Jon Chouinard, Author at 1stVision Inc. - Machine Vision Articles

January 31, 2019March 3, 2023

Basic guidelines in selecting a machine vision camera interface

Industrial machine vision camera interfaces have continued to develop allowing cameras to transfer megapixel images at extremely high frame rates. These advancements are opening up endless applications, however each machine vision camera interface has its own pro’s and con’s.

Selecting the best digital camera interface can be done by taking in several considerations first and in doing so, can window down your selection.

The following are some considerations in making an interface selection.

Bandwidth (Resolution and frame rate)
Cable Length
Cost
Complexity

Bandwidth: This is one of the biggest factors in selecting an interface as it essentially is the size of the pipe to allow data to flow. Bandwidth can be calculated by resolution x frame rate x bit depth. You essentially find out pixels / second x the frame bit depth resulting in your total Megabits / second (Mb/sec). Large frame sizes at high speeds will require a large data pipe! If not, you’ll be bandwidth limited, so one would need to reduce the frame rate and image size or a combination of both.

Cable Length: The application will dictate the distance between the camera and industrial computer. In factory automation applications, the cameras can be located in most cases within meters from the computer vs a stadium sports analytics application requiring 100’s of meters.

Cost: Budgets dictate in most cases, so this must also be considered. Interfaces such as USB are very low cost versus a CoaxPress interface which will require a $2K frame grabber and more expensive cables.

Complexity: Not all interfaces are plug and play and require more complex configuration. If you are leaning towards interfaces using frame grabbers and have no vision experience, you may want to elect using a certified systems integrator.

The considerations above will start to dictate the interface for the machine vision application. The chart below provides an overview to help in the selection process. For a PDF of this chart, please email jonc@1stvision with subject “Interface chart”

Digital machine vision camera interfaces.

The interfaces each have pro’s and con’s aside from the designated bandwidth, cable lengths and costs, and outlined as follows:

USB2.0 is an older standard for machine vision cameras and now superseded by USB3.0 / 3.1 . Early on, this was popular allowing cameras to easily plug and play with standard USB ports. This is still a great option for lower frame rate applications and comes with low cost. Click here for USB2 cameras.

USB3.0 / 3.1 is the next revision of USB2.0 allowing high data rates, plug and play capabilities and is ratified by the AIA standards, being “USB3 Vision” compliant. This allows plug and play with 3rd party software following the GENICAM standards. Cables lengths are limited to 5 meters, but can be overcome with active and optical cables. Click here for USB3 cameras

GigE Vision was introduced in 2006 and is a widely accepted standard following GENICAM standards. This is a the most popular high bandwidth interface allowing plug and play capabilities and allowing long cable lengths. Power Over Ethernet (PoE) will allow 1 cable to be used for data and power making a simpler installation. GigE is still not was fast as USB3.0, but has benefits of 100 meter cable lengths. Click here for GigE cameras.

5 GiGe (aka N-base T) & 10GigE similar to USB2 moving to USB3, is the next iteration of the GigEVision standard providing more bandwidth. Both follow the same GigE Vision standards, but now at higher bandwidths. Specific NIC cards will be required to handle the interface. Click here for 5 GigE cameras.

CoaxPress (CXP) is a relatively new standard released in 2010, supported by GENICAM, utilizing coax cable to transmit data, trigger signals and power using one cable.. It is a scaleable interface via additional coax cables supporting up to 25Gb/s (3125MB/s) and higher now with CXP12. The interface can support extremely high bandwidth as seen in the above chart with long cable lengths to 100+ meters depending on the configuration. This interface requires a frame grabber which adds cost and some complexity in the overall setup. Click here for CoaxPress cameras

Camera link is a well established standard, dedicated machine vision standard released in 2000 allowing high speed communications between cameras and frame grabbers. It includes provisions for data, communications, camera timing and real time signaling to the camera. A frame grabber is required similar to CXP adding cost and some complexity and is limited in cable lengths to 10 meters. Longer cable lengths can be achieved with active and fiber optic cable solutions which additionally add cost. Click here for CameraLink cameras

CameraLink HS is a dedicated machine vision standard taking key aspects of CameraLink and expanding upon it with more features. This is a scaleable high speed interface with reliable data transfer and long cable lengths up to 300+ meters with low cost fiber connections. Similar to CXP and camera link a frame grabber is required adding cost. Click here for Cameralink HS cameras

Some caveats: In calculating bandwidth, you can calculate the theoretical data rate, but in some interfaces, the real world practical will be different. In Camera link and CoaxPress, the theoretical and practical are the same. In Cameralink HS, limits will be set by the computer interface (i.e PCIe x 8, Gen 3 is 6.8 Gigabyte / sec and an Xtium CHLS frame grabber can capture 7 Gigabyte / sec!)

The theoretical and practical limits for USB and Ethernet can be quite different and there will always be some difference. For example, large frames and low frame rate generates less interrupts, providing less overhead to the CPU. A small frame with high frame rate generates more interrupts causing more load to the CPU.

As a note: This blog post covers the basics of each of the interfaces. There is much more information 1stVision can share with you to be sure you are taking all aspects of the vision application into consideration. We have several additional resources we can share to help, so don’t hesitate to contact us for free consultation!

1st Vision’s sales engineers have over 100 years of combined experience to assist in your camera selection. With a large portfolio of lenses, cables, NIC card and industrial computers, we can provide a full vision solution!

Ph: 978-474-0044 / info@1stvision.com / www.1stvision.com

Related Blogs

Learn about the new 5GigE camera interface

Learn about CoaXpress and frame rates achievable with Allied Vision’s “Bonito Pro” cameras with CXP-6

What Is Offered in an Industrial PC Machine Vision Computer?

January 28, 2019March 3, 2023

What are the f-numbers on machine vision camera lenses? f-stop explained!

Why does 1stVision focus (no pun intended) so much on machine vision lenses. As the old saying goes, if you have garbage in, you get garbage out.

The lens is the input to the machine vision system. A low quality lens means that you have already degraded the image coming into the sensor. For instance, let’s say you chose a camera with 5um pixels, which equates to a lens being able to resolve 100 lp/mm. If your lens’ Modular Transform Function (MTF) is only 50 lp/mm, you should have chosen a camera with 10um pixel size, because the lens can’t do any better than that. As a note, don’t infer that a camera with 10um pixels is worse than a camera with 5umpixels from this example, as that is not true. Learn more on MTF here

A machine vision lens gathers light and then focuses it. When we talk about focus, we are talking about the MTF, but when we discuss light gathering properties, we need to discuss the lens f-number.

FUJI -f-stop — FUJI lens showing f-stops

f-number
The f-number is defined as the ratio of the focal length by the aperture width (diameter of the entrance pupil). So a 50mm focal length lens with a f-number of 2 has a 25mm entrance pupil. The lower the f-number, the more light will be allowed into the system, however this equates to more expensive lens as you need more glass to make a wider entrance pupil.

f-stop
Many camera lenses have an adjustable iris that opens and closes at the front of the lens to limit the amount of light coming in. When open all the way, the f-stop is the f-number. From there, each f-stop from wide open halves the amount of light, which corresponds to reducing the size of the aperture by 1/sqrt(2) or about 0.707 and in turn halving the area.

The f-stop is represented by a sequence of these numbers below, each letting in half the light.

Sequence: f/1, f/1.4, f/2, f/2.8, f/4, f/5.6, f/8, f/11, f/16, f/22, f/32, f/45, f/64, f/90, f/128

The sequence is obtained by approximating the geometric sequence

Characteristics of the f-stop

Most lenses are designed to be optimal in the F4-F5.6 range, in which they have the best MTF.
The higher f-number (ie f/8 ) is, or the more closed the aperture is, better the depth of field if achieved
The lower the f-number (ie f/1.4) is, or the aperature being wide open is where you get the least depth of field, but not great MTF.

In a practical application, you need to trade off exposure time, depth of field, and available machine vision lighting. These three variables are always in tension. If you need fast exposure AND depth of field this means very small amounts of light gets to the sensor. If you need high contrast images in this situation, something has to change. Either get more light, accept less depth of field, or have some image blur.

For a full listing of machine vision lenses, click here and use the filter to help in your selection.

Related Blog posts

Not All Lenses are Created Equal! Lens MTF Comparisons

10 key things you need to know when selecting a machine vision lens

Demystifying Lens performance specifications

Information courteous of Wikipedia

December 10, 2018December 17, 2018

Learn about the new 5GigE camera interface

Machine vision interfaces have continued to evolve over the years increasing data throughput and cable lengths. Commonly used interfaces are GigE and USB3. However, 5GigE is an interface now gaining attention in the industrial imaging / machine vision market with some nice advantages.

We will outline the benefits of 5GigE, but first, lets give a brief overview of the commonly used camera interfaces, with their pluses and minuses:

GigE / GigE Vision

110 MB/s of sustainable throughput. In real world terms, a HD, 2MP camera can get 50-55 fps in 8 bit mono or 8 bit color mode. Note, this isn’t real HD, since you need 60 FPS.
Data cable lengths up to 100m using regular CAT 5e/6 cable.
Easy to put multiple cameras on a system.

USB 3 / USB3 Vision

420 MB/s of data throughput. A HD 2MP camera can run 60 fps in 8 bit mono or color and can also run RGB at 60 FPS no problem. With the higher throughput, a 5MP camera can achieve 85 fps in 8 bit mode.
Data cables up to 5 meters and up to 20 meters with active cables. However, active cables can be quite costly, adding up to $200 in cost.
Not as easy as GigE to put multiple cameras on a system, and gets harder with each additional camera, especially if you have limited USB3 controllers.

As a note, there is no cost difference when using cameras with the same sensor from the same manufacturer with USB or GigE! They will cost about the same with no premium for one interface over the other.

What are the limitations of GigE and USB3 now solved by 5GigE?

USB3 is limited in cable length, so going faster than GigE is great, but you can not have long cable lengths.
GigE has cable lengths up to 100 meters, but is limited to ~ 110MB/s of data, so you do not have the high frame rates as in a USB3 camera.
USB3 in 4+ camera systems is not as stable as GigE AND you’re still limited on cable lengths.

Wait! – What about 10GigE?

Up until now, 10G was the next interface. However, the jump to 10G has quite a few limitations as outlined below.

Heat generation is significant, so cameras are large and not in the smaller 29 x 29mm cube form factor.
Not a lot of demand for very high speed 10G, so not a lot of sensors being offered
Minimal number of manufacturers for 10G, higher cost.
Special cabling, either optical or high quality cat 7.

What we have found is that there are several types of applications for 10G cameras and are as follows

Applications where you need 10G of speed of course (high resolution + fast frame rates)
Require greater than 110MB/s of data and need long cable lengths.
Where there is the required combination of 110MB/s for high frame rates, multiple cameras and long cable lengths, 10G is a perfect solution.

We have seen that the need for higher bandwidth + long cable lengths is more prominent vs. the real need for 10GigE!

Introducing 5GigE that provides increased bandwidth, long cable lengths at reasonable prices! or N Base T.

5GigE (also known as N Base T) has become a new standard for industrial, machine vision cameras.

In the general compute world, a much much larger market than vision, there has also been a need to go faster than GigE. However, the issue of replacing the existing cabling is the major issue preventing this. If you think of a big box store, say a Home Depot for instance, the amount of cabling is huge. Ripping that out and rewiring far exceeds the cost of the equipment to use it!

5G was made to go faster, but use existing cabling. Regular cat6e cable can be used, and 5G is a subset of 10G, so all switches etc. can be kept in service.

5G gives users in the vision market USB3 speeds, but with ALL of the regular GigE features, at a very small premium!

December 10, 2018December 19, 2018

Get a 5MP lens for the price of a 3MP one! FUJI’s new XA-5M lenses.

Fujinon introduced their first 5MP series, the 5MP SA-1s over a decade ago. Back then they were the first really high quality lenses for under $500.

Now, they have introduced their next generation 5MP series, the new XA-5M , and they have made them smaller, better, and less expensive! Now you can get a 5MP rated lens for very close to what 2 or 3MP lens cost with the new HF-XA-5M series lenses! 1stVision has preferred pricing and stock! Contact us for a quote

The Fujinon XA-5M are designed to work at 3.45um resolution (and very good at 2.5um), so they are a great choice for the 3.45um pixels found in the Sony Pregius IMX line of sensors. See image below showing a heat map of the resolution.(Darker red is higher resolution, 2.7um, orange is 3.3, yellow is 4um, light blue is 5um, and dark blue is 6um)

Further, the new CMOS sensors have more of a ‘stacked’ architecture, meaning each pixel is tall. Without having a lens that has a small chief ray angle, which keeps the light rays as close to perpendicular as possible to the sensor, each pixel possibly shades its neighbor. The new Fujinon lenses are designed to solve that problem for these sensors. So when comparing this line to their competitors, there is less shading at the edges and more even illumination!

The results can be seen when comparing the center to edges in an application. In the example below, the text in the center and edge on the FUJI “4D High Resolution” lenses is crisp vs the competition on the right.

Finally, Fujinon has dramatically reduced the size of the lens, all the lenses are 29.5mm in diameter except the 6mm, which is 39.5mm. Now you can get a 5MP rated lens for very close to what 2 or 3MP lens cost with the new HF-XA-5M series lenses! 1stVision has preferred pricing and in stock! Contact us for a quote

Watch this video for more details on how the HF-XA-5M lenses compare to the competition

Related Videos – FUJI is now incorporating anti-shock and vibration into their lens series!

Related Posts

Not All Lenses are Created Equal! Lens MTF Comparisons

There is NO such thing as a “Megapixel” machine vision camera lens!.. Say what??