FANUC iRVision Setup: Camera Calibration, Application Frame, and Vision Process That Repeats Across a Shift

A fresh iRVision install is a calibration cascade, not a vision system. The camera calibration feeds the Application Frame. The Application Frame feeds the vision process. The vision process feeds the motion program. Get any layer wrong and the offsets land your TCP a centimeter off the part, or worse, you start chasing intermittent MOTN-074 alarms because the calibration grid was the wrong size. At Probot Systems we deploy iRVision on R-30iB and R-30iB Plus cells regularly, and the same handful of mistakes account for most of the calls we get from plants doing their first vision job.

This post is written for integrators and technicians setting up 2D Single-View iRVision on an R-30iB or R-30iB Plus controller, typically with a SC130 or Basler GigE camera over the JRS27 connector. The principles apply to 2.5D and 3DL as well, but the calibration steps and traps are different at the upper end.

How iRVision actually works mechanically

iRVision builds three coordinate transforms in sequence. First, camera calibration maps pixel coordinates to a flat plane in millimeters, using a printed grid placed at the working distance. Second, Application Frame ties that calibrated plane to a known robot frame, typically taught against a fixture point with a sharp TCP. Third, the vision process locates a feature in the calibrated frame and outputs a position offset that a motion program reads through VOFFSET, VR[x].

If any one of those three is set against the wrong reference, every downstream layer is broken, even though each individual layer reports success. The vision process will happily report a perfect score against a part that is sitting 8mm off because the application frame was taught with the wrong tool active.

When a calibration step fails, the controller often surfaces it as MOTN-074. The official FANUC alarm code manual entry for MOTN-074 reads: “Speed is not within 0 to $speedlim. Remedy: Set speed within 0 to $speedlim.” That looks like a motion-program error, but on the Education Cert Cart and similar cells, MOTN-074 actually fires during the calibration motion when the robot cannot reach the next calibration point at the requested speed. The Education Cert Cart 2D iRvision camera calibration fail MOTN-074 thread documents this exact symptom.

Most common pitfalls

1. Wrong calibration grid for the focal length. FANUC supplies grids in several sizes (15mm, 30mm, 50mm dot spacing). Using the wrong one at your working distance under-fills the image and calibration drifts. The Fanuc irVision part identifying thread mentions sizing mismatches as a recurring cause of poor part-find scores.
2. Application Frame taught with the wrong UTOOL active. UTOOL is the source of truth for where the TCP is. Touch the calibration fixture with the wrong tool active and the Application Frame inherits the wrong tool offset, propagating into every part find for the life of the program.
3. Lighting unstable across a shift. Vision processes tuned at 10am with the skylight bleeding light fail by 3pm when the sun shifts. The Fanuc Project Help pls thread mentions iRVision as a use-case that needs controlled lighting before the threshold values mean anything.
4. Camera not calibrated against a level surface. A grid sheet curling on a wood pallet is enough to throw the calibration off by a couple of millimeters. The grid needs to be flat against the working plane, period.
5. Vision data missing from the backup. Vision setup lives in .VD files that some image backups skip. After a camera replacement or controller swap you find out the hard way. The iRvision mastering issue on Education Cert. Cart thread describes recovery effort after a controller change.
6. Wrong camera pinout on JRS27. The 20-pin Honda connector on R-30iB Mate has a specific pinout for Basler acA1300-75gm and other GigE cameras. The Pinout for FANUC R-30iB Mate JRS27 to Basler GigE Camera thread walks through this. A wrong pin and the camera enumerates but image quality is unusable.

How to set this up correctly

Step 1. Confirm camera type and working distance. Write down focal length, working distance, and field of view at the part plane. Match those numbers against FANUC’s recommended grid size in the iRVision setup manual. If the field of view is 200mm wide, you want the 30mm grid (around 7 dots across) not the 15mm grid (would have 13 dots and dots become noisy at that density).

Step 2. Mount the camera rigidly. iRVision calibration is invalidated the second the camera moves relative to the robot base, even by half a millimeter. If the camera is robot-mounted on the wrist, that is the calibration; if it is fixed in space, lock it down hard.

Step 3. Teach UTOOL precisely with the tip-touch method. Do not use a “close enough” UTOOL for vision work. The Application Frame layer multiplies any UTOOL error.

Step 4. Run Camera Calibration. The vision setup screen walks the robot to multiple positions across the field of view. Watch for any MOTN-074 in the alarm log. If you see one, the issue is usually that your taught reference position is too close to the singularity or the calibration motion crosses an outer envelope limit. Reposition the grid.

Step 5. Teach Application Frame against the same fixture you used for the calibration grid. Use the same UTOOL. The fixture should have a hard scribe mark or a precision-drilled hole that you can revisit later for verification.

Step 6. Build the vision process. Lock lighting first, tune thresholds against multiple sample parts, save, and re-run the same parts to confirm pass rate. If part-find score swings more than 10 points between identical parts under stable lighting, the process is too sensitive.

How to do it right when the cell is already running

If part-find is drifting through the shift: Block the camera from ambient light. Add an LED ring or a dedicated illuminator with a controlled brightness. Re-tune the threshold against the controlled lighting only.

If part position lands a measurable offset off the target: Re-verify Application Frame against the fixture scribe mark. The fastest way to spot a wrong Application Frame is to write a one-line TP program that goes to UFRAME pos 0,0,0 and see whether the TCP touches the scribe mark.

If MOTN-074 fires during calibration: Move the calibration grid so the robot reaches the calibration points well inside its motion envelope. The Education Cert Cart problem mentioned in the forum thread is a textbook case of the cart geometry forcing the robot to reach near its limits.

If you are swapping cameras: Take an AOA backup with the explicit “include vision files” option checked. Save the .VD files separately to a USB stick. The Delete IR vision files thread highlights how easy it is to lose vision configuration permanently if it is not separately preserved.

If frame errors appear when running a vision-aware program: The Fanuc frame error using vision thread is a reminder that not every frame error is iRVision related. Check whether the program references UFRAME[0] (world) while the vision process outputs an offset against UFRAME[5], for example.

When to call a specialist

If you are setting up iRVision for the first time, the lab learning curve is real. A specialist will pay for themselves in commissioning days saved, especially because vision-related rework after a cell is in production is much more disruptive than getting it right the first time.

Two signals that you should bring in help: you have done the cascade three times and the offset is still wrong by an unexplained amount, or you are dealing with a 2.5D / 3DL setup where the calibration involves multiple depths and reference frames. contact us for vision commissioning, or set up a maintenance preventive contract that includes annual camera calibration verification.

Related errors to check

• MOTN-074 Error in speed: as described, often surfaces during vision calibration when the robot cannot reach a calibration point inside speed limits.
• INTP-251 Vision process error: vision process failed to run, usually a stale .VD reference.
• SRVO-068 DTERR alarm: not directly iRVision, but appears when a camera-mounted EOAT moves at high speed and the camera cable pulls.
• MOTN-017 Limit error: appears if the calibration motion would push the robot past a soft limit.

Probot Systems is a FANUC integrator based in Lévis, Quebec. We deploy iRVision on R-30iB and R-30iB Plus cells across Canada and the US, including pick-and-place, depalletizing, and machine tending. If your vision job is drifting through the shift or you are scoping a new vision cell, that is a contact us conversation.