Wetpixel

Canon 5D Mark II underwater housing shootout

Canon 5D Mark II performance:

This camera is the only 21 megapixel camera with 1080/30p/24p/25p video in the market today. With very clean ISO performance up to 800 (my personal preferences range from ISO800-1600, depending on the scene) for stills, it is a top performing camera for high resolution stills for image quality in raw.

Like every other camera though, it is not without its flaws for underwater use. The AF is good in terms of subject acquisition, but tracking and low light performance feels “less brisk” compared to others like the 1 series cameras I use. For macro performance, it does a good job at acquiring higher contrasting subjects but it is a little slow catching the darker objects in poorer light. I tested the various AF patterns and the best was still the single center AF point.

With the Auto AF Pt Selection mode, singling out subjects wasn’t always reliable. It would choose the higher contrast background or the brighter moving object (eg. Photo XX). It would not track a darker colored subject as well due to lack of contrast.


AF tracking sometimes doesn’t track a subject that well

With Center AF point, the camera performs better. It is more difficult to track a quick moving subject with one AF point but it is more reliable to focus then recompose if the subject isn’t too fidgety. The center AF sensor is a cross type, making it the most reliable sensor in the AF sensor array. The 14bit sensor also allows for shooters to expose “to the right” by a stop or more in RAW mode, allowing the user to pull back highlights that would be clipped otherwise.

 


Using the center AF point, one can use the shallow depth of field of the sensor to isolate subjects

The interface is very good especially the INFO mode with has all the shooting information accessible through the Multi Control and dials. The viewfinder while very bright is still only 98% of the image.

The 5D2 has a few glaring flaws. The biggest has to be Auto ISO mode is poorly implemented for video and even more so for stills. The mode does not have on the fly access to Exposure Compensation so you cannot control exposure shooting in video mode if you choose AUTO ISO in manual mode. For stills, it’s even worse in that it’s stuck at ISO400 when you set it while in manual mode. This is a very limiting implementation of a function that can be very useful.

Video Performance:

Besides excellent stills, the camera also shoots H.264 compressed 40mbps 1080/30P video, wrapped into Quicktime .MOV files. In video mode, the ISO performance is excellent even up to ISO1600. Coupled with a fast lens, the lowlight performance beats many higher priced cameras, even production HDCAM ones. However, the video is not professional quality in the strictest sense. The resolution is in the 5-600 line range and the aliasing and moiré is significant. Still to the untrained eye, the aliased image looks sharp because the fake detail from the aliasing because many things in nature do not have fine straight lines which can cause these frequency issues. So the aliasing and the false details work in favor for underwater use, but more for close up subjects than scenery. Combined with the ISO performance, it should be the king of all video cameras ,right? Well not exactly. There are a few important issues that require serious consideration:

1. The camera can only shoot up to 4GB size files. So on average you can only have about 12 minutes of 1080p and then it’ll cut off. It doesn’t continue on automatically but hitting the record button again will quickly start recording again.
2. For video, the optimal ISO performance requires a bit of technical knowledge. In various tests, it’s been found that ISO 160, 320,640 and1250 perform better noisewise vs other ISO levels. The reason for this is complicated but suffice to know that for optimal noise performance, it’s better to use ISO 160, 320, 640 or 1250. Any higher and the noise levels are too high for any reasonable difference in performance. For stills, the trade off using the incremental 1/3 stop settings is loss of dynamic range in the highlights.
3. The same shallow depth of field that cinematographers salivate over the new HD ViDSLR cameras can also be problematic for underwater use. With a 1/2” EX1, you get almost 5 stops more of DOF than with a full frame 35mm sensor at the same field of view. For example, the 2/3” sensor Sony F900 with a 6mm at f5.6 would have the same DOF of the 5D Mark II at 32mm at f15! It makes keeping subjects in focus an important issue.
4. With modern consumer and prosumer cameras, Auto Focus has made focusing easy. The 5D2, however, does not have an active AF (ability to AF while shooting.) The AF it does have is contrast only and is slow and can only work when it’s not recording. Does not having active AF while recording and super shallow DOF disqualify it as a good video camera? No, since many shoot in manual focus mode. But for many, it does limit usefulness. Shooting with manual focus and selecting your shots isn’t easy, especially with macro.


To overcome this issue of DOF and lack of an active AF, the 5D mark II has 2 weapons, the excellent ISO performance and the access to super wide lenses like the 15mm Fisheye and rectilinear lenses like the Sigma 12-24, Canon 14mm and 16-35mm. The wider the lense, the more DOF at the chosen f stop. With underwater, that isn’t always advantageous. With a wide lens, everything looks small, so you have to get closer, except the subject may not cooperate with you. Then there’s still the issue of focus.

 


This pipefish, shot with the 100mm 2.8 macro, was bobbing left and right and the 5D2 tracked it well with Center AF pt.

Professional shooters don’t use AF modes for video because sometimes these modes can lose focus on the main subject by a passing fish, bubbles or dirt on the dome or even particles in the water. Then of course cameras like the Red One do not have AF and use DOF to keep subjects in focus.

The art of focus pulling is one skill that all cinematographers have to learn with manual lenses. It is doubly hard underwater due to the lack of a good focus monitor and difficult placement of the focus dials. With wide angle lenses, combined using a high f stop to create a larger depth of field, focus pulling isn’t as vital.

However, with narrow lenses, especially when you are chasing for a subject close-up and the subject is moving in and out of the focus range, the skill of focus pulling and choosing the right lens for the subject becomes very important. A major part of that is dependent on the housing. Most housings only have one dial on the housing for either zoom or focus control, but not both, Seacam being the only housing with separate controls. With some macro ports, manual focus is possible via a dial on the port. Dome ports don’t have this feature. This is something that housing manufacturers will need to address with future designs if they are to offer customers the ability to use their cameras to the fullest potential, that is to use super wide zooms with manual focus.

I also did try the Magic Lantern firmware v0.16 a couple of times when shooting video but the requirement of removing the battery to reset to the manufacturer firmware made it impractical for testing a housing. Still, having the blown highlight areas depicted onscreen via the highlight warning shading was awesome for video exposure (when I could see it past the reflections of the glare on the rear window plexiglass. The 5D2 could also use audio meters as the Magic Lantern firmware has. It would make audio monitoring much easier.

 

Canon 5D mark 2 test video @ Liberty Wreck Tulamben with 15mm fisheye and magic filter Also shown is non-optimized color graded vs uncorrected video.

Here’s a sample video with the 5D2. You can download a better file on the vimeo site.

 

Color filter with strobes, anyone?

Another potential dilemma for users is how to setup the camera.  Do you first consider for stills, video or both? Obviously with wide angle, the use of a red filter is great for video but what about the still side of things?

 


Shot with 15mm fisheye with Magic Filter.

To that end, I decided to experiment using the magic filter with strobes. The optimum method is to match the color temperature of the strobes to the filters output. This way, the output of the flash matches the color correction that is filtered with the magic filter.  It’s nothing new but now there’s new urgency to get it right. Since I couldn’t find matching the filters, I just went ahead and dived with the magic filter and my 4500°k Hartenberger strobe to see what happens.

 


The same soft coral without filter with the Tokina 17mm. There is less contrast but the light loss is also gone.

As you can see, shooting with a filter can possibly work (but not in all conditions), even with strobes which are not matched in color. It will require that you shoot raw so white balance can be adjusted in post processing. The water is much bluer once proper white balance is adjusted for the foreground. I did find that sunballs were a bit more difficult with the cyan ring cast showing up on a few shots. The color filter causes the blue/green channels to clip as the red channel rises, causing that cyan halo. So as long as one is careful not to shoot a direct sunball/burst, shooting with the Magic Filter installed can work well. It gave the pictures an extra contrasty look with a rosier tint, which could be tempered in RAW conversion.

I also shot quite a few shots in 1/250 with strobes, going past the sync rating of 1/200 for the camera.  At first, I thought it was fine but I realized after looking at the pics with 1/200 that the sync is just off enough that I had to dial in more power for the strobe to illuminate subjects at 1/250 than at 1/200. At 1/320, the shutter shows up in the images.  For those whose housings have all pins connected like the Aquatica TTL board, to attain 1/250, it’s a matter of just taping over the 3 pins on the hot shoe. That will fake out the camera into thinking there is no strobe attached.

 


At 1/320, the strobe synchronization can’t compensate.

Another much talked about test is the use of video lights for shooting stills, especially macro.  I could try extreme macro as I only had the dome port. With my Hartenberger LK1 fitted with 100W HLX (rated 3600 lumens), I could manage ISO 800 at 1/60 and f4 for a still shot. With 2 heads and a guesstimate drop of 1/3 power for inefficiencies, I had over 5k lumens on the subject and could barely light it enough to get higher f stops. One would need a lot more light for wide angle in daylight. I doubt even a 200W HMI could illuminate the test subject to get down to ISO200 and normal shooting speeds.


Shot at ISO800, 1/60 @ f4 using twin Hartenberger 100W HLX halogens.