Using HMI lighting in high speed filming applications

Shooting at high frame rates requires substantial lighting levels which makes HMI sources an ideal choice. However, the process can reveal artefacts in the light which would not be apparent at regular shooting speeds. Dr Phil Ellams of Power Gems gets to the source of the potential issues and offers advice for overcoming problems when shooting high speed with HMI lamps.

Problems of high speed filming with HMI lamps

HMI lamps need to be driven with an alternating current to prevent migration of the constituent chemicals. This is implemented using a square wave current which ensures that the light output is essentially constant, however, the lamp current takes a finite amount of time to change direction, and this means that there is a short dip in light output at each square wave transition. At regular shooting speeds this has no noticeable effect on the recorded image, but at higher shooting speeds this can become visible.

A second issue is the path taken by the arc which can be substantially different on each half of the square wave. This can result in a rhythmic variation in the light output from the fixture – particularly in the case of PAR fixtures which have tightly-coupled optical systems.

A further concern is irregular instability which can occur in the arc due to effects such as turbulence in the gases in the arc chamber. The frequency of the square wave current can have an effect here by stimulating ‘acoustic resonances’ in the arc chamber. The effects of arc instability can be a challenge for both high speed and regular shooting speeds.

Main problems

1. Square wave transition time.
2. Rhythmic variation at the square wave frequency.
3. Irregular arc instability.

Let’s take a look at each of these in turn and see what can be done to eliminate their effects :

1. Square wave transition time

Sometimes, at Power Gems, we are asked ‘what is the maximum frame rate I can shoot with your ballast ?’ The upper limit is determined by the time taken for the square wave current to change direction as there comes a point where the camera will start to ‘see’ the transition. But rather than asking what the maximum frame is, a more accurate question would be ‘What is the minimum capture time or shutter angle’. The advantage of considering minimum capture time is that it is an absolute number, whereas minimum shutter angle requires frame rate also to be taken into account.

The risk of encountering problems with square wave transition increases as the camera capture time approaches the transition time. But how long is the transition time ? Somewhat surprisingly this is determined not by the ballast, but by the inductance of the lamp circuit (predominantly the inductance of the head feeder cable, and of the tesla coils in the ignitor).

Due to this inductance, the higher the lamp current, the greater amount of time it takes to reverse direction. However, typical figures are 20-30 microseconds. This doesn’t mean 20-30 microseconds of complete darkness, as there is lamp current present throughout this period, but the overall light output will dip.

If the transition period takes up a significant portion of the camera capture time, then it can start to become detected in the playback image. This may appear as dark frames, or dark bands, depending on the camera shutter system.

What is the acceptable limit for capture time ? Well, flicker is a subjective issue and can only be satisfactorily determined by eye, and as mentioned, the transition time is a variable quantity dependent on a number of factors. However, tests show that the point where transition time will become an issue is in the region of 100 micro-seconds (100,000 nano-seconds) capture time. So, if you are approaching this capture time, or have a particularly long head feeder, then you should carefully check the images for transition-time effects.

What does this mean for frame rate ? With a 360° shutter, this equates to a maximum frame rate of 10,000 frames per second. Or with a 90° shutter a maximum frame rate of 2,500 frames per second. As mentioned, a long head feeder cable or high lamp current such as found with 18kW or 24kW fixtures may necessitate a wider shutter angle.

2. Rhythmic variation

A substantial difference in the path taken by the arc in the two halves of the square wave can lead to a noticeable rhythmic variation in light output from the fixture, as the light collected by the reflector varies. Increasing the frequency of the square wave current can reduce the variation in the path between the two half cycles.

Reduction in rhythmic variation is particularly pronounced when changing the ballast output frequency from 100Hz to 300Hz. In some cases, there can be further improvement by changing from 300Hz to 1000Hz. But there can be an additional reason to change up to 1000Hz, and that is to move the rhythmic variation to a frequency where there are multiple oscillations of the light per frame which dilutes, or cancels the effect. (For digital sampling fans this is a consequence of Nyquist’s theorem.)

A disadvantage of operating at 1000Hz is that is can stimulate strong acoustic resonances in the lamp, producing light variations at unpredictable frequencies. This means that it is essential to ‘tune’ the ballast frequency to avoid resonant peaks in lamp operation. At Power Gems we developed an auto-scan system to make this process easier. With auto-scan, you can sit back, and the ballast will automatically scan for you, and move to the frequency which best suits stability. This pioneering feature is available as standard on the 9kW, 18kW and 24kW models, to save time and improve stability.

3. Irregular arc instability

The gases inside the arc chamber of the lamp can be quite turbulent and in some cases will disturb the arc, causing movement in the light output pattern. Sometimes the disturbance will happen at a frequency which is not visible to the naked eye, but can appear as ‘breathing’ of the light when played back from a high speed recording.

Lamps positioned with the arc near to the vertical position have worse stability than having the lamp horizontal, so this can be a consideration for fixture type or fixture positioning. Additionally, certain square-wave frequencies can stimulate ‘acoustic resonances’ in the lamp (the same principle as creating sound by blowing across the top of a bottle). This can be a challenge when working in the 1,000Hz mode, and it is necessary to tune the ballast away from resonant peaks.

Lamps fresh out of the box can display some instability and benefit from a few hours of burning in. Similarly, end-of-life lamps can start to run unstable and should be replaced.

Tips for high speed filming with HMIs

  • Use the maximum shutter angle possible. Stay as far as possible away from the 100 micro-second (100,000 nano-second) minimum capture time.
  • Set the ballast to 300Hz mode to reduce rhythmic arc variations.
  • If ‘breathing’ in the light pattern can still be seen on playback, then select 1,000Hz mode to shift this to a non-problematic frequency. Remember to run auto-scan on the ballast, or manually tune to avoid acoustic resonances.


HMI light sources are available with enormous light output making them great tools for capturing images at high speed. With the sort of maximum frame rates typically used in cinematic filming (1,000fps, 2,500fps, 5,000fps) perfect results can be achieved by following the simple steps described above. In most cases, the 300Hz ballast setting will give flawless results, and avoids the need to tune the ballast frequency. The 1,000Hz setting is another ‘tool in the bag’ for difficult situations.

For specialist situations such as scientific analysis with frame rates approaching or exceeding 10,000fps, it may be necessary to use multiple fixtures to dilute the effects of the square-wave transition.

High speed filming – solve problems with Power Gems

Only Power Gems ballasts have the range of frequency options and built-in Auto-scan feature to enable you to get the best high speed shots.

Detailed testing was carried out in conjunction with the Digital Cinema Society to demonstrate how to effectively set up for high speed filming. You can see footage (on Vimeo) with a range of shots to demonstrate how problems can occur, and effective ways of dealing with them :

Footage from tests on 9kW PAR

Example of ‘rolling bands’ at 300Hz, the video shows flicker caused by transition time
An example of the issues of shooting at high speed, here we see bands rolling down the screen. Shot at 2,400 frames per second, with a very short capture time of 5 micro-seconds. Power Gems 9kW ballast set to 300Hz.

Changing the frequency from 300Hz to 1,000Hz merely changes the frequency of the bands
Here, we change the frequency to 1000Hz to try and get rid of the rolling bands visible at 300Hz (see above), but it doesn’t solve the issue – just changes the frequency of the bands. Shot at 2,400 frames per second, with a very short capture time of 5 micro-seconds.

Solution : Shutter at 90° gets rid of rolling bands visible at 300Hz and 1,000Hz
Changing the frequency doesn’t solve the problem, but increasing the shutter angle does. This video shows the solution, still using 300Hz on the Power Gems 9kW ballast, is to widen the shutter angle to 90°, giving capture time greater than 100 micro-seconds.

Footage from tests on 18kW PAR

Example of ‘rhythmic flicker’ at 300Hz
Here’s a flicker problem, but it’s not transition time – we know this as we have a huge capture time. This time it’s an arc movement issue.

Shot at 1,000 frames per second with substantial capture time of 500 micro-seconds. Power Gems 18kW ballast set to 300Hz. There are no rolling bands, but there is a subtle overall flicker due to rhythmic arc movement.

Solution : Changing the frequency to 1,000Hz clears the issue by moving the variation to a different frequency
This shows the issue of rhythmic flicker is removed by using the Power Gems 18kW ballast on 1,000 Hz setting. Shot at 1,000 frames per second with substantial capture time of 500 micro-seconds.

Banding returns at 2,400Hz
Just to prove a point – if we increase frame rate (decreasing capture time) we re-introduce a transition-time problem.

Keeping the Power Gems 18kW ballast settings the same, the frame rate is increased to 2,400 frames per second which reduces the capture time close to the 100 micro-second boundary. Now we see ‘banding’ start to become a problem again.

Source :