Incense

by Bart Arondson

submit your photo


Hall of Fame
View past winners from this year

Please participate in Meta
and help us grow.

Take the 2-minute tour ×
Photography Stack Exchange is a question and answer site for professional, enthusiast and amateur photographers. It's 100% free, no registration required.

I've read some information about sensor sizes here

http://en.wikipedia.org/wiki/Image_sensor_format

according to this, the 35mm ff-CMOS is the sensor with the largest dimensions used in digital cameras. It has a lot of advantages to smaller sensors, caused by its size.

Why are there not even bigger sensors available to force these advantages? 1,5 FF e.g.?

share|improve this question
3  
That page is outdated. Hasselblad launched a medium format CMOS sensor in March. –  Philip Kendall Aug 27 at 10:01
    
The article specifically mentions different medium format chips that are larger than so-called "full frame" (a misnomer). en.wikipedia.org/wiki/… –  his Aug 27 at 10:48
    
@his Which CMOS sensors which are larger than full frame does it mention? –  Philip Kendall Aug 27 at 11:03
    
@fubo Are you specifically interested in CMOS sensors (as opposed to CCD sensors) or do you really mean "what limits the size of digital imaging sensors?" –  Philip Kendall Aug 27 at 11:17
    
@PhilipKendall updated –  fubo Aug 27 at 11:23

4 Answers 4

up vote 23 down vote accepted

You can make some very large CCDs. An older press release talks of a CCD that was made for the US Naval Observatory that is 4" x 4" and 10,560 pixels x 10,560 pixels. Thats 111 megapixels on one sensor. Thats kind of not small.

A 111 megapixel sensor

(From above press release)

The first restriction that the sensor has is that it must be a single wafer of silicon, and thats a fixed price. You can create CCDs that are designed with a three edge CCD (the remaining edge is where you can read the data out) such as:

mosaic CCD

(From http://loel.ucolick.org/manual/deimos_ccd/science/overview/EL3160.html)

These are often used in telescopes to get a larger imaging area with only a smaller increase in price. Note that there is the issue that each CCD needs to be calibrated separately from the others (no two image sensors have exactly the same response) - this is a significant concern for scientific uses (calibration information for one such CCD array).

The mosaic CCD can be scaled up quite significantly. PanSTARRS has a 1.4 gigapixel sensor array that is made up of a massive array of 600x600 pixel CCDs:

8x8 CCD array from PanSTARRS

Above is an 8x8 array of CCDs - each one quite small. This then is part of a larger array of 8x8 of these segments giving an overall 64x64 array of sensors. This was done because of cost savings, speed (its faster to read out four thousand 600x600 pixel CCDs simultaneously than it is to read out one larger CCD), isolation of saturated pixels, and an easier replacement in the case of defects.

The LSST uses more conventional three edge CCDs to reach its goal of 3.2 gigapixels. Each segment there is an 8x2 array of 500x200 pixel sensors. All the same factors mentioned for PanSTARRs is also in place here. It is expected to take 2 seconds to read out 3.2 billion pixels (which is actually quite fast). Going to fewer, larger CCDs would mean its slower - not faster.

LSST sensors

So, while its possible to use multiple sensors in aggregate, they are still composed of rather small individual sensors rather than a large single sensor (as was done with the USNO's 4x4" sensor). In some cases, the CCDs are much smaller than even those used in point and shoot cameras.

Look back to that first image of the 4x4" sensor and then consider the size of regular sensors on there:

sensors on a wafer

This has some additional information on there to consider. There's the maximum yield of how many you can put on a wafer (you just can't fit more on) and the waste. In order to make that 4"x4" sensor they needed an extremely high quality wafer of silicon. On a regular full frame the flaws in the crystal are there no matter how many sensors you put on the wafer. With an 8" silicon wafer (same size as the one on the top - notice that half the diameter is at the 'edge'), there are flaws scattered throughout the wafer. The fewer sensors on the wafer and the higher the chance that there will be a flaw in the sensor making it unusable (the 36% waste on a full frame sensor wafer vs 12.6% waste on the 13.2mm x 8.8mm sensor). This is part of the reason there's often more research done on increasing the density of the chip rather than making it larger (and that density research has other applications like making CPUs go faster).

With a sensor that is intended for a 60mm x 60mm frame, you can only fit about 8 sensors on the wafer and the waste goes up. You can see the economy of scale at work there.

Consider that the 15 or 16 working sensors off of the full frame wafer cost the same to make as the 213 or so smaller sensors... and are priced accordingly. The following image shows the issue with the flaws located in the same places on the wafer for various sized dies.

Sensor yield

(From http://commons.wikimedia.org/wiki/File:Wafer_die%27s_yield_model_(10-20-40mm)_-_Version_2_-_EN.png)

If you are willing to step away from the 'an image in one go' you can get a single array (well, three - one for each color) of sensors that move across the image. These are often found as scanning backs for large format cameras. There, the issue is the precision of the equipment rather than the size of the sensor (memory, data storage, fast i/o become significant). There are some cameras that have this as an integrated unit such as the Seitz 6x17 digital.

Further reading:

share|improve this answer
    
111 megapixels is small compared to the planned LSST (3.2 gigapixel). I think the current largest operating telescope (in terms of pixels) is PanSTARRS, at 1.4 gigapixel. –  Joe Aug 28 at 13:07
    
@Joe the key there is one sensor that is 4" x 4". if you scroll down to the "LSST Focal Plane" section in the link provided you will see explanation of "189 3x3 rafts" where each part of that is a 3 edge mosaic CCD. The mosaic approach can be scaled quite large as you've linked... but its not a single sensor. PanSTARRS uses a similar approach - image-sensors-world.blogspot.com/2007/09/… with an array of CCDs ( pan-starrs.ifa.hawaii.edu/public/design-features/images/… ). For both of those, the sensors are rather small. –  MichaelT Aug 28 at 16:28

The largest CMOS sensors available commercially for photography are "medium format" and measure about 44mm x 33mm. CCDs exist in slightly larger sizes up to 54mm x 40mm. Larger sensors for scientific applications may have been produced.

Sensors are produced by projecting a mask onto a large wafer of silicon using UV light. The wafer is then cut into individual sensors. The absolute size limit of a sensor that could be produced with this method is determined by the size of the image circle produced by the projector (though there may be other concerns with very large sensors, such as power usage and heat dissipation which present a hard limit on size).

The practical limit of sensor size is reached much earlier as it is determined by the yield, that is how many sensors have to be discarded during fabrication due to defects. When making many small sensors on a single wafer a single defect will lead to one sensor being discarded but many more being viable, if one sensor takes up the entire wafer then a single defect will mean no sensors are produced. The yield thus decreases with the square of the sensor size, which makes larger sensors uneconomical.

Economies of scale also apply, 36mm x 24mm "full frame" sensors would be more expensive if produced in the same volume as medium format sensors.

share|improve this answer
2  
Good answer - I appreciate bringing in the realities of both engineering and business –  B Shaw Aug 27 at 11:59

There are even bigger sensors. If you look closely to the image in the top-right corner of that page you'll see that the biggest sensor there is 'Medium Format Kodak KAF' sensor.

Ok, I understand that it isn't quite easy to figure this out because one can easily take that the background of that image is gray while in reality the image has a white background.

See it better here.

Besides that sensor there are another sensors bigger than FF. On the same page scroll to the Table of sensor formats and sizes, click on the 'Crop Factor' column to sort the table and look at the formats with a crop factor smaller than 1. Get out the film formats and you will end up with the following sensors in this order:

  • Phase One P 65+, IQ160, IQ180
  • Leaf AFi 10
  • Medium-format (Hasselblad H5D-60)
  • Kodak KAF 39000 CCD
  • Pentax 645D
  • Leica S

But beware: there are also disadvantages for such sensors: big, heavy cameras and lenses. Much more difficult to build a lens for such a sensor (bigger image circle) and... ...of course... ...price.

share|improve this answer
    
But the Kodak sensor is CCD, not CMOS. –  Philip Kendall Aug 27 at 11:02
    
+1 good sensor list –  fubo Aug 27 at 12:49

Some more things that can limit what's practical to below what can be manufactured:

  1. weight (of the resulting system). A very large sensor needs a very large image circle, which means large lenses, and a large camera.
  2. power consumption. A large sensor needs more power than a small one, thus battery life is decreased (unless you again increase the size and weight of the camera to house a larger battery).
  3. speed. It takes longer to read out a larger sensor than it does to read out a smaller one with the same sensor element density. So your "shutter speed" goes down.
  4. cost (hinted at, but comes into play at several levels). A larger sensor of course costs more than a small one, not only because it needs more raw materials but also the amount of discarded products goes up, the cost of which all has to be recapped from the smaller number that are sold.
share|improve this answer
    
I'm surprised no one else mentioned the speed issue. It's also worth mentioning that the larger you get (in inches or cm), the more distortion you get at the edges. There are astronomy papers describing how to describe the projection of the image so that others can understand how the image is distorted so they can re-project it to co-align multiple images. Scaling up in pixels without scaling up in physical size is also a speed issue, as it requires longer exposures for sufficient signal to noise. –  Joe Aug 28 at 13:15
    
@Joe that's a side effect of the lens you put in front of the sensor not generating perfectly parallel rays all across the sensor face, not a problem with the sensor itself. You could get around that by making your lenses (and your image circle) that much wider, increasing weight, size, and thus cost of your system even more. –  jwenting Aug 28 at 14:02

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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