A GL-2 Imaging Bug / Attribute

[Ken Tanaka]

You may recall that I recently reported in another thread that my GL-2 appeared to have a bug in its imaging system. Having now received and tested a second camera I can report that this "bug" does, in fact, seem to be intrinsic to the camera's design and was not an isolated manufacturing flaw in my first camera.

In brief, the GL-2's imaging can show a distinct "vignetting" or "key-hole" effect under certain conditions. That is, the corners and edges of the image will noticeably and increasingly darken as the lens' zoom is pushed towards a subject. I've not yet established a definite pattern for this phenomenon but my tests seem to indicate that the effect is most severe when using Tv (shutter-priority) mode under slightly under-exposed conditions. It does not appear at all under "Auto" mode or when shooting well-lit, brightly-exposed scenes.

This attribute seems to be a bug rather than an explicit engineering attribute, since the second unit I've tested displays a much less severe effect than my first. If it were an engineered attribute you would expect the effect to be consistent across each camera.

This vignetting generally remains outside the normal ("action-safe") viewing area of a regular television, although the effect with the first camera I tested did infringe into the action-safe area under extreme conditions. NOTE, however, that this also means that YOU CANNOT GENERALLY SEE THIS EFFECT while you are shooting since neither the lcd nor the viewfinder display much of the image outside the "action-safe" area. Only after capturing your footage or while monitoring the camera with a professional monitor are you likely to see this bug. Of course anyone targeting their footage for full-frame displays (ex: Web streaming or other computer-based displays) should take special note of this attribute.

Since I am not a camera technician I don't know what's causing this problem. But I have reason to speculate that it's not simply the GL-2's lens but rather a combination of the lens and the GL-2's new CCD block. I reviewed some footage I shot with my GL-1 over 2 years ago and, lo and behold, I could detect darkening corners on several shots, although not nearly as noticeable as with the GL-2. This leads me to suspect that the new CCD has a hand in the problem along with the lens' characteristics.

On the whole, the GL-2 produces outstanding images and is the best of its class today, in my opinion, when all of its features and performance are weighed in toto. But if you've purchased a GL-2 you should take time to shoot a test tape specifically to determine if yours exhibits the same vignetting under low-light or slight under-exposure conditions (especially in Tv mode). Unless the problem is very severe on your camera you're probably best advised to simply learn to manage and avoid it since it appears to be the nature of the beast at this stage in the product's life.

If you have a chance to test your GL-2 please take a moment to report your results here. Two cameras do not constitute a categorical finding. Two dozen or more might.

Side note: Kudos are once again due to the folks at ZGC. Upon learning of my discovery last week, they immediately sent a replacement camera to me. No hassle. No yang-yang. Just plain ol' fashioned good customer service. Thank you ZGC!

[Chris Hurd]

An interesting, detailed, and well written report, Ken. Many thanks,

[Frank Granovski]

Oh well..., that's what you get when you're the first kid on the block to buy one of these new toys. Remember the PD100? And then the new and improved, PD100A?? Yup. I remember...as I kiss my MX300s good night.

[Barry Goyette]

Ken

Can you get the camera to do "it" in manual...so we'd know what fstop and gain settings cause the problems? I'd like to test my camera to help confirm or deny what you're seeing. So far I haven't noticed anything like it, and I've shot a lot of low light stuff....but I haven't captured much yet either.

Is this something that you notice "while zooming", as in a change over time that makes it more obvious, or is it something you can hold, and see the vignetting.

"as the lens' zoom is pushed towards a subject"--- does this mean the extreme end of the zoom range, or does it happen in the middle of the range or all the way through?

This sound's a little like a similar issue with the xl1s that was reported on earlier in the year, in that it happens at the most severe end of the exposure spectrum. (although I believe then it was the center of the image was darkening). It was most noticeable "while zooming", and was eventually concluded to be an limitation of the zoom in low light, low aperature, high gain situations.

Barry

[Jeff Donald]

The GL-1 and GL-2 have the longest zoom ranges in the industry (20X). The lens is a variable aperture design. In shutter priority the aperture is forced to change as light values change (which frequently happens when zooming the 20X range). My gut instinct says its a combination of extreme zoom range (lens design), variable aperture design and changing aperture that combine to produce the effect. The amount of gain applied only helps make the effect more apparent by making the scene brighter. Old zoom lenses ('60s and '70s) showed this aberration very frequently.

Jeff

[Ken Tanaka]

Barry:
Try as I might, I could not get the camera to key-hole in manual mode! Nor could I induce the effect in Av (aperture-priority) mode. I could only induce it from Tv (shutter-priority) mode...and rather easily at that. Here are the steps I follow:

1. To avoid the hassle of unloading and capturing from tape connect the camera to a professional monitor that features underscan display. I use a Sony 8045.

2. Set the shutter to, say, 1/60 and aim it into a moderately lit or slightly dim room.

3. Zoom-in towards a subject as if you were snapping focus.

4. If the room is sufficiently dim (but not extremely dark) you should begin to see the effect on your monitor. Tromboning the zoom will reveal that the darkness incroaches further into the image as the lens is pushed closer. If you do not yet see it, try engaging the ND filter.

Jeff:
The fact that I can only reproduce the effect in Tv reinforces your supposition that the lens' automatic aperture adjustment is the agent. When I engage Manual mode right after observing the effect the lens tends to be wide-open and the gain has cranked up as high as 18dB. The Gl-2's 20x lens is, indeed, one long honkin' zoom. Most ENG lenses stop at 19x. So I suppose that some side effects are to be expected, even from the lens gurus at Canon.

All in all, this is probably a very moderate, manageable problem on most GL-2's. But as I noted earlier, the effect is not consistent across all cams. So, under the heading of getting to know your equipment's characteristics and limitations, I encourage GL-2 owners to check their cams for this attribute.

[Jeff Donald]

I think the reason that the key-hole effect varies from camera to camera is variables in the manufacture of the lenses themselves. When lenses are tested and evaluated it is not uncommon to see variances in focal length. The 84mm lens might be 83.5mm or 85mm etc. Because the grinding and assembling of the lenses is not a perfect process Canon has design tolerances that allow for some variations. I believe it is these variations that make the effect more pronounced on some models than others. Since it occurs in the overscan (in most cases), I doubt Canon will do anything considering it is an NTSC video camera. I wonder if the increased resolution of the new chips may also play a part?

Jeff

[Ken Tanaka]

<< Jeff: "Since it occurs in the overscan (in most cases), I doubt Canon will do anything considering it is an NTSC video camera." >>

But the good news is that they've eliminated their horizontal blanking problem from the underscan area! <g>

<< Jeff: "I wonder if the increased resolution of the new chips may also play a part?" >>
After seeing hints of this key-holing on my old GL-1 footage I believe it probably does play a part.

[Steve McDonald]

It seems the vignetting is likely caused by the aperture not opening wide enough to keep up with the zoomed-in image, that needs more light. That you can accentuate the effect with the ND-filter, makes this appear to be even more likely. Putting it in aperture-priority eliminates it, as you say. I'd say be careful about zooming in too rapidly or not recording during zooming, until after the exposure is adjusted, unless necessary to catch the content.

Steve McDonald

[Ken Tanaka]

Steve,
Once this "key-holing" effect kicks-in it is persistent and increases as the lens zooms closer. It does not dissipate when the zoom stops; it remains static in the image. To shake the effect off you have to change to a lighter subject or increase the exposure (shutter speed).

[Steve McDonald]

Ken, you'd be inclined to think this vignetting bug might be embedded in the processer's programming, if others reported the same thing. But, as it is, the mystery has yet to be flushed from its lair. Remember the flutter during zoom-in on some XL1s and the viewfinder burn-out that many suffered?

[Ken Tanaka]

Yes, I sure do remember those bugs. Actually, the 16x on my XL1s still flutters from time to time when auto focus is engaged. I think it's just the nature of Canon's auto-focus processor.

[Jeff Donald]

Ken,

Does the GL-2 have an Exposure Compensation dial? What happens if you zoom with the dial engaged (+/- 1/2 EV).

Jeff

[Ken Tanaka]

Jeff,
Unlike the XL1s the GL-2's AE adjustment is in a menu, although it behaves the same. Making adjustments to it does have a slight impact on the effect but it does not vanquish it. In fact, it looks to me like AE shifts merely accentuate or decentuate the key-hole. It really looks like a mechanical effect rather than a pure electronic/signal effect.

[Randy Wisman]

Ken,

To confirm your findings I encounter exactly the same 'feature' with my brand new XM-2.

I noticed it in faily dim lit rooms. It becomes stronger somewhere in the middle of the zoom range.

Since it's not always there, I'm still doing some tests to exactly figure out when it happens. Can you confirm the following:

- it's absent (or minimal) when the whole image (corners and center) is more or less equally lit (dim light).

- It's becomes clearer when the center of the image is lighter than the corners. (e.g. spotlight causing a darker ceiling)

It feels like that the CCD are non-linear. The exposure suddenly drops, whenever the light goes beyond a certain threshold level.

In a certain area of the zoom range, the exposure at the corners drop below this level while the center stays above.

This might be a combination of lens and CCD.

Please let me know your latest conclusions.