Hacker News Reader

Comment by Tempest1981 | Hacker News Reader

Tempest1981

And apparently why we have LED car taillights that flicker annoyingly.

If anyone here works at a car manufacturer, please try to get this fixed. Apparently not everyone notices.

It's PWM dimming. I see it most on Cadillac SUVs. Drives me nuts too.

They could design it not to be that low of a frequency but apparently someone thought that 40-50hz was imperceptible to humans and went with it

I find it most annoying when moving your eyes. Anyone will notice the flicker if their eyes aren't stationary since then each "flick" will end up on a different part of the retina. So, instead of a normal motion blur you see a dashed line which is hard to ignore.

What's the cause? Alternators run at very high frequencies with good rectifiers, so I'm guessing the flicker is introduced by PWM dimming, but why would that be a low enough frequency to bother people?

I'm sensitive to flicker myself, but only on the more extreme half of the spectrum. For example, half rectified LED drivers on 60 Hz AC drive me nuts, but full rectified (120 Hz) I very rarely notice. I don't notice any problem with car tail lights, except in the case of a video camera where the flicker and the frame rate are beating. The beating tends to be on the order of 10 Hz (just shooting from the hip here) so if frame rates are 30/60/120 then I guess the PWM frequency is something like 110 or 130 Hz?

> introduced by PWM dimming, but why would that be a low enough frequency to bother people?

The human fovea has a much lower effective refresh rate than your peripheral vision. So you might notice the flickering of tail lights (and daytime running lights) seen out of the corner of your eye even though you can't notice when looking directly at them.

For sure. I summarized my particular sensitivity too aggressively earlier, but I tend to see flicker straight on up to 65 Hz and peripherally up to 120 Hz if it's particularly egregious (i.e., long valleys) but usually up to something less. In any case, I've never noticed car tail lights flickering even peripherally, despite video artifacts revealing that they do flicker.

Automotive engineer here! I don’t do lights, so I have no idea!

BUT… I kinda do. You want the lowest PWM frequency you can get away with. In this case, at the back of the car, furthest from the battery, you really don’t want a 8kHz PWM nor do you need it. It costs money to isolate the supply demand, so you don’t want a noisy field for no reason. The “good enough” frame rate is 60Hz static, not moving, no other flashing lights, not using a camera, etc.

60Hz or 60fps has issues though. If you expose a PWM LED to another flashing light or movement you get really bad imaging. Imagine you took an LED in your hand and shook it, at 60 Hz you will see snapshots of where the LED was as you’re shaking it. At 240Hz you will see a blur. Guess which is better for a vehicle?

I figure most car LED taillights internal to their housing would be 200-1000Hz depending on factors but I haven’t ever measured.

200Hz PWM is a really common value. No need for Samaritan base-12 here.

For halogen and incandescent, we use PWM, fun fact. Low Hz though! About 88Hz, depending on voltage. You might wonder why. We can get 98% of the light output with 85% of the required wire. It’s all about weight and cost. Although not many vehicles use this anymore.

Fascinating! I've played with VW's BCM stuff, but for some reason everything except the lighting is translated into English on the common factory equivalent scan tools, so it's a matter of sorting though a bunch of "lichtfunktion" this and "dimmwert" that.

> We can get 98% of the light output with 85% of the required wire.

I guess a 12v bulb can survive a much higher voltage than it's rated for (which needs fewer amps for the same watts, so thinner wire) if it's pulsed, like 120v assuming 10% duty cycle, as long as the frequency is dialed in such that the filament is just about ramped up to its rated temperature when power is cut. Very clever!

> For halogen and incandescent, we use PWM,

Must be fancy cars then. All I've ever owned used relays to switch the lights, just straight 12v onto it no pwm whatsoever.

That used to be the case, of course.

We PWM those now because the PWM drivers are there anyhow, and with a filament, it’s still lit and bright during the off pulses. It’s cooling, but you really can’t detect it.

Imagine your eye's effective integration time (the period over which it "samples" light) shrinks during a rapid saccade—say down to about 5–10 ms. Under steady conditions, our flicker fusion threshold is around 60 Hz, but that’s because our retina integrates over a longer window.

If we want to “smooth out” the PWM cycles so we don’t see discrete pulses, we need a few cycles (say, 3–5) within that 5–10 ms window. In other words:

    For a 10 ms integration window:
        3 cycles → f≥30.01 s=300 Hzf≥0.01s3 =300Hz
        5 cycles → f≥50.01 s=500 Hzf≥0.01s5 =500Hz
    For a 5 ms window:
        3 cycles → f≥30.005 s=600 Hzf≥0.005s3 =600Hz
        5 cycles → f≥50.005 s=1000 Hzf≥0.005s5 =1000Hz

So, to cover worst-case scenarios (rapid eye movement, bright conditions where the eye’s temporal resolution is higher), the PWM systems for LED lights should be rather in the ballpark of 300–1000 Hz than 200 Hz. Given that one would be viewing headlights some 2 meters away (impacts the what is the view angle change and speed).

And yes, what we are now seeing in cars is super annoying. Similar disregardful to user comfort can be seen with HUD displays (at least with Volvo).

Actually, you want to use a DC-to-DC converter that properly delivers the constant desired reduced voltage, rather than making a ridiculous stroboscopic light (aka PWM).

Constant current is more expensive and does not scale to a lot of LEDs or high power LEDs well. The scale wouldn’t be an issue with the tail lights, but the power kind of is. Take a look at modern taillights sometime, they all have heatsinks on them. I’m sure there are some; but I am not aware of any constant current taillights.

I see it, but only in my peripheral vision. Drives me nuts.

For me, it's also noticeable when my eyes are sweeping from one object to another. Like you should be doing when driving. I guess that's somewhat peripheral, but not entirely.

Find a Cadillac Escalade with the vertical LED lights, and give it a try.

What's a good way to capture this on video? Or to measure the frequency?

That's a different thing. Your eyes automatically filter out the blurry signal during a saccade (ie. while they're doing a fast pan from one location to another.)

Rapidly strobing lights (eg. LEDs being PWM'd) mess with this process and cause you to perceive the intermediate images as your eye moves, leading to some weird visual effects. This was a big thing in the early days of DMD projectors, where they projected sequential red/green/blue frames, and while it looked great when you stared at it, any time your eyes moved the whole image strobed with overlaid RGB fields. (more info: https://www.prodigitalweb.com/rainbow-effect/)

Or for some reason in the rear view and side mirrors!

This is because your cones (colour-specific 'pixels' in your retina) cluster around your fovea (high-resolution area around the center of focus) while the rods (black-and-white, bigger, dumber, but much more sensitive and therefore faster-responding 'pixels') fill the rest of your visual field. Rods respond fast enough to see the flicker but cones don't.

You might also notice in light that's almost too dim to see, you can see things better if you look a bit to the side of them, for the same reason.

Reply on news.ycombinator.com