Testing LEDs and Screens (archive)

My experience detecting flicker and correlating light with symptoms prior to purchasing a flicker meter in 2023.

How do my "LED" symptoms correlate with light and screen flicker and is there any safe flicker? 

See Testing LEDs and Screens 

Stimuli that do not trigger or intensify LED symptoms for me:

Stimuli that trigger LED symptoms for me (notes from before obtaining my own flicker meter in 2023)

LED lights with ≥120 Hz flicker

Conclusion: This lighting setup causes symptoms that are too severe to be compatible with working in the light. The loss of short-term memory and severely compromised ability to concentrate make me unable to do my job.

Data for Waveform 4000K Centric Daylight strip lights being controlled by a Waveform CENTRIC SERIES "Flicker-Free" Nondimmable power supply 

Flicker frequency: 120 Hz 

% flicker (for magenta region highlighted in graph):  1.27%

% flicker (overall, determined by meter software):  2.325%; Note: I was told verbally that in a repetition of this test, the lighting consultant obtained a reading of 1.3% flicker with the meter software.

Calculation of % flicker:

Distance from min to max point measured in Photoshop: 689.60 px

Height of indicated portion of wave (magenta line) measured in Photoshop: 375 px

If the minimum of the magenta line was set at 0.2260, the maximum of the magenta line would be at 0.2318. [(0.2367-0.2260) *(375 px/689.60 px)+0.2260 = 0.2318]

% flicker = 100% * (max-min)/(max+min) = 100% * (0.2318-0.2260)/(0.2318+0.2260) = 1.27%

It is unclear to me what creates the broad peaks and valleys in the graph of luminosity vs. time. I asked this question of the lighting consultant over the phone the day the testing was being done. They didn't have a definitive answer, other than to say that the readings were made with a handheld meter that wasn't necessarily kept completely steady. Theoretically, the amount of light observed by the meter would depend on the exact distance of the meter's sensor from the light source. I was told verbally that after our conversation, they repeated the test, obtaining a meter reading of 1.3% for the flicker of these lights, but I was also told that the same broad fluctuations happened even when the meter was held steady against the wall (I'm still confused as to how both statements can be true). I wonder to what extent there really is sub-120Hz variation in light intensity and I wonder to what extent it might contribute to my symptoms. 

This strip light was installed on a wall and was the only light source in the room during testing with the meter and during testing on me. The flicker was not visible on a smart phone slow motion video.

5. LED T8 tube lights: Waveform 4000K CENTRIC DAYLIGHT™ Full Spectrum Flicker-Free T8 LED Tube Light. Two of these T8 tube lights were installed in a new fixture from PLT Lighting in which the tubes were directly connected to the line power, as directed by Waveform to achieve flicker-free light. There was no ballast attached. These have 2.679% flicker. There are 120 peaks and valleys in the wave per second, suggesting the flicker frequency is 120Hz as the computer software determined. However, there are also 60 sharp valleys per second, indicating that this pattern in the flicker has a frequency of 60 Hz (see measurements below). Since I've been regularly using flicker-free light bulbs from this same company since May 2021 without symptoms (Waveform Flicker-Free A19 bulbs), it was immensely disappointing to learn that these LED tube lights advertised as "flicker-free" causes symptoms for me and are not actually flicker-free. These were the last lights I tested on 3/23/22, after I already had a headache with pressure in my right temple that followed testing of a first kind of light and after I had already tested the second kind of light, the strip lights controlled by the Waveform Centric Series Non-dimmable Power Supply, for about 1.5 hours. When the Waveform T8 LED Tube Lights were turned on, I immediately felt a new slight twinge of pain behind my right eye. This pain intensified over about the next 10 minutes. At this point the pain was moderately intense and I decided to abort the test and to turn off the Waveform T8 LED Tube Lights for safety reasons. The sharp pain behind my right eye stopped as soon as the Waveform T8 lights were turned off. The sharp pain behind my eye did not start again when the lights with the Waveform Centric Series Non-dimmable Power Supply were turned back on. The sharp pain behind my right eye is a symptom I experience only in the presence of flickering LED lights. The intensity of this pain tends to be correlated with the severity of the flicker, the length of time I have been in that flickering light, and it tends to start more quickly if I have been recently sensitized by other flicker. Given how obviously painful the Waveform T8 Tube Lights were after only a few minutes, I decided it was not worth the risk to my health to keep them on longer to try to learn what kinds of long-term symptoms they might cause for me.

Data for Waveform 4000K CENTRIC DAYLIGHT™ Full Spectrum Flicker-Free T8 LED Tube Lights

Flicker frequency: 60 Hz

% flicker:  2.679%

The overall luminosity oscillation is at 120 Hz, as indicated by the flicker meter software, but there are extra-low points occurring with 60 Hz regularity. I count 30 of these extra-low valleys in the 0.5 seconds of time displayed in the graph of luminosity vs. time.

[No additional calculations were done for a subset of this graph since the original readings seemed relatively steady].

Two T8 tube lights were installed in a new overhead fixture, connected directly to the mains power supply. They were the only light source in the room during testing with the meter and during testing on me. Their flicker was visible to a small degree on a smart phone slow motion video.

6. LED T8 tube light replacements in old fluorescent T8 fixtures; 120 Hz and an estimated 20-100% (probably closer to 100% flicker) based on a smart phone slow-motion video. A few minutes of exposure caused symptoms lasting weeks. Details of my symptoms and images are provided below in the section "How I test for flicker." These lights were installed in the hallways of a New York City Public School building on 2/16/22, making me quite concerned for the health and ability to learn for the potentially sensitive fraction of the million NYC Public School children if these lights are being installed throughout the system. 

7. Recessed "2x2" LED lights with a striped pattern; 120 Hz flicker and additional ~2000 Hz flicker on top of that based on my slow-motion smart phone videos. Flicker percent from 4.3% at full brightness to 9.8% when dimmed based on lighting consultant measurements. I suspect that part of the reason that I found these lights to be extremely painful might have been because, in addition to the temporal flicker, they had a spatially repetitive pattern of lines of light (see photos below) that was not obscured by any diffuser. The pattern was very obvious when looking at the light - the brightness and contrast of in the bottom photo (vertical recessed fixture section of the top photo) were adjusted to approximate how the light looked to the naked eye.

2. Clusters or 2-dimensional arrays of flickering LEDs seem to be worse than single point sources when considering normal light bulbs.

3. Industrial-style LED strip lighting or 2-dimensional LED arrays seem to much more quickly trigger initial pain and subsequent symptoms than single household bulbs, even when the flicker of the industrial LED strips is much faster than 120 Hz. I don't know why these lights are so triggering for me. Perhaps it has something to do with the percent of the retina exposed to the flicker or it’s something about the spatial arrangement of the LEDs. Perhaps it’s related to the brightness of the LEDs or some aspect of the exact frequency or shape of the waveform. Perhaps it has something to do with there being multiple LEDs in a line - even though they are covered by diffuser panels, the sharp pain behind my right eye tends to spike intensely if I walk directly under one of these fixtures, following the path of the fixture. Perhaps the lights have been white-tunable and thus have included color flicker. I only know that this kind of light has been linked to triggering all of my worst episodes of symptoms, including my initial months-long 2018 symptoms (only LED strip lights in the workplace), the episode at the neuro-ophthalmologist's office in January 2019 that triggered peripheral blindness and symptoms lasting 8 days (45 minutes of only LED strip lights), and my months-long April 2021 symptoms (3 hours of mostly LED strip lights along with some LED bulbs and 2x2 LED lights).

4. Dimmed LED lights that use PWM are very likely to quickly trigger symptoms, either beginning as pain behind the right eye or beginning as intense nausea and spatial disorientation. When nausea and disorientation are the first symptoms, I suspect that the LED flicker may be interfering with my normal perception of the environment in a way that makes me feel motion-sick.

Click here for a pdf containing the above graph and a spreadsheet containing the data and sources of statistics for flickering LEDs that cause my neurological symptoms (plotted above) as well as data and sources of statistics for completely flicker-free LED lights that do not cause any neurological symptoms for me. See Background: LED Lights for a discussion of the IEEE recommended limits on lighting flicker. See Background: Health Effects of Flicker 100 Hz for a discussion of the referenced flicker visibility studies.

Other non-LED lights

Screens (especially LED screens)

Flicker in videos

Flicker on an external eInk computer monitor that has no backlight. 

I keep the optional front-light on my Dasung Not eReader off and view the screen in sunlight. The screen has a 45 Hz refresh rate and much lower density of pixels than a normal monitor so there is a choppier appearance to the graphics and the grayscale rendering. The eInk monitor does not appear to flicker and does not trigger LED symptoms when used with most software. The list below is the subset of instances where there is flicker.

Repetitive patterns, especially when moving across a screen

For example, I am sensitive to the image of escalator treads in Figure 15 of the IEEE Std. 1789-2015 report on flicker that is included as an example of the kinds of patterns to which people with photosensitive epilepsy may be sensitive. When viewed on an LED monitor, I feel strong aversion and moderate nausea and spatial disorientation. Also, on the LED monitor, scrolling the page made the image appear to flicker, which restarted a new round of LED symptoms lasting 3 hours 

Other triggers of the sharp initial pain behind my right eye that were not observed longer than a few seconds:

Other triggers of the initial spatial disorientation/nausea feeling that were only observed for a few seconds:

Flicker exposure time and symptom onset for me:

Strategies that help to prevent LED symptoms for me:

Maximizing my exposure to continuous light sources is my preferred method for preventing LED symptoms and the method that I implement most of the time. In addition to working best, it has the added advantage of not involving the use of tinted lenses that could potentially increase my photosensitivity over time.

Strategies that have been tried that do not prevent LED symptoms for me:

How I tested for flicker prior to purchasing a flicker meter:

Smartphone slow motion 240 frames per second video can detect 120 Hz flicker in the Americas. Note that some phones may now require a 3rd-party app to record true 240 fps video. My iPhone stutters the recording, obscuring flicker in the built-in Camera app and the info for the video says ~180 fps instead of 240 fps even if the app is set to 240 fps. Instead, I use the Moment app to record 240 fps video, but play it back in the built-in Photos app. Extreme brightness flicker and extreme color-to-color flicker can be detected.

(1/2 cycle per frame) x (240 frames per second) = 120 cycles per second = 120 Hz

The image below shows 6 consecutive frames from a slow-motion smart phone video filmed at 240 fps of the LED fixture that was in the common hallway of my apartment building from 2016 until we replaced the fixtures with nearly flicker-free LEDs in the summer of 2022 to eliminate the harmful health impact of the flickering LEDs; the entire frame is shown for each image. This pattern is typical of 120 Hz flicker, with a full cycle taking 2 frames, although the contrast between bright and dim varies for different lights with 120 Hz flicker. More subtle flicker may be difficult to notice when consecutive frames are placed side-by-side. The flicker for the light below is obviously visible if the slow motion video is played at the typical playback speed of 30 fps. I find the flicker when such videos are played to be quite painful and triggering of symptoms, so I have chosen to show consecutive frames, rather than the actual video, since the difference in brightness in consecutive frames is visible enough in this format.

Similar images and videos from an NYC Public school, the NYC Subway system, and a grocery store are shown on the Public Health Risks page.

The image below shows 4 consecutive frames from a slow-motion smart phone video filmed at 240 fps of the LED lights at the Blick Art Materials store at 6th Ave and 20th in NYC, filmed on April 23, 2022. These lights appear to be white when observed normally, but actually flicker between slightly red and slightly green whitish light at 120 Hz. I'm curious about whether these might be an example of color-tunable LEDs or whether they were just designed to always have the same degree of flicker between reddish and greenish white light. Less than a minute in this light while wearing a hat and shade 5 welding glasses triggered a headache behind my right eye (mostly pain without much pressure) and nausea with loss of appetite that lasted at least 6 hours, until I fell asleep.

The flicker of the color-changing LED device below is even more complex. The flicker seems slower than 120 Hz. although it's hard for me to figure out how many frames constitute a cycle, partly because the device is continually changing colors, so the time that individual colors are displayed varies. This device was very triggering of my symptoms. The image shows 21 consecutive frames from a slow-motion smart phone video filmed at 240 fps; frames are slightly cropped around the device. Notice how the frames with 2 colors show how the smart phone records each slow-motion frame by scanning from one long side of the image to the other long side, with the color changing while the scan is in progress in some cases.

(6.5 cycles per frame) x (240 frames per second) = 1560 cycles per second = 1560 Hz

The images below show the very subtle flicker from the LED strip lighting that triggered my months-long symptoms in 2018. This was the first time, at age 42, that lighting had caused severe headaches and other neurological symptoms for me. Each of the 2 panels shows 6 consecutive frames from a slow-motion smart phone video filmed at 240 fps. The entire image for the first frame is shown, then only the rightmost portion of each of the next 5 frames is shown to the right side of the preceding frame. The vertical lines between the 2 panels indicate the divisions between the frames. The top panel shows the original images. The left portion of the frame shows the LED strip through a gap in the covering diffuser panels. The right portion of the frame shows a diffuser panel that is covering additional LEDs in the strip. The bottom panel shows the same images as the top panel, except that the images have been enhanced together in Photoshop to maximize the visibility of the horizontal bands that are parallel to the long edges of the frames. In the original video, very subtle bands of increased brightness seem to move across the screen as the slow motion video is played back. This is an example of the flicker percent being about at the limit of what can be detected using a smart phone. I see between 6 and 7 bands across the frame, which would indicate that the frequency of the flicker is about 1500 Hz, as calculated above. However, the data I provide for these lights in the graph above and in the linked spreadsheet lists 1000 Hz because that was the highest frequency setting for these lights for which a lighting analyst provided statistics. There are other examples of LED strip lighting that cause my symptoms, but for which I haven't been able to detect the flicker with a smart phone, although measurements from lighting analysts show flicker.

Sometimes the banding may be much more obvious if the dim phase of the flicker is dimmer, such as in this image of a color-changing LED device. This following image shows one complete frame from a 240 fps smartphone video. This is the original image except that a black rectangle covers a portion of the image; the colors and contrast have not been enhanced. There are 5 cycles of banding visible over the device and probably a similar number of bands over the other parts of the frame. So if there are about 10 cycles of banding, (10 cycles/frame) x (240 frames/second) = about 2400 cycles/second = ~2400 Hz for the flicker frequency.

Note that if you are attempting to detect very fast flicker with an iPhone, the banding will be parallel to the longer side of the phone. I didn't realize this when I first started testing lights in 2018, but one should ideally orient the phone so that an LED strip light spans the short distance across the phone (from left to right, rather than from top to bottom) so that you have a chance of observing the banding pattern as a variation in brightness, either across the strip light or across a part of the ceiling that is approximately evenly illuminated. Orienting the phone in the other direction, which is the more intuitive way that maximizes the length of the light in the video, makes it nearly impossible to observe the banding because there isn't a region of approximately uniform brightness in which to look for brightness variation. Focusing the camera on the brightest part of the light can help to maximize the contrast in the images. The banding pattern is generally more visible for a dimmed strip light than for a strip light at full brightness; see Background: LED Lights for a discussion of dimming strategies for LED strip lights, which often introduce or enhance flicker.

See more examples of flickering lights being installed in NYC Public Schools and throughout the NYC public transit system on the Public Health Risks page.

Lessons from mistakes in flicker testing:

From my perspective as a scientist, this seems obvious. However, when working with a lighting consultant, I realized that this was not their standard practice. They tended to have all of the lights on and then they'd hold a handhold meter up near individual light sources. This probably is fine under most circumstances, but when trying to obtain data about lights that are flicker-free or have very low flicker, other lights with higher flicker in the vicinity can contaminate the tests of the lights with low flicker. This was the case when a lighting technician tested Waveform 3000K A19 Centric Home Flicker-Free 10W bulbs. I had been using 2700K and 3000K versions of these light bulbs without symptoms since May 2021 at home and at work. The 3000K bulbs that were tested 3/23/22 were installed in hanging globes in a room with refractive glass walls that curved around the the two long sides of the room. Hallway lights on either side with 7.88% flicker and the lights from rooms across the hall on either side with 9.19% flicker also shone into the room where the Waveform bulbs were tested. A lighting consultant sent me the images below of the flicker meter readings. It didn't occur to me to ask at the time whether they had turned off the nearby flickering lights hallway lights or the flickering lights in other rooms when taking the readings. Later that day, I went to the facility to test various lights on myself and a lighting consultant led me in with a blindfold. We started in the room with the Waveform 3000K A19 Centric Home bulbs, since those had the lowest flicker readings of the 3 kinds of lights we would be testing. I was told that all of the other nearby lights were off and it was safe to remove the blindfold. I could see that the hallway lights were off through the open doorways. During the testing I thought that the only lights were the globe lights with the Waveform bulbs, so it was unexpected that almost immediately after arriving I was feeling tingles of pain behind my right eye (a typical first symptom in the presence of flickering light), but if I’d look directly at the Waveform A19 globe lights, that pain would go away. It was strange that I was feeling pain in the presence of a style of light bulb that I’ve used extensively in the past without issue. It was also initially confusing that looking directly at the lights would make the symptoms better instead of worse if those were the only light sources. Those feelings of pain behind my right eye gradually became more pronounced. Eventually after about 35 minutes, I noticed that there were lights on in a room across the hall that were barely in my peripheral vision and when I turned to put those lights further behind me, I didn’t feel the tingles of pain behind my eye. In retrospect, I also think that there may have been other light sources compromising this test too since the surrounding refractive glass walls seemed fairly bright this first day of testing - especially compared to the next day. On this first day, 3/23/22, after about 40 minutes in this room I felt strong headache pain and pressure behind my right eye in the right temple area. This headache pain and pressure lessened slightly after about 45 minutes in the next testing area, but lingered the rest of the evening. 

The next day, other people helped me to repeat tests and they carefully turned off all of the other lights on the floor that it was possible for them to turn off. They found that one light in a nearby room across the hall wouldn't turn off, so they positioned me looking away from it with a pillar behind me, blocking light from that room. This second day, the glass walls of the room with the A19 bulbs looked very dark. There was much less illumination in the room overall - the light from the Waveform bulbs wasn't particularly bright, and the light generally had a warmer feel than it had the day before. I suspect that the previous day, a significant amount of the room's illumination came from other lights on the floor with cooler color temperatures. On the second day of testing, the light from the Waveform A19 bulbs felt good, like how sunlight feels for me, and I didn't develop any symptoms at all after an hour and a half in the light. My conclusion was that these bulbs are still safe for me when installed in this facility.

It only occurred to me a few days later that other lights might have been on when the flicker meter reading was made for the Waveform A19 bulbs - especially since there had been enough flickering lights left on in the vicinity by the lighting consultants to start my symptoms on the first day of testing and one nearby light couldn't be turned off at all. I asked the lighting consultant, who couldn't recall having turned off either hallway lights or the lights in rooms across the hall when the flicker meter readings were made for the Waveform A19 bulbs. They just said that the technician had held up the meter by hand near the hanging globes. Given how difficult it was for others to get all of the floor lights off the next day - since a special lighting control system was involved and the lights could be triggered back on by motion sensors and one light wouldn't even go off - I believe it would have been remembered if this had been done when the flicker meter readings were taken for the Waveform A19 bulbs. I especially think this was likely not done since not all of the other lights were turned off when testing the effect of the Waveform bulbs on me during the first day of testing.

So given the above, there's no way to know how much, if any, of the flicker in the below graph is from the Waveform A19 bulbs and how much is from the known flicker of the surrounding lights. For example, if hypothetically the Waveform A19 bulbs didn't flicker and if 5% of the light reaching the meter came from other light sources and that other light had 8% flicker, the amount of flicker recorded by the meter would be 0.4%, which is about what was recorded. Therefore, nothing can be concluded from this test about whether any of the recorded flicker actually came from the Waveform A19 bulbs. It's very disappointing not to have data from only the Waveform A19 bulbs. In order to collect data on what kind of flicker profiles can be tolerated by sensitive people, it's very important to obtain flicker readings that are definitively only from the light source being tested by turning off all other lights.

Combination of nearby lights with 7.88% and 9.19% flicker and Waveform A19 Centric Home 3000K 10W bulbs in hanging globes:

Flicker frequency: ? - I count 60 wave peaks in the graph's 0.5 seconds, meaning that the frequency would be 120 Hz. However the meter software says the frequency is 2 Hz and the left histogram with flicker frequency on the x axis has a scattering of frequencies between 2 Hz and 16000 Hz. 

% flicker (for magenta region highlighted in graph):  0.44%

% flicker (overall, determined by meter software): 0.8943%; Note: this was a hand-held reading.

Calculation of % flicker:

Distance from min to max point in Photoshop: 617.85 px

Height of indicated portion of wave (magenta line) in Photoshop: 293 px

If the minimum of the magenta line was set at 0.2283, the maximum of the magenta line would be at 0.2303. [(0.2324-0.2283) *(293 px/617.85 px)+0.2283 = 0.2303]

% flicker = 100% * (max-min)/(max+min) = 100% * (0.2303-0.2283)/(0.2303+0.2283) = 0.44%

Is the irregularity in the luminosity vs. time graph not all due to the technician taking a handheld measurement, but somewhat due to the actual light output of the bulbs being slightly irregular or from irregularity in how light from nearby lights is reaching the meter? The Waveform statistics for the A19 bulbs say "0 Hz" for their flicker frequency, suggesting that there is perhaps not enough flicker from the bulbs themselves to measure?