Upgrading from DV to HDV

Image Comparisons of Canon GL1 with JVC GY-HD100U

Posted Feb 13, 2006 by Earl R. Thurston

As I write this, heated debates are taking place all over the Internet as to which of the new "baby" HD cameras is best. The level-headed consensus is that all of them are pretty darned good image-wise, so it really boils down to which capabilities and features are best for any given purpose or preference.

But for some people the question is about whether to move to HD at all. Is the difference in image quality really worth the investment? For us, the answer is a resounding "yes!". However, one has to consider the specifics of the sitution.

Our Situation

We have been using a Canon GL1 for the past five years. While certainly not the most current or highest-quality MiniDV camera around, it has served us very well. It was most recently used for a respected educational video [link broken!] about beverage container recycling, which was in turn edited into several public service announcments ( [Links Updated:] 1, 2, 3, 4, 5 ) for airing on province-wide television in BC and Alberta. Not too bad for a little camera competing in an industry dominated by BetaSP (and a production originally intended for CD-ROMs!)

An upcoming project has now given us the opportunity to upgrade. Although it will intially be distributed in NTSC, it is expected to have a three- or four-year life span, so there is the potential for the footage to be shown in HD eventually. With that in mind, and the new availability of affordable HD cameras, we decided to take the plunge and move from standard definition to high definition.

After a lot of research, we narrowed down our choices to the Panasonic HVX200 and JVC GY-HD100U. My initial preference was the Panasonic mainly because it has a wider range of features we could utilize, including native 4:2:2 colour sampling (important for bluescreen/greenscreen work). Unfortunately, the scheduling of the project and slow roll-out of the HVX meant we couldn't wait for it, so we opted for the HD100 instead.

Note that we bought the camera without testing it first. I was confident about its capabilities and image quality based on reports from sites like DV Info Net and DVXuser that we simply placed our order sight unseen. However, what wasn't clear at the time was how it compared directly against the GL1.

Hence the purpose of this page. Once we had the HD100, we were eager to conduct some tests to see how the two cameras compare. Perhaps these tests will be useful to other GL1 owners who are considering making the move to HD themselves.

SD versus HD

First off, it would be prudent to address the simple fact that we're dealing with two entirely different formats. An SD camera shooting 4:3 aspect at 720 x 480 is going to produce a substantially different image than an HD camera at 16:9 aspect 1280 x 720. Therefore, this page attempts to compare the images in both domains. Each still below is provided in an SD-resolution version, with HD scaled down and letterboxed to compare against the GL1; and an HD-resolution version, with SD scaled up and cropped to match the HD100.

Methodology

Each camera was set up on its own tripod, taking the same image as closely as possible at the same point in time. (Note: There are some minor variations.)
Exposure was based on highlights. Each camera had zebra bars set for 100 IRE, and each iris was adjusted so the same amount of zebra appeared in both images.
The GL1, which does not have many image adjustment options, was left in its factory default settings. This is how it has been used for all of our productions.
The HD100 has many image adjustments, so the following adjustments were used based on some earlier tests to obtain the widest latitude:
- Master Black: -1
- Detail: MINIMUM
- Black: STRETCH 3
- White Clip: 108%
- Knee: MANUAL; Level: 80%
- Cinema Mode: OFF
- Color Matrix: STANDARD
- Gamma: CINEMA; Level: NORMAL
Each tape was captured in Premiere Pro 1.5.1 with the GL1 footage using the DV NTSC 4:3 preset and HD100 footage using the Cineform Aspect HD 1280x720 24p preset.
Stills were captured in TIFF format, and then resized, cropped and letterboxed where necessary in Photoshop 5. These were then saved in both TIFF and JPG (maximum quality) versions.

			HD Resolution 16:9 @ 1280x720 (1.0 aspect)
1	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up/Cropped JPG / TIFF	HD100 Native JPG / TIFF
2	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up/Cropped JPG / TIFF	HD100 Native JPG / TIFF
3	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up/Cropped JPG / TIFF	HD100 Native JPG / TIFF
4	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up/Cropped JPG / TIFF	HD100 Native JPG / TIFF
5	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up ** JPG / TIFF	HD100 Native JPG / TIFF
6a	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up/Cropped JPG / TIFF	HD100 Native JPG / TIFF
6b	GL1 Native JPG / TIFF	HD100 Down/Letter JPG / TIFF	GL1 Up/Cropped JPG / TIFF	HD100 Native JPG / TIFF

Notes and Observations

These were taken on a sunny winter day around 3:00pm with very sharp contrast levels. These have always been the most challenging shots for the GL1 to capture well. As expected, the HD100 has a much wider latitude, so it handled these scenes wonderfully.
Images 4 and 5 demonstrate the range of zoom on both cameras. The cameras were situated in the same spot for image 4 and zoomed all the way across the lake for image 5. Because the coverage of the GL1 (with a 20x zoom) is tigher than the HD100 (with a 16x zoom), it was up-scaled to a 960 x 720 4:3 image rather than widescreen to make a better comparison at a 1:1 image size. However, what is very clear in these pictures is how bad the HD100 stock 16x lens is at full telephoto. Tons of chromatic abberation and softness in the far edges. Even at SD resolution, the GL1 is sharper.
Images 6a and 6b are meant as exposure pairs. As noted above, the GL1 has trouble with contrasty scenes. This is readily apparent in image 6a where the shaded side of the house is far too dark on the GL1. (Remember, both cameras were exposed for highlights at the same 100 IRE zebra levels.) So, in image 6b, we changed our tack and exposed for the GL1's shadows instead (but now with the same amount of "over exposed" zebra in each camera). This resulted in an okay image for the HD100, but one could easily consider it over exposed.
The down-scaled, SD images from the HD100 were done in Photoshop, thus they represent an "ideal" conversion from HD to SD letterboxed. Naturally, it would even be better to convert to non-letterboxed 16:9 anamorphic. But also keep in mind a true down-conversion using the DV codec will not be quite as sharp in either case. Something I'm planning for later tests is finding the best way to down-convert HDV to DVD.

If you have any comments, questions or suggestions about these tests, feel free to post them to either DV Info Net or DVXuser (both have free registration). I have an account on both of these sites and would be eager to hear from you. You can also email me here.