There are so many options when it comes to shopping for LED lights, but no really good way of telling how the lights will look on camera. Will a unit look warm? Cool? Green? Magenta? I needed to find out more before I put down cash.
While I've seen tons of reviews on individual LED lights, I've found very little actual measurements or direct comparisons—especially when it comes to evaluating color (which is the big question mark when it comes to LEDs). Recent advances in color meter technology allowed me to conduct a scientific shootout. Read on for the methodology of our 11-light LED shootout, watch the 4K video, and download full-resolution R3D and TIFF files.
In this 4K video, you will see all the footage, first full screen individually so you can just form an opinion on what light you think looks good in terms of color. Then the second half of the video features the LED's directly compared to the control plates, side by side so you can see how close or far they fall from the Tungsten and Daylight sources. This video is not a standalone, self-contained review. It is meant to be viewed with this article and the in-depth analysis that follows. You may want to continue reading for more background and then come back to watch:
Why this test is only possible with a Sekonic C700 color spectrometer
I called my friends at Sekonic to see if I could borrow one of their latest color spectrometers, the C700. A color spectrometer is like a light meter, except for color. The new technology in the C700 Meter makes metering LED lights possible, because it can measure color spikes.
Older generation color meters functioned on an RGB sensor that measured the difference between the three colors. Think of it as three monochrome luminance meters with Red, Blue and Green dyed filters over them, sort of like how technicolor works. The information gathered from the three sensors was interpreted, then a value was displayed. It worked really well on film, where the three RGB dye layers functioned essentially the same. This was back when the only light sources you would see were tungsten, the sun, an electric arc lamp, fluorescent and HMI. The advent of LED and its nature of emitting imbalanced spectrums of light, could give false readings on the older RGB spectrometers, similar to fluorescent and low quality HMI's.
Sekonic has used a new design with the C700, a CMOS-based system that is sensitive to far smaller changes in the the wavelengths of light, essentially instead of RGB, its sensor's photosites are more like a rainbow made up of a couple dozen subtle shades ranging across the visible spectrum. What this means is the C700 can see spikes in color. If a light has a relatively smooth spectrum, but a gap in one wavelengths and a spike in another, a coarser RGB reading of older meters essentially averages them together and it doesn't really reflect the imbalance. The C700 with its finer sensor detail can detect the spikes, and references that value in its results. This is one of the only meters that can accurately measure the color quality of non continuous spectrum LED, HMI, fluorescent and neon lights. Since everyone these days loves LED lights, I felt it was important to know what the units I was going to buy were actually doing.
Thanks to B&H Photo Video
Jump forward a couple weeks, and with C700 in hand, I visited the only place I knew of that would have more than a couple LED brands on hand: B&H Photo Video. I brought along a scrap of duvetyne so I could shield the meter from the overhead and as many other light sources as I could. The staff looked at me funny at first. Here was a guy draping every light on display with a black cloth then diving under it and turning the light on. Once I finally explained what I was doing, and showed them the meter, it clicked. They let me have run of the store. I chose a set of LEDs based on my findings and I will get into what I chose later on. I learned so much that day, that I felt like I should share my results, as unscientific as they were... but totally unscientific just doesn't work for me. Since I used the C700 to such good effect, I called Sekonic to see if they would be interested in making this test possible. Luckily they said yes! So here we are. After many months of work, I present the LED color comparison test. I also must stress a big shout out to B&H for supplying us with the lights for the test.
We deduced that the best way to measure the color rendition was to have two control groups, Daylight and Tungsten—the two standards which film specific LED lights are trying to emulate.
The comparison process
My team consisting of:
- Ab Sesay - Photographer and producer who helped me organize and find a location to shoot.
- Nicholas Wise - Most amazing DP who graciously lent me his RED Dragon when mine was in the shop, and helped me sort out the technical details of the shoot. The camera was equipped with Skintone High light OLPF ( best color rendition).
- Ali Cengiz - my ride or die 1st AC.
- Peter Gagnon - DIT, notetaker and BTS photographer
- Geoff Gresh - Editor who assembled the video and made it all make sense visually.
We had a few goals we wanted to achieve. First, we needed to remove as many variables as possible within reason. We deduced that the best way to measure the color rendition was to have two control groups, Daylight and Tungsten—the two standards which film specific LED lights are trying to emulate. Then shoot the LED lights as a sole source in a pitch black room at their corresponding"Daylight or Tungsten" settings, so no contamination from any other units would skew the results. We filmed the Daylight control at midday when the sun was at its highest arc, on a cloudless day, with the sun to our backs so the blue North sky would have minimal effect. We shot the Tungsten control in the blackened room with the ARRI 1K studio fresnel as the sole source of light in the room. We then white balanced the RED dragon set to RedGamma4, and DragonColor2, to the same grey card used in all the setups. We performed several manual white balance operations back to back to ensure consistent results on the grey card in the sun and under tungsten. We then set each of the lights on the same scene, with two models, one pale caucasian and one olive skinned model. The lens used was a Sigma 50mm F1.4 ART set to F2.8. Exposure was dictated by the distance of the lighting unit from the grey card. For the Sunlight control shot, exposure was controlled with shutter speed at a fixed aperture of F2.8 with no ND to skew results. The grey card is the important element and where you, the reader will be able to discern the most information from the test. Each scene will be available to download as a R3D still, and TIFF file for further examination in either Photoshop or REDcineX.
What we are looking at is color reproduction: how far from the standard daylight and tungsten are the LED lights? We are not looking for light quality or usability in this test, though some notes were taken and will be discussed in each of the sections with the corresponding lights. Each scene was metered with the C700 and its reading recorded to help us understand what's happening with the spectrum and CRI rating. Color spikes, and gaps will be clearly visible on the graph.
Important notes on color temp and "Ra"
- The color temperature on the slate is what the camera read as "white" under the control lighting. The Tungsten light read as 2956K on camera (the meter read 3055K) and the sun read 5375K on camera (the meter read 5464K). This Kelvin number is not absolute on the camera. The camera is taking the visual information of what's actually hitting the sensor, and generating a "white balance". The discrepancy of Kelvin ratings is essentially the OLPF, Lens, imperfections in grey on the card and the internal algorithms of the camera itself. The meter is reading the photons coming off the source without any barriers. However, once this Kelvin rating was established on camera as "white", it was locked for the remainder of the Daylight or Tungsten portions of the shoot. This way what you will see visually is how far off of the control the LED lights will be in terms of color
- Lights that are BiColor were tested both at the Daylight setting, and the Tungsten setting. We set the units to what the manufacturer deemed "Tungsten" and "Daylight". If the unit has a numerical read out we set it to 3200 / 5600 depending on which group it was in. If the light was daylight only and came with a correction gel from the manufacturer it was then inserted and tested in the appropriate group. This only applies to the Dracast LED 1000 D (daylight only) as it was the only light that had a "stock" CTO filter and was tested in the Tungsten group alongside the bi-color units.
- The CRI index is represented on the meter as the Ra number. An Ra of 99 is a CRI of 99.
Download the full-quality R3D and TIFF stills for yourself
These images were extracted from the source footage. I am including them so you can take them into your NLE, Photoshop or RedCineX and you can balance the images to the grey card to see exactly how much tint, warmth, or coolness is present versus the control. For each unit you can determine for yourself: should you ever need to mix LED's into an existing tungsten lighting package, or need one as a fill light in a daylight scene, how far from neutral will the LED be?
Analysis of the control plates
Well what can be said about the sun? It is bright and perfect. Notice how dense the spectrogram on the meter is. Nearly every wavelength is fully saturated. This is what the Daylight LED's should mimic.
Tungsten performing perfectly as expected. Notice the nearly straight line of its spectrograph. That is how a tungsten source should look. This is a great way to compare the performance of the LEDs in Tungsten. How even is the spectrum of color output on the meter? Spikes are where things go awry to your camera: you want the smoothest band possible. The human eye can compensate, but the camera cannot. This is why the "fluorescent" look is so green on camera, but not nearly so bad to your eye. Your brain can filter our the green spike.
Analysis of each LED light
I chose a wide spectrum of lights to test from popular manufacturers. I took everything that was available from B&H. The selection covers all bases from entry level to top shelf (prices listed are at time of publishing). The lights were tested in ascending price order.
The Ikan PL90 unit is basically an eyelight. It measured 89.4 CRI and 5374K. It's small, dim and the most inexpensive of the group. We had to push the exposure up to 1600 ISO to get an exposure at the wider frame. In the video, a close up where the light is moved much further is available to view. The close up is also included in the stills package for closer inspection. It was also only available in daylight so is tested accordingly. It has a handy circular shape so it's catch lights will be circular, and its shadow quality will be smooth. Respectable considering the price!
The Cineo Matchbox is a remote phosphor LED. This is a new technology as its relies on internal LED's operating at a specific Wavelength, activating a phosphorescent sheet at the face of the light. This multiplies its light output for given wattage and because it is a chemical process, it actually produces a rich, high quality light. The Area 48 lights operate similarly. The Matchbox is very small, and thus its output was low, the wider shot had to be pushed up in ISO to match the others. The closeup in the video shows it at proper key intensity. This would be a killer light for tabletop where accurate color is a must, but heat is bad. Think food, product and potentially small controlled wildlife, think insects or small animals on a set.
The Generay SpectroLED 500 is a decent light: bi-color, decent color reproduction. A bit dim at full output, as its bi-color with only 500 LED's, meaning only 250 LED's are lit up at a time unless you blend them into a middle color temperature between Daylight and Tungsten. One of the more affordable units, but limiting. With a fast camera, it will feel less limiting. Does not come with any accessories, Barndoors, filter holders etc...
The Ikan IFB576 was a bit of a surprise. It's Daylight performance was a bit underwhelming, but its Tungsten performance was quite good as it sported a 3157K meter reading, and 96.2 CRI. Up there with the top echelon of LEDs. It's not the brightest light ever, it suffers from the same problem as the Generay, in that only half its LED's are dedicated to one color so its going to be only half power at either extreme of the color knob. It also has a strange loud beeping sound when you change its brightness setting. Another odd quirk is that the barndoors are gold/silver on the inside. It causes the edges of the beam to be a touch warmer than the center. If you can mitigate its quirks, it is potentially a lot of light for the money.
At time of test, we had incorrect information on its price: we had a figure of $825, hence its place in line.
The Limelight Studio Light was only available in Daylight, so it was omitted from the Tungsten test. It looks average, with a slight green tint, pretty much exactly what LED's are stereotypically known for. Its control panel is interesting. Its an 8 bit light with 255 steps of brightness.
These are the LED's I wound up buying. The Dracast LED 1000 D are simple panel lights, with built in DMX control, all metal housings, and the color reproduction in Daylight is superb. The 1000 all daylight diodes really throw out a lot of light. If you pay attention, the distance of the light from the grey card distance is listed in the slate in each frame. This can give you a sense of how bright the unit is. The bigger the number the more output it has. This LED can register in broad daylight as a fill light. It is available in a Bi color, but I personally prefer the punch of having all 1000 LED's pushing daylight, when I likely need the intensity. The CTO corrected frame is very warm, however note the CRI is still pretty high. If you were to color correct the R3D still frame using the WB dropper tool I bet it would lock onto neutral very well. Perhaps with better CTO gel the results would be better initially. This to me struck the best balance between price, power, and light quality.
The Westcott Flex Light 2x1 is a very interesting light. In addition to being a great performer, especially in Daylight, it is literally a flexible shape holding mat. It is definitely a specialty light. We found it awkward to use and rig, but when you need a light to wrap around something, or bend into a small space, the Wescott Flex is going to get the job done. One of its odd quirks is that its LED's are spaced very far apart, and without diffusion, it causes a strange checkerboard shadow to form. This light needs diffusion. (ed. note: at No Film School we have some of these in our kit, because of their extreme portability)
The Lite Panel Astra Bi color is the output king: by far the brightest light we tested. In Daylight, I find it very green, however in Tungsten, it's the champ. Nearly exactly 3200K on the nose, and a 97.3 CRI to boot. Its also not terribly expensive, but has lots of plastic parts. If the Dracast can be all metal at $400 cheaper I think the Astra can be metal too. One great feature is that the power adapter block is built into the light, no more AC power block dangling when the light is raised on a stand.
The BBS Lighting Area 48 Soft Light is another Remote Phosphor Light. The unit only came with the Daylight Phosphor insert so we only tested for daylight. This is where the price takes the big leap up. I was honestly surprised the CRI rating wasn't higher as its image looks great. Anything above 90 looks pretty good to camera, but they really lock on above 95CRI and enter a very natural looking territory. I find the Area48 to look quite fantastic despite its metered reading. The spectrograph also looks rather nicely balanced and saturated. The build quality and fit and finish are truly something to behold.
I find the KinoFlo Celeb light to be the overall most pleasant in terms of color. It has sense of balance where, despite its slight warmth in Daylight and slight coolness in Tungsten, the colors seem to track together with little in the way of magenta or green. If money were no object, the Celeb is what I would use on a daily basis. On an aesthetic note, it's beautifully built, has built in Wifi for control, and a VERY cool looking red LED display on the back. Also, it has the most "soft" native output. Its front face is a sheet of milk glass-looking material.
The ARRI Sky Panel is by far the most expensive unit we tested. Its performance was very good: solid CRI numbers and "accurate" color readings that come very close to both 5600K and 3200K standards, yet I found it very warm. However, like the Celeb, its overall balance is good. No major green or magenta shifts. On paper it comes close to the mark, but visually it's not particularly close to what the control plates looked like. I think with a simple click of the WB dropper on the grey card in post everything will line up perfectly. Exquisitely crafted light dripping with quality and features. Given the price I think these features and light quality should be expected.
The purpose of a test like this is to help the cinematographer choose the best tool for the job. Seeing first-hand the performance of the various lights, my own opinion formed rather quickly. I like the Dracast LED D and Kino Flow Celeb. My needs are for a daylight unit, so naturally I'm attracted to the excellent daylight performance of the Dracast and Celeb, while the Dracast having a relatively low price and more output. Some other team members loved the look of the Wescott Flex, while others thought the Area 48 light was the top performer.
I hope seeing the images posted in this article give you a better understanding of what you can expect when buying or renting LED lights. I felt there was a missing element to help people choosing between different LEDs. I was once in that position and it was overwhelming how many different price points and feature sets are available. Hopefully this shed some clarity and can help when the time comes to make a decision.