8
\$\begingroup\$

So I just watched a review on the Sony NEX-7, which is 24MP in APS-C size.... I am a Canon user and I immediately think about the announced Canon Powershot G1X with almost APS-C sensor but at a lower resolution.

One question suddenly springs to mind :

Why has there not been a full-frame but low resolution sensor?

I have no knowledge of imaging sensors manufacturing, but I asked myself, "Would it make sense to create a cheap full frame sensor that is very low resolution?"

I figured it somewhat would make sense. For starter, great ISO performance, secondly, greater control of DOF.

These are not available to consumers without spending a significant amount into top-level photographic gears.

For example, if I create a full frame sensor that is 10MP in resolution, would it be cheap to manufacture? If not, what are the reasons that make full frame sensors so expensive? Would it still be expensive to manufacture such sensor if I make its resolution very low, like 10MP or even 8MP etc?

I know it is a theoretical question, but if Canon can offer a Powershot compact camera with full-frame sensor at 8MP for below $1000 (USD), I would definitely buy it!

\$\endgroup\$
1
  • \$\begingroup\$ An earlier question of mine (on ISO vs post processing) provoked a description of sensor chip design. The new Sony's have internal amplifiers that eliminate a source of noise, and the same reasoning suggests that there is now no downside to batching the pixels to acheive the same result. Your low-light high-iso image can combine (add, not average) adjecent pixels and not have extra per-cell noise floor that would be different from having larger cells. \$\endgroup\$
    – JDługosz
    Sep 1, 2015 at 5:18

7 Answers 7

13
\$\begingroup\$

As mentioned by the answers by @matt and @rowland, the price is directly linked to the the area of silicon used to create the sensor. Ideally a sensor with twice the area should cost about twice as much. Since all production of electronics on silicon (and other substrates) will have flaws, not all the produced chips/sensors will work. The yield rate (as it's called) is lower when the sensor is bigger, using the same production process.

Imagine a sensor A that is twice as large in both directions compared to another sensor B. That means you can make 4 times as many of sensor B in the same area of sensor A. But if you have one flaw in that area, you're still left with 3 usable sensor Bs. If you were producing sensor A you'd have to scrap that sensor. This means that the yield rate is much greater for smaller sensors, which adds to the price differences.

The smaller the chip/sensor the less area and higher yield rate, which means a much lower price.

\$\endgroup\$
11
\$\begingroup\$

Sensor price is more proportional to the physical size of the sensor, than the number of pixels within it. There are full frame sensor with lower pixel counts on some of the older models (for instance, the first Canon 1Ds). It's worth noticing that the sensitivity is lower than modern sensors - not because the pixels are larger, but due to other advances.

There may be scope to make larger pixels, but it wouldn't necessarily be any cheaper.

\$\endgroup\$
7
\$\begingroup\$

Look at Nikon D3X vs D3S. They both have the same sensor size but the D3X has twice the resolution (25 MP vs 12 MP). The cameras are otherwise almost identical but the lowest resolution one is $5200 USD vs $8000 USD for the higher-resolution one.

The 25 MP sensor requires finer circuitry and therefore will have lower yields. At the same time there is a market for both because the D3S can produce images that are much cleaner but it wont print them as large. Its standard ISO range reaches 12800 (with boost to 102400) while the D3X has a standard range that maxes out at 1600 (with boost to 6400).

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Yeah, the D3s is amazing. It practically sees in the dark, yet the resolution and noise are still substantially better than what you could scan film at. For most "normal" films, if you scan them to get the 12 MPix of the D3s, you will see a lot of grain noise. \$\endgroup\$ Aug 31, 2015 at 20:26
1
\$\begingroup\$

Just to add one more interesting bit: once upon a time there actually was a relatively low resolution full-frame sensor. The Contax N was a six megapixel, full-frame design.

Unfortunately, despite the low resolution its low-light performance was pretty lousy (even compared to other cameras of the time). They seem to have more or less given up on getting it to autofocus well, and built in a focus-bracketing mode instead. Being a Contax, it was also quite expensive.

On the plus side, at ISO 100 or below it had probably the best 6 megapixel sensor anybody ever built, and the Zeiss lenses really are extremely good.

Bottom line: Contax dropped the N after less than a year on the market. Shortly after that, Contax dropped out of the market entirely.

\$\endgroup\$
0
\$\begingroup\$

This is basically Moore's Law at work. Manufacturing technology for sensors follows the same basic rule as for any other chip: over time, the number of elements that can be put on a chip doubles for a roughly-similar cost. It may be slightly cheaper to continue using an established technology level since there are some sunk costs, but in general fabrication facilities get upgraded as new tech comes along. There's no big savings in doing it the "old way". The primary distinguishing factor is size, and that scales with area, and worse, scales nonlinearly because making a bigger area flawless is harder than making many smaller chips in the same space. So, bigger sensors are always going to be more expensive.

\$\endgroup\$
3
  • \$\begingroup\$ This is a wrng application of moores law. In case of the sensor, the main price divers are surface (as in: more silicone costs more money) and surface (larger surface = bigger chance of a fault). Contrary to chips (Like Intel multi core) you ahve no way to "burn out" part of a chip (say 1 or 12 cores is bad, block 2 and sell it as a 10 core processor) - because the whole surface must work. No long tail, significantly reducing yield. \$\endgroup\$
    – TomTom
    Nov 29, 2016 at 17:36
  • \$\begingroup\$ On top, things like sensor sensitifiy depend to a large degree on the surface of the photo trap (pixel size). WHile finter structures make certain things easier - at the end the surface size IS significant. This is not the case for a processor. \$\endgroup\$
    – TomTom
    Nov 29, 2016 at 17:37
  • 1
    \$\begingroup\$ @TomTom Nonetheless, what I wrote almost five years ago basically continues to be true — sensor density has and still continues to increase. I don't think you're right about requiring perfection, either — dead pixels are masked out, and the smaller they are, the more you can mask out without having visible effect. \$\endgroup\$
    – mattdm
    Nov 29, 2016 at 17:40
0
\$\begingroup\$

Some full frame digital cameras were and are relatively low resolution: The Canon 5D was 12.8 MP and was released after the 17MP 1Ds Mk II. The Nikon D3 and D3s were 12.1 MP. The D3s was introduced a year later than the 24.5MP D3X with twice the resolution.

As of late 2015 the highest resolution full frame models are the 50MP Canon 5Ds (and 5DS R variant) and the 36.3MP Nikon D810, yet both manufacturers still offer full frame models such as the 20MP 6D and the 24MP D600. The Canon 6D has only forty percent as many pixels as the 5Ds, and the D610 only has two-thirds as many pixels as the D810.

Sony currently offers the mirrorless α7 in three variants: The 12.2MP α7s, the 24.3MP α7 II, and the 47.4MP α7R II. The α7s has roughly one-fourth the pixel count of the α7R II.

\$\endgroup\$
-4
\$\begingroup\$

There seems to be a misunderstanding that lower resolution means better low light performance. However, as long as the gaps between individual photosites are small enough, increasing the resolution barely decreases its low light performance (it gathers similar amount of light).

\$\endgroup\$
4
  • 1
    \$\begingroup\$ While potentially true, this isn't an answer to the question as asked. \$\endgroup\$
    – Philip Kendall
    Nov 29, 2016 at 16:57
  • 1
    \$\begingroup\$ This is incorrect. Lower resolution with the same sensor size means more area for each photosite. This leads to lower signal to noise ratio for each photosite, given the same total light hitting the overall sensor. You can't get the same thing by averaging the values from multiple photosites of a higher resolution sensor due to the quantization noise introduced by the A/D. As a real world case, consider the Nikon D3x with twice the resolution of the D3s. The D3s has better low light performance, *even after filtering a D3x picture to the same resolution". \$\endgroup\$ Nov 29, 2016 at 18:15
  • \$\begingroup\$ No, you can't get more information from the same amount of light. If the gaps between individual photosites are small enough, the light gathered would be similar. Although the SNR would be higher for individual photosites in lower resolution sensors, the greater number of photosites of a higher resolution sensor can be 'binned' together to generate a SNR similar to the lower resolution (or you can do it digitally by downsizing the picture). This has been a myth for ages and just recently DPReview finally acknowledged it. \$\endgroup\$
    – Michael
    Dec 4, 2016 at 3:31
  • \$\begingroup\$ To compare the noise of sensor with different resolutions, you have to resize them to the same resolution and compare their apparent noise. You would realise that the difference between the amount of noise of both sensors are negligible given similar sensor design. \$\endgroup\$
    – Michael
    Dec 4, 2016 at 3:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.