Crib sheet: ND filter effect on shutter speed

This week I had more numbers floating about in my head, this time about ND filters.

About ND filters

'ND' refers to 'neutral density', meaning that the filter reduces all wavelengths (i.e. colours) from passing through in equal amounts. Looking through the viewfinder the photographer sees a darkened image with no particular colour cast. ND filters are available in various strengths. Below I consider the effect of plain ND filters, rather than the graduated ones (which are used when you want to darken a particular part of the image, such as the sky). More info in the Wikipedia entry.

Stops, light reduction, filter numbers

I've always found it difficult to grasp the concept of 'stops' of light. When looking at various ND filter strengths I was getting confused by how much a stop was, when referring to shutter speed. The numbers below explain how to figure out what effect a given ND filter will have on shutter speed.

Daylight conditions

On a recent time-lapse shoot in changeable daylight conditions, I found when shooting in Aperture Priority mode at f/8, the shutter speed would vary from 1/640 in bright sun to 1/320 when cloud passed over. Afterwards I considered what effect on shutter speed and whether it would introduce any motion blur.

ND filter numbers

Hoya and many manufacturers quantify their filter strength by using a set of numbers that double: 2, 4, 8, 16, etc. Tiffen uses a different scale of numbers: .3, .6, .9. I was researching the Hoya style when I put together the numbers below.

• ND2, ND4, ND8, etc refer to how much light passes through, as a fraction. So ND4 allows 1/4 of light to pass through. ND16 allows 1/16 through.
• If you test your exposure without a filter, you can calculate what effect the ND will have. In my example above, an ND8 filter (in bright conditions, exposure normally 1/640) will slow the shutter to 1/80. ND8 allows 1/8th of the light through, which would extend the shutter time by 8 times: 640 / 8 = 80. In the lower light conditions, normally 1/320, with the ND the shutter speed would be 1/40: 320 / 8 = 40.
• If you want to measure the effect of the filters in the units of stops of light, this is also easily figured out: 2 to the power of [stops] equals the ND number (and therefore division of light), for example ND8 is 3 stops: 2^3stops = ND8. ND32 is 5 stops: 2^5 = 32.

In practice

When shooting you need to know what the normal shutter speed for exposure would be. Then you can calculate what effect a given ND filter would have on the shutter speed. You're unlikely to have a vast array of filters so you'd have to make a judgement call on what is to hand.

What if you want to hit a target shutter speed? Say you want a shutter speed of 1/2 and you only have a ND16 filter. From the calculation above you can see that normal exposure should occur at a shutter speed of 1/32, which is quite slow for daylight. Perhaps you could achieve that with a tiny aperture, or if you were exposing for a dark area of the scene and didn't mind over-exposed areas elsewhere in the picture.

Maybe you're shopping for an ND filter to use in daylight conditions. Using the same f/8 aperture as above, you'd need the equivalent of ND320, something you can't buy off the shelf. There are ND128 filters available, but in square sheet form rather than screw-in filters, so you could use two of those stacked together (equivalent of ND256) and close the aperture slightly. However using multiple filters can introduce ugly reflections between the elements if the sun catches in a certain way. And using very strong ND can introduce a colour cast, which is more prevalent in cheaper items.


I hope this guide is useful. If you are looking for professional film making services (including time-lapse shooting of any duration from one day to several years), see the website for my new company: Construct Films.

Crib sheet: Time-lapse calculation

While on a day long time-lapse shoot today, I wondered about how long my various cameras' card space would last and how much video I'd end up with. I had various memory card sizes. In the past I'd just work it out on the fly but today I thought a quick crib sheet would be useful.

Calculating video length

(assuming playback is 25fps, as it always is here in Europe)

• 30 seconds interval = 5 seconds of video per hour of shooting
• 10 sec interval = 14 seconds of video/hr shooting
• 5 sec interval = 29 seconds of video/hr shooting
• 1 sec interval = 144 seconds (2min 24secs) of video/hr shooting

How long before the memory card is full?

(assuming 1.7Mb per picture and actual card capacity is ~90% of nominal capacity)
(1.7Mb is the average file size of 'Medium Normal' quality on the Canon 600D, YMMV etc)

• 30 seconds interval: 4Gb card in 17 hours, 8Gb card in 35 hours, 16Gb in 70 hours
• 10 sec interval: 4Gb in 5.8 hrs, 8Gb in 11.6 hrs, 16Gb in 23.2 hrs
• 5 sec interval: 4Gb in 2.9 hrs, 8Gb in 5.8 hrs, 16Gb in 11.6 hrs
• 1 sec interval: 4Gb in 35 minutes, 8Gb in 70 mins, 16Gb in 141 mins (2hrs 20mins)

How long before the battery is flat?

Canon (or your camera manufacturer) will have their own claims about the number of shots the camera will make on a single battery. The typical battery capacity in my various Canon SLRs is 1200mAh.
In recent months I've been experimenting with battery technology, and I have a custom-modified battery pack which has a capacity of 6500mAh, and in my tests will last for 3700 shots. I also have a 9800mAh pack which I will test soon.

Fun with a variable ND filter

A little while ago I set out to do a bit of casual filming using my stabilised 18-55mm lens. I wanted to shoot wide open for narrow depth-of-field so I screwed on a ND8 filter (which blocks 8 stops of light), but at the maximum aperture of f/3.5 the picture was underexposed. The weather was a little overcast but mostly bright. So I swapped the lens for my 50mm f/1.8 and the picture was exposed enough, but I was shooting hand-held so the unstabilised 50mm wouldn't do.

Back on the computer, I ordered a variable ND filter. Comprising two polarisers that darken considerably when their orientations are crossed (see video at the bottom), these filters are great when you want fine control over exposure in bright environments. As soon as it arrived I tried some experiments in my back yard.

f/3.5 with variable ND filter

f/18 without filter

These two stills are grabs from video clips I shot with 1/50th shutter. In the first picture I set the aperture first, at the maximum of f/3.5, and then adjusted the filter so that the highlights were just below clipping. In the second picture the filter was removed, and then I adjusted the aperture to retain the highlights, which turned out to be at f/18. Quite a difference, which is demonstrated in the noticeably narrower depth-of-field in the first picture.

One other noticable difference is the colour tone. With the filter, the picture looks cooler with a blue tint, and possibly desaturated. I have found with cheap ND filters that the optical quality is not perfect and colour shifts can occur, usually pushing towards blue. The effect varies with the severity of the light loss, and the quality of materials used -- one time I tried a plastic filter and the effect was something akin to picture taken with a Lomo/Diana toy camera, or something tweaked with Instagram. In the case of this variable ND filter, the colour shift is not so bad that it can't be corrected during editing.

A short video to demonstrate:

Lessons from The Wheels: Auto WB

Back when I bought my Canon 600D specifically for film making, I had everything set to manual.  Most of the settings had to be locked anyway; 25fps, 1/50th shutter, ISO 100 unless you're indoors, and so on.  This bit me in the bum during a shoot for a friend's film when I realised my first half-dozen shots of the day were at the wrong white balance setting.

So white balance went back to Auto and this served me well for a while, until recently when I was shooting The Wheels.  Once again when I was reviewing footage (back home at the computer this time, far too late to do anything about it) I found it to be all wrong.

The cause this time was the mixture of light sources.  At the location I had mistakenly assumed that the outdoor light coming into the room from two sides would be sufficient, but we ended up using a couple of domestic incandescent lamps as well.  I knew the footage would look a bit orange, but what I didn't realise was that the Auto WB setting is taken from the brightest part of the picture, rather than the centre of the frame (as auto exposure is, when taking stills).

When Auto WB gets it wrong

The pictures above show Dave at the table, mostly lit by the incandescent lamps, but the brightest part of the picture is the outdoor scene beyond the patio doors.  The Auto WB has set itself to suit the bright outdoor scene, and as a result Dave and much of the interior is a red/orange hue.  This can be corrected in post but I found the shot above was so pushed to red it was tricky to bring it back to resemble something approaching normal without ruining the rest of the frame, especially when trying to brighten the noisy shadows. (My discoveries about noise in shadows is a whole other post.)

When Auto WB gets it right

This picture shows a different composition, which does not have an outdoor scene in the background; on this occasion the Auto WB has set itself to suit Dave and the interior rather than the outdoors.  As per my usual workflow, some colour balancing was made to match the shot with the others, but you can see it required far less work, and therefore the image wouldn't need to be 'pushed' so far and degraded so much to achieve this.

Learning a lesson
When you're on an important shoot (as opposed to grabbing some shots on a day out with the family), white balance should enter your consciousness as one of those things you just need to check, like adjusting ISO to get the histogram looking right.  Unfortunately white balance isn't something that's so visible on the camera, but luckily allowances can be made in post if the results aren't terrible.  Perhaps a reminder on the shot list / storyboard would do the trick.