I first noticed slow PWM in factory tail lights on Cadillacs somewhere around the turn of the century. I pointed it out to my dad one night when we were out for a walk, and told him how to see it on a car that was disappearing down the road.
It took a few seconds for him to understand and to observe it himself, and then he said "I can accept that you're annoyed by that, but maybe they just build them that way because they think it looks cool."
Ever since that night, I have always assumed that slow PWM was used as a deliberate visual effect: After all, General Motors isn't really trying to squeeze a few pennies out of a new Cadillac [that's what Chevrolet is for!].
(I also think it's a dick move on the manufacturer's part, whether it is to save forty-seven cents, or if it is to look cool, or even if it is both.)
This prompted me to go look for a better answer, and I found a good candidate: pulsed current at a relative low frequency (~60hZ) allows higher perceived brightness from smaller/cheaper LEDs. From my perspective, it is undesirable to increase the brightness of car taillights relative to what was typical with incandescent bulbs or to make them flicker.
The PWM function is built-in into the electronic switches that drive the light outputs, so you can programmatically control the brightness, it also accepts different switching frequencies so it’s just a matter of increasing the frequency for the “strobe effect” to be unnoticeable. The advantage of these switches over analog are several, including output diagnostic and lower power dissipation.
Keeping in mind that it's on a moving vehicle, and road users often pan their eyes around rapidly, the frequency would need to be higher than for many other situations.
Or they could use a variable linear regulator with no flickering at all.
Is a higher switching frequency problematic, somehow? Is there an upper bound on switching frequency that makes LEDs difficult or unnecessarily inefficient or something?
There's nothing I'm aware of relating to the electronics that drive them that would place such a low bound on frequency that the resulting pulses would be visibly distinct.
MOSFETs are generally good for > tens-of-KHz frequencies, and so are the hardware PWM channels in every-day MCUs.
The effect described in the StackExchange answer to increase perceived brightness at a lower power level seems to rely on a low frequency and short duty cycle. It's not necessary to use that effect given that LEDs rated for higher power levels are available, and the amount of power is insignificant at the scale of a car, but it does allow the use of cheaper LEDs.
A series of web searches found one of the driver boards often used in LED tail lights: the Valeo b003809. Some images of it show an identifiable TLE4242 linear driver IC.
I was starting to suspect that would be the case after reading the explanation about slow PWM and a short duty cycle resulting in greater perceived brightness from cheap LEDs that aren't rated for much power. Direct drive from the car's electrical system wouldn't result in the constant brightness seen in LED tail lights because the voltage is lower at idle, so there has to be some kind of regulation.
There would be little reason to use a more efficient, more expensive switched-mode power supply, though an aftermarket replacement for the Valeo board does use one. Even an inefficient LED design will be far more efficient than an incandescent bulb, and the power requirement of any taillight is trivial relative to a modern car's electrical system. Linear makes the most sense for the application.
Linear regulators aren't rare in LED flashlights nor is direct drive with PWM on a FET for low modes. Neither is good in flashlights, but that only matters to the manufacturer if their customers know the difference. Most customers don't; lights marketed to discerning customers use a switched-mode power supply.
The problem could be the LED driver itself rather than the vehicle driver output. Most LED lights use some kind of PWM or multplexing, but people only notice the bad or rather the slow) ones. It’s easy to find out, just take a few videos at different frame rates an you’ll see.
There are two ways to implement a constant current LED driver; a linear regulator is one of them. The other is a switched-mode power supply. A linear regulator is the cheaper of the two.
Is it possible to control LED brightness with a linear regulator? I am surprised to hear it - when I was making light effects with LEDs years ago, I understood PWM to be the only practical mechanism, due to the strongly non-linear conductivity curve.
But maybe it's possible, if you just want to fix your LEDs at a specific brightness which never changes, that you could find exactly the right current level...?
As a flashlight nerd and reviewer, I know a number of ways to control the brightness of an LED, and linear regulators are a fairly common one. It's important to note that a linear regulator can be designed to target a specific voltage, or a specific current. LEDs should be regulated with current for the reason you mentioned.
One that I think would work well in cars is a fixed-current linear regulator like the AMC7135. There's no need to do anything fancy to control it; put a 7135 in the circuit and the LED gets 350mA. Two in parallel and it gets 700mA, and so on. More fancy is a linear regulator with variable current output.
Linear regulators are a low-end option for flashlights, with switched-mode power supplies being preferred for their superior efficiency. A car has a lot more energy to work with and more mass with which to sink and radiate heat.
You can but it isn’t efficient. You lose (V_in - V_out)*I_led Watts to heat if you use a linear regulator.
You might ask why not connect them in series and get the voltage difference as small as possible. But the “forward voltages” of LEDs are highly temperature dependent and car battery voltage is (somewhat) engine rpm dependent (might swing between 12 and 13.8V between no rpm and some rpm. It’s kept flat at 13.8V)
Linear regulators can do it easily, if used in a constant-current driver configuration. That constant can be variable, so regulated-current might be a better name.
Your classic LM317 linear regulator can do it pretty easily, with a potentiometer instead of a fixed resistor to set the output current. Section 9.3.3[1] of the datasheet describes the circuit, though in practice you might add a couple of capacitors for stability.
There are also a number of discrete-transistor circuits. A current mirror with a rheostat is probably the simplest, though not generally ideal performance.
Interesting! I have used plenty of LM317s with LEDs, but I never thought of using them to regulate brightness - I would just pick whichever resistor corresponded to the forward current specified in the emitter datasheet. Didn't realize that there was enough room on the curve to use current control as a meaningfully configurable brightness setting.
