The RAW factor – Is it better to shoot overexposed or underexposed?

Studying Steven Ascher’s The Filmmaker’s Handbook has left me somewhat perplexed and made me reflect on the issue of exposure when shooting RAW. Or rather, what lattitude of error exists in wrong exposure. Is it better to overexpose or underexpose RAW footage? Exact and balanced exposure is of course always the preferrable, but what is worse when shooting RAW, to overexpose or to underexpose? Doing a Google search only made me confused as I heard different explanations and preferences. Some like to treat RAW as Video, i.e. never overexposing for the fear of blowing out highlights and preferring to underexpose to be able to recover details in shadows. Others treat RAW as old negative film stock, i.e. preferring to overexpose as there is supposed to be more information in the highlights compared to shadows, when it comes to the Log curve profile, and thus fearing crushed blacks. Steven Ascher, in his otherwise excellent book The Filmmaker’s Handbook, never mentions RAW per se in this instance, but does say that a Log capture should never be underexposed. Ascher often compares RAW to Log so I guess the same goes for RAW, which I surmise shouldn’t be underexposed. Intuitively, I see a logic in treating RAW as film, i.e. never underexposing and to be forgiving in overexposure. RAW behaves as film in many respects, such as responding to light metering (again referred to Ascher). But I cannot base my cinematography on a hinch, so I went over to the Blackmagic Forum and BMCuser and asked the forum members on their opinions on this matter, based on their own experience. Luckily for me, there were some answers which I want to share with you in a paraphrased and somwhat reinterpreted format.

The Blackmagic Pocket Cinema Camera histogram
The Blackmagic Pocket Cinema Camera histogram exposing to the right

There seems to be a general consensus not to underexpose RAW footage, in particular when shot through Blackmagic Design Cinema Cameras (i.e. the BMCC, BMPCC and BMMCC) which generally performs best in the range of midtones and highlights, rarely experiencing highlight clipping (except on spectuacular reflections). When it comes to overexposure there seems to be a tentative agreement that it is better and safer to overexpose than to underexpose in RAW (as underexposing two stops, the footage will still be darker than expected at ISO 1600, and super noisy), however with caution; one should only overexpose slightly to feed the sensor. Many cinematographers use the formula of exposing to the right (ETTR), or in other words allowing the sensor to receive as much light as possible without clipping the highlights. This is done by pushing the peak of the histogram (as presented on the camera screen) off centre and somewhat to the right. The reasons behind this is to provide as much information as possible into post where one may make all of the aesthetic choices concerning exposure when grading the image. Shooting at ISO 800 (being the native setting for the Blackmagic Cinema Cameras and a good baseline to start off) and overexposing, in grading one can easily reduce one or two stops in the Camera RAW setting, with good results. Many others pull one stop by rating the ISO of the camera one level lower, i.e. 400 instead of 800  which overexposes the image. Let me quote the BMCuser Howie Roll (typo corrected and duly edited):

On daytime exterior shots I’ll ETTR. I’ll put clouds at about 95% on the zebras, or if there are no clouds, 80% on a clear blue sky assuming the subject isn’t in deep shade in which case lights may be necessary otherwise I might only use some bounce. For interiors and nightime exteriors, after much testing, I found a happy balance rating the Pocket sensor at about ISO 540, the mnemonic was 40 footcandles at an f/4.0 which would put middle grey at 50 IRE. I use a lightmeter for interiors so having a footcandle rating is critical for making exposure decisions. Again it’s not simple, if I have a high key setup I’ll use 40 FC for the fill side, if I have a low key setup I’ll use 40 FC for the key side at f/4.0. The advantage of figuring out the footcandles is that you don’t really need zebras or histograms, you know it works and you can bop around the set with a lightmeter and know you’ve got it covered. Shooting an f/2.0? No problem now it’s 10 FC.

The screen and exposure tools on the [Blackmagic] Pocket [Cinema Camera] are primitive at best and not really any indicator of what the final image will look like. I know that several things have changed on the Pocket since my last deep dive including the way [DaVinci] Resolve interprets the raw footage. I’d advocate getting a lightmeter and testing the camera for yourself. Without a lightmeter you really don’t have any idea of what the camera’s response is, the histogram is too vague and gives no reference to the light input. I’ve had the same old Sekonic L-398 that I used in school 20 years ago and magically it’s still going strong. [I use exactly that same meter myself ☺ – Ed.]

Old school cinematographers, experienced with emulsion film stock, regard overexposing the sensor using ETTR as a generally bad practice which should be avoided as a general rule, although it may be unavoidable in certain situations; it’s better knowing your stuff and to expose the image correctly in the first place by placing the desired shadows and setting ratios, as too overexposed images may make it more difficult to grade the image in post. The golden rule is to try keeping exposures of skin tones consistent from shot to shot, which makes grading much easier, as skin can be more important than highlights. In this respect, using “false colour” on your monitor and watching pink and green levels, is a good formula to keep this skin exposure consistancy. Speaking of methods of measuring exposure, the entire topic largely relies on the camera model chosen as each one hadles zebras, lightmeter, false colour, and in particular display colour/space and gamma somewhat differently (where the Blackmagic Pocket Cinema Camera seems to be pretty accurate). Cinematographer and BMCuser Dan Kanes has the following personal advice to give:

As far as where to set your ISO, I like to do it by where my mid tones look correctly but never to clip highlights. Once highlights are gone it’s usually harder to pull highlights back into a useable space because of digital clipping. Generally if it looks very very clipped you probably are going to have a harder time getting it back. You may often find that whether you’re at [ISO] 200, 400 or 800 – your highlights will still be clipped, so if you keep your highlights and set your ISO to make your midtones look how you want with nothing too hidden in shadow that you hope to see later, you will probably find yourself a happy person.

So in conclusion, digital RAW footage tolerates overexposure errors more than underexposure. Old school negative film stock is also more allowing for overexposure compared to underexposure, i.e. RAW (digital negative) more behaves as film negative, not as video, as digital cinema cameras that shoot RAW use a linear tonal “curve” that represents a perfect straight diagonal line. A line that you can bend in any way you want in post, it seems. The classical negative filmstock uses a slightly “S” like curve with a “toe” and a “shoulder” which goes across the entire spectrum, from the lowest left corner to the highest right corner. The RAW diagonal line also covers the entire spectrum from corner to corner but without the small “toe” and equally small “shoulder” of the film negative. But much of the straight line of the film curve coincides with the RAW feed, which probably accounts for the similarities in the behaviour of both film negative stock and RAW negative digital film. However, the straight linear “curve” is not as steep as the straight line of the film negative curve (provided by the “S” shape), which explaines the less contrasty and flat image of RAW. But the thing is that you can basically do whatever you like in post with the information contained in the digital negative RAW format.

The RAW linear tonal curve as presented by Adobe and compared to the film negative curve
The RAW linear tonal curve as presented by Adobe (left) compared to the film negative curve (right)

The Log (logarithmic) curve is another mode of creating a film-like image which captures highlights in a similar way as negative RAW and negative film, but differs somewhat from the latter. Actually, the log curve follows the f-stops of the iris which emulates how the human eye experiences light. Log encoding is used in the Apple ProRes family of codecs, also featured in the BMCC, BMPCC and BMMCC. Kurt Lancaster, in his Cinema Raw: Shooting and Color Grading with the Ikonoskop, Digital Bolex and Blackmagic Cinema Cameras, states that the BMCC converts 16-bit linear to 12-bit log uncompressed RAW, which probably applies to the BMPCC as well (being the upgraded kid brother of the BMCC with a slightly smaller sensor) as the RAW image coming from the BMCC and BMPCC looks alot like the flat log profile, different compared to the Blackmagic URSA RAW which lacks the flat image of its predecessors and provides more colour prior to grading. So, it seems that the world between RAW and Logarithmic processing is akin in the early Blackmagic Design cameras. What emulsion negative film, digital negative RAW and Log encoding share together is poor handling of underexposure which should be avoided.

The Logarithmic curve with evenly spaces stops throughout the scene exposure range
The Logarithmic curve with evenly spaced stops throughout the scene exposure range

Underexposed digital footage may also look worse than underexposed emulsion film footage because of the digital noise behaving differently and being more distracting than film grain, especially if it creates fixed patterns (which is an issue with all CMOS cameras, including the sCMOS Blackmagic Design cameras). Raising underexposed footage, such as skin tones, to middle grey creates noise that doesn’t match with the normally (or over)exposed shots. Cinematographer Rakesh Malik has further brought to my attention the fact that keeping consistancy in skin tones may also mean that I have to expose skin at two stops over middle grey, pulling it down in post to maintain a single look that matches up colour, noise, shadows, ratios, etc. With a wider shadow ratio of three or four stops rather than one or two, I might have to overexpose the skin highlights by a stop or two and underexpose skin tone shadows at one or two stops below middle grey to ensure the retention of detail and texture of skin, as well as the desired dramatic look. In these instances, a field monitor with false colour would really help, a good reason to aquire one (such as the affordable 5″ Blackmagic Video Assist) in the foreseeable future.

The 5
The 5″ Blackmagic Video Assist with its False Colour feature activated

Please, read my follow-up post on the subject of overexposing RAW.

Advertisements

One thought on “The RAW factor – Is it better to shoot overexposed or underexposed?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s