WP's HP: HOME | SloMo: Links ~ FAQ ~ [Info] ~ Tips ~ Data ~ Clips ~ HYCAM ~ Freq. ~ f-Calculator ~ Trigger ~ Image RAM | ?Help? | Deutsch

WP's SloMo CCD and CMOS Sensor Info


Sensor techniques

Resolution - black-and-white versus color cameras

Evidently most of the digital cameras with semiconductor imaging sensors, so-called solid state cameras (in contrast to bulb or film based ones) are single-chip camera systems. This is the reason for the lower light sensitivity of color cameras. Because compared with the naked black-and-white (better: monochrome) version the surface of the according color sensor is coated with a Red-Green-Blue (short: RGB) layer. And this swallows about 50% intensity. Often even much more, because for correct color separation and for sufficient modulation range (dynamic) the near infrared (NIR, near IR) part of the spectrum must be suppressed by an additional IR-cut filter (also IR-stop filter) up to about 1.1 micron wave-length. The reason is the transparency of the color mask for light with longer wave-length and especially in this spectral range the sensors are rather efficient. (Filterless black-and-white cameras could easily be used as cheap substitutes for night goggles.)

Filter lines
RGB stripe filter


Bayer pattern
Bayer mosaic filter

Due to the filter coating in stripe or mosaic pattern an additional disadvantage arises: In a black-and-white sensor each light sensitive cell contributes to the resolution directly. Each single cell can cover a tint by its gray scale value between black and white, thus each single cell represents one pixel (= picture element). A color sensor of the same quantity of cells needs one red, one (or two) green and one blue masked cell to represent a tint. Thus each pixel consists of a RGB sub-pixel triplet of three cells. Just arithmetical the resolution in relation to the quantity of cells is reduced to one third only. These facts are held back when talking about resolution and the quantity of cells is told to be equal with the quantity of pixels. However, through complex algorithms the loss in resolution caused by the color coating can be enhanced even in real time. Nevertheless, hardly more than 2/3 resolution of the according black-and-white sensor is achieved.

The images on the left show parts of color filters. Each cell represents a (sub) pixel. The twice as much green cells are used to enhance the resolution due to human perception capabilities (middle of the spectrum; sensitivity). The electronics for control and address circuits are usually placed in the black, here only thin drawn pixel outlines. Depending on the type of sensor such a blind area then can reach and even surpass the area of the valid active pixel. Mosaic filters offer a higher resolution than stripe filters, the color interpolation, however, is more complex.

Therefore the pixel count is not the main issue. At best it is as evident as the processor clock as sole criterion of PC efficiency. What matters is also contrast, sensitivity, dynamics, color separation and many more. A megapixel monster with slack images just inflates memory demands with a lot of useless data in the end.
Besides more pixels per square means smaller cell sizes paired with reduced sensitivity and increased noise tendency.
In this context one can just check what camera and sensor suppliers expect what number of defective pixels their customers have to tolerate. Usually one does not see them, because they are »mapped« by using the interpolated values of pixels in their neighborhood. Concerning quality suppliers one has to search for them, whereas others force sensors with dozens and dozens of faults upon you.

There are various color filter patterns and arrangements, not only stripe and Bayer patterns (named after Dr. Bayer, who invented it at Kodak in the 1960s). Usually this coating is done by photo lithography. Also one finds other color sets than RGB, e.g. magenta instead of red and yellow instead of blue or two different shades of green. Also additional uncoated pixels can be integrated in the matrix and so on.
In fact each electronic color camera gains its images by interpolation, because there is no colored current. (Not even from Yello ;-). Because the color algorithm knows the corresponding filter color above each individual grayscale cell it is able to calculate the color at this location. As a rule several neighboring cells are used for the calculation.

Just to mention: Expensive three-chip cameras (also named 3 CCD) as used in standard video applications, however, have a uniform masked red, green and blue sensor for each color channel. The beam split is done by a prism arrangement.
In the meantime for still cameras there are also sensors with vertical instead of lateral structure, i.e. one uses the dependence between penetration depth and wave-length of the light to separate the colors, refer to Foveon (www.foveon.com).
If the application is not pressed for time one can use color filters or color wheels similar to those which are common sight with professional spotlights. Some studio still cameras shoot so three successive photos each with a different color filter in front of the sensor or lens.


Filter effects - moiré, aliasing and fix pattern noise

Effects of filters are responsible for color artifacts at edges or lattices. If the image on the sensor is about the same size as its filter grid (or its cell structure, resp.), moiré and aliasing effects will possibly be the result. Good to visualize by error colors in shots of venetian blinds or fan grills. Just make a survey with your camera. Also the effect is well-known when scanning rastered newspaper photos.
Here on principle the Nyquist-Shannon-Kotelnikov theorem is valid. One can easily check it: When one shoots e.g. a white (or gray) fence, the image of each slat will have to cover at least one row of red, green and blue pixels otherwise the white tint cannot be »mixed«. It is created when all color channels are illuminated with the same intensity. Due to the fix pixel arrangement even monocolor cameras can suffer from similar errors. Due to different partial coverage/eclipse of pixel, »beats«, ghost images, occur, simply structures are shown where none of them really is - thus aliasing. (This effect often occurs when scanning newspaper or magazine photos. Depending on resolution and magnification suddenly harsh patterns appear due to the rastered printing (~ grid) of the source material.)
Moiré will occurs, when two grids in the optical path are shifted to each other, especially distorted. Then wide, format filling streaks occur with their numbers and directions depending of the angular the position between them. This could even happen to human eyes: Just lie two finely woven fly screens over each other and slowly change rotation angle between them.

Because each pixel provides a different dark current (i.e. does not show »0« when not exposed), below a certain threshold level all pixels are compulsorily pulled down to black. Otherwise one would find scattered colored pixels in a dark area. That limits dynamic and leads to an effect called »drown in black« - dark surfaces and shadow areas show no structures, inhomogeneities are wiped-out. In the beginning a major part of digital cameras suffer from this typical disease. Currently especially (cheap) CMOS cameras and notably those with small pixel sizes are affected.

One has even tried to use sensors with not regularly but randomly arranged color pixels, which means without fix scheme for positioning the single RGB color filters.
Even three-chip cameras can show colored lines, speckles or wedges when beam splitting prisms are not well adjusted or separated or under intensive illumination. Rotating color wheels can cause unstable brightness and wrong colors when transmission differs or synchronicity is not perfect.


Spectral sensitivity (or spectral response)

Spectral sensitivity: response as function of wave-length
Spectral sensitivity; comparison
human eye and CCD/CMOS sensors

Spectral quantum efficiency
Color separation of the CMOS sensor channels;
why an IR attenuation filter is necessary

Emitted color spectrum visible for human eyes (VIS) lies best case between about 380 nanometers (violet) and 780 nanometers (dark red). With the highest sensitivity in green-yellow at about 550 nanometers. CCD and CMOS sensors show a broader spectrum. Especially they work in the near infrared region beyond 780 nanometers up to the so-called bandgap of the basic material silicon at about 1 100 nanometers with a maximum sensitivity between 600 and 900 nanometers. Their maximum sensitivity is more pushed to the red compared with the human eye. (By the way: For wave-lengths beyond the bandgap silicon turns to »transparent«.)

The figure on the upper right shows the spectral sensitivity of the human eye in bright light mode in comparison with that of uncoated monolithic silicon, the base material for CCD and CMOS sensors in the visible and near infrared spectrum. (Each curve is normalized on its maximum.)
The characteristics of CCD sensors are well given by the typical bare silicon curve. CMOS sensors show a wider maximum pushed expanded to shorter wave-length due to their flatter structures supporting the reduced penetration depth of light with smaller wave-length.

The figure below shows the spectral sensitivity of the three RGB channels of a accordingly masked (= coated with red-green-blue filter pattern) color CMOS sensor. One clearly perceives the transparency of the polymer in the red and especially in the IR region and caused by that the demand to attenuate or respectively to block this part of the spectrum, in order to avoid overstress of red and to reduce noise, wrong level and overdriving.

When using color sensors one can manipulate this within certain margins by selecting transmission specifications of the color filter pattern and the arithmetical weighting of the color channels. Additional color conversion filters (reduction of sensitivity in red; for color video cameras often made of Schott BG 39 or BG 40 glass) help here, too.
In the short wave-length region of the spectrum (UV, blue) the sensors are comparatively insensitive. Besides here the glasses add further restrictions.

The spectral sensitivity of a camera, however, is not only limited by the sensor or the film and the filters, but also by the optics, especially lenses and the IR-cut filters mentioned above. Because using the full spectral sensitivity can lead to dull/blurred images. Due to focal length is a function of wave-length, the single color images arise in different distances to the lens (chromatic aberration). Achromats (lens systems in layer structure) help to avoid this effect.

Often filter (and lenses) are additionally coated in order to reduce reflection in the visible spectrum. The reflection loss at each air/glass interface amounts to about 4% to 5%:

ΔR = (1 - n)² / (1 + n)²

with the refraction index n of the lens material in air (nair ≡ 1).

Thus about 8% to 10% per transmission through a glass plane. Sometimes one happens to discover that by the iridescent colored surfaces when light incidents diagonally because ΔR is a weak function of the angle of incidence.
Lenses of higher quality sometimes offer an IR switch for operating exclusively with wave-lengths in the near infrared of about 780 nanometers and more. Then the inscription on the distance ring will fit again. (The IR image is displayed deeper inside the camera, due to smaller refraction of longer wave-lengths.)

By the way: The transmission of standard lens glass badly decreases with wave-lengths of 320 nanometers and less. In the long wave section the drop lies far away from what a sensor made of silicon perceives. Just to mention the coating layer of a sensor is made of quartz glass (silicon di-oxide). It is covered by a thin layer of silicon nitride.
If one wants to make shots in another region of the spectrum, one has to use different sensor and even lens materials. For instance IR cameras and especially thermographic cameras, which work in the wave length region of 3 micrometer and above, partly use sensors made of Indium-Gallium-Arsenide (InGaAs).


CCD versus CMOS sensors

Function principles

Each of both present solid state sensor technologies CCD (Charge Coupled Device) and CMOS (Complementary Metal Oxide Semiconductor) are designed with silicon as raw material. Their advantages and disadvantages depend on their basic functions. They use the inner photoelectric effect - in both the pixel work like an array of solar cells: more light = more charge or current, resp.

CCD and CMOS cell structure
CCD and CMOS circuit principles

The image on the left shows the basic principles of a CCD cell in comparison with a CMOS cell.
The charge generated in a CCD cell by incident light is directly read out from each cell. This charges are moved step by step outside of the photoactive area. Outside they are converted and amplified, resp.
In the CMOS cell the incident light generates a photo current proportional to its intensity in each cell and decreases the reverse resistance of each photo diode. These reverse currents through the photodiodes (i.e. the generated charges) are processed.

The cells of a CCD sensor operate as exposition meters accumulating charge and are read out in certain intervals. The »fast« CCD sensors exist in three types: FT ((Full) Frame Transfer), ILT (Interline Transfer) and FIT (Frame Interline Transfer). For still photography the full frame principle is sufficient enough: The charges remain in the active area until read-out shaded by a central mechanical shutter.
In FTs the complete frame is moved through the photo active cells in a light shielded area and then it is processed. Due to the different structures often it is possible to identify both regions by one's naked eyes. ILTs show alternating photosensitive and read-out lines. The charge of each photoelectric cell is directly pushed to the according light shielded cell and this line is then read-out. FITs are a combination of both designs.
FTs reach a fill factor of almost 100%, but suffer of a second exposure during the read-out phase (so-called smear). ILTs and FITs show a reduced fill factor, but are less sensitive in read-out phase.
The expression fill factor gives the ratio between optical sensitive area and area occupied by addressing circuits of each cell. Thus it allows to compare sensitivity under the precondition, however, using the same basic technology.

CMOS sensors continuously measure the light induced photo current. One uses the proportional relation between reverse current or induced charges, resp., and exposure of a photo diode. (Expressed in correct terms one measures the current to re-charge the reverse capacity at the pn junction inside the photo diode. One can imagine a capacity parallel-connected to the diode. There are different possibilities for appropriate circuits.)
The structure is rather similar to the layout of a DRAM, and so experimental shots were already made with opened memory circuits.
A PPS (Passive Pixel Sensor) operates like a CCD ILT. By illumination caused charges generated in a photo sensitive cell, mainly a photo diode, are read out cell by cell directly and are amplified and converted outside the photoactive area.

CMOS APS pixel cell
CMOS APS cell with global shutter

Common nowadays is the APS principle.(Active Pixel Sensor. The acronym has nothing to do with the APS-C format.) Here photo sensitive cells just work with the indirect principle of a CCD sensor: The exposure of a photo diode controls the connection to the operating voltage Vcc using a CMOS transistor (amplifier). Each cell just has its integrated amplifier circuit.
Because illumination reduces the charge over the photocell one can compare it with the negative film inside a camera.

The figure on the left shows the structure of a CMOS APS cell. Description of one cycle: Over the reset switch a pre-charge is brought in. After that the switch opens again and over the photo diode D related to illumination charge floats off. With that the control voltage at the amplifier T changes and with it the voltage V. Using the optional switch t shutter (if available) one can interrupt the discharge process, then the capacity C controls the amplifier all alone.

The pre-charge (realized by the reset switch) each time during exposure reduced by the photo diode is the reason for CMOS APS being insensitive to saturation effects unlike CCD sensors. More than complete discharge is not possible.
The simplest cell is made of three transistors, namely the switch for the initiation before starting the frame/exposure (reset), the switch for read-out process (read) and the real amplifier (T). This circuit offers a rolling shutter only. If one adds some kind of sample and hold circuit (t shutter and capacity C) in the amplifier control one will get a global shutter. This is an electronic shutter operating on the whole sensor at one given time.

With more transistors one receives additional opportunities for controlling and improving the signal quality. Thus using a 5 or 6 transistor cell one can decouple recording and read-out, so already start recording or exposure, resp. again during the read-out process is still running.


CCD sensors

CCD sensor
EG&G Reticon Frame-Transfer CCD sensor
HS0512JAQ; 512x512 pixel in 8.19x8.19 mm²;
ceramic DIP housing; >1 000 fps, monochrome


CMOS sensors

150 mm wafer on blue tape
6 inch wafer with 45 CMOS sensors on blue tape


Color filter Housed sensor
Bare CMOS sensor die for COB mounting
1536x1024 pixel on 16,9x11,3 mm²; >1 000 fps, color


Diagonal RGB color filter pattern
Color filter in diagonal arrangement (magn. 1 000 x)

State of the art

Evidently still many of the photo and surveillance cameras use CCD sensors while in video cameras the CMOS sensors are coming. In mobile phones with camera function CMOS sensors are used almost exclusively. The trend moves to CMOS technology, to the camera-on-a-chip.
In high-end cameras yet one sometimes tends towards low integration to optimize each part like analogue, digital electronics, control, power supply, ... individually. In standard use hardly one technique shows an advantage. In noise critical special applications, however, the CCD sensors lie in front. Cooled, if necessary. For shots against illumination sources and possible overexposure effects (shiny surfaces, ...) CMOS sensors may be the better choice.
(The often made assertion CCD sensors are in principle slower than CMOS sensors is not true so far, because there are CCD high-speed cameras as well.)

Images on the upper right: Silicon die without housing, about 20 mm x 15 mm x 0.5 mm in 0.5 micron technology. The greenish area is the mosaic filter. In the framing dark border one recognizes 137 small squares, so-called bond lands or pads, for the later electrical contacting with bond wires.
This die is directly, thus without housing, glued on the printed circuit board and is bonded then chip on board (COB). After that for protection reasons a cover with glass bottom is put over and is glued to the board. (Of course, traditionally housed chips are available as well.)

Image on the lower right: 1 000 times detail magnification out of the filter matrix of the CMOS sensor. The light sensitive area of the photo cell, marked yellow in the image, shows with 11 microns half of the cell size of 22 microns. The chessboard pattern tilt by 45° built with »honeycomb« cells like Fuji and Sony successfully use it in their »Super-CCD« and »ClearVid« arrangement resp. allows a higher resolution especially of vertical and horizontal structures compared with the standard Bayer pattern.

Also according high-speed cameras this change is obvious. Whereas typical systems of the 1990s still use modified CCD sensors in general, systems since the end of the 1990s mostly operate with CMOS sensors. Nevertheless CCD sensor systems exist further on.
Well known high-speed sensor sources are Dalsa/Teledyne (Canada), EG&G Reticon (USA), Fillfactory/Cypress (Belgium), Photobit/Micron/Aptina (USA) or CSEM (Switzerland).


Improvement techniques


Everything counts in large amounts. Because of a fill factor, i.e. the ratio photoactive area to control electronics area, smaller than one area and therefore sensitivity is lost.
The structure of CMOS sensors makes it worth to try microlenses in order to compensate this lost at least partly. The lenses should transport the light falling on the »blind« sticks to the light sensitive parts of each cell. Especially one like to use them for very small cell structures and fill factors (10 micrometers per square cell size and 30% fill factor and less). The lenses are directly deposit on the sensor surface by photo lithography. Due to the small size of the lenses, diameter a few hundredths millimeters each, and the large quantity - one lens for each cell (= pixel) - it is rather tricky and the optical qualities and the uniformity are not very high. This causes an underproportional increase of sensitivity. The price is higher expense and possibly reduced image quality due to optical faults and other parasitic effects.


Back-side illuminated CMOS (BSI)

Eyes and especially CMOS sensors show a quite similar layer structure. In both the light must pass a shading supply layer - blood vessels or electric circuits, resp. - before received in the photoactive layer. Fill factor and image quality are reduced. Moreover with further reduced structures in semiconductor design a trade-off between what is good for electronics and what is necessary for optics. Shrinking and higher clock rates and reduction of power consumption as well cause noisy small photocells. For instance standard design rules (structure size, width of conductor paths, gate width) of present DRAM and CPU techniques do not match CMOS sensor demands by a factor 50 or more. (Present structure width 28 nm!) Besides the transistors in logic circuits are trimmed for speed and digitally switch with (super-) saturation (bang-bang switch), while photoelectric elements should be read out analogue and a saturation should be preferably avoided.)

Therefore the idea to grind the sensor bottom and then to put the sensor die upside down. The supply electronics are at the bottom, the middle is filled by photocells and on the upper side color filter pattern and perhaps microlenses are placed.
The gap between CMOS standard designs and sensor demands can be reduced by that, what makes the sensors cheaper especially by integrating further electronics and conceals deficiencies like e.g. small penetration depth for light beams.
Not really a quite simple semiconductor process (among other massive grinding of the wafer is necessary), but similar technique was used with special CCD sensors, too.
Fore more info refer to:


Curved sensors

Existing lenses throw their images on a flat area, the image plane. The mathematical correct projection, however, takes place on a slightly curved area the so called Petzval Panel, especially with simple lenses (astigmatism and image field curvature). In order to improve among others depth of field presently circular arc shaped sensors are in the market.
Maybe that the sensor is really curved or a location-dependent path extending coating or according optics are added. (Anytime then there will be saddle shaped and conical sensors... ;-)


Lenses and optics

Lens mount, flange-back (back focus), focal length and crop factor

Elementary criterion for selecting lenses is to check which mount is given by the camera body. The mount determines the flange-back or back focus distance. It is the distance measured between the last cylindrical landing of the lens housing before the thread and the film or sensor plane. Thus the distance between optics and image plane is mount specific. With a lens out of a different mount family possibly the focal length inscription of the lens is not correct, as far as one can adjust its focus and fix it at all. The flange-back e.g. with C-mount is 17.562 mm and with CS-mount it is 12.562 mm. When one happens to get a more or less sharp image of the iris aperture and nothing else, one has screwed a CS-mount lens on a C-mount camera. (With a 5 mm spacer a C-mount lens fits onto a CS-mount camera.)

Crop factor
Image circle, format and crop factor

Within a mount family there are not such problems. All adequate lenses show the same flange-back, thus each image is displayed at the same location. The only difference is the varied size of the circular images the lenses offer at the image plane. Given is the format of the sensor that is illuminated sufficient enough and without too strong optical distortions. If the sensor completely lies inside this image circle, everything will be all right. If it is larger, the shot images will be considerable sharp, but their corners tend to be underexposed (keyhole effect, vignetting). Thus especially one can usually select without problems C-mount lenses of larger formats for a C-mount camera with a smaller one, e.g. an 1 inch lens on a 2/3 inch camera. In the other way, however, with restrictions only, refer to sensor C in the figure on the left.

The figure on the left shows the crop factor, also named focal length extension. This is in fact a pseudo effect, because focal length is an invariable attribute of each lens. The different image impression, especially the angle of view (also field of view), is only caused by showing the image of a smaller sensor (B) with a greater magnification on a screen (or a print). Its pixels lie closer to each other than those of a bigger one (A) and are displayed somehow stretched, nevertheless without negative effects on quality. Adequate resolution (pixel count) assumed.
At least the sensor must lie inside the image circle of the lens, given as »format«, otherwise vignetting takes place (C).

If one compares two format filling images of an object, e.g. one of a C-mount camera with a 2/3 inch sensor and one with an 1 inch sensor, usually one will not see a difference. Displayed on the monitor screen they are of the same size.
But due to the fact that they come from areas of different sensor sizes, the magnification must be smaller on the smaller sensor (here B), in order to make the image fit to it. Therefore the same »XY Millimeter« lens in front of a larger sensor (here A) will work at a camera with smaller sensor like one with a larger focal length and vice versa. If, however, one compares the shots based on the angles of view (= what fits onto the respective sensor using the same lens) only. So far focal length and image format belong together and give the imaging scale. The image circle, however, is a lens parameter only. And the focal length, too.

Perhaps one could introduce the term »effective focal length«, but related to what? Often in common usage this dummy effect of focal length extension, when screwing the same lens on a camera with smaller sensor, is called (digital) crop factor:

crop factor = 43.3 mm / sensor diagonal

The value (about 1.5 ... 2 comparing typical digital SLR cameras with miniature 35 mm; format cameras of 24 mm x 36 mm; therefore sometimes the expression »XY mm 35 mm format equivalent«) states the apparent increase of the focal length finally due to the decreased sensor diagonal. Thus the expression »crop« meets the facts better. Much of the image just misses the smaller sensor. One will see the center only and will interpret this as tele/zoom effect, when the monitor is displaying the sensor full screen, refer to the figure above.

Hooked on high quality demands one has to take into consideration, however, that the area with optimal corrected lens errors is smaller than the maximum illuminated one and that smaller formats claim lenses of better qualities because of the higher pixel density. Besides one needn't spoil the image quality of a good camera by using just a better bottom of a bottle instead of a lens. Then the image remains sharp to the corners.

In professional still imaging widespread use of the according possibilities is made.
If the focal length of a lens is of about the same value as the diameter of the format, one will often call it the »normal lens« for this format. Concerning miniature or 35 mm cameras this would be about 50 mm ±5 mm, refer to next chapter and [SloMo Tips]. Then the angle of view equals that of humans roundabout.



Elementary feature, too: Usually the format or film gate, i.e. the size of the optical active area on the CCD or CMOS sensor, is given for CCTV cameras in inches. The somehow arbitrary dimensions are from the Vidicon valve era. The data derive from the size of their (socket!) dimensions. As base for the inch class of a sensor the 1 inch sensor with a diagonal of about 16 mm is used:

Inch format ≈ sensor diagonal [mm] / 16 mm

or when using inch measurement:

inch format ≈ sensor diagonal [inch] / 1.5875

although in real one inch measures 25.4 mm.
(Legend: fps = frames per second of some selected high-speed camera matrix sensors at full resolution I have worked with.)

Format [inch] Width [mm] Height [mm] Diagonal [mm] Remark
1/10 ~ 1.44 ~ 1.08 ~ 1.8 autonomous CCD camera for medical »in corpore« use (to swallow like a pill)
1/8 ~ 1.626 ~ 1.219 ~ 2.032 CMOS cameras for special applications
1/7 ~ 2.055 ~ 1.624 ~ 2.619 CMOS cameras for Pads, Smart and Mobile Phones, ...
[128² pixels] 2.048 2.048 2.896 Dalsa CA-D1-0128A CCD, 736 fps (tweaked 2 020 fps)
1/6 2.4 1.8 3.0 single board cameras for special applications
1/5 2.88 2.16 3.6 single board or simple cameras, webcams, ...
1/4 3.6 2.7 4.5 cameras for observance (typical with pinhole lens)
[256² pixels] 4.096 4.096 5.793 Dalsa CA-D1-0256A CCD, 203 fps (tweaked 531 fps)
1/3 4.8 3.6 6.0 cameras for automation (»machine vision«)
1/2 6.4 4.8 8.0 above all portable TV video cameras, often 3 chip version
1/1.8 7.2 5.4 9.0 Standard still (CCD/CMOS) cameras
2/3 8.8 6.6 10.991 standard in the image processing sector
[512² pixels] 8.19 8.19 11.582 EG&G Reticon HS0512JAQ CCD, 1 000 fps (tweaked 1 094 fps)
16 mm film (movie) 10.3 7.5 12.7 16 mm film cameras (Stalex, HYCAM), 3 000 fps, 10 000 fps
1 12.7 9.525 15.875 professional (studio) TV video cameras
[1280x1024 pixels] 15.36 12.29 19.67 Photobit MV-13 CMOS (Aptina MT9M413C36ST C/M), 500 fps
[1536x1024 pixels] 16.896 11.264 20.306 FhG/CSEM Cam 2000 Visario CMOS, 1 106 fps
1¼ (30 mm) 17.1 12.8 21.4 professional (studio) TV video cameras
4/3 (»Four Thirds«) ~ 18.0 (not def.) ~ 13.5 (not def.) 22.5 semi professional digital SLR (photo) cameras
35 mm film (movie) 22.05 16.03 27.28 professional 35 mm film (movie) cameras
APS-C (24.0) 25.1 ... 25.5 (16.0) 16.7 ... 17.0 ~ 30.15 APS-C film and sensor cameras
miniature or 35 mm 36.0 24.0 43.3 35 mm (photo film) cameras and »full format« digital cameras


Aperture (f-stop)

Entrance for light: The aperture or f-stop reduces the incident light intensity falling on the sensor or film. The f-stop number is defined as k = focal length / effective opening diameter. On the f-stop adjustment ring one finds the f-stops inscription as multiples of the square root of 2 (1.4 - 2 - 2.8 - 4 - 5.6 - 8 - 11 - 16 - 22 - 32 - 64). With increasing f-stops the intensity let through is reduced in a way that one receives half of the intensity from step to step.
The reciprocal of the smallest f-stop is often called »transmission/light strength«. But if one wants two lenses offering the same brightness, one will have to take two with identical t-stops, defined as t = 10 × f-stop / √ transmission [in %], as it is usually provided by lenses for movie cameras.
When using a sensor with the same size a smaller aperture (= high f-stop number) enlarges the depth of view, refer to [SloMo Tips].


Zoom lenses, digital lenses and resolution

Expression zoom stands for lenses, which permit to tear closer an object by using a stepless focal length increase within a certain range of given focal length. Usually they are large, heavy, expensive and of poor light sensitivity. One should prefer to use exchangeable lenses of different but fixed focal length. There are available real zoom lenses where the image stays sharp during zooming and vario lenses where one can change the focal length, too, but has to adjust focus as well.
»Digital zoom« or »electronic zoom« stands for an enlargement of a part of an image by means of a (simple) imaging program. High zoom levels cause pixeled images. Normally electronic zoom is not a worthy substitute of an appropriate zoom lens, even if marketing likes to claim it so.
By the way, similar case: »5x zoom« does not say anything about the magnification of a lens, but merely that its smallest adjustable focal length is five times smaller than its biggest.

In the meantime »digital« lenses or sometimes also called telecentric lenses are offered. They are designed so that the beams at the image side hit the sensor in an angle as perpendicular as possible. This is done to accommodate the shaft-like structure, which especially has a bad influence on small sensors with high pixel density causing faults like diffraction, shadows, color errors, flattening, ... (Just the small, cheap sensors need complex and expensive lenses.)
To call such lenses telecentric, however, is wrong. This type of special lenses images an object in the same size disregarding its distance. These optics are especially used in image processing for measurement applications. For instance when one views in a tube it does not become narrower with increasing distance, but it looks like a ring washer and it is no problem to measure the inner diameter and the thickness of the wall. Used with an usual landscape scenery it shows the strange effect that e.g. roads running to the horizon do not get smaller.

When talking about resolution it is highly recommended to get an idea of the MTF modulation transfer function, find an excellent explanation e.g. at Optics for Digital Photography (www.schneiderkreuznach.com/knowhow/digfoto_e.htm).


Glass delay (= z-axis image shift)

Exotic in most cases: When one screws a thin filter in front of the lens, this will rarely have any effects on the imaging position of the optical system. But this will drastically change, if one adds a protection glass, a filter or a LC shutter behind the lens. These plane-parallel plates cause a kind of glass delay effect

Δd = (1 - 1/n) × d

with the refraction index n and the thickness d of the plate (nair ≡ 1).

With a glass plate the image would appear about 1/3 of its thickness behind the sensor, due to the increase of the effective flange-back. A mechanical solution must be designed in order to adjust this certain additional distance between lens and camera body, exact the sensor.
Therefore the threaded pins or special mechanics, resp. in the lens adapters e.g. for cameras prepared for optionally mounting an internal LC shutter or an additional internal (IR-cut) filter or which are designed for IR shots.


Tips and tricks around making sequences you will find here:





©WP (1998 -) 2012
Update: V8.4, 2012-03-02