Bit Error Test Takes Guesswork Out of Camera Link® Cables

Sometimes, ‘good’ still isn’t ‘good enough.’ 

That was the summation of Components Express (Woodridge, IL) president Ray Berst after an OEM client came to him asking for samples of his Camera Link^® cables for testing with a high-speed 80-MHz camera.

The bad news for Components Express was that almost all of their cables failed, despite being sampled on a regular basis and tested against the industry standard Eye Pattern tests for operation up to the highest speed covered by the Camera Link^® standard: 85 MHz.

The worst news for the machine vision industry was that the OEM had tested cables at 85 MHz from all the major manufacturers and they’d fallen before a bit-error test like sheaves of wheat before the harvester. The OEM wasn’t happy, and neither was Ray Berst.

‘‘The OEM came back to us, told us what was up, and said some of our cables had worked, while others hadn’t,’‘ Berst explained. ‘‘They wanted to know what was unique about those that worked, and why.’‘

That was the beginning of a journey that started with a conundrum and has led to a new way to production test and certify 100% of Camera Link^® cables, while possibly extending the performance of these cables and providing a path for even higher performance.

Not Just a Cable Problem
‘‘We’ve had customers bring back an entire system because something wasn’t working right, and it turns out to be the cable,’‘ explains Reynold Dodson, President and Director of Software Development at frame grabber manufacturer BitFlow  (Woburn, MA). ‘‘One day I heard a support technician suggest turning the cable around and it worked. It was a colossal waste of time, and we really wanted those problems to go away.’‘

Frame grabber makers regularly use bit-error tests to check their products rather than Eye Pattern Tests. Eye Pattern tests use an oscilloscope and standard electrical signal pattern to check a Camera Link^® cable design, however, the tests takes approximately 2.5 hours per cable, so the method isn’t used in production for 100% quality verification. Eye Pattern tests also only check one pair of conductors per cable at a time, and so aren’t affected by cross-talk or other potential interference generated by energizing all five pairs simultaneously in a Camera Link^® cable.

‘‘Eye diagrams are the traditional way of characterizing electrical performance, and they’re great for analyzing a problem with a cable and for defining a specification, but they’re not fast enough for production,’‘ Berst says.

Bit error tests mimic the way a cable, camera or frame grabber will operate in the ‘real world’ by using a signal generator or camera emulator to send a known image across a cable, and then check the received data against what was sent.

‘‘In the past, we’d send the cable to JAI for Eye Pattern tests, but we figured OEMs are using bit error tests when they check a cable before putting it into one of their pieces of equipment, so let’s mirror what they’re doing,’‘ explains Components Express’ Berst.

Taking the Fall 
Today, high-speed cameras operate at 100 MHz or beyond, past the 85 MHz limit of Camera Link^®. For Berst and Dodson, the question wasn’t just how to validate the existing Camera Link^® cables to make sure they met customer needs, but how to provide a way to extend Camera Link^® speeds to levels system designers were already asking for.
 
‘‘We had two major problems: some Camera Link^® cables weren’t working even at slow speeds, while National Semiconductor was releasing a 112 MHz channel link chipset,’‘ Berst explains. ‘‘We needed a solution that could verify the slower speeds, and help us reach new speeds with Camera Link^® . The bit error test was the solution.’‘

Berst contacted BitFlow’s Dodson and Camera Link^® specialist David Johnson at Vivid Engineering  (Shrewsbury, MA). Dodson volunteered an R64e-CL PCI Express, 8-tap frame grabber, while Johnson contributed a programmable camera emulator that could generate 256x256 images at 64-bits. Connecting the two was a Components Express Camera Link^® cable. Wrapping up the package was a segment of the bit-error test routine used by BitFlow to analyze their frame grabbers. ‘‘We gave it to him at no cost because we wanted a solution as bad as they did,’‘ said BitFlow’s Dodson. ‘‘It was a 90% solution for them, and we were happy to help.’‘

‘‘I figured the best thing to do was throw myself on the sword and show the bit-error test using one of my cables to the Automated Imaging Association’s (AIA) Cable committee,’‘ Berst concluded.

Check the Label
AIA’s Cable committee formed a Camera Link^® sub-committee to study the issue with input from makers of all major machine vision component types. For the test to be valid, it would take more than a Components Express solution, it would take an industry-wide agreement, and AIA’s Cable committee was the place to start.

Market Intelligence News & Insights: 

How Will Apple’s $500 Billion Investment for US Manufacturing, Education, and AI Play Out?

On February 24^th, Apple announced a “commitment… to spend and invest more than $500 billion in the U.S. over the next four years”, with plans for new manufacturing plants, enhanced support for manufacturing partners via their Advanced Manufacturing Fund, and a myriad of other initiatives. As part of this plan, Apple projects 20,000 new hires, with roles focused on R&D, silicon engineering, software development, and AI/ML engineering.

Read More

Berst and Dodson – with assistance from AIA’s Director of Standards Development, Jeff Fryman - made their pitch to the receptive Camera Link^® subcommittee at the recent VISION show held in Stuttgart, Germany in November. ‘‘Now the subcommittee is gathering data and hopes to rewrite the Eye Pattern test in Appendix D of the Camera Link^® standard based on the data from the bit error tests,’‘ Berst says. 

If approved by the AIA committees, cable manufacturers will still submit their cable designs to third party labs for verification before achieving AIA Camera Link^® certification using the new Eye Diagrams. In addition, manufacturers would have the option of testing each cable using the bit error test and a standard image. Each cable could then be labeled with a ‘‘minimum speed’‘ in both directions at the standard Camera Link^® clock speed offerings of 40, 66, and 85 MHz. Production testing and labeling would be at the discretion of each individual manufacturer.
 
Bit Error Test Bed 
Although Components Express used a BitFlow frame grabber and Vivid Engineering emulator, the new Camera Link^® test could use any emulator and frame grabber that meets the Camera Link^® specification. To make it easier on cable manufacturers and promote the new standard to the betterment of the vision industry, BitFlow’s Dodson says he is happy to make the same bit error test routine software available to anyone that needs it. Berst’s Components Express team has put a graphic user interface on the front-end to make it easier to use.

In operation, the bit error test routine triggers the emulator which sends the 256x256, 64-bit test image from the emulator, across the cable under test, to the frame grabber, which checks the received image against the stored test image. Testing begins at 88 MHz, and if one bit is lost, the test is repeated at 69 MHz, and if one bit is lost again, it is tested once more at 43 MHz. Each test level is 3 MHz higher than the Camera Link^® standard speeds of 85, 66, and 40 MHz, to give manufactures a buffer so that they know the cable doesn’t just meet the standard. The subcommittee is considering changing the 3 MHz to a percentage of the base speed so that the buffer increases as speeds increase. A full system will require a 10-tap frame grabber, Berst adds. Today, several frame grabber manufacturers offer 10-tap frame grabbers, including BitFlow’s new Karbon-CL model. An important consideration is to select a frame grabber that will easily change from one Camera Link^® speed to the next.

At the subcommittee’s next meeting in June, the group is expected to consider recommending the test method for formal adoption, including a rewrite of Appendix D to include new Eye Diagrams for testing based on bit error test results. The voluntary labeling method could also be confirmed.

‘‘Testing each cable could really help everyone in the vision industry by proving that a particular cable with a specific length can operate in both directions at a minimum speed.  Not every one needs a 12-meter Camera Link^® cable that operates at 85 MHz. For many customers, 40 MHz is fine,’‘ concludes Components Express’ Berst. ‘‘At the same time, camera manufacturers are making cameras that go to 100 MHz and higher. The same cable that makes that application possible is also a cable that will give a robotic vision customer operating at 85 MHz longer intervals between cable replacement as cable flexing inevitably reduces performance. This could help keep Camera Link^® relevant as a standard longer into the future.’‘

Bitmap image of BAD 10 Meter CL @ 85 MHz
 

Bitmap image of GOOD 10 Meter CL @ 85 MHz

Example of binary file comparison between a good 10 meter V.S. bad 10 meter.

Comparing files 10M.bmp and 10Mg.BMP
000004D6: B7 97
000004DA: BB 9B
000004EA: BB AB
000004EB: BC AC
000005D3: B3 93
000005D7: B7 97
000005DB: BB 9B
000005EB: BB AB
000006DC: BB 9B
000006EC: BB AB
000006ED: BC AC
000007D5: B3 93
000007D9: B7 97
000007DD: BB 9B
000007ED: BB AB
000008DE: BB 9B
000008EE: BB AB
000008EF: BC AC
000009D7: B3 93
000009DB: B7 97
000009DF: BB 9B
000009EF: BB AB
00000ADC: B7 97
00000AE0: BB 9B
00000AF0: BB AB
00000AF1: BC AC
00000BD9: B3 93
                

Bad 10 Meter Row          Good 10 Meter Row
(binary difference)            (Pass base)

Courtesy of Components Express, Inc.