The scientific reason you don’t like LED bulbs – and the simple way to fix them | Hacker News Reader

The scientific reason you don’t like LED bulbs – and the simple way to fix them

theconversation.com

80 pointsValentineC67 comments

What gets me are side-lights on cars that flicker. Cars have DC electric systems in the first place. Where did the Audi engineers find 100 Hz to run the LEDs from?

I guess there's a switch mode power supply to convert down from 12 V to something lower that the LEDs need. But AFAIK, switch mode power supplies don't switch at 100 Hz.

I'm pretty sure you could show by experiment that flickering car side-lights are distracting. For example, in a situation where you are pulling out of a T junction in the dark, you're less likely to see a motorbike when it is in the same part of your visual field as a car with flickering lights. Particularly when you are moving your head from left to right, as you do when at a T junction.

I'm not sure how to do the experiment though, since using TV screens, monitors or projectors wouldn't work.

That is seriously annoying. Not only do they flicker, they are slightly blue as well. I have mistaken an Audi for an emergency vehicle in my rear view mirror on multiple occasions. I am sure that this is their plan. Such lights should be outlawed.

Now I have found that some mirrors on cars and even motorcycles can distort the reflected light to give blue points. combine with polarizing glasses and it becomes even more difficult to find the exact cause.

now if viewing straight on without another element to distort the light, like polarized lenses; I would be interested if the reflector built into the headlamp assembly can cause an issue.

I've found the colour to be too close to emergency-vehicle blue as well, particularly if the car behind hits a bump and momentarily brings their headlights into my mirrors, it's almost exactly like the flash of emergency lights. I shouldn't have to be second-guessing whether there's an emergency vehicle behind me, I should be following my first instinct which is to ensure I'm not obstructing traffic. Not clearing out of the way for an Audi pulling a tonne to rocket past me.

I suspect its not the PSU, its the PWM for the dimming.

compared to actually lowering the voltage/current of a LED/light/other its far simpler and cheaper just to turn it off and on again at varying duty cycles/ frequency.

Correct me if I'm wrong, but isn't it just the de facto way of dimming LEDs? Since as opposed to incandescent bulbs, they require certain voltage/current to be lit, if you pass them less, they simply won't emit light.

You can feed it DC and vary the current, for example by filtering the pwm.

They require a certain voltage (to a first approximation). Their light output is (again, to a first approximation) directly proportional to the applied current.

PWM control is just a cheap way of reducing the current.

Good point. But there's no reason for the PWM to be in the 100 Hz range either. Even the crappest possible microcontroller using timer interrupts could handle 1000 Hz. Hardware PWM peripherals (that can switch at 500 KHz) have been trivially cheap additions to microcontrollers for decades.

It's likely not for dimming but for reducing power usage without reducing apparent brightness. See my comment sibling to yours.

May not be for dimming. PWM control is used also to reduce the current (and therefore power, and therefore heat) required to achieve a particular apparent brightness. This utilizes the persistence of vision effect.

My guess is that, above a certain frequency, persistence of vision no longer applies, and the light is in fact seen simply as being dimmer (which is undesirable in this case). Hence ~100 Hz.

Another technical consideration would be to have grid-based scanning of the LEDs. That is you have some sort of row output, one row at a time, and the columns are multiple parallel outputs. This inherently pulses the LEDs as the rows aren't all lit simultaneously. The persistence-of-vision effect is a great side effect and can be capitalized on. The grid is best for limiting the number of wires to access a substantial number of LEDs (like 16 or so) in a group. I don't think a grid for the turn signals of a car would be a great idea, at least not at the level of the individual turn signal diodes. Perhaps within a single corner's turn signal assembly.

Many rear car LED lights flicker because their full brightness is used as brake lights, and they are dimmed via pulse width modulation for normal lighting. This is simply a bit cheaper, which is ridiculous on an expensive car.

I don't actually see the flicker normally, but when they quickly move accross my visual field I sometimes see a line of dots from the flicker.

Cadillacs are even worse. I could swear their brake lights run at 20Hz.

I've had this problem and decided to test the problem away. Used a Vishay visible spectrum phototransistor and my scope. Think it was a BPV11 transistor.

Philips bulbs are 100% clean. Poundland ones are not. Cheap no brand ones from Amazon are not. The latter two have a 100Hz output.

I suspect the hatred for LED bulbs stems from the cheap ones which use the lowest cost solution to get mains voltage into a constant current which is usually a capacitive dropper and rectifier with an inadequate smoothing cap.

Really as always with these things, buy the best you can.

Also the cheaper ones seem to have a blue/white tone whereas the expensive ones have a yellow/white tone and are warmer. This helps the eyes.

I'd be shocked if the flicker went to 0% as this article seems to imply ("because LED flickering is even more pronounced, with the light dimming by 100% rather than the roughly 35% of fluorescent lamps") -- and I doubt the flicker for fluorescent lamps is at RMS.

There's _gotta_ be at least a smoothing capacitor after the bridge rectifier (are they actually that cheap to omit one?!). I can't find a single LED driver circuit schematic online without the low pass filter.

Often the first things to be cut in cut-rate gear are comfort/longevity features such as smoothing capacitors and protection circuits. Generally if it doesn't have an effect on the return rate, it gets cut out to slash costs.

the cheaper ones seem to have a blue/white tone whereas the expensive ones have a yellow/white tone and are warmer.

This is just normal color temperature and both come in all price points. I have several warm bulbs I got at Target on clearance for $2/2. I think it's just that the warm cheap LEDs are usually sold out, so all that's left are the prison-colored ones.

It's not quite that. The cheap ones use white LEDs which are basically phosphor doped blue LEDs. They have a distinct blue peak which is far brighter than the overall output. The more expensive ones use blue LEDs with separate phosphor outside the LED chip. Philips do that. Seems to result in far better filtering of the blue part of the spectrum.

Thanks, I never looked into the technical reasons behind this.

Another issue is that many LEDs have peaky spectra. This makes cool white harsh and warm white dingy. This isn't limited to cheap products, as producing a fuller spectrum with LEDs is less efficient.

I'll take the efficiency hit for better light; the most common measure used, printed on the packaging of some bulbs is CRI (color rendering index). It's based on blackbody radiation, so incandescent bulbs and sunlight both have the maximum score of 100. The average unspecified LED is in the 60s and sodium vapor lamps are negative. 90 is good for an LED, and 97 is the highest I've heard of someone testing in an LED[0].

CRI isn't the whole story on color quality, and it's possible to have high CRI with off tints or even a peaky spectrum outside of the specific colors the CRI standard tests, though the latter doesn't happen in practice. Several newer standards for measuring color quality exist, but are unlikely to be listed on a light bulb label.

[0] http://budgetlightforum.com/comment/1096106#comment-1096106

Agreed! It took several tries to get LEDs which worked well for vanity (i.e. Mirror) lighting. I thought it was a simple matter of color temperature - so wrong!

High CRI strip LED arrays are surprisingly hard to find. The end result was worth it though - skin looks natural, and the nearest commercial version is upwards of $800 - and I'm not entirely sure they got the CRI right.

Which solution did you end up with?

I ended up with a strip from colordiode, they're available on ebay. Was described as CRI>95 3000~3200K SMD2835, 90/m. Came with a pretty decent datasheet, seemed well-characterized. Cost was ~ $10/meter. They were responsive enough to include a few samples of other color temperatures with the order when I asked nicely.

The housing used was extruded aluminum u-channel by Torchstar (U05, semicircular), very nice clean finish. Available on Amazon.

Like I mentioned, this was for vanity lighting - there's a full length mirror in the closet that's wedged on the wall between a closet door and a window. Not much room for anything else. Hung two vertical strips, ~ 4ft long, on each side of the mirror.

Light ended up soft, diffuse, and flattering - but still directional enough for good modeling. Significant other couldn't have been happier.

Yeah. The best example was lighting all my clothes with LED lights. Colours seem obviously messed up. Some colours are emphasized while others are suppressed.

Low pressure sodium lamps (SOX) are negative, -44. High pressure sodium lamps (SON) are positive, +20.

By "100% clean" do you mean no flicker, ie. they use the more expensive components? I invested in Philips LED bulbs about a year ago and honestly I can't fault them at all. I consider myself to be sensitive to things like flickering. It doesn't usually give me a headache, but it will make me notice it to the point of distracting me. I was so happy to find that the Philips LEDs were almost just like old-fashioned bulbs (instant on, no flicker, warmer colour), but with virtually no heat output and vastly reduced power consumption.

Yes 100% free of flicker. I actually measured modulation technically speaking so even brightness modulation of a small amount would have been picked up.

Same experience with the Philips bulbs here.

This has also been my experience with Phillips bulbs.

It's not just cheap LEDs unfortunately. I bought some expensive recessed lights with Osram drivers (~€80/light), and even those flicker a little.

This is important to know. Thanks for posting your experience.

I have a different experience. Osrams dont flicker (measured using the cheapo way: deferent types of cameras), whereas Philips LEDs do.

But otoh the osram ones have all failed or started blinking due to what I think is over heating.

Yeah, adequate cooling is actually a big deal with LEDs, and the problem is compounded in cheap bulbs with minimal cooling and lower-quality components all around. These things are supposed to last decades. http://www.ledsmagazine.com/articles/print/volume-11/issue-7... talks about this with a specific focus on the fixture, since it affects thermals so much. The LED automotive headlights I purchased recently have substantial aluminum fins and cooling fans ( https://www.amazon.com/dp/B00WVJQO58 ) plus external drivers also with aluminum cases. There are tons of cheaper options on Amazon with less thermal consideration. Given the 80w rating (for the pair) compared with the kind of heat sinking you need for a CPU with a ~80w TDP to keep it happy and below 90*C, these looked to be at about the lower end of what would do the job. The fans definitely spin fast and move a bit of air; you can hear them quite well with the ignition on and the motor off.

All the cheap ones I've had that blew up were either an SMD resistor blowing out or one of the LEDs blowing out. When the LEDs blow out if you peel the board off the heatsink you will always find a missing bit of goop under the LED that blew.

Interestingly I stuck one on my bench supply (30v 2a) and I couldn't get one to blow out at all with that!

I have multiple Osram drivers (different models) and some flicker, some don't - so I don't think as a general rule you can say "X brand doesn't flicker". I'd assume the choice of LED, and how many you have makes a difference as well, as it might be based on power draw.

I found the same. The ones from Screwfix work very well and only cost £10 for 5 of them: http://www.screwfix.com/p/lap-gls-led-light-bulbs-white-bc-9...

Thanks for sharing, especially as it confirms my suspicions!

A few years back we changed every single bulb in our house (including the GU10s in the kitchen) to Philips CorePro LEDs. I've never seen this "flicker" or had any complaints about it from visitors (nobody even seems to notice that anything is different until I point it out). I also recently fitted out an office with SkyTile LED panels. Again, no flicker.

When I'm in a house or an office with cheaper LEDs for more than a few hours I find it very straining on my eyes. Then again, I had the same issue with CFLs, which is why I made the decision to change everything to LED in the first place.

Yeah it's the CorePro or Master ones I used. Recent ones have been these units: http://uk.rs-online.com/web/p/products/1244319/

CFLs are awful. I got rid of all of mine years ago back when LED bulbs were £20 a piece. Was worth it even at that cost!

When I bought a pack of cheap LED bulbs, the harsh flicker really surprised me. I thought cheap would mean that the diodes would be uneven in brightness and have a very narrow spectrum. I was OK with that. It had not crossed my mind that they'd decide to slice a nickel off the cost of each bulb by leaving out capacitors. Whoever designed those bulbs needs to spend a week living under their horrible 60Hz flicker as punishment.

I don't think they leave out capacitors. The typical Mylar units they use are cheap. They only put enough capacity in to get them out of the door.

I hate it when LED light output isn't continuous or visibly flickery, especially when it's from LED lights that are on (or are visible from) moving objects -- such as LED taillights on automobiles or in-road lane lights. All the eye movement (or object movement) makes the flicker even more flagrant :(

I remember extremely flickery lane lights on the leftmost lane markers on California's infamous SR 110, on the northbound direction (towards Pasadena) in the tunnels (the ones right before the left-exit to I-5 north).

Good fluorescent bulbs will have electronic ballasts (the driving circuit) that supply current to the bulb with a higher frequency (in the low kilohertz) than the time response of the bulb's phosphor. Thus, any sort of flicker gets low-pass filtered and isn't significantly noticeable in the output light.

Even with a subpar magnetic ballast that drives the bulb with line frequency, the phosphor never gets close to fully extinguishing -- so there is flicker, but the instantaneous light output never gets less than, say, half the maximum instantaneous light output. Look on page 3 of http://www.usailighting.com/stuff/contentmgr/files/1/e18f88d... for some instantaneous light vs time graphs for fluorescent bulbs.

LEDs have a much higher frequency response (in the tens of megahertz, easily) -- they respond quicker and more faithfully replicate their electrical drive signal as light output. You can see on page 4 of that report how violently jagged the LED's light/time plots can get.

If the manufacturer didn't want to include adequate filter capacitors in the LED's driver because of cost/reliability concerns, well, I hope your eyes enjoy that signal because the LED, unlike the fluorescent bulb, isn't going to smooth it out.

Those LED plots are pretty crazy. Surprising how much incandescent lights vary though (10-20% according to the plots) — would have it would have been much less than that.

I find this a problem with my LED-backlit LG monitor. It's ridiculously (by which I mean, blindingly) bright at full brightness, but anything below and the imperceptible flicker gives me migraines. So I have the choice of impossible brightness with no flicker, or acceptable brightness with headaches.

I don't get why LED bulbs need to flicker though - the bulbs don't dim, and if they're providing steady light, they should be operating permanently. If it's a heat production problem, the bulbs need heatsinks, not PWM.

I've found a vast array of modern tech to be completely unusable for me because of this. I can stare at the CCFL-backlit screens on my workstation all day, but a few minutes of my home monitor is enough to send me straight to bed in darkness.

> I don't get why LED bulbs need to flicker though

It's _much_ cheaper/easier to blink an LED on/off for 75% of the time than to drive a constant but lower current to produce 75% of the brightness. Most products are cost optimized.

From experience there's a massive difference in the quality of light from a cheap LED bulb to an expensive one which I don't think is something that the bulk of consumers are aware of yet.

There's also a massive difference in how long they last for. The cheap ones use cheap components that fail after a year or so. The LEDs might last a decade but the cheap drivers won't.

For those in the UK, I can recommend Screwfix LEDs. They seem to be lasting for me, don't flicker and are relatively cheap at £2/bulb http://www.screwfix.com/p/lap-gls-led-light-bulbs-white-bc-9...

You can see the flicker of your light sources for yourself if you have a slow speed camera. iPhone 6s/SE/7 can record at 240FPS. Point that at your LEDs, fluorescents and incandescents. Try your computer screen as well. I did this half a year ago and can confirm. Cheap LEDs are really bad, some more expensive ones don't flicker at all. Cheap computer screens flicker as well. Whenever I find a light source that I don't like somehow but can't quite say why, I record it slowmo and usually it turns out it flickers.

Most of my led lighting is dimmable, and takes noticeable time to go dark after I flip the switch to off. The ones that actually have dimmer switches installed also take noticeable time to come on when set as dim as possible. I seriously doubt that those bulbs have flicker problems.

It's quite likely that that's intentional 'soft on/off' behavior that's been engineered into the dimmer. My lights at home do this too, but I can turn it on/off. Some LED bulbs don't deal with it very well - depending on whether the driver in the bulb themselves expect leading or trailing edge dim signals (in the case where you send mains voltage to the bulb and the bulb has in internal driver; this is different when you have a full DC circuit).

Say you have an LED which can sustain 1 Watt of continuous power. If you feed it 10 Watt in fast pulses with a pulse/pause ratio of 1:9 it would still be 1 Watt on average, but the light intensity impression to the human eye would be higher.

Weird, I've never had issues with LED bulbs, but I routinely have headaches from fluorescent. Maybe I've just lucked out with my specific bulbs being higher quality (Ikea, Philips Hue). I imagine color spectrum is a large part too.

Me too - constant headaches from fluorescent lights, none from LED. I buy the less efficient full-spectrum LED lights which have a phosphor. The phosphor probably filters the underlying LED flicker.

Yeah, I think my Ikea ones are phosphor, the Philips Hue are RGB though and still fine.

Anyone know what the LiFX bulbs use? We have a dozen of them and the adjustable color temp is great. I also have a box of smaller LED bulbs from Home Depot. Neither hurt my eyes and I don't (didn't, will see now that I'm paying attention) notice a difference, but I'm sure there is one.

I also wonder about e.g. The Anker desk lamps, where the bulb is built in. I have one behind my monitor with the head turned round (it's designed to do this) to throw light up the wall behind my monitor as bias lighting, and another (their clamp lamp) on my electronics desk. They're great lamps with annoying touch power buttons, but both are adjustable color temp and have built-in LEDs which never require changing.

LiFX let you set the color right? So by definition they're going to be RGB (have three colored LEDs inside). Generally, LEDs that are single color are more likely to be phosphor (blue LED inside with a phosphor that converts to yellow).

I went to LEDs in my apartment about two years ago. 3 Hue bulbs + hub, some 48" T8 LEDs, and some E26 Ikea bulbs. The Ikea bulbs are great, power on instantly and slowly fade off. The T8s take about 3/4 second to turn on, and are nice and bright (and a far cry from the power consumption of the T12 florescent bulbs they replaced).

The Hue bulbs... well, one's developed a noisy transformer. Replacements are, what, $40? Meh. I like the Zigbee integration, but not $40 in two years worth of like. And the CRI of the Hue is just as bad (actually worse) than the other bulbs.

Lesson learned, those Ikea bulbs are great.

Not trying to sell Hue here, but their bulbs have gone through several iterations by now. They were one of the first big names to push consumer LED bulbs, I wouldn't be surprised if their earlier versions had flaws just by being new. But maybe not, I don't know.

(Also their color bulbs are $40, but they have much cheaper white ones, and also an in-between model they call ambiance which are white but allow you to adjust color temperature)

Yes, you can tell the Ikea ones aren't flickery because they have huge phosphor persistence - several seconds after being turned off.

I've had the opposite, no problem with florescent, but many LED's drive me crazy - I keep seeing a "speckle pattern" like what a laser makes.

If I don't move it's ok, but as soon as I look around I see a pattern of black and white areas from LED light.

Reminds me of the effect of using an electric toothbrush, and simultaneously looking at a LED indicator of e.g. a washing machine. It is explained here: [1].

[1] https://electronics.stackexchange.com/questions/32675/why-do...

I recently got myself some LIFX smart bulbs and I don't notice any flicker or discomfort. So maybe the morale of the story is to stop buying shit bulbs?

But then wouldn't our 60Hz screens have a much bigger impact on office workers than the lights above them?

Screen backlights aren't turned on and off at 60 Hz.

The screens have separate backlight which is done with other technologies and if with leds, definitely with higher frequencies than 60 Hz.

Screen refresh rate (rate that each image is replaced) is a lot different to the frequency of the PWM applied to the backlight. If the backlight were pulsing at 60Hz people would be _very_ upset.

On CRT screens yes, but backlight in LCD screens is LED/fluorescent driven by much higher frequencies. Though some LCD screens still flicker, but it is visible only on special patterns: http://www.lagom.nl/lcd-test/inversion.php

Not necessary much higher. I've measured screen flickering in Lenovo T500 (probably one of the oldest laptops with LED backlight) with small solar cell and oscilloscope and it's 220Hz, with lowest brightness setting only 100Hz (although this brighness level is not too useful): http://commonemitter.blogspot.com/2017/06/lenovo-t500-measur...

Reply on news.ycombinator.com