The fix for a dead PCI analog frame grabber is one device at the PC end: an external converter that takes the mux's composite NTSC output and enumerates on the host as a USB3 Vision or GigE Vision camera at 720 × 480, 29.97 fps, about 10.4 MB/s of payload. The 55 cameras, the coax and the 64:1 mux stay exactly as wired; only the card in the PC is replaced. The alternative is a current-production PCIe analog frame grabber, and it is the right pick only while your acquisition software stays locked to a card SDK. This guide gives the decision between the two paths, the numbers an NTSC channel actually delivers, a worked sweep-time example for a 55-camera mux, and the checks to run before you order. You will finish knowing what to buy, what timing to expect from the mux, and when replacing the card is the wrong move altogether.
The decision: convert once, at the PC end
Whatever sits upstream, the mux output is one composite NTSC channel. Digitize that channel once, next to the host, and hand the software a standard digital stream. Two paths do that today:
- An external converter to GigE Vision or USB3 Vision. A small box takes the BNC composite input and enumerates on the host as a standards-compliant digital camera. This is the right default here, because you are free to change software and PC hardware, it needs no slot, and no analog driver ever touches the new PC.
- A current-production PCIe analog frame grabber. A few makers still build analog capture cards with drivers for current operating systems. Choose this only while the acquisition software stays locked to a card SDK.
What never works: digitizing 55 cameras individually, or chasing a driver for the discontinued card on a new operating system. If you are not certain what the dead card was, identify a frame grabber from its part numbers before you quote a replacement. For the wider task of living with analog sources in modern software, start from analog frame grabber integration with vision software.
CoaXPress, Camera Link and GigE cards
58 models · from $189.00What one NTSC channel delivers
NTSC-M carries 525 lines per frame, 2:1 interlaced, at 29.97 frames per second. Digitized at the standard 13.5 MHz studio sampling (ITU-R BT.601), a frame is 720 × 480, and the two fields land 16.7 ms apart.
| Quantity | Value |
|---|---|
| Frame rate | 29.97 fps |
| Field rate | 59.94 Hz |
| Frame interval | 33.4 ms |
| Digitized frame | 720 × 480 |
| 8-bit payload | about 10.4 MB/s |
That payload is trivial for modern transports. A well-set 1 GbE link moves roughly 90 MB/s of image payload and USB 3.0 several times that, so even three converters sharing one GigE link, about 31 MB/s, stay inside budget. The wire is never the bottleneck here. The bottleneck is that a 64:1 mux serves one camera at a time, which sets the sweep arithmetic further down.
The channel itself is 1 V peak-to-peak composite into 75 Ω, and the digitizer provides that single termination.
Option A: an external analog-to-GigE or USB3 Vision converter
Sold as external frame grabbers, these boxes take composite in and present a GigE Vision or USB3 Vision device with a GenICam feature list. To the host, the converter is a camera: same discovery, same acquisition calls, same feature tree. Any vision package with standards support acquires it, and what GigE Vision guarantees for software compatibility carries across to USB3 Vision.
It fits this case well: no PCIe slot, so any current workstation hosts it; software freedom, since a software update was planned anyway; and a clean exit, because the converter retires together with the analog cameras while the host stays.
Three practical points:
- One converter per mux output. A second converter only helps once you split the camera group across a second mux or a second independent output.
- Deinterlace mode must be settable per acquisition: full frame for detail, single field for motion.
- USB3 versions want one real port each; a rear port on an add-in card is more dependable than a front-panel extension.
For hosts short on ports, a quad USB3.0 host adaptor gives one port per converter and room to grow:
- Interface
- USB3
- Ports
- 4 ports
The MV-GU2204 carries 4 USB3 ports on one card.
Plan the host-side cable with the converter, not from a drawer:
- Type
- Camera data cables
- Length
- 15 m
For USB3 converters, the LS-CBL-0219 is a 3 m USB3.0 camera cable in high-flex drag-chain grade. On the GigE side, the LS-CBL-0117 is a 15 m camera data cable with locking-stud RJ45 plugs.
If one viewer sees the converter and another refuses it, that is an SDK support question rather than a hardware fault: why one viewer works and another fails.
Option B: a current-production PCIe analog frame grabber
Analog capture cards are still made, mainly for exactly this installed base. They work when three things line up:
- Driver support for your exact OS build. The classic way a working card stops working is a software or OS update underneath it, covered in frame grabber driver mismatch after a software upgrade.
- Mechanics that match the PC. Some current analog cards ship only in laptop module formats and need an adapter to sit in a desktop slot; checking the full interface chain before ordering saves a return.
- Mux behavior you have tested. Cards differ in how quickly sync relocks after the input switches, and that difference lands directly in your sweep time.
For selection criteria on the analog card itself, this walkthrough of selecting an analog frame grabber for a legacy vision package still holds. If the card goes into a crowded older PC, resolving frame grabber resource conflicts is worth ten minutes up front.
Decision: take Option B only while the software cannot leave a card SDK. The moment software is being rewritten anyway, Option A wins on host freedom and on the exit path.
Worked example: sweeping 55 cameras through one 64:1 mux
Inputs, each with its source:
- Frame interval 33.4 ms, from the NTSC 29.97 fps line standard.
- Settling after a mux switch: 2 frames, about 67 ms. That is the typical relock of a digitizer sync separator and pixel clock after its input changes underneath it; treat it as a rule of thumb and measure yours once.
- Capture: 1 full frame, 33.4 ms.
- Cameras in the sweep: 55, on a 64:1 mux with inputs to spare.
Rule in words: sweep time equals camera count times the sum of settling and capture per camera.
Each camera gets a fresh frame about every 5.5 s, an update rate near 0.2 Hz per station.
Decision, with the threshold: if the process needs a complete sweep faster than about 5 s, one converter on one mux output cannot deliver it. Split the 55 cameras across two muxes, or two independent outputs if the mux has them, with two converters, and the sweep drops to about 2.8 s. Capturing single fields trims each camera to roughly 83 ms and the sweep to about 4.6 s at half the vertical resolution; take that trade only if one second matters more than the lost lines.
What interlace costs you
The two fields of a frame are one field interval apart, so a full frame is two instants glued together. On a static scene that is free vertical resolution. On anything moving, edges comb vertically, and a dimensional measurement taken on such a frame mixes both instants.
The working rules:
- Static or slow scenes: full frames, both fields, maximum vertical detail.
- Moving product: single-field capture, half the lines, at the 59.94 Hz field rate.
- Never measure dimensions on a full interlaced frame of a moving part.
If the cameras are color NTSC, expect cross-color fringes on fine edges from composite decoding. If they are 525-line mono, set the converter to mono decoding so it stops hunting for chroma that is not there. When vertical edges jitter or lines tear, suspect the digitizer line lock and sampling phase before the camera: that failure pattern is covered in analog video scanline distortion and frame grabber settings.
Planning the digital migration
The converter is a bridge, so plan the far side now. When the NTSC cameras retire, each position becomes a digital camera on GigE with PoE. The port arithmetic for 55 positions: 55 ÷ 4 = 13.75, so 14 quad-port NICs, which is more slots than one PC offers. Plan three or four hosts, each carrying a few PoE cards, or mix port counts to fit the racks you have.
Host cards that cover most of that build:
- Interface
- GigE
- Interface
- GigE
- Ports
- 2 ports
The LS-FGB-0004 is a PCIe x4, 4-port PoE Gigabit Ethernet card built around the Intel i210AT. The MV-GE2002P is a dual-port PoE NIC in full-height form where fewer ports per host are needed. At 55 cameras, port assignment and power-up discovery become their own engineering: read assigning cameras to ports reliably in a multi-camera system and camera discovery contention in multi-camera systems before committing to a topology. Camera choice can wait for the schedule: browse area scan cameras when the budget year is known.
Checks before you order
- Signal standard. NTSC-M color or 525-line mono composite; mono sources need a converter that decodes mono, not only color NTSC.
- Termination. Exactly one 75 Ω termination on the line, at the converter. A second one halves the level and none doubles it; both smear sync.
- Mux control path. Keep the switch interface, serial or GPIO, and add its command latency to the settle budget. A slow mux answer can cost more than the relock.
- Standards identity. The converter must enumerate as a GigE Vision or USB3 Vision device with a GenICam feature list. Anything driven only by a private SDK rebuilds the obsolescence you are escaping.
- Deinterlace control. Frame and field mode settable per acquisition, not fixed in firmware.
- Host ports. One port per converter, fitted before the old card is pulled.
- Coax health. Re-terminate any BNC older than the application; keep single RG-59 runs under about 100 m.
- Evaluation. Qualify one converter against your real mux and software before the old card leaves the machine.
When replacing the card is the wrong move
- Several cameras must be watched live at once. One mux output is one view, and no grabber changes that. Migrate the stations that need simultaneity to digital cameras first, and let the sweep cover the rest.
- The analog cameras are near retirement. If spares are gone and image quality is already marginal, the converter buys a bridge, not a destination. Spend on the camera-by-camera migration instead.
- The inspection needs more than 480 lines across the field of view. A converter changes the plug, not the physics: the channel carries what it carries, and no digitizer adds a line.
Quick checklist
- Converter type chosen: external GigE Vision or USB3 Vision, or a current PCIe analog card.
- Signal confirmed: NTSC-M or 525-line mono, one 75 Ω termination at the converter.
- Settle budget measured: frames counted after a real mux switch.
- Sweep time computed: cameras × 100 ms, compared with the line requirement.
- Deinterlace policy set: fields for motion, full frames for static detail.
- Host card and cables ordered with the converter, not after it.
- One converter qualified on the real mux before the old card is removed.
Pick the host card and cables in the configurator accessories step, browse frame grabbers, cables or the accessories finder, or send the mux model and your sweep-time target through the contact page and we will size the replacement path with you.
Questions
Can one USB3 Vision converter carry all 55 cameras at the same time?
No. The converter digitizes one composite input, and a 64:1 mux routes one camera to that input at a time. Watching several cameras live means switching the mux faster or splitting the group across more muxes and converters. Genuinely simultaneous views from all 55 cameras are possible only with a digital camera at each position, which is the migration this replacement defers.
How far can the coax run from the mux to the converter?
At NTSC frequencies the loss is modest: RG-59 coax loses roughly 2-3 dB per 100 m, so keep single runs under about 100 m, and move to RG-6 or an amplifier beyond that. The converter must see composite at about 1 V peak-to-peak into 75 Ω, with exactly one termination on the line. Ghosting or rolling on one channel is almost always a connector or termination fault, not the converter.
Do the 55 analog cameras need to be synchronized?
They free-run and the mux never aligns them, so frame timing belongs to the host, not the cameras. Stamp each image with the mux channel and the host clock at the moment of the switch; that is the only trustworthy timestamp in the chain. If a station truly needs exposure-synchronized images, that requirement alone justifies moving that station to a triggerable digital camera.
Will the converter work in software that only knew the old analog card?
Not through the legacy analog acquisition calls. A converter that enumerates as a GigE Vision or USB3 Vision device looks like any digital camera, so your package needs its standards-based acquisition path, which current versions of most vision tools include. Before committing, ask the converter vendor for the GenICam feature description and borrow an evaluation unit to test discovery, deinterlace control and frame rate with your mux.
Is it worth sourcing a spare of the discontinued card instead?
As a bridge, yes; as a plan, no. A spare of the discontinued card keeps you on drivers the vendor no longer rebuilds for new operating systems, and used cards fail with the same age-related modes as yours did. Keep one only to hold the line for the weeks a converter or a current-production analog card needs to be installed and qualified.