EX3's have any problems with 5600k Flourescent Lighting?

[Alister Chapman]

What normally happens in industrial and office installations is that lights on different circuits are on different mains phases in order to balance the buildings total power load evenly across all 3 phases. As a result the lights in different parts of an office will have different phases, so even if you get the camera in phase with one bank, others will be out. Even if the lights have high frequency ballasts you will still get phase shifting and colour banding as the lights go in and out of phase with each other.

The only cure is to have all lights running on the same phase with identical ballasts and this is very rare in European installations. Changing shutter speed on the camera will not eliminate the problem in many cases as different parts of the office are running on different phases.

Introducing your own fluorescent lights, plugged in to the offices ring main can make the problem even worse as this may be on a different phase to the office lighting and the ballast may be running at a different mains multiple to the ceiling fixtures, so unless you turn off all the office lights you may get bad strobing.

This becomes a massive problem if you want to start shooting at speed higher than 50/60 fps. FS700 users are going to be in for a fun time!

[Derek Reich]

Alister, this makes perfect sense! It certainly explains why one room would be stabilized, and an adjacent classroom across the hall would not. Where the problem was the most noticeable, was indeed an older building which was actually in the process of being renovated and updated. I was later shooting in a very new university building and had no problem whatsoever. Thanks for the great explanation.

[David Heath]

Originally Posted by Alister Chapman

What normally happens in industrial and office installations is that lights on different circuits are on different mains phases in order to balance the buildings total power load evenly across all 3 phases. As a result the lights in different parts of an office will have different phases,

That should not cause any problems. Key is that in a 3-phase system, all phases are at precisely the same frequency, but one phase is 120 degrees ahead of reference, one is 120 degrees behind. Point is that if you lock the camera to any one of the phases, it MUST be locked to the other two - albeit with a phase difference.

That won't make any difference in this context. I'd typically expect the shutter to be open for 180 degrees (a half cycle) of the 50Hz field rate of the camera. The fluorescents I'd expect to dim/brighten at 100Hz rate, as they are effectively full wave rectifying the mains. (So you get a peak at each positive peak of the mains AND each negative peak.)

Consequently, the camera will always capture a half cycle of the mains - it won't matter if it's 0-180, 120-300, or 240-60 degrees in respect to phase.

The issue has pretty well gone away with LCD screens, but filming (CRT) computer minitors always used to be tricky, and special boxes bacame an add-on in Beta days. The screens typically ran at 60-75Hz and it was necessary to adjust the shuttering to correspond, such that the shutter was only open for one single scan - more and you got a white bar, less and a black one. But note it was purely the FREQUENCY that mattered. The phase would correspond to where down the computer screen the camera shutter opened - which could be anywhere, as long as it closed at the same point exactly one scan later.

Problem was when two screens may have been in shot at the same time, and if they ran at different frequencies. You could get either OK - but not both at the same time.

[Alister Chapman]

While it is true that the lights are synchronised, all be it 120 degrees apart, the problem is that with a CMOS sensor, the exact image brightness and colour seen by the camera depends on where you are on the sensors read cycle. So perhaps as the sensor scan is at the top the nearest and furthest lights in a room may be producing maximum output, while the middle lights may be at minimum. when you get to the bottom of the scan a different pair of lights may be on while another is off. Also the colour temperature of a fluorescent tube changes through the strike period depending on many factors including the voltage, tube age etc. This leads to both brightness and colour banding in the image, that may be different from one room to another. You must remember that there is a distinct time difference between exposure of different parts of a CMOS sensor for any one given frame. The most obvious consequence of this is the banding you see from a flash gun. Even if the camera is in sync with the lights you will often still see colour and brightness changes as depending on which combination of lamps on which phase are "on" the true colour and illumination levels will be cycling as one combination of lamps will have slightly different outputs and colour to other combinations. With lamps on all three phases you will have 9 possible combinations of colour and brightness with every mains cycle and these will cycle up and down a CMOS sensor image due to the scanning process. As I said the only way to guarantee a clean image is to have all lamps on one single phase so they all come on and off together, that way you don't get different light output conditions due to differing combinations of light output.

[David Heath]

Originally Posted by Alister Chapman

........the problem is that with a CMOS sensor, the exact image brightness and colour seen by the camera depends on where you are on the sensors read cycle. So perhaps as the sensor scan is at the top the nearest and furthest lights in a room may be producing maximum output, while the middle lights may be at minimum. when you get to the bottom of the scan a different pair of lights may be on while another is off. Also the colour temperature ...........

But with a 50 field/s system, 50Hz mains and 180 degree shutter, that still shouldn't make a difference. All positions on the sensor will integrate for a 1/100 of a second, albeit doing it at somewhat different times - but in each case the integration will be of the light output from a complete rectified half cycle of mains.

It shoudn't matter if the integration is from zero through peak to zero (0 deg), or peak through zero to peak (90deg) or anywhere else - point is that the values from an entire rectified cycle - no more, no less - are being captured (and integrated). It's like throwing a big, medium and small ball into a bucket in sequence, one ball per second. Leave the bucket there for three seconds before taking it away and you'll have one of each size - the order in which they arrived (the phase!!) is irrelevant. It doesn't matter if the balls arrive small medium big or medium big small - the contents of the bucket will always be the same. (Obviously, there WILL be a big difference if the bucket is held there for a different time period, 2 or 4 seconds, say.......!)

In this context, whether the phasing is due to differences in supply phase or CMOS read likewise doesn't matter. On a big studio rig it's quite likely that the lighting may be three phase anyway.

What certainly WILL give strange results is when the camera shutter is set to something other than the above, or if the mains frequency doesn't correspond to system frequency.

I do remember one occasion when a camera had it's shutter set to make a computer (CRT) monitor look stable, In the background was a medical lightbox powered by fluorescents. All looked fine on the camera viewfinder during shooting - get it into the edit suite and the light box was seen to be cycling through all the colours of the rainbow. The camera was only capturing light from a shorter period than a single duty cycle. The perils of b/w viewfinding. :-)

[Derek Reich]

um..... say what now?

can either of you guys put that into a perspective for a simple photog who doesn't have an engineering degree? (or, more importantly, relate how to avoid this problem in the future?)

thanks. I wish you guys had sat next to me in math class......

[Alister Chapman]

David, your argument is correct, provided the camera sensor is phase locked to the lights and starting it's refresh cycle in sync with the lights, you need to take into account the fact that the flash of light from the tube doesn't last very long, so where the flash occurs relative to the read/reset point on the sensor will determine the exposure for that part of the sensor.

I've come across this phenomenon so many times on corporate shoots, so it is a very real problem and even if you are running at the frequency of the lights, you can get banding. Plus you must consider that the "shutter" isn't open for the full 1/50 1/60 of a second due to the duration of the read/refresh process.

In the case of a flash gun going off the sensor is integrating for much longer than the flash, but what you see is a narrow band of light across the image because some parts of the sensor will be getting reset/refreshed much closer to the time of the actual flash trigger point than others, so when you allow for the refresh delay, some parts of the sensor, a narrow band are exposed to the flash for a shorter period than others. If you consider that a fluorescent light is much like a flash gun, only flashing 100 times a second, then it's not unreasonable to expect each flash of the lamp to generate a light band across the image, similar to a camera flash, but broader as the flash duration is longer (but still not the full shutter duration).

As none of us shoot locked to the mains frequency the phase relationship between the camera and these, for want of a better description "flash bands" will slowly change and thus the bands will roll up or down the image.

The trick I use to minimise flash banding with flash guns is to use the slowest possible shutter speed to maximise the the period between the refresh cycles in the hope that the flashes will occur cleanly between cycles. However with a fluorescent lights flashing continuously you'll find it near impossible to get the CMOS camera refresh cycle timed to occur when the lamps are of on a single phase, let alone all 3 phases.

I'm thankful that mainstream CMOS sensors didn't arrive before LCD computer monitors, trying to shoot a CRT with CMOS is a complete nightmare, much worse than CCD as you have to match not only the frequency, but also the phase.

[David Heath]

Originally Posted by Alister Chapman

David, your argument is correct, provided the camera sensor is phase locked to the lights and starting it's refresh cycle in sync with the lights, you need to take into account the fact that the flash of light from the tube doesn't last very long, so where the flash occurs relative to the read/reset point on the sensor will determine the exposure for that part of the sensor.

Honestly, that's not so. The PHASE locking is not relevant (but see caveat below!) under the circumstances described above. If the camera FREQUENCY is locked to mains, the PHASE is irrelevant *if we're talking about a shutter opening for exactly half a cycle of mains* - 1/100 sec in the case of the UK. Just think again about my analogy with catching the balls.........

I've come across this phenomenon so many times on corporate shoots, so it is a very real problem and even if you are running at the frequency of the lights, you can get banding. Plus you must consider that the "shutter" isn't open for the full 1/50 1/60 of a second due to the duration of the read/refresh process.

I don't doubt the reality of the problem, it's the explanation that I disagree with, the bit that refers to 3 phase supplies and the CMOS cycle. Key is that the lights are flickering at 100Hz – not 50 - and the shutter needs to be open for 1/100 sec. (But again see caveat below!)

In the case of a flash gun going off the sensor is integrating for much longer than the flash, but what you see is a narrow band of light across the image because some parts of the sensor will be getting reset/refreshed much closer to the time of the actual flash trigger point than others, .......

Agreed, the timing "phase" of the CMOS scan will vary with position on the sensor.

......so when you allow for the refresh delay, some parts of the sensor, a narrow band are exposed to the flash for a shorter period than others. If you consider that a fluorescent light is much like a flash gun, only flashing 100 times a second, then it's not unreasonable to expect each flash of the lamp to generate a light band across the image, similar to a camera flash, but broader as the flash duration is longer (but still not the full shutter duration).

Now here's where I disagree. If we assume the CMOS is refreshing top-bottom, and the top flash goes off when the read is a third of the way down. It will only register on the part of the chip that is still integrating. (We'll assume the top third is in it's "shuttered" state.) Hence the flash registers below the top third line - not above.) But if the flash repeats EXACTLY 1/100 of a second later, it MUST fall within the integration period of the top third.

Put another way, if the flash rate is exactly twice the field rate, every bit of the sensor must get illuminated by one of two flashes - if a portion misses the first, it follows it must be affected by the next one. Then it starts again for the next field 1/50 sec later - one flash registers on one portion of the field - next registers on the remainder.

As none of us shoot locked to the mains frequency the phase relationship between the camera and these, for want of a better description "flash bands" will slowly change and thus the bands will roll up or down the image.

Once the system frequency is not locked to mains, that's a completely different story - but not one that's related to single or three phase.

CAVEAT All that said, it does rely on each flash being the same in intensity. But what if the 100Hz flashes aren't like that? What if half are at one intensity, with alternate ones at a lower intensity?

It follows that in the example above, the brightness of part of the frame will be greater than the other part - depending on whether it was an "even" or "odd" flash (or "bright"/"dimmer") that occurred when any part of the frame was integrating.

And I've since learnt after my last post that this is the real problem. My earlier assumptions for the fluorescent tube output assumed 100 pulses of light every second - but assumed that all would be of equal characteristic. That tends not to be the case with tubes as they age. They have an electron emitting cathode at each end, so in any mains cycle the current pulse will travel one way through the tube in the first half cycle (1/100 sec), then reverse for the other half cycle. And the point is that if the cathodes wear asymmetrically the pulses in one direction won't be the same as in the other - and that may not be just an overall brightness difference, there may be colouring differences one pulse to the next.

Under these circumstances, what I said in the previous post – “ It shouldn't matter if the integration is from zero through peak to zero (0 deg), or peak through zero to peak (90deg).....” – won’t apply, not if alternate pulses aren’t the same.

Originally Posted by Derek Reich

can either of you guys put that into a perspective for a simple photog who doesn't have an engineering degree? (or, more importantly, relate how to avoid this problem in the future?)

The $64,000 question.

1. Use fluorescents with high frequency operation.
2. Otherwise : Use a shutter speed which corresponds to mains frequency
3. Use 180 deg shutter
4. Test!
5. If a problem, suspect what I said above. Are the tubes old? Try new replacements. Phase differences/3 phase/CMOS scan won’t give a problem necessarily – but from the theory above may if a tube is behaving asymmetrically.

Derek – you earlier said “ Where the problem was the most noticeable, was indeed an older building which was actually in the process of being renovated and updated. I was later shooting in a very new university building and had no problem whatsoever.” I suspect you experienced exactly what I’m referring to – in the first case the problem was the average age of the tubes.

Does that help?

[John DuMontelle]

A great educational read for me.

I have yet to go into any office setting where all the fluorescent bulbs were the same.

Instead of trying vainly to match settings for an impossible perfection...I turn them all off and light for tungsten.

The exception is when I use my Diva lights for interviews and I let the chips fall where they may. ;)

[Dave Sperling]

As long as were concatenating our list of problems...
I've run across this one more than a few times ---
Offices with several laser printers and/or copy machines --
Many of the newer printer/copier designs use numerous short on/off cycles for their drum heating elements - this tends to produce a voltage drop induced flicker in the lights - often even when they aren't printing of copying, but worse while they are. Of course these flickers are totally random, and often hard to trace down...