EV Autoplay logo EV AutoplayElectric driving, plainly explained
Driver Assist

Automatic High Beam Assist and Why It Dazzles Drivers

Driving home on a two lane road north of Trondheim last November, I got flashed three times in eleven minutes. Not by police. By oncoming drivers telling me my high beams were in their eyes.

An electric car on an unlit rural road at night with its high beams lit and an oncoming car appearing over a hill crest

Driving home on a two lane road north of Trondheim last November, I got flashed three times in eleven minutes. Not by police. By oncoming drivers telling me my high beams were in their eyes. The car was a 2025 model with automatic high beam assist switched on, and the system was working exactly as its engineers intended, which is the part owners find hardest to accept.

Auto high beam promises something simple. Full brightness when the road ahead is empty, dipped light the moment it is not, and no stalk work from you. What actually happens is that a camera behind the windshield makes a judgement call several times a second, and every judgement carries a delay. I logged 140 dipping events over six weeks across three electric cars, timing each one against a dash cam clock, because chasing intermittent failures was my job for years before I started reviewing car screens. The delay was never zero. On some roads it was long enough to put real light into somebody else's eyes.

What the camera is looking for, and what it ignores

The sensor is almost always a single forward camera mounted high on the windshield, usually the same unit that feeds lane centering and forward collision warning. It does not measure brightness the way a light meter does. It hunts for patterns: a pair of white point sources sitting roughly level and growing, which reads as an oncoming car; a dimmer red pair moving away at a similar speed, which reads as a car you are following; a regular rhythm of warm points overhead, which reads as street lighting.

Manufacturer specifications usually claim detection of oncoming headlights at around 600 metres and of taillights ahead at around 400 metres. Those figures come from straight, flat test tracks. On a crowned rural road with hedges, my measured first detection of an oncoming car sat closer to 320 metres, and on a left hand bend it dropped under 150.

There are also gates that surprise people. Most systems refuse to raise the beams below roughly 40 km/h, and several will not activate at all above a modest ambient light level, which is why the function sits idle on a lit ring road and then wakes up the instant you turn onto an unlit lane. A motorcycle with one headlight breaks the paired pattern the camera wants. So does a cyclist with a single blinking LED, a tractor with a work lamp on the cab, and a horse rider holding a torch.

Where the delay comes from

This is the part no brochure explains, and it is the reason the feature gets a bad reputation. The camera runs at about 30 frames per second, so one frame is 33 milliseconds. The software will not act on a single frame, because a wet reflective sign or a puddle flaring under your own beams would trigger constant flicker. It waits for confirmation across several frames.

Breaking down what I timed, the chain looks like this. Confirmation across three to five frames costs 100 to 170 ms. A smoothing filter that stops the beams from hunting adds another 200 to 400 ms. Then the lights themselves have to change, and that is not instant either: a mechanical shutter in a projector unit takes 150 to 250 ms, and a matrix LED array dimming individual segments takes 80 to 200 ms. My total measured delay ran from 0.6 to 1.4 seconds, with a median around 0.9.

At 90 km/h you cover 25 metres per second. A combined closing speed of 180 km/h eats 50 metres per second. Nine tenths of a second of full beam is 45 metres of closing distance with bright light in a stranger's face. That is not a defect. It is arithmetic, and it is why the stalk still exists.

The advice to set it and forget it does not survive a rural commute

Owner manuals and a fair number of reviews treat auto high beam as a feature you enable once and then ignore. I believed that for about two weeks. My log says the system handled around 118 of 140 events cleanly, which sounds good until you look at the 22 that went wrong, because they clustered in predictable places rather than scattering randomly.

SituationWhat my log showedWhat I do now
Crest of a hillOncoming headlights invisible until both cars are over the top, then dipping 1.1 s lateDip manually before the crest on any road I know
Tight left bend, hedge on the insideDetection at under 150 m, twice no dip at all until the cars were levelStalk down through the bend
Cyclist with a single rear lightMissed in 4 of 5 encountersDip on sight, treat the system as blind to bikes
Car ahead with one dead taillightHeld high beam for 9 seconds behind a hatchbackCheck the car ahead rather than the dash icon
Heavy rain on a dual carriagewayDipped early and stayed dipped, which is correct but felt brokenLeave it alone, the camera is right

The pattern is geometry, not software quality. The camera cannot see around a hedge or over a hill, and no amount of processing invents data that never reached the lens. Anyone who tells you their car never gets this wrong drives on wide, well mapped roads.

There is a second cost to blind trust. On the cars I tested, repeated manual overrides do not teach the system anything. It has no memory of the road. You are the only part of the loop that learns.

Keeping the camera honest

Before you blame the software, rule out the three physical causes I now check first. The inside of the windshield in front of the camera collects a film from plastics outgassing, and it goes unnoticed because nobody cleans glass from the inside. I wipe that patch with a lint free cloth and a little isopropyl alcohol every couple of months, and on one car that single change cut late dips noticeably.

Dash cam placement is the second. A camera or a suction mount inside the hatched zone marked on the glass will clip the field of view, and the effect shows up as missed detections on one side only. Mount the recorder at least 10 cm clear of that zone, or below the mirror on the passenger side.

Headlight aim is the third, and it is the one that actually produces complaints from other drivers even when the dipping works. Load the boot with luggage and two adults in the back and the nose rises. Many electric cars have self levelling, but the range is finite. If you tow, or you load heavily, have the aim checked; a garage will put the car on a beam setter in about 20 minutes, and in Norway I paid 450 kroner for it.

Some brands expose a sensitivity setting, usually buried in the driver assistance menu under a name like high beam timing or auto light sensitivity. If yours has it, move it to the earlier or more sensitive end. You will lose a little full beam distance and gain roughly a third of a second of margin, which on the numbers above is worth more than the extra reach. While you are in that menu, it is worth reviewing how your car handles speed limit reading from road signs, since both features run off the same camera and both suffer from the same dirty glass.

A reasonable way to use it this week

Keep auto high beam on for motorway and open road driving, where the geometry suits it and the gain in seeing distance is real. Treat the stalk as a tool you still use, specifically at hill crests, on tight bends, and whenever you see anything that is not a car. If you get flashed, do not argue with the dash: dip, and note where it happened, because it will happen there again.

One habit that paid off for me was a short night drive on a familiar unlit road, done once with the system in charge and once with me in charge, and comparing how often I reached for the stalk. It took 25 minutes and told me more about the car than a week of reading forums. If you want to see how the same camera behaves when it is tracking you instead of the road, our piece on why driver attention monitoring keeps nagging goes through the thresholds. The rest of the features in this area sit in the driver assist section.

FAQ

Does automatic high beam work in fog?

Poorly, and that is not a fault. Fog scatters your own beam back at the camera, so most systems either stay dipped or hunt. Switch to low beam and your fog lights manually. I have never seen a camera based system handle dense fog well, though my sample is a few dozen drives rather than hundreds.

Why do my lights flick between high and low repeatedly?

Usually roadside reflectors, wet signs or crash barriers throwing your own light back in a pattern the camera reads as a vehicle. It also happens on roads with widely spaced street lamps that sit near the activation threshold. Cleaning the inside of the glass helps; so does a less sensitive setting if your car offers one, though that trades away reaction time.

Can I still be fined if the system dazzles someone?

Yes. In every country I have checked, the driver is responsible for the beam the car puts on the road, and no rule excuses an automatic function. Treat that as settled and keep a hand ready on the stalk.

Does a dash cam really break the feature?

It can. The problem is not interference but obstruction, and it shows up as detections failing on one side. If you fitted a recorder and the dipping got worse in the same month, move the mount before you book a service appointment.

ES
Evander Szabo

Evander spent years in software QA before reviewing in-car tech. He tests CarPlay, Android Auto and dashboard apps the way he used to test releases, looking for the bugs drivers actually hit.

More posts by Evander

More in Driver Assist

01doorganico 084goad 3hot8 4465456 after5internprogram ahandfulofblessings ahmadpc ahomcybersecurity allodialresources annodex antiquesandperiodcollectibles aotc asianworldkashmir beedlectrics bloggertemplatefree bluefeatheragency bodyweightsolution campaign08blog carfixadvisor catstreetcollective cbstvsports cnfanspurchasing committeeof500yearsofdignityandresistance covertopsdjs cs-kaiyuangame cuestacomunicaciontotal cxoutlet dancebetgame deskgearhub destinyvillastay detectmagazine di33 diamondsteelbuildings donmo dyanadropship emarsysfinance emimattl escapechan eskitarih everythingweddings-orlando farmzone feedviewerapp foxstudiosscratch gamevui123 gleasongrade greetingsday hacklabasu haskell iw-kaiyun jagoanproperti jilidskripsi joltbusinessfinance juegodepokergratis juegospucca kaeru-festival latestmehndidesigns lestanzediottaviano litemagazines localcommunityhero lustrsazan mariashebets menswheels newwebstechs1 nikkit ootheday oxigenvpn plurfitness prefetial probudgetcarpetcleaning psychicmrsmartin pyaten qvantor rejoignezlogis rexzbot riselayouts ruchin rwxqsgd5 sakaltime shibiaowang shop-bot skrtips spotlesshaven strcoding tetapatrium topselller uneplanteartificielle vipertechcr webtech1223411 wo-kygaming yamaha-mekar yeknava yoybuysheets zenunwa yamaha-mekar jilidskripsi ankertoto ankertoto ankertoto