Hacker News Reader

The article is worse than confused. It's a marketing piece written in a way that sounds vaguely science-y but with only a tenuous basis in real research.

I wrote more here: https://news.ycombinator.com/item?id=44312224

> I'm willing to entertain the idea that LED flicker is actually problematic, but I wish essays like this would be honest about the degree of confidence we have given the current state of the evidence.

I (and it seems others too) are very interested in this topic. I would appreciate if you could write an aritcle with "less confusion" so I can save it in my tumblog for future reference.

Yep, it's obvious that a lot of people are interested because junk articles like this usually get penalized harder. That interest is what this piece is preying on. Unfortunately it's easier to write a piece that tells people what they want to hear and spreads FUD than it is to write a piece that corrects the misinformation.

Much more rewarding too, because "we really don't know very much about this yet" is hard to expand to a full click-worthy essay and less likely to move product.

I think it’s pretty common for people to be able to perceive PWM flicker in their peripheral vision that they can’t when looking directly at the source. I encounter this fairly regularly myself.

Nah, LED lighting generally uses at least 200 Hz at a minimum. Some up to kHz. You can't perceive that. Older stuff or cheap quality might be using un-rectified AC/DC which you can see. Like cheap Xmas lights.

> Nah, LED lighting generally uses at least 200 Hz at a minimum.

Eh, they use what they can get away with. Nobody is out there policing flicker rates. Especially when you add a dimmer into the mix, there's a lot of room between good and bad, and when you're at the hardware store buying bulbs, there's not much to indicate which bulbs are terrible.

Lots of people don't seem to notice, so the terrible ones don't get returned often enough to get unstocked, and anyway, when you come back for more in 6 months, everything is different even if it has the same sku.

> there's not much to indicate which bulbs are terrible

https://lamptest.ru

Not only flickering, but lots of other information about internationally available brands, including cheap Chinese stuff: CRI, real power use, etc.

Use your favorite online translator.

> Nobody is out there policing flicker rates.

Actually, Energy Star and California's Title 24 have flicker standards. They may not go as far as some people like, but you can look for these certifications to know that a bulb at least meets a certain minimum standard.

> at least 200 Hz at a minimum

> You can't perceive that

I very easily can. I had to get rid of an otherwise good monitor a few years ago before I knew it used PWM to control the backlight (and before I even knew PWM was used at all for this functionality — I only had experience with CCFL backlight before that).

It was really annoying to look at, like looking directly at cheap fluorescent lighting. Miraculously, setting brightness to 100% fixed the issue.

By googling around, I found that it used PWM with a modulation frequency of 240 Hz, with a duty cycle of 100% at full brightness, which explained everything.

I can also easily perceive flickering of one of my flashlights, the only one that uses PWM at a frequency of a few hundred hertz. Other flashlights either run at multiple KHz, or don't use PWM at all, and either one is much easier on the eyes.

Some of us really do perceive this stuff, which can be hard to believe for some reason.

Like back when CRTs were mainstream you'd have those computer labs with monitors set to 60-85Hz and most people wouldn't notice, but some would. I definitely did, I couldn't stand looking at a CRT set to less than 100 Hz for more than an hour.

Absolutely, I also had massive issues with them, ending with red eyes and headaches within an hour of use. Getting my first LED monitor (with a CCFL backlight) was out of this world.

Considering that people use screens that are 360-500fps for noticeable improvements in video games, people can definitely perceive that.

Next time you see a high refresh screen, move the cursor around rapidly. It's very easy to tell.

That's 2~3ms per frame.

In gaming situations what they perceive may not be the actual "flicker" of frames but the input->to->display latency, which is a very different thing to notice.

But people don't get noticeable improvements from that.

The jumps from 30-40-60-72-144 are all pretty noticeable, but 144-240 is already very minimal and 240-360+ is pretty much quackery.

just shake your hand between the lights and your eye(s).

I can see it with one lamp I own, and in one shop that I've noticed that is using LED strip lights

> Nah, LED lighting generally uses at least 200 Hz at a minimum. Some up to kHz. You can't perceive that.

My partner and me both notice the difference between cheap LEDs and expensive ones (hue). Whe both cannot pinpoint it down.

This is due to different CRI (basically, how even light output is distributed across wavelengths), not due to PWM.

Depends on the bulb. For those of us with "fast" eyes, some LED bulbs that are just fine for others are a subtly flickering infuriation generator.

You may not believe that people that can see 120->240Hz flicker exist, but we do. In this era of frequently-cheap-ass LED lighting, it's a goddamn curse.

I had to shop around a LOT before I could find LED headlight bulbs with just a fan and a bare-ass resistor for current control so I wouldn't see that forsaken flicker.

It's really bad for me as I work in an LED and LASER facility. I handle ALL the PWM stuff while everyone else handles the simple led/resistor/connector board assemblies. EVERYTHING FLICKERS.

My thanks to you for shopping around for non-shit headlights. For a long, long while it seems like every third car had strobing headlights. (Now, the primary problem is INTENSELY bright and poorly-aimed headlights. I'm not sure which is worse, to be honest... but both SHOULD be super illegal.)

> EVERYTHING FLICKERS.

I absolutely could not handle that. My sincerest condolences.

The OP was specifically talking about hue. Hue does not change due to PWM.

> The OP was specifically talking about hue.

I presume this is why you think that?

> ...between cheap LEDs and expensive ones (hue)...

If so, they're referring to the Hue brand of bulbs, rather than the color property. More evidence for the fact that they're talking about flicker is that they quoted this to indicate that they were replying to it:

> > Nah, LED lighting generally uses at least 200 Hz at a minimum. Some up to kHz. You can't perceive that.

Hue bulbs have pretty bad CRI actually. They only claim >80, which almost any LED bulb is capable of these days. A good LED bulb (include those made by Philips these days) have a CRI >95.

200 Hz PWM on lighting is very noticeable.

Hi there, flashlight nerd here.

There's a third way: a switched-mode power supply with regulated output current. This is used in most better-designed flashlights (which doesn't always correlate to price) and can be used by anything else that needs to drive an LED as well.

The article doesn't discuss what technique should be used for "constant current reduction"; it probably shouldn't be a linear regulator where efficiency is a priority.

PWM is less annoying if the frequency is very high (several kHz), though I'll leave it to people who research the topic to speak to health effects.

In the world of LEDs, a switched-mode power supply with regulated output current is called a "constant current driver". I assume that's what this calls "DC dimming".

A linear regulator might also reasonably be described as "DC dimming" or "constant current". The article is only concerned with flicker so it doesn't discuss efficiency.

It could be, but I doubt you’ll find any commercial LED driver that is a linear regulator, so I don’t think that’s what they mean.

It causes physical pain for many. You can't always see the flicker but your eye muscles react faster than your perception, and they get sore.

I had to get rid of a Samsung TV (120hz backlight, replaced with linear dimmed Sony, nicer anyway..) due to this, and I can only use modern phones at 100% brightness which disables PWM.

When I moved into my current residence I couldn't figure out why my eyes were always sore. I realized my landlord put in cheap LED can lights. I swapped them out for nicer ones, pain gone. People need to stop being cheap AF.

It makes me wonder how differently we all perceive the world around us, and if that impacts our decisions in more ways than we realize.

I've had to drastically change my devices after learning PWM was causing my vision issues and eye pain. My partner has no issues using those same devices.

Basically the same. I know this because it only affects my left eye :)

> There are two ways to dim LEDs: linear regulation and some sort of pulse modulation. Linear regulation is wasteful and you're pretty unlikely to encounter it. Pulse modulation is common.

Within the pulse modulation case, though, there are two important subcases. You can PWM a load that consists basically of just the LED itself, which acts as a resistive load, and will flash on and off at high rate (potentially too fast to be noticeable, as you say). But you can also PWM an LED load with an inductor added, converting the system into a (potentially open loop) buck converter. And this allows you to choose both the brightness ripple and the PWM frequency, not just have 100% ripple. Taking ripple down to 5%, or 1%, or less, is perfectly straightforward… but inductors are expensive, and large, so shortcuts are taken.

Higher PWM frequency will let you use smaller inductors.

Yes, but you need to at some point decide it’s worth adding an inductor to the BOM — trace inductance isn’t enough at any sane switching frequency (and FET switching losses kick in at the insane frequencies). And BOM item costs more than no BOM item.

>There are two ways to dim LEDs: linear regulation and some sort of pulse modulation. Linear regulation is wasteful and you're pretty unlikely to encounter it, especially in battery-powered devices such as phones or laptops. Pulse modulation is common.

The other reason is LED brightness and color is quite non-linear with current, so PWM gives a much more straightforward dim than adjusting a current smoothly.

There are other ways to dim LEDs. Voltage stepping to control how much current can actually flow through the diode is one method (and this is how we test LEDs where I work, a 3V LED gets tested at ~2.4-2.6V. It will barely draw a couple milliamps despite the power supply set to allow an amp of draw.) The LED will light up, enough to see that it is working well, but not eye-searingly bright.

Voltage control of LEDs is extremely unreliable. The amount of current consumed at a specific voltage will vary wildly depending on temperature, process, etc. Nobody does voltage control for LEDs unless you do not care about consistent brightness at all.

"Voltage control of LEDs is extremely unreliable."

No, no it isn't. We can keep units within half a percent of starting output all day long.

Voltage control is done explicitly on laser diodes, to boot. And those are WAY MORE FINICKY than an LED.

> No, no it isn't. We can keep units within half a percent of starting output all day long.

On what, exactly? How can you possibly guarantee output using purely voltage control of an LED? LEDs (and laser diodes) are fundamentally current controlled devices. You need current feedback to set the output brightness operating points.

> Voltage control is done explicitly on laser diodes, to boot. And those are WAY MORE FINICKY than an LED.

Maybe if you don't care about the output power of the laser diode. Just not practical, and will change output power at the same voltage as temperature changes.

"How can you possibly guarantee output using purely voltage control of an LED?"

They don't tend to emit much heat when underdriven using voltage control because they can't pass high amounts of current at low voltage. Run an LED from 2.4-2.7V in discrete hundredths of a volt steps. You can get almost nothing to getting close to drawing 50mA doing discrete steps like this. This is how we also characterize and bin LEDs. As long as the LED is on a thermal mass, it isn't going anywhere near thermal runaway.

I've done this for almost 2 decades, now.

> They don't tend to emit much heat when underdriven using voltage control because they can't pass high amounts of current at low voltage.

No LED used for lighting in households is operated this way. You run them hot, because they can take it and you aren't wasting costly LED die area.

> Run an LED from 2.4-2.7V in discrete hundredths of a volt steps. You can get almost nothing to getting close to drawing 50mA doing discrete steps like this. This is how we also characterize and bin LEDs.

Discrete hundredths of volt steps? Maybe for characterization you can do this, but you expect residential LED bulbs to use circuitry able to precisely output in 1mV steps? Not happening.

I'll concede the point you can likely do characterization of LEDs using voltage control, but when you get to actually wanting to drive them for real usage in a lighting environment (i.e., not for some indicating function), voltage control isn't going to be useful. You don't have that precision, you don't have that thermal mass, and you certainly don't have the ability to drive them that weakly.

"No LED used for lighting in households is operated this way. You run them hot, because they can take it and you aren't wasting costly LED die area."

I literally work in this field, manufacturing thousands of LED boards of all types every single day.

Yes, they absolutely do. Especially under-cabinet lighting - it's deliberately underdriven via voltage because the thin strips of aluminum are basically a backing support strip and nothing more.

Let's grab what under my cabinets, for example. 4 LEDs in series. Okay, assuming 3V that's a 12V chain - the included wall-wart power supply is a 10.5V output supply at 200mA.

Please stop.

> Yes, they absolutely do. Especially under-cabinet lighting - it's deliberately underdriven via voltage because the thin strips of aluminum are basically a backing support strip and nothing more.

Under cabinet lighting and more generic cosmetic lighting tends to use constant voltage for the input supply, but is regulating its current elsewhere, within the cabinet module or LED itself. This is literally how the most common example of RGB LED strips (WS2812s) work. They have current limiting circuitry within the LED module itself. Getting the most brightness isn’t the most important thing, unlike in actual bulbs.

> Let's grab what under my cabinets, for example. 4 LEDs in series. Okay, assuming 3V that's a 12V chain - the included wall-wart power supply is a 10.5V output supply at 200mA.

Nobody is driving a standard white LED at 2.6V for lighting. Please, show me a datasheet of a white LED that has a reasonable lumen output at 2.6V forward voltage drop. That also doesn’t include any voltage drop allowance in any wiring, boards, or any circuitry, which can be significant in low voltage applications. Especially when cabinet lights tend to be daisy chained together. I’d guess your cabinet lighting is actually driving a string of 3 white LEDs in series.

I wonder if it’s possible to have randomized sub-cycle interval variation so as to “spread out the spectrum” of the PWM signal while preserving the same resulting integrated brightness over the observable timescale.

Basically like how Apple laptop fans work, but for temporal modulation of signal instead of spatial modulation of fan blade gaps.

"Human vision has a pretty limited response speed, so it seems pretty unlikely that PWM at a reasonable speed (hundreds of hertz to tens of kilohertz) can be directly perceived."

I make power drivers. I have this ultra-tiny one, with output scope captures. It produces ~825kHz PWM output, single-digit mV and mA ripple, 94+% efficiency depending on input voltage (output is input minus 0.2V)

I can induce sacchades in my eyes at will and at high speeds. Couple that with waving my hand in front of my face as I do that, and add in human vision persistence, and I can get artifacting that reveals flicker even at that high of a rate of PWM. Only direct battery power fails to induce that artifacting effect in my vision when I do that combination of movement.

I personally doubt PWM is visible at all at high enough speeds (kilohertz+) even in peripheral vision. I would love to see a good study on this.

I can't see PWM flickering even at 30Hz. But it gives me bloodshot, twitching eyes even at 20kHz.

no, 2 main methods to change led brightness: Changing duty cycle, or changing current. PWM is way to change duty cycle, and Linear is 1 way to change current. There are other ways to change duty cycle or current.

Even if your PWM is faster than your eye's response it's trivial to see how one is still affected, namely, the stop motion effect.

That effectively lowers the frequency of the lcd.

PWM is not to be confused with Switch Mode Power Supplies in the general case