Does Moore's law apply to digital photography?

[JohanCruyff]

https://en.wikipedia.org/wiki/Moore%27s_law
As far as I know, (computer) hardware geeks are upset by the delays of the releases from Intel and by the weak competition by AMD: the latest news seem to begin to contradict the holy Moore's law.
I would like to know whether an electronic-statistic-photography expert has ever tried to check whether this law applies to photography too (with respect to frames per seconds in pictures and videos, usable ISOs, resolution etc.).
I would also like to know whether the conjunction "whether" can be used twice in a sentence or not, but that's a bit off-topic. :'(

 

[hendrik-sg]

Of course not.....

There is no single physical law, of how powerful a computer can be. Reaching the limit of one technology, developping a better one is the way to go. BUT as all exponential functions, Moore's law can not be valid for ever, but nobody would have thought how long it would last, say 50 years ago.

Photograpy ist completely different. At very best, a sensor can collect ALL the photons coming,and measure their color simultanously. so sensors can converge to this limit, and moore's law could be modified and it could be that the losses will be overcome at a constant rate, means 32% 16% 8% 4% 2% 1%...... loss. means efficiency will be 68% 84% 92% 96% 98% 99%, so the rate of imoprovement slows down quickly.

Lenses are similar in general, but there are more degrees of freedom and all designs are compromised in many ways, so that all possible aberations are as small as possible and well balanced. They can be manufactured more precisely, manufactured cheaper (so a more complex design fits into the budget), they can be optimized with more computer power, but i am sure, all the easy designs are on market already and can be optimised only in detail

And maybe most important: for most of us, photographic skills are tightly limited, as is the potential to improve skills. Without talent and lots of effort, we will never become good fotographers, so most of us will continue to produce 1000's of bad shots, even with the best equipment (this appies specially to me, so please nobody be upset)

 

[Hillsilly]

Per Google- The observation made in 1965 by Gordon Moore, co-founder of Intel, that the number of transistors per square inch on integrated circuits had doubled every year since the integrated circuit was invented. Moore predicted that this trend would continue for the foreseeable future.

How can it be a "law" when it is so vague? How long is a forseeable future? Two years? Five years? Forty Years?

Re whethers, I'm cool with that. There are no laws in English (at least for 99.99% of us).

I don't know much about cameras, but I'd say that "useable" (ie generally accepted high quaility) ISO and resolution seems to double every three years. This will continue for the foreseeable future. The Phase One wikipedia site gives a good history of their sensor resolutions and progression over time.

But is this a good thing?? No! I want a sensor that is awesome at ISO 100, not one that performs "OK" to ISO 6400+. Bring back the CCD!

 

[kubelik]

Johan, unless you were being tongue-in-cheek, you're very much mistaken in calling it "the holy Moore's law." Moore was simply basing it on observational data and very clear that it was speculative when used to project into the future.

Moore's law has only ever been an intellectual curiosity. it doesn't take into account any of the practical factors associated with anything, it purely states that from a manufacturing perspective, we seem to be able to double the amount of transistors on a reference sized circuit board every couple of years (not every one year as is sometimes misstated).

this doesn't account for heat dissipation, energy consumption, limits to data pipelines, material properties, cost of R&D, cost of manufacturing, stability and other performance characteristics of the circuits... basically, any of the things that actually matter in any product. it's an amusing observation that has in some ways perhaps become a self-fulfilling prophecy - you can see how Intel attempted to structure their "tick-tock" release cycle to match Moore's law, but how in only a few cycles they've already hit a wall with reality and are morphing to a "tick-tick-tock". where we go from here is where things really start to get interesting.

 

[Antono Refa]

I think you're making a few mistakes interpreting Moore's law.

Moore's law is about how many transistors can be fit per area. This easily translates to memory size, but not necessarily memory speed, which is one factor that factors into the frames-per-second.

AFAIK, there is no connection between usable ISO and number of transistors per area. As noted above, its a matter of pixel efficiency.

I suppose there is a direct connection between transistors per area and resolution, lens resolving power is a different story, which was discussed on this forum in the past.

 

[scyrene]

It does tangentially. Of course in-camera processors are faster, but more importantly, computers have benefitted from Moore's Law, and this has opened up a lot of interesting avenues for photography. Aside from better noise reduction and quicker/finer postprocessing, things like image stacking in astrophotography (for lower noise or for better sharpness), focus stacking in macro work, panorama stitching, HDR, etc. have enabled us to get around optical or other physical limitations. With a standard laptop, you can produce astro images (especially of Solar System objects) as good as professional astronomers' pictures twenty years ago (with a good telescope, patience, and knowhow); macro photographs that don't have to trade off sharpness against depth of field; images with more dynamic range than any sensor could produce; massive panoramas with more resolution than any sensor has (and subsets of these, such as Brenizer shots that simulate lenses with apertures far wider than any that physically exist).

 

[kaihp]

Moore's law is about how many transistors can be fit per area.

Allow me to pick nits here: Moore's original observation was that the number of transistors on a single die seemed to double every 12 months (which was later changed to 18 months). So the observation took into account the fact that yield was going up, which in turn made it economically viable to manufacture larger dies (chips).

The number of transistors on a single die are driven by (A) the geometry used and (B) how big you can economically allow your die to be for the particular product.

As geometry to a large part also drives the capacitances involved, this also drives the speed and power consumption of the devices. There are a number of important second-order effects I'm going to leave out of this (leakage in particular).

Moore's Law has been actively used to predict and drive the growth in the (digital) semiconductor industry.

So how does this apply to digital photography: Memory capacity, memory speed and processing speed - all for a given power consumption budget.

I would like to know whether an electronic-statistic-photography expert has ever tried to check whether this law applies to photography too (with respect to frames per seconds in pictures and videos, usable ISOs, resolution etc.).

Frames per second is limited by how fast you can read out the pixels on the sensor (and readout speed impacts noise), but also on mechanical things like how fast the mirror and the shutter can be safely operated.

Video: here Moore's Law helps us by having a faster interface to the (flash) storage, but also that more advanced processing can help us to get visually better looking movies for the same amount of data/second. As an example of use of more processing power, the new H.265 standard codec for 4K video claims to deliver similar visual performance at the same bandwidth, as a H.264 codec requires for a Full-HD (1/4th the pixels) stream.

 

[neonlight]

The issue I see is that the biggest problem for CMOS sensors is noise. Pixels need to be larger, in general, to reduce noise, which is moving in the opposite direction. Pixels come in various flavours with different numbers of transistors (the basic pixel uses three) and making these smaller might leave more room for the photodetector diode, but they will contribute increasing leakage and noise (again)...
Moore's law might give us the means to process photos in the camera ... as mentioned above, focus & image stacking, HDR etc. instead of having to use the computer?