The Truth About High End Cameras
A number of more expensive cameras can be fitted to FTA instruments. The inducements for alternative cameras include
higher speed, up to a practical limit of 10,000 frames per second.
higher pixel count (mega-pixel)
alternative electrical and mechanical interfaces, such as USB
There are trade-offs, of course, as no choices are free of side-effects. The purpose of this note is to explain some of these.
High Speed Cameras
Current camera sensor technology can provide reasonable sized (say 320 x 240 pixels) images up to 10,000 frames per second. This is 100 microseconds per frame. All high speed cameras provide their highest frame rates by limiting image size, because internally they are limited to some total number of pixels per second. Full size frames are possible up to about 1000 frames per second.
High frame rates require very intense illumination because of the short exposure time for each frame and, also, because we are working with optical magnification that necessarily reduces light intensity. 10,000 frames per second will require a 150 or 300 watt illuminator with fiber optic coupling. An infrared filter may be required to reduce heat on the sample.
High speed cameras are expensive. The end-user cost of the best cameras can easily be $100,000. That is in addition to the cost of the instrument. The minimum cost for cameras producing 1000 frames per second is $10,000. Anything slower than 1000 frames per second is merely a "fast" camera. The top end frame rate of these is about 500 frames per second. The end-user cost drops rapidly as you dip below 1000 frames per second, down to the $3000 to $5000 range.
Cameras with more than the NTSC, PAL, or VGA pixel count (all nominally 640 x 480) are classified as megapixel cameras. Potentially these provide more spatial resolution, which would be useful. However, this resolution is often traded for color, so you wind up with a camera with VGA monochrome resolution (true 640 x 480), but with color pixels. Note these color cameras are advertised with higher pixel counts, but the pixels are used in sets of four to make monochrome pixels, so the apparent gain in spatial resolution is not there even though there is a gain in color resolution. See the discussion which follows on color.
Megapixel cameras do offer electronic zoom and pan, which is useful. A monochrome megapixel camera does offer greater spatial resolution and this is powerful when combined with zoom and pan. The FTA ArtCAM USB 2.0 camera is an example of such a true monochrome camera.
In most cases, color is good when taking pictures of people and nature and is not good in the laboratory. Excuse the generalization. But software such as drop shape analysis works on finding edges (a glorified ruler) and this is completely different from pleasing color representation.
Color is achieved by using color dyes over quads of pixels. Each group of four, sometimes called a 2x2 bin, has two green, one red and one blue filtered pixel. This is called Bayer or raw color coding.
Bayer coding makes for very pretty images. It was developed at Kodak for just this purpose. And it is factual to say a megapixel color camera has, say, 1280 x 960 pixels. But it is misleading to suggest these give you spatial resolution of 1280 x 960. Because they are binned 2 x 2, they give you the spatial resolution of 640 x 480 and very fine, very good, color resolution. They make very pretty pictures. They do not find edges any better.
USB, IEEE-1394 Firewire, Camera Link and 100BaseT Ethernet Cameras
The title lists four different ways beyond classical frame grabbers by which images can be transferred to a computer in more or less real time. None of these transfer images at "high speed" rates of 1000 frames and above. High speed images are captured in memory local to the camera and then slowly sent across to the host computer.
None of these interfaces makes installation and integration of a new camera painless. These interfaces specify electronic and mechanical aspects of the interface, but not high level software aspects (indeed, there is no way they could). They do offer varying degrees of convenience compared to frame grabbers.
USB and Ethernet connections are built-in to new computers, so no card has to be added in these cases. IEEE-1394 Firewire and Camera Link interfaces require adding a card to the computer, so they are much like frame grabbers. They do offer the advantage of better cross-platform support (think, for example Linux) than do frame grabbers.
USB cameras do offer advantages for laptop users who want portable operation, since the camera draws power through the USB connector.
USB and Firewire interfaces are only marginally faster than frame grabbers. It turns out that it is difficult to beat an analog signal on a coax cable for bandwidth. USB and Firewire interfaces are chosen not for speed but for convenience and portability.
Ethernet speed is only similar to frame grabbers. Ethernet is chosen for universality and the ability to move images long distance. It is probably the interface of the future for instruments. Wireless, the only other new competitor, is great for operating your laptop at a cafe, but offers nothing to the instrument designer.
Camera Link can be thought of as a super frame grabber. Or a next generation frame grabber. Irrespective of who makes it, a Camera Link frame grabber is supposed to offer standardized interfaces. Camera Link provides the fastest real time frame rates into the computer, remembering that "high speed" cameras have local memories and dribble their images across to the host computer. Camera Link cameras can get into the 1000 frame per second class, so there is a performance advantage over USB, Firewire, and legacy frame grabbers.
Monochrome Image Quality
First, notice we are speaking of monochrome (black and white) image quality. This is what you need for finding edges (determining shapes). While the other cameras may offer features and capabilities that you like, or feel necessary, none will produce images superior in quality to legacy frame grabbers. Sometimes the degradation is slight, sometimes it is significant, but do not be under the illusion the alternative camera monochrome image is superior to a frame grabber for edges.
If you put aside USB 1.1 cameras, used mainly for portable applications, all of the options described above are more expensive, and some much more expensive, than legacy frame grabbers.
All cameras require some level of special software to bring their images into the FTA program for analysis and instrument control. There is no "interface-less" camera, just like there is no "free lunch." FTA provides the following interfaces as part of its standard product line. These are already developed and paid for.
MuTech and Leutron Vision (Lutrex) legacy frame grabbers for the PCI bus. These support NTSC and PAL cameras, of which there are many. Color cameras are treated as monochrome, without any consideration of Bayer coding as the incoming signal is analog from the camera. Full size images are nominally 640 x 480.
Windows "Direct Show" cameras, typically used by USB 1.1, but also supported by some IEEE-1394 Firewire cameras with their own intermediate software drivers.
ArtCAM USB 2.0 megapixel cameras, both monochrome and color. These monochrome cameras are true monochrome, so the full pixel resolution is available for edges.
JAI M30 fast camera, capable of 360 reduced-size frames per second.
Some of the standard FTA camera options carry a premium cost, but this is for the camera and its mechanics, not for special software development.
Some of the optional non-standard cameras, particularly the high speed cameras, are quite expensive. We are quite up-front that we will mark up any special camera we buy for you. Such cameras are not part of the standard instrument cost. This mark up is to pay for two significant costs: the cost of software integration and the marketing cost of this non-standard camera. Both of these are tangible and quantifiable. We can not absorb these into the cost of the standard instrument; to do so would be unfair to buyers of the standard instrument as it raises their price.
The end-user, retail cost of special cameras will be approximately twice what the camera cost us. You can consider this the "integration cost" of the camera. For this you get an integrated, working package with a guarantee.
You may also choose to buy the camera separately, or maybe you already own the camera. FTA will integrate your camera into our system for a flat fee which is typically $5000. If we can not do this, or you are unsatisfied with the result, there is no charge for the integration and the instrument is run with its standard camera. Many special cameras have "features" which prevent them from operating satisfactorily within an instrument. FTA can not guarantee the suitability of any of these cameras. We have our own war-stories to tell in this regard. All FTA can do is promise its best efforts and there will be no charge if the interface is not accepted by you the customer (of course, you do not keep it if you reject it). FTA must also have access to the camera for several weeks to test the software interface. We have interfaced six camera/frame grabber systems over the last several years and only one of the six worked as advertised straight out of the box. All others required communication with the supplier's factory and work-arounds or fixes. This is just normal, but it says you can not design and implement these systems "on paper" and expect them to work without substantial debugging. This is what costs money.
All of this notwithstanding, you may find it economical to purchase the camera yourself and pay us to integrate it for a flat fee.
There is a final way, and that is for you to install your camera and use external files it may produce, say AVI files, to transfer images into the FTA software. You must work out how to provide enough real time video that you can control the instrument. FTA will be helpful in such efforts, but can offer no guarantee the results will be successful.