,600. But they might be nice to have...

bclaff Loc: Sherborn, MA (18mi SW of Boston)

Pixels don't record color. They record (in a linear fashion) the amount of light that gets through a color filter.
Color is calculated later in a process called demosaicing.

Inaccuracy (noise) in one pixel can affect the demosaicing and result in a less accurate hue (color).

bclaff Loc: Sherborn, MA (18mi SW of Boston)

Your faith in future technology conflicts with the laws of physics, not just our current understanding.
Ultimately, even if read noise is brought down to zero; we are limited by photon noise, and we're not all that far from that limit now.
The "ideal" lines on the PhotonsToPhotos Photographic Dynamic Range (PDR) chart are drawn at the limit of photon noise.

In the near term the most improvement is in computational imaging which is currently being implemented in mobile phone cameras.

tdekany Loc: Oregon

Neither did your post did it? If the op looked at some test shots from the a7sII, his opinion may change.