Analog camera on a Camera Link frame grabber: why adapters fail

You inherit an analog CCD camera and a Camera Link frame grabber, measure the connectors, and the idea writes itself: the camera's cable ends in 26 pins, the digitizer card wants 68, so build the adapter and be done. The pin count is the only thing the two links share, and that is the trap. One side speaks composite video, the other speaks a digital LVDS standard, and no harness translates between them. The sections below give you the numbers on each side of the connector, show where the adapter idea breaks, and lay out the three purchases that actually put an image on screen — with the checks that stop you ordering the wrong half twice.

Will an adapter make an analog camera work on a Camera Link frame grabber?

No — the two sides do not exchange pins, they exchange voltages: the analog camera drives one composite video signal at about 1 Vpp into 75 Ω, and every Camera Link input pair expects LVDS digital data swinging about 350 mV, so no pin map can turn one into the other.

An adapter is copper routing. It lands each conductor against a conductor on the other side and passes whatever voltage arrives unchanged; it cannot change what a voltage means. The deserializer on the grabber watches its clock pair for regular LVDS transitions and only then reads the data pairs as pixels. A continuous video waveform gives it nothing to lock onto, so the link stays dead.

Run the idea the other way round and it fails identically. A Camera Link camera pushes 350 mV differential bits at a video decoder built to find 0.3 V sync tips on a 75 Ω line, and the decoder never finds synchronisation. When an analog card of this generation reports no detectable video source, the frame grabber signal checks apply — but a standards mismatch like this one sits a layer below those checks, which assume the camera and the card already speak the same family.

What signal does an analog CCD camera send down its cable?

An analog machine-vision camera sends broadcast-style composite video: RS-170 defines 525 lines delivered as 60 interlaced fields per second (CCIR uses 625 lines at 50 fields), with the synchronisation pulses and the image brightness riding together on one coaxial line at about 1 Vpp into 75 Ω.

The arithmetic: 525 lines × 30 frames per second = 15,750 lines per second, so one line every 1 ÷ 15,750 Hz ≈ 63.5 µs, of which roughly 52.5 µs carries picture. CCIR runs 625 × 25 = 15,625 lines per second and a flat 64 µs line period. Composite means single-conductor: about 0.7 V of video stacked above roughly 0.3 V of sync.

The camera never marks where a pixel starts. The digitizer invents a sampling clock: to place 640 pixels across 52.5 µs of active video it samples at 640 ÷ 52.5 µs ≈ 12.2 MHz, and every pixel boundary is a reconstruction rather than a transmitted fact. Fields arrive interlaced — even lines, then odd lines one field later — so any moving scene carries a half-frame time skew inside each frame.

What does a Camera Link frame grabber expect to receive?

A Camera Link frame grabber expects a parallel digital pixel stream re-serialised over LVDS differential pairs: about 350 mV of swing into 100 Ω, five pairs in the base configuration carrying 28 bits — 24 of pixel data plus frame-valid, line-valid and data-valid flags — at pixel clocks up to 66 MHz.

Camera Link is the serial standard that packages parallel digital pixel buses for transmission over cable, in base, medium and full sizes that differ in how many data pairs cross the connector. Base is enough for the argument here: four data pairs and one clock pair, with two further pairs running a low-rate serial channel for camera control.

The deserializer locks to the clock pair first; only then do the other pairs carry meaning. The connector is a 26-pin MDR on most cards, with compact SDR/HDR variants on later hardware. Behind that MDR there is no video amplifier, no sync separator and no analog-to-digital converter — the board was never built to read a waveform.

Why can't a 26-to-68-pin adapter cable map one onto the other?

It cannot work, because an adapter only reroutes conductors — it cannot turn a 1 Vpp, 75 Ω composite waveform into the 350 mV differential bit stream the Camera Link deserializer needs to lock a clock and count valid pixels.

Even with both pinouts printed on the bench, the best possible mapping fails at the receiver. In the analog-to-digital direction, the deserializer sees slow analog ramps where constant-rate bit transitions belong; a 15.75 kHz line structure spread across 52.5 µs is not a clock. In the digital-to-analog direction, the video decoder receives about 350 mV of data noise — a third of the composite level it expects, and without sync tips at all.

Translation takes silicon, not copper: a sync separator and an analog-to-digital converter on one end, a serializer on the other. That combination is called a frame grabber, and no harness is one.

And once camera and grabber genuinely speak one standard, a black screen moves the hunt to conductors and terminations — the camera connected but no image wiring checks cover that layer.

Can firmware or a configuration file fix the mismatch?

No configuration file, driver setting or firmware update can bridge the gap, because the mismatch sits below everything software can reach: the deserializer never receives a valid LVDS stream to decode, so there is nothing for the configuration to tune.

Everything configuration does — pixel clock, tap geometry, flag polarity, region of interest — shapes how a grabber interprets a compatible stream that has already arrived electrically. When the hardware matches and frames still fail, the causes become specific and fixable, as in the camera configuration revision mismatch that produces no-frame errors. With mismatched standards, no diagnostic gets past the connector.

Which side should you replace: the camera or the grabber?

Replace whichever side is easier to lose, and for most teams that is the camera: an analog RS-170 camera yields about 9.2 MB/s of image data, while even base Camera Link moves roughly 198 MB/s, so the digital hardware is worth keeping.

What do the throughput numbers say?

Worked example, analog side first. A typical machine-vision digitisation of RS-170 runs 640 × 480 pixels at 30 frames per second: 640 × 480 × 30 = 9,216,000 pixels per second. At 8-bit depth that is 9,216,000 × 8 = 73.7 Mbit/s, or 9.2 MB/s. Camera Link base at its 66 MHz maximum carries 24 × 66 MHz = 1,584 Mbit/s of pixel data, which is 198 MB/s. The ratio is 1,584 ÷ 73.7 ≈ 21. Widen the analog assumption to a 768 × 494 digitisation and the payload rises to 768 × 494 × 30 ≈ 11.4 MB/s — the digital side still leads by about 17×.

Capacity, then, was never the problem; vocabulary is. The decision turns on scarcity. New analog CCD cameras have mostly disappeared from current machine-vision ranges, so replace the camera when its claim to the bench is only that it exists. Replace the grabber when the camera carries the value — a sensor behaviour your measurement depends on, or optics and calibration already tuned around it — and buy an analog digitizer card instead; selecting an analog frame grabber is the practical starting point for that hunt. The broader trade-offs are set out in analog vs digital cameras for machine vision, and the same decode-or-replace call on a compact host is worked through in an analog camera on an embedded vision controller.

What do you buy for each path?

Three purchases produce a working link: keep the camera and buy an analog frame grabber; keep the Camera Link card and buy a Camera Link camera; or modernise with a GigE or USB3 camera, which needs no frame grabber at all.

Stay with the analog camera

Order a digitizer card whose data sheet names composite, RS-170 or CCIR support and a pixel rate above the 12.2 MHz your raster needs. These cards have mostly left current product ranges, so the selection guide above is the practical route; wiring one into your acquisition software afterwards follows the same pattern as analog frame grabber integration with vision software.

Stay with Camera Link

Any camera in the area scan cameras category that lists Camera Link pairs with your card once the configuration (base, medium, full) and connector generation match — the Camera Link frame grabber compatibility checks walk that comparison, and the acquisition bring-up itself follows the same steps for Camera Link and CoaXPress, as in frame grabber image acquisition. If the existing card is gone or untrusted, the MV-GC1102IOL Camera Link frame grabber takes the host side.

After a correct match, the failure modes change shape: instead of silence you chase lost frames and timeouts, and electrical noise joins the suspect list — the frame timeout and noise diagnosis guide covers that work-up.

Replace both sides with GigE or USB3

For a new build, this route skips the frame grabber entirely. A 2.5 MP GigE camera with a global shutter, such as the MV-CU020-90GC, replaces the whole analog chain:

That sensor holds roughly eight times the pixels of a 640 × 480 analog raster and reads them out progressively, so no field skew. In the host, a dual-port PoE network interface such as the MV-GE2002P feeds several GigE cameras at once and carries power over the same cable:

If the Camera Link card itself is due for retirement, replacing an obsolete analog NTSC frame grabber with USB3 Vision maps the full swap — and the same logic applies when the analog camera is the half that has aged out.

When the inspection outgrows one GigE link

The frame-grabber route returns with CoaXPress: a CXP-6 frame grabber such as the MV-GX1008 pairs with CoaXPress cameras for rates no analog-era card ever approached.

How do you check the signal standard before ordering?

Before you order anything, read the connector and the label: a BNC or a single 75 Ω coax means composite analog, a 26-pin MDR means Camera Link, and the words RS-170, CCIR, EIA-170 or NTSC/PAL on a camera mean analog no matter what cable fits it.

• Connector first. BNC or single coax: composite analog. 26-pin MDR: Camera Link. RJ45: GigE. USB or M12 port: USB3. Connectors track their standards closely enough to shortlist on sight.
• Label second. RS-170, EIA-170, CCIR, NTSC or PAL mean analog. Camera Link, base configuration or LVDS mean digital. Sales listings often omit the interface entirely, which is how mismatched pairs get bought.
• Data sheet third. A video decoder or composite digitizer names the analog family; LVDS pair counts and pixel-clock limits name Camera Link. A sheet that names neither family is a question, not a purchase.
• Match the configuration, not just the family. Within Camera Link, base, medium and full and the connector variants must line up; the compatibility checks for resolution upgrades extend the same method to pixel counts.
• Confirm the software layer. A card is only as usable as its driver on your host; the frame grabber finder filters by interface and port count and the camera finder does the same on the sensor side.

From there the system configurator walks grabber, camera and cabling in dependency order. If you would rather have a second pair of eyes on a camera-plus-card pair before money moves, send both model numbers through the contact page.