Today I want to talk briefly about dynamic range, but not that of a camera sensor. Instead I want to talk about the dynamic range of a light.

As I mentioned in a previous blog post, I’m working on designing an LED Light Panel controller and part of designing such a thing is figuring out what the operating range of the light should be. Specifically I need to make decisions on two key factors, how far the light can be dimmed, and what kind of resolution I want to have in the dimming (e.g. 1/3-stop steps or 1/10-stop steps).

Now this may sound like an obvious answer? More is better right?

But this is going to be digitally controlled, and that means that it is subject to a whole host of “limitations”.

Anybody who’s done any research into digital cameras, expose to the right, and so forth will be familiar with the relationship between bits in a raw file to stops that can be recorded.

For those that aren’t, let me briefly summarize.

Camera sensors, and most raw formats, are linear in nature, while most photographic things are exponential (specifically powers of 2). So when we talk about something being a stop brighter, we’re talking about it being twice as bright, and on that sensor or in that raw file, it uses 2 times more values.

So if you start at 1, and you go a stop brighter, you have 2, and a stop brighter than that is 4. And a stop further still you have a value of 8. And so on until you’ve doubled as many times as there are bits in the raw file (so 14 times, or 2¹⁴ for a 14-bit file).

There’s a number of interesting things that happen with this. As you move up the scale, you get more values between each stop. Between 1 and 2 there are only 1 intermediate value, but between 2 and 4 there are 2, and so on and so forth.

What this ultimately means is that when you map an exponential system to a linear one the way digital values map to light levels, the resolution changes with the brightness. At very bright levels, there’s a lot of resolution for what ultimately is largely unimportant variations. And at low brightness levels, there’s very little resolution and potentially even major variations in brightness get tossed into a few buckets.

The same applies to a digitally controlled light. Whether the light is controlled with pulse-width modulation (PWM) or via current and/or voltage regulation, every time you dim the light by a stop you lose half the possible range of values that you can use to adjust the system.

So what are the implications of this?

For starters, there are hard limits to the ranges of PWM counters, digital potentiometers, and DACs. Many digi-pots are limited to 7 or 8-bits worth of wiper positions. Most PWM output on inexpensive micro controllers are likewise limited to 8-bits. And going beyond that, on some systems the resolution of the PWM affects the frequency of the PWM output.

The other aspect is that the resolution of the control becomes increasingly course.

Assume for a moment that a control system has 8-bits worth of values. For the first stop of dimming there are 127 intermediate settings, but nobody is going to really use them. I mean what’s the point? You simply can’t tell in a picture or by eye the difference between 100% power and 99.6% output. However, on the bottom end of the spectrum, dimmed 8-stops, you have a choice between on and off, and nothing else. If you want a system that can support, say, 1/3-stop adjustments over the whole range, 8-bits doesn’t provide enough resolution to do without jumping though hurdles in the output state.

There’s another angle to this problem too. The actual light output, and how much light is needed by the camera under reasonable conditions.

Does it really make sense to have a 500 lumen LED panel that can be dimmed by 8 or 10 stops? I’d suggest that less than 2 lumens aren’t actually useful for video even with a modern video camera. However, if we’re talking about a 6000 lumen LED panel, being able to dim it to 100 or 200 lumens — almost 13 stops is probably a useful feature.

So what’s the big point here?

Well to start with, if you’ve ever wondered why say a Canon Speedlite 600Ex of Nikon SB–910 can stop down further than a Canon 430Ex III or Nikon SB–700. This is a part of why. It’s not that Canon or Nikon couldn’t make the flash go lower, it’s that there’s doing so wouldn’t be useful — or at least not useful for the additional costs to do so.

The second point is something I never really thought about when it came to cameras and digital files; that’s the reduction in resolution as you move down the scale. Not having enough intermediate values between two stops is a big deal, or at least it can be.

As for answering any of the questions I’ve attempted to raise in this piece. The biggest answer is lies in testing. And right now I simply haven’t done enough testing to be able to say how low you can go and how fine of resolution you actually need. I have some guesses, but getting into those is something for another more focused article.

With that said, I’m going to cut this off here, though I’ll certainly be diving deeper into the issues I’ve raised and a lot more testing in future posts. In the mean time, this is something to think about.