Minor-Street Green Times Out Every Cycle with No Traffic
A retimed eight-phase NEMA TS-2 intersection runs a 100 s coordinated cycle from 6 a.m. to 7 p.m. The arterial is phases 2/6 (two through lanes each direction, 1,500 vph); the minor street is phases 4/8 (one lane each direction, 90 vph combined). After the retiming, the public reports that the minor street receives 35 s of green at 6:15 a.m. with no vehicles present, while arterial queues reach 15 vehicles. The controller is in coordination with phase 2 as the coordinated phase. The timing sheet shows phase 2 split 65 s, phase 4 split 35 s, and MAXIMUM RECALL checked on phases 2, 4, 6, and 8. Minor-street detection is one 6 ft by 6 ft loop at the stop line in each lane; arterial detection is video, recently installed and aimed low toward the east at 6 a.m.
1. What is the most likely cause of the 35 s minor-street green with no demand, and how would you confirm it at the controller rather than guessing?
Read the worked answer
Most likely cause: maximum recall on phases 4 and 8. Maximum recall forces the phase to time its full split every cycle whether or not a call exists. It is the classic explanation for a fixed, identical green on an empty approach.
Confirm on the status screens, do not guess:
- Phase status: phase 4 will show recall active and no detector call. If instead the detector call is asserted continuously, the cause is a stuck call, not recall.
- Detector status screen: a stuck loop or a mis-zoned video detector shows a call present with no vehicle and often on adjacent phases.
- Timing plan table for the active TOD plan: verify recall bits, and check the coordination settings (force off, yield, offset) for the same plan.
- Confirm the controller is actually in the plan you think it is; a plan mismatch between the intersection and the central system produces exactly this complaint.
Decision rule: recall with no call means the timing table is wrong. A call with no vehicle means the detection is wrong. Those are different repairs and different responsible parties.
2. Given the volumes, is a 35 s minor-street split justified? Recommend splits and state the check you must run before shortening it.
Read the worked answer
Capacity check. At a 100 s cycle, minor-street arrivals are 90 x 100/3600 = 2.5 vehicles per cycle. With 1,900 vphgpl saturation and about 2.0 s of startup lost time:
g = 2.0 + 2.5 x (3600/1900) = 2.0 + 4.7 = about 6.7 s
Add yellow plus red clearance of 5.5 s and the split need is roughly 12 s. The 35 s split is about three times the need and is being taken directly from the arterial, which explains the 15-vehicle queues.
Recommendation: phase 4/8 split of about 12 s with a 7 s minimum green, remove maximum recall on 4 and 8, and let the phases gap out to the coordinated phase. Phase 2 then picks up roughly 88 s.
Mandatory check before shortening: pedestrian timing. If the minor-street crosswalk is, say, 24 ft, FDW = 24/3.5 = 6.9 s and WALK + FDW = 13.9 s. The green must be at least 13.9 - 5.5 = 8.4 s, which a 12 s split satisfies. If the crosswalk is longer, or a school route uses it, the split must be driven by the pedestrian clearance, not by vehicle demand.
3. The arterial video detection was recently installed and is aimed low toward the east. What would you check, given that the complaints cluster at 6:15 a.m.?
Read the worked answer
The clustering is the clue. At 6 a.m. the camera is looking east into low sun; the image saturates and the detector drops vehicles for seconds at a time. That produces gap-out on the arterial, which hands time to the minor street and looks identical to a bad timing table.
Checks:
- View the camera image and overlay at 6 a.m. from the controller or the detection unit's diagnostic port. Confirm whether zones are live during glare.
- Verify zone geometry against the stop line and lane lines after any arm work or re-aiming.
- Review detector call history for drop-outs clustered at the same hour.
Fixes: reposition or re-aim the camera, add a sun shroud, or switch that approach to radar if the geometry cannot be corrected. Add a detector delay of about 2 s to bridge brief drop-outs, and verify the passage and extend settings. Then re-run the timing analysis; if the arterial was gapping out, correct detection alone may restore the queue, but the maximum recall on 4/8 still has to come out.