# LED flicker, has technology eliminated it?

Based on this question - Two pictures with the same setting come out with different lighting - and the answers & comments I did a quick test on one of my LED continuous lighting panels.

I'm aware of the potential for flicker on LEDs, as with fluorescents etc but I realised I'd never noticed it when shooting using only two of these panels. They're Excelvan GK-J-1040AS which I admit I just found on eBay the same time I bought my first DSLR last November before I knew anything about the subject at all, so let's call it a fairly random newbie choice.

I did a quick test on a lit subject, trying faster & faster exposure, to see if I could catch it 'in between' pulses... but I couldn't.

As a final test I rather uncomfortably pointed the camera straight at the light & shot it at my fastest exposure, 1/4000s The result is a bit blurry, because a) there's a diffuser on it & b) it was too squinty to really want to gaze at more than necessary.

However, the salient point is that even at 1/4000 I cannot find any dark LEDs or sign that they are flickering at all, even with multiple shots.

This picture with the light intensity set to minimum, but I saw the same results at various brightnesses, no hint of flicker or 'scanning'.

Have we reached the point where we can have fully dimmable [& had I paid extra, temperature variable] LED panels, without any flicker, or at least flicker that can't be seen at 1/4000s?
Or is there error in my method?

Histogram from the picture

• What are the R,G,B values for the lights? Just because they appear the same in every exposure does not mean they aren't flickering. It just means even at their dimmest they may still be causing full saturation. Twice as bright, or 200 times as bright looks exactly the same if the dimmer value is still enough to fully saturate the pixels they cover. – Michael C Mar 7 '17 at 9:56
• I added the histogram. Testing randomly around the RAW image, I'm seeing general peaks around 235 -245 [in all 3 RG&B] with the very occasional 250. No noticeable patches of 255s. – Tetsujin Mar 7 '17 at 10:04
• That's not how LEDs work. The ones w/ flicker are driven w/ a pulsed power source to reduce heat load. It's not a physical law that LEDs have to be driven intermittently. – Carl Witthoft Mar 7 '17 at 12:28
• @CarlWitthoft No one is saying they have to flicker. I was just postulating that they may be flickering even if the recorded value was constant at full saturation. – Michael C Mar 7 '17 at 15:11
• The ones that flicker worst of all are not the ones driven by PWM (which usually uses high frequencies), but rather the ones where they don't fully rectify the power before using it to drive the bulb, resulting in it being on for 1/120th of a second and then off for 1/120th of a second (60 cycles per second). The color temperature of white LEDs is largely driven by phosphors, rather than by the gap emissions, so the temperature should be mostly independent of voltage, but YMMV. – dgatwood Mar 8 '17 at 19:51

As you already figured out, your LEDs run of a DC voltage. If that's supplied by a battery, you will have absolutely no flicker. But if they're powered by a AC-DC converter, you might find that the produced DC voltage is not perfect. Usually, there is some small ripple. That is, small "left-overs" of the original AC voltage.

That is, in theory that's there. Practically, there are some differences:

To reduce this ripple, your AC-DC converter contains some capacitors. They store energy to bridge the rising/falling part of the rectified AC wave. For some applications that ripple should be minimal, in the mV range (for example in a computer power supply). Additionally, your normal AC flickering is at 50Hz (that's 50 times per second, for those who are younger or just not that much into physics). A switching power supplies ripple is roughly at its switching frequency. For a cheap one, that's about 100kHz = 100,000 times per second.

A cheap manufacturer will probably save money by using too small capacitors, resulting in a lot of ripple.

But the good thing is: LEDs can withstand even a strong ripple. As long as the peak voltage is not above their specs, they only flicker at the ripple frequency. And even if the ripple is really huge, that flicker will be difficult to capture by your camera: A LEDs brightness is a non-linear function of the applied voltage (the higher the voltage, the less efficient), and your camera's sensor is non-linear, too.

I can't give a reliable estimate, but a wild guess: Even a utterly crap cheap 19V AC-DC switching power supply with 1V ripple at 50kHz should produce a DC voltage that's good enough for photography. Like, the odds are way less than 1:100,000 that you can take a picture with slightly (<1%) dimmer light.

And I suppose your normal AC-DC converter is more close to 0.1V ripple at 120kHz, reducing the odds a lot more - even for moving targets/light sources.

(To give some relations: I once measured my [expensive] computer PSU with a professional oscilloscope, and even under highest load the ripple was way below 0.05V at over 240kHz).

Disclaimer: I am neither professional photographer nor a electrical engineer. And your lights are totally fine! :)

• Thanks. I do understand Hz [I'm a sound engineer] & rectifiers/switch-mode 'transformers' [very very broadly] but this tied in some bits I didn't know. The main initial revelation, as you suggest, was that they run DC which I simply hadn't grokked. The rest now falls nicely into place. – Tetsujin Mar 8 '17 at 17:35
• You're welcome. btw: The main reasons the high freq. switching stuff is so broadly used today are not the nice effects for photography. The higher the switching frequency, the less metal you need in the transformer => the cheaper they get. But SMPS tend to "broadcast" noise in the range of the switching frequency + harmonics. Good for audio (>100kHz), bad for some radio transmission bands if not decently filtered (google keyword: EMV and PFC) ;) – ArchimedesMP Mar 8 '17 at 17:46
• tbh, I arrived at the small knowledge I have on switch-mode supplies because of the myriad "my laptop tingles when I touch it" questions around SE & the interwebz in general, & delved slightly deeper to try debunk some of the "you need to plug it into an earthed[grounded] socket" badvice that goes with it. – Tetsujin Mar 8 '17 at 17:50
• I chose this as the accepted answer as it concatenates salient information already contained in the other equally good answers & presents it in an 'easy-to-swallow' form for someone like me who understands elements of electronics but doesn't posses a detailed knowledge. – Tetsujin Mar 9 '17 at 10:59

Those are LED panels specifically made for photo or video lighting. They are spec'd as "flicker-free" and use DC, so there's no AC flicker, and i assume dimming isn't performed via PWM, but reduced current.

That's very different to a simple and cheap household LED lamp, which uses much simpler electronics and works off AC.

• You could also use very fast PWM - 10kHz or more, or conversely PWM followed by a low-pass filter to take your back to pseudo-DC. – Chris H Mar 7 '17 at 12:48
• I bet those panels also work off AC (there seems to be a power chord included in the kit) using an AC-DC converter. The key difference is really in the word cheap: a $2 LED lamp can't afford to have the same electronics as a$30 light panel. – Dmitry Grigoryev Mar 7 '17 at 17:46
• @DmitryGrigoryev - they do indeed come with 19v DC mains adaptors, or they will run from Sony V BP batteries [not supplied]. I'd never realised they were DC, until today. – Tetsujin Mar 7 '17 at 18:06
• @Tetsujin cheap AC/DC supplies can still have significant 60 Hz current ripple or in higher frequencies depending on the switching controller. Depending on the LED light, it may or may not have it's own power supply that's capable of removing the variation, or it may introduce its own. A quality "constant-current" switching power supply could switch at 200 kHz to 2 MHz. If absolutely constant lighting is needed from an LED (maybe for a high-speed camera and you don't want to use a halogen for some reason), a linear supply could be used, but they're less efficient. – Nick T Mar 7 '17 at 22:33
• Ideally, you should use something faster than 30 kHz, because anything under about 25 kHz will produce an sound that may be audible, depending on how good your ears are. 60 kHz is often chosen because it is well outside the hearing range of dogs (though still barely in range for some cats). :-) – dgatwood Mar 8 '17 at 19:54

LEDs don't create flicker from nothingness. They only flicker if their drive current varies.

It's not really a matter of new technology. The technology to make constant current drivers with reasonable efficiency has existed far longer than LEDs suitable for illumination purposes have existed.

The reason many LED lights flicker is because it is cheaper to build the driver circuits that way.

• My favorites are the LED Christmas lights that I'm pretty sure literally use the diodes in series as a half-wave bridge. – dgatwood Mar 8 '17 at 2:35

LED flicker

There is no such thing as "LED" flicker. Every light source can flicker (although not every can flicker fast enough). Flicker is created by turning a light source on and off, it's a property of a power source, not a light source. Take same LED, on DC it won't flicker, on PWM it will. Make the PWM too slow or too fast and it appears to be gone again. Flicker is a property of a particular setup, not a category of lights.

has technology eliminated it?

No, why would it? The technology introduced flicker (in form of PWM) as highly efficient way of regulating and energy saving measure. The oldest and simplest, resistor-regulated LEDs never flickered. Flicker is the new thing.

However, the salient point is that even at 1/4000 I cannot find any dark LEDs or sign that they are flickering at all, even with multiple shots.

This picture with the light intensity set to minimum, but I saw the same results at various brightnesses, no hint of flicker or 'scanning'.

I think you're mixing up flicker with scanning. If the panel needs to display a picture (eg a tv screen) then scanning makes sense, because one needs to address every single pixel separately. But for a flickering lighting panel, all LEDs would flicker in the same phase.

Have we reached the point where we can have fully dimmable LED panels, without any flicker, or at least flicker that can't be seen at 1/4000s? Or is there error in my method?

We had the technology since 1843. It's called rheostat, but it means lot of heat and lot of bulk to handle that heat so it never was practical. Today you can do it with linear regulator, but the heat issue remains. On the other hand, high-speed power electronics became cheaper, so "flicker too fast to be noticeable" is more easily achievable now. But, it's nothing new as well. Higher frequency usually means higher loses, so expect popular dimmers to stay just outside human eye range, not outside high speed camera range.

You're using panels that are specifically designed for photography. So they can be "flicker-free" by either having lossy DC regulation (thus are truly flicker-free even if you film them with a 1'000'000 FPS camera), or by simply flickering so fast that it won't matter at typical photography 1/4000. Once you get a camera capable of 1/1000000 shutter, you may notice the flicker again.

The best way to catch flicker is to lock exposure, set the fastest shutter and auto-fire dozens of shots of exactly same scene, lit only by your suspected source set at minimum (you were at the right track here!). Flicker will result in inconsistent exposure. One picture doesn't mean much, it's the picture-to-picture consistency that's lost with flicker. (Sometimes, just sometimes the shutter may catch flicker, which results in horizontal banding. But most consumer-grade LEDs, especially the "white" ones, are too slow for that, just as incandescent bulb is too inert to flicker at 50-60Hz.)

• Dimmable DC current drivers are not necessarily less energy-efficient than PWM. PWM sure is the cheapest way to make a constant-current DC driver dimmable, but modern current drivers are often dimmable by design, and the extra cost is often negated by the reduced effort it takes to get the FCC certification. – Dmitry Grigoryev Mar 7 '17 at 17:16
• i like the bit about the new technology introducing flicker, that's how it historically happened ;-) but linear regulators or high frequency pwm are not the only two options. now that we know we need it, we can use switching mode constant current regulators to smoothly control the led brightness, practically flicker-free and still energy efficient. – szulat Mar 7 '17 at 18:04
• PWM in this context is still DC, so that distinction isn't strictly accurate... just sayin'. ;) – junkyardsparkle Mar 7 '17 at 18:41
• For that matter, unless I'm missing something, a switched mode regulator technically is PWM, just with filtering on the output to smooth it into DC. – dgatwood Mar 8 '17 at 21:41
• @dgatwood That's bit oversimplification, but if we look at the final result alone (output voltage/current) it's close enough. Introduce energy loses in iron/winding/ESR compared to pure PWM and the waveform is still jagged, never battery-like. – Agent_L Mar 9 '17 at 8:39

Most of the time any noticeable flicker can be traced to 50-60Hz coming from AC. This happens when a so-called transformerless power supply is used to power the LEDs, which as you can guess from its name is cheaper than a proper power supply which includes a transformer.

However, once a LED lamp is made decent enough to eliminate 50-60Hz flicker, it's very unlikely it will flicker at all. PWM dimming could be used in the cheapest varieties, but that usually means the power supply will also be cheap, and you'll see the 50-60Hz flicker as well.

Of course, there's no such thing as a 100% constant current, and even good LED drivers will have a variation of a couple of percent (usually with a frequency around 50-100 kHz), which will produce a variation in LED brightness of a similar magnitude and frequency. This variation can be measured using a fast light sensor and a spectrum analyzer, but it's unlikely you'll be able to capture it with a camera.

• There's nothing “improper” about a well-designed transformerless power supply. In fact a good switching power supply can much easier eliminate 50/60 Hz flicker than one using a transformer based on that same frequency. LF power supplies can only get rid of that flicker with capacitors fat enough to supply the LEDs through half a cycle, plus a lossy CCS stage, whereas a switching supply only needs reasonably small capacitors to smoother the much higher frequency switching ripples. – leftaroundabout Mar 8 '17 at 15:39
• @leftaroundabout All good switching power supplies I've seen had a transformer in it. You'll absolutely certainly find a transformer in a USB wall wart, for both efficiency and safety reasons. Transformerless PSUs are limited to devices like cheap LED lamps which on one hand don't need much current and on the other hand are completely finger-proof, since there's no transformer to isolate your from the 120V or 230V AC. – Dmitry Grigoryev Mar 8 '17 at 16:14
• Yes, but the galvanic decoupling property of transformers has little to do with any flicker tendency. My point is: whether a PSU has a transformer is orthogonal to whether a LED lamp using it will flicker – a good old 50 Hz trafo with giant capacitors will not flicker, nor will a switching supply with 100 kHz transformer or something based on a transformerless buck converter. A “PSU” that simply uses a triac to keep the voltage in the right range and then some simple transient filtering will of course flicker a lot, but so will a “proper” 50 Hz trafo with insufficient smoothing. – leftaroundabout Mar 8 '17 at 16:25
• @leftaroundabout By transformerless power supply I really mean this, not a 100 kHz switcher or a buck converter. That's what you're likely to find in cheap LED lights. – Dmitry Grigoryev Mar 8 '17 at 16:48

Not an answer but adding knowledge: To capture LED flicker, try moving the the camera during the short exposure. Flickering LEDs will appear as dotted lines while non-flickering sources will be solid. There's a picture of a scanner LED (using PWM) in this page that shows what I mean: https://www.metabunk.org/dashed-lights-coming-from-the-sky-likely-long-exposure-flash-flickering-lights.t2639/

• I shall try that tomorrow & report back. I also have other domestic & industrial LED lights elsewhere in the house, all on 'actual 50Hz AC' & will do a few experiments. – Tetsujin Mar 8 '17 at 17:42
• OK, I've tried every LED in the house & cannot get that effect to happen. I'll blame my technique before anything else, but all I was getting were faster & faster stripes.. & dizzy ;) – Tetsujin Mar 9 '17 at 9:52
• You can practice using just your eyes - scan them back and forth or in circles. Our coffeemaker and some electronics have flickering LEDs I can see this way. Some cars taillights flicker too (some more than others; Cadillac is pretty bad). Also, you can still get dizzy this way so only do it as a passenger! – Dithermaster Mar 9 '17 at 22:54

I looked into building a panel using cheap ribbon-mounted LEDs.

To control the brightness, there are two general ways.

Varying the voltage is easy with a manual knob, but wastes power (bad if (using batteries) and generates more heat in those components that must be dealt with.

Blinking the LED at nominal voltage instead is more efficient. But it's blinking. Classic clocks and such power one segment at a time and some people (like me) can see that, and it shows up in photos. But now there are brightness control chips that are much faster, and you can make it fast enough to not show up in your fast exposures. You can check on the electronics.se as to tradeoffs of using the fastest ones: higher price, good quality additional components needed, power usage efficiency?

It's also reasonable to set up the duty cycle of each row to be staggared, so the totql brightness (and power draw) is constant so yku won’t notice the flickering in a scene illuminated with the panel. This would, I suppose, be more “soft” than simply turning off some of the LEDs and require fewer separate lines for the same amount of control.

It’s also possible to produce a lower voltage in an efficient way now. Note that the “wall warts” containing transformers are nearly extinct, having been replaced by very tiny switching power supplies. This might not be so good for a powerful panel, but note that variable power supplies do exist.

Checking for LED flicker is easily done.

You could use a simple electronic light measurement device. I'll invent one on the spot for you:

• Purchase a small solar cell. This converts illumination to voltage.
• Put a capacitor in series (like 10µF).
• Hook it up to the audio input of your smartphone
• Install an audio oscilloscope application which will allow you to view the waveform.

This should be quite cheap. Now, there are other solutions. For example, this flashlight has a PWM dimmer:

Waving the flashlight rapidly (as shown in picture) will make the PWM flicker VERY visible to the naked eye (or the camera).

If the light is too large for this to be practical, you can stick an opaque sheet in front with a tiny hole, and quickly scan your eyes left to right. Or use your smartphone's camera mode, and wobble the phone to make a motion blur.

If the light is constant, it will be blurred into a line. If it flickers, the line will be dashed.

So, when buying lights, if you can try them in the shop, whip out your cellphone and do the test!

A bit of electronics:

PWM dimming is the simplest (ie, cheapest) way to dim a LED.

• However, it flickers
• It is accurate (on average).
• It increases switching driver efficiency (because it always works at its maximum efficiency point)
• It decreases LED efficiency (because LED efficiency in lm/W decreases at higher currents)
• Overall, it is less efficient than proper analog dimming
• It keeps color temperature constant (this also depends slightly on current)

You can modify a flickering LED light by adding Caps and resistors, but please don't attempt hacking mains-powered (and/or expensive) equipment if you have no idea of how it works...

Have we reached the point where we can have fully dimmable LED panels, without any flicker

Yes, it is doable, although a little more expensive. Manufacturers don't like extra costs, and this is why I explain how to check.

Please note that color temperature will vary slightly depending on light output, unless...

[& had I paid extra, temperature variable]

Yes, however this will likely be done with several types of white LEDs (each dimmable), so the diffuser will need to be good enough to ensure no colored shadows.

This type of panel could compensate LED color drift versus current and temperature via software calibration, IF the manufacturer implements it.

Also, since different LEDs have different spectra and CRI, there is no way to guarantee two different panel models will have the same color rendition. Even if they have the same color temperature, their spectra could differ.

Since only photographers are interested in this, expect to pay the photography equipment premium...

• Nice explanation, but it doesn't answer the question – Olivier Mar 8 '17 at 19:59
• Okay, I added an edit ;) – bobflux Mar 8 '17 at 20:27

High quality LED light sources "flicker" in the tens of kHz or above frequencies, so the flicker will be undetectable unless you are doing really high-speed photography.