Low Light Canon XF100 XF105 Test - 4:2:2 Codec Shines

[Philip Lipetz]

Shivaratri 2011 - Canon XF100 XF105 Test on Vimeo

Shivaratri is the day that Shiva, the aspect of God that breaks the illusion, comes out of meditation and aids the world. A major Hindhu holiday, it is traditionally celebrated by repeating “Om Namah Shivaya” and with light displays. This was videoed at the Saibaba Temple of Ohio, located in Columbus Ohio. My love and thanks to this welcoming community.

This is a low light test of the Canon XF100, all shots were done at +6dB, and was edited for test purposes, not narrative flow. Lighting was dim and harsh florescent lights, and South Asian people have skin tones that are very hard to maintain under adverse light; previously found that out the hard way. I have shot in the same room with a Panasonic GH1, and these conditions are at the limits of that camera. A real test. The results AFTER post processing are better in the XF100 – much to my surprise, dSLRs are supposed to king of low light.

You will notice that the instant autofocus “wobbles” subtly during low light pans. This has been explored by Erik Norgaard who finds it occurs in most low light conditions with autofocus (http://www.dvinfo.net/forum/canon-xf...-limits-2.html). About one fifth of my footage suffered from this problem, and the worse examples were not shown. BTW, the 4x magnify button on the XF100 makes manual focusing simple even in low light. Other people suggest simply turning off the instant autofocus.

Previously I have complained about how the XF100 falls apart when used outside of its boundaries (http://www.dvinfo.net/forum/canon-xf...-limits-2.html). This test was to explore the limits of those boundaries.

All of this footage has a red channel increase of 32%, a blue channel decrease of 15%, a 4% saturation increase, and no sharpening. I purposely avoided using sophisticated color correction tools, just three way color. All footage was shot at 24p 1/24 shutter (least sharp you can reasonably use but best for low light) auto WB in Video C preset mode.

While the auto white balance of the camera clearly could not cope with this lighting, and I deliberately did not manually set the WB, the footage was easily corrected in post processing. The auto white balance produced varying Kelvin temperatures throughout the room, but every shot but one (included in test) produced the same color balance, indicating that the color WB mechanism is consistent and therefore easily correctible in either pre or post processes. As long as I can recover easily with batch processing I am happy.

When viewed straight from the camera the color was muddy, but then we discovered the strength of the XF100 MPEG2 4:2:2 codec; it has plenty of room for post processing color correction. Shots that would be unrecoverable with any of my previous cameras, including the hacked Panasonic GH1(3), easily correct without showing signs of color breakup.

The XF100 and XF105 are professional cameras, designed for people who expect to make adjustments while editing. The 50/mbs MPEG2 codec gives much more latitude for correction than the 24/mbs AVCHD codec of the Panasonic GH1(3).

There has been much talk about the advantages of RAW footage, where all adjustments are made in post processing; the thought being that such system allows you to exert total control after shooting and devote your attention to shooting during the acquisition phase of videography. The problem with such a system is that it is very expensive and the footage takes up absurd amounts of digital storage space. The Canon MPEG2 4:2:2 codec seems to be a wonderful halfway step between compact footage as most lower cost pro cams record, and RAW footage. It allows much more control while editing, but without the expensive overhead of RAW footage. Think of it as intermediate between normal footage and RAW.

In my previous review of the Canon XF100 XF105 limits I complained about the way the camera seemed to suddenly fall apart when it reach its limits; now I must retract that statement. I suspect that the camera does not apply normal amounts of in camera correction, but instead concentrates on recording the most information and then letting you make better informed corrections while editing. The initial result may look like it is falling apart, but there is so much hidden information that you can do better while editing than do cameras that rely on in camera noise reduction and color correction. I find such cameras frequently produce plastic or watercolor looking images, not suitable for my needs. My mistake, for which I apologize, was assuming that the XF100 images were as fragile under correction as are the images from other cameras under low light.

THIS IS A CAMERA THAT REWARDS YOU BY GIVING YOU MORE CONTROL OVER YOUR FINISHED IMAGE THAN ANY OTHER CAMERA IN ITS PRICE RANGE HAS EVER DONE. TRULY REVOLUTIONARY.

What I do not know is how well the Canon XF100 does after post processing at 9dB, and above, gain under low light. I intend to find out.

Like any revolutionary advance, it will take a while until we know how to use what has been unleashed. If you are not into making the effort during editing then there is another version of this camera, the lower cost Canon XA10, which uses AVCHD to make corrections in the camera, giving an initially more pleasing image but one that does not respond as well to corrections while editing. By providing both cameras Canon is making a statement that the MPEG 4:2:2 codec is the key differentiator between the two cameras, and now we see that the XF100 gives more, but asks you to take control.

[Dom Stevenson]

Great piece Philip. I;m loving this camera the more i hear about it.

Bom Shankar!

[Philip Lipetz]

Thanks, and Shiva thanks you.

Forgot to say that all exposures were manually set using the waveform monitor.

[Charles W. Hull]

Good work Philip - and nice video. The woman with the bottled water shows an interesting mix of old and new traditions!

I usually see a little noise in shadows with a gain of 6dB, but I don't see this in your video. Did the post somehow correct this, or are you just not seeing it? Or maybe you are controlling exposure better??

I agree the codec responds wonderfully in post processing. On to 9dB! Waiting for your results.

[Noa Put]

Thx philip for posting another video, I must say that even though the footage looks very sharp the colours are quite pale and look a bit washed out on my pc. I really would like to see a side by side caomparison with a canon xh-a1 f.i., especially in the low light area's. I do like the form factor of this little cam and if, like you said, the 4:2:2 codec gives more room in editing to manipulate before it breaks up it woud be interesting to see how far you could push underexposed footage to a better exposed image when no camera gain was applied.

I have shot in the same room with a Panasonic GH1, and these conditions are at the limits of that camera. A real test. The results AFTER post processing are better in the XF100 – much to my surprise, dSLRs are supposed to king of low light.

Just had to say this but that is not a real test, the GH1 doesn't perform well in low light and you had a F4 lens on it. Take a GH2 and slap a F1.4 lens on it and it will blow the xf100 and any other small sensor camera away in low light conditions, no doubt about that.

[Philip Lipetz]

Colors look vibrant on my monitor, perhaps I would lower the brightness a little. Among us GH1(3) hands we feel that the low light reputation of the GH1 was unwarranted, not a 5d but still pretty good -especially when you consider that Canon rates its ISO settings differently than Panasonic did with the GH1. Agreed that a GH2 will blow away other cams at low light, but the point is that at 6dB the XF100 is delivering roughly 400 ISO results with acceptable grain.

Your idea about under exposing and bring it up in post is excellent. I will try it in the next week. Either way it will be interesting.

[Philip Lipetz]

Originally Posted by Charles W. Hull

Good work Philip - and nice video. The woman with the bottled water shows an interesting mix of old and new traditions!

I usually see a little noise in shadows with a gain of 6dB, but I don't see this in your video. Did the post somehow correct this, or are you just not seeing it? Or maybe you are controlling exposure better??

I agree the codec responds wonderfully in post processing. On to 9dB! Waiting for your results.

Exposure was controlled manually via waveform monitor settings. I did nothing in post other than color and a tiny bit of saturation. One reason that I bought a Xf100 was that shadows seem to be better controlled and with more detail than other cams in this price class - that was huge with me after living through shadows producing bands and stripes in the GH1.

[Noa Put]

but the point is that at 6dB the XF100 is delivering roughly 400 ISO results with acceptable grain.

That depends what type of lens you are comparing a 400iso setting with, a dslr lens set at F4 or at F1.4 with iso 400 is light night and day.

It would be great if you could see how well the xf100 footage holds up when pushed. I know that my xh-a1 looses color quite quickly in low light situations, pushing the image then would only work quite well if you'd make it black and white.

[Philip Lipetz]

If F4 at 1600 with the GH1 is barely acceptable then F2 on the XF100 with the same or better results is the equivalent of a ISO 400 or better rating.

And I agree that I HAVE TO TEST the results from pushing underexposed shots.

[Charles W. Hull]

Originally Posted by Philip Lipetz

And I agree that I HAVE TO TEST the results from pushing underexposed shots.

I have pushed 0dB and 6dB with VERY good color results. At 12dB I got awful color results that would be very difficult to correct in post, and I have not tried 9dB. My current default is to stay at 6dB in low light, and to push the exposure in post. I get better results with both color and noise by pushing low gain video than by upping the camera gain. I sure would like to see others results.

[Charles W. Hull]

Originally Posted by Noa Put

That depends what type of lens you are comparing a 400iso setting with, a dslr lens set at F4 or at F1.4 with iso 400 is light night and day.

I have a 5Dmk2 with a f/1.4 lens and the low light performance is amazing. I haven't been too concerned about the XF100 in low light because I can always fall back to the 5D. But the XF100 is actually pretty good after you learn what it can and can't do; I'm cheering Philip on as he explores the limits of the camera.

[Rob Katz]

philip-

on my 27" imac i7, the xf100 hd footage looked stunning.

and at +6db...wow!

i would greatly appreciate if perhaps u can share how you used used the wave form monitor as a guide to your exposure settings.

as others have mentioned, and it can not be mentioned often enough, thank you for sharing your efforts with all of us.

be well

rob
smalltalk productions

[Philip Lipetz]

Originally Posted by Charles W. Hull

Good work Philip - and nice video. The woman with the bottled water shows an interesting mix of old and new traditions!

I usually see a little noise in shadows with a gain of 6dB, but I don't see this in your video. Did the post somehow correct this, or are you just not seeing it? Or maybe you are controlling exposure better??

I agree the codec responds wonderfully in post processing. On to 9dB! Waiting for your results.

Moving to new house so new tess delayed. I am finding that noise is chroma and can be reduced by careful color channel manlipuation. Noise also varies by gamma curve, best results sometimes with flatter curve, cine, and boost in post. However, the video C preset also responds well. I am looking at which gamma curve responds best to underexposure and boost in post.

[Flemming Bo Hansen]

Night shots from Earth Hour 2011 in Shanghai

www.vimeo.com/21575605

[Spiros Zaharakis]

Being a photographer used to ISO speed ratings, Gain has always looked strange to me.
I've seen most video sensors being rated at ISO800 (at 0db) and that means that for every +3db of gain you get one more stop.
That means the +6db setting should be ISO1600 assume the sensor is rated at ISO800.
An easy way to find out is to put a DSLR next to the video camera and shoot the same scene using the same aperture and shutter speed. Then the ISO used in the DSLR will show the relationship between the ISO speed and Gain of the camera.