Canon 5D Mark II - Over and Under Exposure Tests

Canon 5D Mark II - Over and Under Exposure Tests

Canon 5D Mark II - Over and Under Exposure Tests from Ben Cain / Negative Spaces on Vimeo.

I really like this little camera but the latitude feels a little thin to me. I want to know how far I can push it and how many stops over and under I can record before running into problems so I did this test exposing the gray card / chip chart to key and then over and under exposing by 5 stops each.

This is a full resolution 1080 video file so watch it full screen in HD to really see what's happening in the images. 

ISO 200, bracketing was done with the iris and shutter. The charts are slightly hotter in the upper center. A soft, even source like a Diva would have been better for this test but all I had at the house was a tweenie. The difference is about 3-5%, not enough to compromise the test so I'm just going to go with it. 

Key Light: Gray card is at 50 IRE. The gray card is even across the parade indicating correct white balance at 3000k. We have some highlight and shadow information in the grayscale and a full range of hues in the Macbeth. Even at optimum exposure, you can see the codec is having trouble with the shadows. The last 4 swatches on the grayscale are clipped down to one. It's important to also include a color reference in tests like these because you want to see how much chroma information is retained in the highlight and shadow portion of the signal.

1 Stop Over: Nothing to worry about.

2 Stops Over: We're still good though we're starting to lose our peak highlights. Look at the skin tone swatches though, they're getting close to the edge. This means you really need to watch the contrast on the face or else the limitations of the codec become apparent. Skin tones even at 90 IRE are crippled even though they aren't clipped. You'll never get the full range of tone back that you could have captured with a correct exposure.  

3 Stops Over: Highlights are gone. A lot of our chroma information is gone as well though we're still retaining some of the darker values. Gray card is almost at 100 IRE. 3 stops over and we're right on the edge. 

4 Stops Over: Gone

5 Stops Over: Even more gone.

1 Stop Under: Nothing to worry about.

2 Stops Under: Still in the zone though we're starting to see some noise in our darker values, darker colors, and skin tone swatches. If you don't want a lot of noise in your faces, you can't be too contrasty. Based off these test, anything more than a 3:1 is compromised. 

3 Stops Under: Dark values are crushed. Still retaining some color information but noise is getting worse. 

4 Stops Under: Gone. There's no way to get any useable information out of this. 

5 Stops Under: Even more gone. 

In my assessment, the Canon 5D Mark II has about 5 stops of solid latitude and about another stop where you're retaining some information but from there it falls off very fast. Let's call it 6 stops dynamic range to be fair. 

HDSLR and Waveforms - overview

HDSLR and Waveforms - overview

I have a much larger project related to this in the works but I just wanted to quickly show you a little tidbit. 

converter.jpg

Using a HDMI to SDI Mini Converter from either Blackmagic Designs or AJA Video, you can send a quality video signal to your SDI based video engineering equipment such as Waveform / Vectorscope from Leader. I've used both and I think that they're pretty much the same. Abel's version that's powered with Canon batteries uses the Blackmagic version I believe. 

For these little tests, I used the AJA. From the 7D, using the proprietary mini HDMI to full size, I hit the mini converter, from there I sent 1080 60i video from the box to my scope via SDI. 

This is a frame capture from the live video raster. For ease of illustration, the scene is a simple Kodak Gray Card in my living room. The raster captures from the Leader are always a little darker and a little warmer than what you see on the actual video. 

1.jpg

As you well know, there's no way to output clean, display free video from Canon SLR cameras. Let's hope they fix this in the near future. 

This is the waveform of the scene. The semi-opaque 16:9 matting and the white center marker are outlined with in red below. That is not part of the raster that's being recorded to card so if you use this method, you need to know what to look for on the waveform so that you can disregard it. 

2.jpg
2_2.jpg

Here is the scene on the vectorscope. Though the screen captures always look a little warm, we know our white balance is set correctly because our picture information is centered on the scope. 

4.jpg

Now let's have a look at the actual raster that was recorded to the CF card. 

CAP.jpg

And the accompanying waveform:

leader3.jpg

And side by side:

comp.jpg

Quite close but let's have a look when they're on top of each other. The blue waveform is from the field and the red waveform is from the CF card:

comp2.jpg

Interesting eh? The codec seems to be compressing it a bit compared to the live video. It's very close though and if anything, it's better that your files are a little clamped down from the monitor rather than expanded. These cameras pretty much shoot WYSIWYG and with their reduced latitude and thin codec, all the more reason to use the right tools in my opinion. That is if you have access to them. If not, I've found that if I depress the shutter a little bit while shooting vid and call up the camera's built in light meter, that setting it right in the middle like you're supposed to results in pretty nice looking video. Pretty convenient actually.. Less stuff to carry and the results are just as good. Nice when that works out. There are plenty of situations where you're going to want a proper monitoring setup but for the average run and gunner, I think the built in light meter gets the job done. 

Next time there will be some much prettier examples but that's all for now ;)