WP's SloMo: High-speed Cameras

Environment illumination provided as standard in an office or in a production hall is sufficient for most black and white video high-speed camera systems. (They have sensitivities somewhere between about several hundred and several thousand ASA at nominal frame rate depending on amplification. Color versions usually show about 25% of that only.)
But it is valid: More helps more. The more light is offered the more the aperture can be closed, i.e. a higher f-stop number can be set, enhancing the depth of field. At f-8 or higher it is well done, but 5.6 is sufficient enough.
Usually there is not too much light - it is possible to shoot the glowing filament of a 500 W tungsten bulb, the hot spot during laser welding or an electric discharge. Only the brightness dynamic of the camera is problematic (that's just valid for all electronic cameras). The 8 to 10 Bit or 256 to 1 024 gray tones (i.e. shades of gray), resp., should be optimal used, even if 64 gray tones offer a sufficient quality. If one wants to image the less illuminated environment, one will have to light it up. Comparable to the additional (background) light-up flash in photography.
Just to mention, when it is too bright at all, (neutral density, short: ND; i.e. neutral, colorless) gray filters will offer help. Many lenses provide threads for that reason.

Practice-oriented example: If one wants to image the sparkles of a cigarette lighter together with the holding hand, about 1 000 W will be necessary (refer to sample 2 with and 3 without lighting up in [SloMo AVI]).

By the way: The dynamic range is often given in units of Dezibel [dB]. The connection to Bits and Bytes gets clearer with the following formulas:

In which x represents the value in dB and y the number of sensitivity steps.

Thus 8 Bit (= 2⁸ = 256) make so 20 × log 256 = 48 dB, while 60 dB stand for 10^60/20 = 1 000, so about 10 Bit (2¹⁰ = 1 024) dynamic range. Thus 12 Bit would be about 72 dB.
These are typical values.

Professionals use at least halogen spots in studio quality (defined color temperature, ...), cool beam reflectors (with reduced IR spectrum), or even HMI spots (daylight spectrum) and DC illumination devices (rectified) to avoid the jitter of the 50 Hertz or 60 Hertz, resp. (1 Hz (Hertz) = 1/sec), power supply frequency. A sequence mainly illuminated by common fluorescent tubes may show distinct pumping brightness, refer to figure on the left.

Due to the different color temperatures of light sources, e.g. the tungsten filaments of halogen lights shine with longer wave-length or warmer (i.e. more reddish or yellow) than daylight, electronic cameras offer a so-called white balance feature. Then white will become pure white again.

And caution: A halogen spotlight with a nominal power of 1 000 W heats with about 900 W. This might be sufficient enough to melt plastics within a distance less than 1 m after a while!

Exposure and aperture (f-stop)

Exposure diagramm
Function of shutter and aperture (f-stop)

External shutter in thread housing, ©WP 2001 Internal shutter in printed circuit board
Liquid Crystal shutters using Displaytech's FLC types.
Designed for external or internal mount.
Supply and control by the camera electronics.

Exposure, better the maximum time of exposure of most high-speed cameras, is nearly 1/frame rate. The sensor is integrating over the incoming illumination during this span of time. The gray areas under the curves in the upper image on the right give a measure for it. The narrow gap between the frames is the read-out and/or initiation time of the sensor, duration in the range of some microseconds or even less.

Closing the aperture of the lens (f-stop) reduces the amount of incident light global for all frames. The bigger the f-stop number the smaller the iris aperture gets. Thereby the depth of field increases.

A shutter reduces the incident light, too, but by repeatedly reducing the (fixed) time of exposure of each single frame. Technical spoken it changes the duty cycle mainly in order to reduce distortion caused by motion (motion blur).
Because if the object is moving during the time of exposure with more than 10% of its size, this will be bothering usually. (In still photography one accepts just up to 3%.)

There are three possibilities for reducing motion blur:

Many up-to-date systems offer as standard a more than ten times faster electronic shutter in their chip design and therefore do not need an additional LC shutter. Exposure times of microseconds and less are not technical problems, but export topics addressed by dual use (military use) restrictions.
For those who want or need more - in this case shorter exposure times - Kerr or Pockels cells can help. Of course, there are mechanical shutters and choppers in the shape of slit or hole wheels, too.

Annotation: In contrast to the mentioned above global shutter (or snap-shot shutter, freeze-frame shutter) simultaneously operating on the whole sensor there is also a linewise operating rolling or slit shutter. Concerning movement analyzing one should prefer cameras with the first one, because the rolling does not guarantee to take the resulting frame at a defined moment.

Changing the shutter time one can of course vary the illumination, if one is not able or does not want to adjust the aperture (f-stop) ring. The latest concerns more the artistic approach, because the depth of field depends on the aperture. Who ever may need it...

Depth of field (or depth of focus)

Expression for the distance region where objects appear sharp on the image. The reason why it is not just a simple point like the inscription on the distance ring (focus) seems to claim, is that within a tolerance area the film, the sensor or one's eyes do not show resolution enough. When a point is displayed within the so-called circle of confusion (about 0.01 mm to 0.025 mm diameter with common sensor formats and 0.033 mm with 35 mm format cameras), one will not perceive the blur at all.

The position and expansion is mainly dependent on the aperture: small aperture (= higher f-stop) number - large depth of field. Where it is larger behind the calculated position of focal distance than in front of it.
Depth of field increases with shorter focal length, longer distance to object and larger film/sensor format exact larger pixels, resp. (Due to crop factor sensor shrinking seems to increases(!) depth of field.)

There is a rather bulky formula, refer to f-calculator [SloMo f = ∞]. Anyway depth of field is a no-word in high-speed imaging in industrial environment (mostly the aperture of the »Dark-o-matic« is opened like a barn door;-).

Using the so-called hyperfocal distance h the region starting at h/2 to infinity appears in focus. It is also called fix focus setting or near adjustment and is especially used with simple cameras.

Cheap lenses, especially for surveillance cameras, often do not even offer a focus ring. They are focused by the aperture only. Here the electronic amplification (gain) of the cameras has to provide the required brightness.

Field of view and lens

Endeavor to do a good job: Image the interesting scene as screen filling as possible. With C-Mount it is not too tricky to use spacers (extension tubes in a set for less than € 50.-) and a fixed focus lens or an additional short distance lens for covering the entire screen with an object of 5 mm (= 1/5 inch) in diameter. Do not choose a wide-angle lens with a focal length smaller than 6.5 mm (concerning C-Mount 2/3 inch format) for applications using automatic image processing (e.g. with object tracking), otherwise distortion would reach a level not to be tolerated. All lenses and equipment suitable for the according mount can be used, even filters, macro, zoom and fish-eye lenses and with appropriate adapters photo lenses, bellows, microscopes, endoscopes, boroscopes, fiber optics, image enhancers (night goggles), ... But pay attention to the fact the more complicate the optics are the more illumination is often necessary.

Professional photo stores offer adapters to fit common lens mounts, e.g. C-mount to Nikon bayonet (Nikon F). Some high-speed cameras (like many professional photo cameras, too) additionally allow due to a mounting plate/reference plane the choice for a wide range of (customer) specific adapters including Nikon F, C, Kinoptik, Stalex, Voigtländer mounts and others.

Actually the focal length inscription on a lens is the focal length at the image side, when imaging an object in infinite distance with a wave-length of 546 nanometers. Then the image appears in the focal point located at the image side of the lens.
The definition according to DIN 4521 standard is:

With the half field angle ω and the half image diagonal y'.
(Subsequent the field angle is heavily simplified to be the angle of view due to format filling setup and f' and f are synonymously taken as the focal length inscription of the lens.)

The rule of thumb at a rough calculation of the focal length for a format or a sensor filling image S is

focal length = distance to object / (1 + object size / image size); [all values in mm]

distance to object = focal length × (1 + object size / image size); [all values in mm]

The angular field of view FOV = 2ω (≡ 2ω', refer to figure above) is given by

The magnification M derives from the imaging through the lens onto the sensor and the display of the image on the monitor

M = focal length / (distance to object - focal length) × diagonal of monitor image / diagonal of sensor; [all values in mm]

One must pay attention to the restrictions of standard lenses. In a distance less than 0.3 m (approx. 1 foot; sometimes even 1 m) to the object they cannot provide sharp images. In such cases one needs a spacer (= extension tube), an additional short distance lens or a macro or micro(scope) lens, etc.

The thickness t of the spacer, which is screwed in between lens and camera body, is given by

in which the relation image size / object size is called imaging scale. But caution, just despite of their simplicity - extension tubes swallow light.

Fine adjustment of flange-back (back focus lens adjustment)

Evidently a wrong adjustment of flange-back derives from manufacture tolerances, mismatch (incompatibilities of components) and media brought in the beam area between lens and film or sensor, resp. And concerning crash-proofed cameras mechanical stability is above all, even above (too sensitive) adjustment mechanics.
Provided that the suitable adapter is selected using a lens of fixed focal length one easily receives sharp images by turning the distance ring (focus). Maybe the distance inscription and the magnification are slightly wrong then. Usually this does not bother one further more.
Using a zoom lens, however, one looses the sharpness of the image during zooming process. As a rule it should be steady and only magnification (and thus the angle of view) should change. Then the zoom lens is only usable in a restricted manner as vario lens. One is ought to adjust sharpness simultaneously during zooming all the time. High-speed cameras, however, are rarely used for such zoom shots. Nevertheless in order to take full advantage of the zoom feature the flange-back has to be adjusted accurately. Using simple tools proceed as follows:

To adjust flange-back camera housings offer either a thread tube to be moved back and forth (like e.g. SpeedCam +500/+2000, SpeedCam 512 and SpeedCam PRO cameras) or metal pads (washers) are fed in (e.g. Stalex and SpeedCam Visario cameras). Some camera models provide mechanics to change the sensor position. There are also lenses, especially C-mount ones, with a cylinder housing at the camera side, which can be shifted. Look for a small depth bold at the circumference.

Evaluation

Eminent: Please do not forget you receive a lot of data. A megapixel resolution at 1 000 frames/sec and more just leads to data rates in the gigabyte/sec range. Thus more than a complete CD-R per second or a big DVD-R per sequence would be filled. Per each camera, mind you. So do not wonder why the high-speed camera system is rather busy, when storing and viewing these files. And - it is highly recommended to have a storage/backup concept for the files.

The huge data amounts cause the cameras to be used offline. Thus they are not immediately used for controlling and they are not directly integrated in a superior machine control circuit. The image processing would be just too costly and too slow. One watches the scene and analyzes afterwards.
(Today slower cameras of the image processing sector - machine vision sensors - can already be equipped with a calculating power allowing them to work like a sensor only providing a simple good/bad signal for the control unit - e.g. »Label position on the bottle was all right - Yes/No?« - and not sending image data for further computing.)

For controlling high-speed cameras even multi-channel systems of different suppliers special software is available. The evaluation is carried out either directly visual or using motion analyzing software packages, so-called motion trackers. Refer e.g. to the links given in [SloMo Links].
To make it easier for automatic motion tracking by software one should pay attention to ensure a homogenous background and to avoid gratings, chessboard pattern or something like a wallpaper with flowers, if possible. Thus reducing the calculating time and prevents the tracking algorithm getting stuck with the attractors in the background instead of pursuing the desired target.

If you intend to archive your image files in AVI format, consider making use of the compression tools Intel Indeo or DivX saving up to 90% memory capacity without loosing too much quality. DivX often provides smaller files, especially when less movement happens in the scene. In contrast Indeo commends itself for automatic image processing, because it manipulates object position in a smaller extent.
AVI files can be processed (i.e. changing frame format, replay speed, ...) with e.g. VidEdit!. For these and other tools just have a look here in [D-Load], the download center, refer to button on the left.

Namely one shoots sequences with super slow motion to slow down fast movement for visual inspection, nevertheless, one should create some faster replays, e.g. with 25 to 100 frames/sec (of an original with 1 000 frames/sec), otherwise the impression of movement is lost. This will become important then, if one wants to show the sequences to some outsiders who are not so familiar to the scenery.

Because the high-speed cameras run also with a normal speed of 50 or 60 frames/sec it is also possible to shoot true video clips. And copied on a CD or a DVD with the same replay speed one can hand out it to the customer the machine has been built for with the words »So your machine has been working during the final tests«. Just generating a sophisticated impression, not only with the high resolution cameras. Thus the test becomes a advertising movie.

Tricks for system integration

Remote control

Exercise: A camera head should be adjusted (scene, aperture, sharpness, shutter, trigger, format, ...), but the host PC is in a distance, e.g. in a control room behind a safety door, so that no live image is available near the camera. Or one wants to control the camera system from an extended distance.

The common interfaces (GigaBit Ethernet, etc.) are sufficient fast enough, but one is often dependent on expensive control software of the camera manufacturer or a third party provider.
Why you just don't try an inexpensive KVM? KVM extender (= Keyboard-VGA/Video-Mouse extension) offer an efficient remote control without interfering with the PC. (Even additional driver/software are not necessary.) The KVM consists of a transmitter and a receiver in the simplest case connected with a standard Cat 5 UTP Ethernet cable. The transmitter is plugged to the control host PC instead of keyboard, mouse and VGA monitor. The real peripheral devices are connected to the receiver. Now one can operate at the receiver as if the host PC would stand beside one's knees. Depending on the chosen system/transmitting technology distances of several ten meters up to some hundred meters and even more are possible.

Trigger and synchronize

Easy to understand - to trigger means nothing else but to start something caused by a single event. For instance the recording will be started, when the crash test vehicle hits the wall. The trigger device is often just a simple closing contact at the bumper providing a short circuit in the impact moment or a light barrier.
To synchronize means to stabilize the frame rate in a defined relation and a fixed lapse towards a repeatedly happening event during a period of time. Ideally this is done by a recurring control signal, which causes a frame to be captured each time. So using a stroboscope as illumination source one will adjust the record phase of a camera so that it will be in recording mode, when the stroboscope flashes. So one equals the frame rate and the flash rate and adjusts the image capture to ensure the camera is active during the flash, not that its shutter is just closed, thus it is able to catch the rather short flash.
Sometimes cameras offer a so-called »strobe« control signal. It marks the phase of exposure of each frame; during the cameras is in exposure mode it is set.

Triggering is not as simple as it seems at all. Because the synchronously operating camera has to react on a sudden asynchronous trigger impulse after all. Therefore a capture gap can easily occur, because the camera has to finish the previous image capture before. Due to possibly already writing in the image ring memory when the trigger impulse comes high-speed cameras often have not a so-called restart capability like video cameras without image memory can offer. The latter are able to start a new frame almost in time with the trigger.

If one wants to trigger or synchronize various devices, one will face the problem of different signal levels, impulse durations and phase shifts not willing to work together.
Therefore here on the left side a very simple and cheap circuit which may be helpful for some adjusting jobs.

(Because not everywhere »Trigger« is labeled, trigger is really inside ;-)

Just in multi channel and especially in 3-D measurement applications synchronization between the cameras and their triggering in general gets extraordinary important.

Single shots of a sequence

Exercise: Generate single images of a sequence, e.g. for inserting in a Microsoft Word document or another DTP or imaging program.

If this does not work there will be another way, however, not so elegant: Just make a hard copy of the complete screen contents with [Shift] [Print Screen], insert it as image in an imaging program (e.g. Microsoft Paint) and cut/crop it.

Autologon and automatic start

Exercise: A Windows PC with or without Ethernet connection should start alone. In a special case even without connected keyboard, mouse and monitor.

Procedure: Using the autologon feature of Windows by the tool Tweak UI of the Windows installation CD (or as download of the Microsoft site as part of the Powertoys). (Attention: After that the password for the system and the network stands readable on the harddisk!)

Result: The system runs up to the Windows Desktop and with your desired program in the Autostart folder even up to your application.

(By the way: Under Windows XP and Vista the magic word is control userpasswords2. One must key it in at run in the Start menu. Concerning computers not being member of the same domain one can select user accounts without password query.)

Cleaning lenses, sensors and LC shutters

Experts at work - of course, nobody has grasped on the lens and even less one has left the lens uncovered until a regular dust hill has gathered on the top... Blow off wouldn't help jet and one has too often the impression to distribute the dirt only during a cleaning attempt at all. Not to mention the risk of scratching the comparatively sensitive anti-reflex coating, if using a dry duster.
Alcohol (isopropanole) drenched cotton pads, damp cleaning rags for spectacles or chamois leather with water thinned dishwasher or window cleaning detergents (and soft wipe paper to dry up) are much better. The ultimate tool against fingerprints and dirt on glass, however, is Opticlean Polymer e.g. from Dantronix - not quite cheap, but final (or just even use the lens hoods instead ;-).

In SpeedCam cameras, like usual for video cameras, the sensors are protected by a sometimes coated, i.e. with optical layers, glass plate. Especially in still cameras, however, the chip can be plane open. Then caution should be exercised when cleaning because of the sensitive color pattern film (polymer).

When no voltage is supplied to the LC shutter, it may show spots and blots making it look spoiled. Do not worry, the operating voltage will erase them all. You can easily check this by operating it at the camera with low frequency and you try to look through it. Or by putting alternating DC voltage (±5 V? refer to manual!) to the shutter and looking through it.) Then it completely opens and completely closes.
It is highly recommended cleaning the LC shutter with much more sensitivity than one is used to do so with glass.

Here images, info and technical data (specs) of the SpeedCam systems
as sample for the features of digital high-speed cameras:

TOUR

WP's SloMo - Tips and Tricks

Tips for shooting

Illumination and aperture (f-stop)