(Excerpted from the December 2016 Digital Cinema Society eNewsletter)
Whether you call it flicker, pulsing, or strobing, the artifacts caused by artificial electric lighting being out of sync with the frame rate and shutter angle of the motion picture camera is almost never desired, unless perhaps you want it to look like “Saturday Night at the Disco.”
We’ve all seen it, (hopefully not unexpectedly in our dailies), what we will collectively refer to as “Flicker.” One example is when there are uncontrolled light sources in the background. This can even occur when shooting our normal 24 fps in countries, such as in Europe, that use a 50 hertz power cycle system compared to the 60 hertz power we use in the US, (and vice versa for our European friends when they shoot 25 fps).
The problem is that what we generally think of as continuous light sources are actually pulsing with the alternating current of their power source. When this pulsing is out of phase with the motion picture camera’s shutter and frame rate, all sorts of flicker, pulsing and banding can unintentionally show up in the picture.
Cinematographer Conrad Hunziker, who formerly served as DIT for most of the 7 season run of the extremely successful series, Pretty Little Liars had this to say: “Even in our ‘safe’ 24 fps land, flicker was definitely a problem to be reckoned with and watched out for. LED based systems were common offenders and would sometimes cause horizontal banding, with the possibility of it happening directly proportional to the cost of the unit. Other types of lighting instruments would also have issues, everything from neon lights to practical mercury lamps (particularly if they were on any kind of dimmers). So precisely dialing in the shutter down to hundredths of a degree, while watching on a large studio monitor to get rid of the flicker kept me pretty busy.” However, the cause of flicker that I will be concentrating on for this essay results from shooting High Frame Rates illuminated by incompatible light sources.
For my non-cinematography conversant audience, I would like to offer a quick primer on slow motion: Normally, motion pictures are shot and played back at 24 frames per second. Motion captured at a higher frame rate and played back at 24 fps will appear as slow motion. For example, images captured at 48 fps will appear to be twice as slow, and as you increase the frame rate, the slow motion effect also increases. If you doubled again from 48 fps to 96 fps and then played back at 24 fps it would be four times as slow since 96 is divisible by 24 four times.
Now there is no rule that says you have to work in 2 x increments, but also note that each time the frame rate doubles it will require another “stop,” or twice the amount of light to get the same exposure. In short, shooting higher frame rates require more light, and unless you’re shooting a day exterior, very high frame rates require a lot of electrically generated illumination.
Back in the film days, there was an established set of guidelines to deal with the issue. We would simply stick with certain frame rate and shutter combinations designed to avoid flicker that had been long established and published in places like the American Cinematographer Manual or by companies like Cinema Electronics, an Academy Award winning manufacturer of precision speed, crystal-controlled devices for motion picture film cameras; (I still have one of their little “cheat sheet” frame rate charts tucked in next to my meter.)
They still publish these charts, one for 50 Hz and another for 60 Hz, as well as having created a free downloadable iPhone app available here: https://itunes.apple.com/us/app/flickerfree-calculator/id447247056?ls=1&mt=8?ls=1&mt=8
It also helped that in the film days, we never used to shoot more than about 120 frames per second, partly because most pin registered film cameras simply didn’t go too much faster, and with film going through the gate that quickly, it made for a very expensive proposition. In addition, film emulsions were not as sensitive as today’s high-end digital cameras, and necessary exposure requirements with such short exposure intervals meant that extremely high frame rate shots were usually worked into day exterior scenes where flicker was no issue.
The move from film to digital technology has greatly reduced the difficulty of high-frame rate cinematography. Even the iPhone now records up to 120 fps, and high end Digital Cinema cameras such as the Vision Research Phantom and Flex 4K can easily reach into the thousands of frames per second range. The higher sensitivity of these cameras help make this possible, and the media is also so much less costly. Although the cost of stock is not usually much of a factor on a major feature, when you’re shooting thousands of frames to capture just seconds of action, it can be very meaningful. In fact, digital is so superior for high-speed shooting that even motion pictures primarily captured on film regularly use digital cameras for their super slow motion shots such as the very memorable super slow-motion on The Wolf of Wall Street.
Another helpful digital feature is “buffering” which allows continuous short term capture of the action so the operator can “post-trigger” the recorder rather than pre-rolling and trying to anticipate the right split second before the action occurs. Such tricky timing previously left many a high-speed film camera to run out before the action they were supposed to cover ever took place. Now, when shooting high speed, the Operator can quickly review the action right on the camera after the cut, then back up to the start of the critical action, and only save the pertinent frames to permanent storage. This not only makes sure you catch the action, but also saves a tremendous amount of media. However, just as utility has increased, so has complexity, and that old problem of flicker has been rearing it’s ugly head all too often.
Before I use the World Series as an example, I must first congratulate all you Cubs fans out there; it was a long time coming, and certainly an exciting series. The diehard fan was likely not bothered by it, but some of my colleagues were probably as distracted as I was every time the network coverage would cut to Slo-Mo for instant replay. There was major flicker and also significant color shifts from the regular coverage.
My friend Jim DeFilippis, CEO of consultancy firm, Technology Made Simple, knows a thing or two about the situation. He was formerly EVP of Digital Television Technologies and Standards at Fox, the network that provided TV coverage of the World Series. According to Jim, “High speed capture for live television can provide dramatic slow motion images. However, the tradeoff is loss of light as well as the potential for flicker due to artificial lighting at stadiums and arenas where the fixtures are not equally distributed over the three phase power distribution. In addition, gas discharge lamps (such as mercury vapor) only produce white light at the top of the discharge curve, thus exhibiting color changes during the ramp up and down, which can be seen as colored flicker.”
Issues also exist in my world of motion picture cinematography. My friends who run the allied lighting manufacturing companies Sunray and FilmGear, Ron Dahlquist and Hans Lau, approached me recently and said they were getting a lot of questions from their customers in regard to what units, frame rates, and shutter speeds were advisable to avoid flicker. They suggested I conduct some tests and publish our findings via the Digital Cinema Society. I explained that since DCS always needs to remain impartial, I couldn’t just test their lights, but would need to look at a broad cross section of other manufacturers’ products.
I figured an easy way to gather such a cross section would be to ask a couple of our other supporters, and maybe a few would be willing to participate. I was surprised when every single one of our Lighting related DCS sponsors enthusiastically jumped on board, thereby turning it into a much bigger effort than I had originally imagined. It seems that they had all been getting such queries from their customers and didn’t know exactly where to turn for answers.
Being that Vision Research, the leading manufacturer of high-speed cameras, is also a DCS supporter it made getting a hold of their Phantom and 4K Flex cameras much easier. (Note: Even though we were only recording HD for these tests, we shot with both the HD and 4K cameras since one has a global shutter while the other employs a rolling shutter. We discovered that if there was an issue it would show up either way, just in a little different form, such as a rolling horizontal bar as opposed to a fluttering in the highlights.)
Since one of Vision Research’s top software consultants, Bob Monaghan of Glue Tools, is a longtime DCS member, he was graciously willing to donate his time as a Phantom expert, (pictured left with Kevan Barsky). The only thing lacking was a location to shoot these tests and Hollywood Rentals stepped up to answer the call by letting us shoot for three full days in the spacious stage located at their Sylmar, California headquarters.
In all, companies supporting the effort included, (in alphabetical order): ARRI – BBS – Cineo Lighting – DADCO – Fiilex – FilmGear – Glue Tools – Hollywood Rentals – K 5600 – Kino Flo – Light & Motion – Litepanels – Luminys – MACCAM – Mole Richardson – OConnor – Power Gems – Nila – Red Scorpion LED – Rosco – Zylight. Most Lighting Manufacturers supplied two units for testing and we shot tests for each at 600, 1000, and 5000 frames per second; everything from a battery powered LED from Light & Motion that fits in the palm of your hand to a 24K HMI powered by a variety of ballasts.
In all, we captured over 180 separate, but fairly identical shots. Half the frame was composed of a partially filled aquarium into which we dropped a plump apple. The idea was to give some sense of the slow motion effect as the apple slowly floated down into the liquid making a bouquet of displaced water droplets. The other half of the frame was a whiteboard with our tech notes: frame rate, shutter angle/speed, and the unit being tested. This not only served as our slate for each shot, but also a large white surface is the easiest way to detect flicker. You can usually see such issues by eye on the monitor as a pulsing, strobing, or even as a slow bar rolling through frame, but to get a more analytical view we used a wave form monitor where it is even easier to see on a static background as flutter in that range of the picture.
We could publish the completed hours of footage, but the only good purpose I could think of to ever go through and view all that footage would be as an effective sleep aid. So, instead, we have selected samples that will illustrate our findings and give the broad strokes about what we discovered in the process; (this streaming presentation is still in post, but should be available soon).
Except for the “Saturday Night At the Disco” shot, where we tried to create flicker by shooting “off speed” using known culprits like an HMI with a magnetic ballast, a small tungsten unit, and a fluorescent with a legacy ballast, we were mostly trying to pick settings that would avoid the problem. 600 fps was selected as a multiple of the 60 cycle power we were using, as was 1000 and 5000 frames per second. We selected a 180 degree shutter on the camera for the 600 fps, since it is the most common in normal cinematography.
When we bumped up to 1000 fps, we went to a 90 degree shutter to give a little variety and figured this might also be a common choice DPs would make shooting action scenes. For the 5000 fps, we sometimes had to rack the shutter angle as close to 360 degrees as we could. The resulting longer exposure helped give us enough illumination with the smaller units, and it also helped to blend the light cycles and hide any visible flicker. So one take away is that if you can live with the motion blur, a wider shutter angle can be one trick to help avoid flicker.
Although LEDs still have some degree of problems delivering a full color spectrum, they have tremendous advantages in that they draw very little power, stay pretty cool, and are generally easy to use. Since they are rapidly gaining acceptance for motion picture lighting, we thought it worth testing them for flicker. In broad terms, LEDs performed very well in this regard. However, I do need to note a couple of exceptions.
One would be when LED lights running on AC power are dimmed. For example, the Mole-Richardson Senior LED and Tenner LED both performed flawlessly all the way up to 5000 fps until they were dimmed. So another lesson learned is that when using LEDs for High Speed photography, keep your fingers off the dimmer. When you get up to higher frame rates, you are usually fighting for more exposure anyway, and not as likely to be dimming your lights, so this may not be an issue.
In fact, shooting at 5000 fps with a 90 degree shutter, as we did with the Mole LEDs, requires about 8 1/2 times the light to get the same exposure; or in stops, an f22 becomes an f1.2, the full range of a super speed lens. With a couple of exceptions like the Red Scorpion LED, which puts out a blinding amount of flicker free light, (and some units from Luminys, which I’ll cover later in more detail), most LEDs were not able to give us an exposure of the magnitude of illumination needed for the higher frame rates. We were not able to get an exposure even on the white board, much less being able to see our apple falling and splashing. Instead, for the smaller units, we aimed the lights directly into the lens.
Although we did in fact occasionally see some minor level of flicker at the higher speeds with a number of the smaller LED units running on AC power, they are really not the kind of lights one would choose for extremely high frame rate shots anyway. For that kind of light output one would typically turn to a larger unit.
There was one other type of LED unit that worked pretty well, even coming up with enough exposure to shoot 5000fps, but it comes with a caveat. FilmGear brought a unit that takes advantage of their LED Tube Lighting system. These are strips of LEDs packed into a housing similar to a fluorescent tube. Like fluorescent tubes, they come in 2’ or 4’ lengths and have two power pins on each end, however they do not require a ballast. The biggest unit, called the Translight, can be loaded with 24 x 4’ tubes, and when they are of the unfrosted variety, the raw light output is enormous. We had no issues going all the way up to 5000 fps, however, when we tried using the same fixture with a slightly different type of tube, known as “Quasar,” there was lots of strobing. The conclusion is that LED tubes can work pretty well for extreme high frame rates, but be careful as to the exact type of tube you are using, and of course prior testing for critical shots is always advised when possible.
New ballast technology is helping, and the smaller HMIs, including a 1600 watt Joker from K 5600 and a 400 watt 400D from Dedolight both utilized 1000 Hz ballasts and did really well in regard to flicker. However, it’s the big guns like the 24K or 18K Sunray or the 9K ARRI we tested that were capable of delivering the substantial lighting levels we required for the higher speed shots.
It gets considerably more complex when entering the world of large HMIs. For example, it’s more than just using a square wave ballast; the power frequency of the ballast must sometimes be carefully adjusted down to the decimal point to get rid of any flicker. ARRI and Power Gems both make such adjustable ballasts specially designed to avoid flicker. However, it can even sometimes depend on whether the globe is single or double ended and whether it is vertically or horizontally mounted. You also need to be sure to give the unit ample time to warm up, (a half an hour or more is recommended). If you really want to dig into the science of HMI flicker, please see a white paper written by Dr. Phil Ellams, the Engineering Director for Power Gems, a company specializing in the manufacture of HMI ballasts. (This link follows.)
Even given the complexities, large HMIs remain a popular choice for shooting extremely high frame rates for motion pictures, and this may be partially due to the fact that they are more readily available in most productions’ lighting packages. However, if you really need a bullet proof ultra-high-speed lighting source, you can turn to lights from Luminys, the same folks who created the Lightning Strikes line. Much as we tried, we were not able to get their 30K LabLight or even the more manageable Sunsource LED 1500 to flicker, (even when we dimmed it).
Now the 30K is a monster of a light and can only run about 10 seconds at a time before needing to “recharge” for a minute or so before it’s ready for the next round. However, 10 seconds of run time at 10,000 fps is about 12 minutes duration when played back at 24 fps. It would have to be a pretty interesting shot to play continuously for 12 minutes. This is not a light most Cinematographers would carry in a standard feature package, but if you are concentrating on extreme high frame rates and don’t want to worry about flicker, these are the units to seek out.
There are, however, some situations where you might want flicker, such as to simulate firelight or subtle the varying glow coming off of a TV. Over the course of this testing, I became aware of a terrific LED unit from Zylight which is known as their “IS3”. The unit features an array of 22 separate 4-color light engines, which can be combined to match standard daylight, tungsten, or any variety of specific digitally tunable color temperatures from a warm 2500K all the way up to a very saturated blue at 10,000K.
However, the 22 LEDs can also be fired independently to create various flicker effects which can be wirelessly controlled from an iOS device such as an iPhone or iPad. Many common effects have been pre-programmed into the free iOS app such as a very believable police light effect or a nightclub strobe. I should also mention that this unit performed very well when we were trying to avoid flicker with no issues all the way up to 5,000 fps, and it was still bright enough to light the shot, where most other smaller LEDs were not.
So, to summarize our conclusions and recommendations:
1) Most LEDs perform pretty well in regard to avoiding flicker, at least up to the speeds where their output provides adequate illumination. They all played pretty well at 600 fps with a 180 degree shutter, but occasionally we would see some flicker at 5000 fps; (which is where you would probably want to think about a larger unit anyway).
2) Don’t try to dim LED lights running on AC; (Some of the smaller units are OK because they convert to DC Power, as do the aforementioned Luminys units.)
3) The easiest place to detect flicker is against a static light background. You can usually see it by eye on a regular monitor, but to get a more discerning view, consult a wave form monitor and look for the lines, which should be static on a locked shot, to flutter in that range of the picture.
4) If you’re experiencing flicker and can deal with the motion blur, open up the shutter angle, or in other words, reduce the shutter speed. Get as close to 360 degree as you can.
5) Try to shoot at frame rates that are multiples of the power cycles of the mains, (for example with 60 hertz power, as in the US, shooting 120 fps, 240 fps, or 600 fps should work better than say 537 fps).
6) If you do try to use the LED Tube Lighting Systems for the higher output they provide, stay away from the “Quasar” tubes.
7) Using a camera with a global shutter as opposed to a rolling shutter didn’t really solve any issues, flicker would show up either way, albeit in a little different form. It might be a slow horizontal bar rolling through frame as opposed to a fluttering in the highlights, but it was still a problem to be avoided.
8) Make sure to give your HMI ample time to warm up when shooting high frame rates.
9) Try to test in advance with the exact units and at the settings you’ll be using for any critical shot.
10) If you discover issues or solutions regarding flicker, please share your findings. It is only through freely sharing this kind of information, as we have tried to do here, that we can all avoid such problems in the future and be able to concentrate our energy on the creative process.
I learned a lot in the process of conducting these tests, and I appreciate that Ron Dahlquist and Hans Lau suggested it to me. I also very much appreciate all the companies that donated their time to participate and their equipment to be tested. I also need to thank Hollywood Rentals for opening up their stage to us, as well as for allowing us to test their “Reel Lite,” an LED space light that performed flawlessly in regard to flicker. Of course, we could not have performed these test without the loan of the Vision Research Phantom and 4K Flex cameras.
Most of all I need to thank the volunteer crew who worked tirelessly with me for several days of shooting to capture footage with so many different lights and settings. The crew included:
–Bob Monaghan, our Technical Supervisor and CEO of Glue Tools
–Kevan Barsky , our High-Speed Camera Operator also with Glue Tools
–David Mahlmann, our Producer, who was also slinging gear around to help make everything happen smoothly.
–Mitch Gulbin, serving as Key Grip
–Christopher Knell, who documented the proceedings, helped with set up and Post, and was even our On-camera Talent for the “Disco” scene.
Again, the companies without whose support we could not have pulled this off, (in alphabetical order): ARRI – BBS – Cineo Lighting – DADCO – Fiilex – FilmGear – Glue Tools – Hollywood Rentals – K 5600 – Kino Flo – Light & Motion – Litepanels – Luminys – MACCAM – Mole Richardson – OConnor – Power Gems – Nila – Red Scorpion LED – Rosco – Zylight.
Link to Dr. Phil Ellams’ white paper on how to avoid flicker with HMI ballasts:
All comments are welcome; please write or leave a comment on our DCS Facebook page: http://www.facebook.com/DigitalCinemaSociety