Scratchpad

Pete's scratchpad is a place to share video and other multimedia clips with other hobbyists. My intent here is to stimulate discussion and solve technical problems in the creation of multimedia content. Feel free to e-mail me with comments or questions, or better yet, post on the great forums at DVinfo.net...I check there daily and am a moderator for the XL2 board. See you there!

The XL H1 is here!

Sorry for no post yesterday (Dec 7)...I did the below tests last evening and then fought unsuccessfully with FrontPage to make the Auto-Thumbnails feature work properly with my large BMP files. My wife finally broke the code this AM, so here you go! (And I guess now I'll have to go buy her a Christmas present after all!)

Someone on DVinfo noticed that the Coin Twirl WMV shows repeat frames; I had forgotten, but I've seen this before and it is a function of either transcoding to WMV (perhaps incorrect pull-down setting -- haven't checked back on that yet), or a bug in playback of 24fps WMV in Windows Media Player. The repeat frames do NOT appear within the 24fps PPro project, and this 1920x1080p24 mpg2 export also plays properly -- no repeat frames:

There was concern on DVinfo that interlace artifact was causing the spinning coin to look blurry. Well, a rapidly moving object is going to have MOTION BLUR at 1/48th second shutter speed. I can't pick out anything more than that...if there is interlace artifact there, I can't see it. This brings me to my Wheel Test; I originally intended to use it to elicit GOP compression artifacts; it didn't, but the exercise did serve to illustrate how well whatever secret process Canon's uses to create a 24fps or 30fps full-frame image from an interlaced CCD block works.

How I did it: My wife still has an old turntable and let me "customize" it (alas, yet more Christmas presents...) to spin at very close to 60RPM...actually about 62. I then hand-made a 24-slice circle from two sheets of colored card stock and Sharpie markers; that's my "wheel." Then I just gave the wheel as spin and recorded at various settings.

I also wanted to determine how the standard definition lenses stack up against the new HD-20x lens. Canon makes it very clear that they consider the SD lenses inadequate for HD use, albeit compatible with the XL H1. There is a BIG warning "HD Incompatible Lens" in the viewfinder when you power-on with one of the SD lenses. Although I don't have a 16x lens to compare, given that it is supposed to be the sharpest of the SD XL lenses, the results below should make you 16x owners happy.

Please don't make anything of the coloration on some of these clips. I was only concerned about doing resolution tests and did not white balance for the various lighting configurations that I used to simply get the shutter and F-stop in the ballpark I wanted. It did seem, though that a tendency toward red followed the SD lenses, as compared to the HD-20x.

I shot these at about 82% zoom for the 20x lenses, using auto focus, either 1/48 or 1/60 shutter and for most of the shots, about mid-range F-stops (f4 to about f5.6).

I was quite amazed to see relatively little difference between the resolution of the various lenses. Especially so the 3x, which has been widely accused of being soft even in SD shooting. I actually repeated some of the HD-20x tests to make sure I hadn't screwed them up, but I got the same results and they are about what I would have expected of the lens. The surprise is the good performance of the SD lenses. If my testing is borne out, I will be changing my mind about selling the 3x along with my XL2!

That said, I'm not a test engineer and my setup here is a bit makeshift. If someone sees how I may have significantly erred in my testing, or does similar tests but gets different results, PLEASE let me and everyone on DVinfo.net know! We all want The Truth -- we can handle it.

My wife wasn't keen on me releasing shots of her getting worked on at the orthodontist's office and the weather sucked, so you'll have to wait a little longer for "real world footage." Today it is chilly in Houston but the rain supposedly is done. So maybe after work I can shoot some HD-20x and some 3x real-world shots.

Finally, here is a little video sample of my H1's 24F mode. When I e-mailed him for information, David Taylor of Cineform was kind enough to share a non-public beta trial of Aspect HD 3.4 with me. I got the impression that it'll be released to the public pretty soon...so be patient. This version will specifically provide support for the Canon XL H1, including the 24F mode. Despite having limited time this evening, I eagerly fired up the H1 to shoot a short clip, loaded Aspect HD onto my editing box, and captured. This was shot using Autofocus, Auto White balance, no custom presets, gain -3. Shutter was 1/48, aperture was about f=5.6...oh, yeah, that reminds me to say that Canon's return to the aperture wheel, rather than the XL2's flip-switch, ROCKS! What a difference! I captured into PPro using the Cineform 1440x1080p24 preset (beta) and exported from the timeline to WMV 1920x1080p24 without any added effects:

I've got ambitious plans for the next couple of days -- probably more than I can realistically get done. But for my DVinfo friends, I'll do my best! I hope to compare the standard 20x lens, 3x lens, and the 20x HD lens on the rez chart; then, I want to see if I can get the GOP compression to break down. Test videos are boring so I'm going to get out the door and shoot a little "slice of life" stuff, too.

Look what I got for my birthday. And I wasn't even a good boy!

The box

All the documents.

Encased in polyethylene were the big pieces:

camera body, lens, viewfinder.

Plastic wrapped.

The bottom layer held the microphone, remote, and other smaller parts.

All of the parts.

Close-up of 20x HD lens and camera body.

Component cable.

Initial impression is very favorable. If you're a previous XL owner, you'll feel right at home. The camera is just enough heavier than the XL2 to notice. Probably no big deal for most situations, but it might be a factor during a long handheld shoot. I really like the textured matte black finish; aside from what I guess most people will feel is a more professional look, it seems to give one's hand a more positive grip. As previously reported on DVinfo.net, there is one small disappointment: Canon went to the trouble to create a new viewfinder for the camera that includes pro features like peaking and magnification to help with focusing, but ignored pleas to make sure it shows 100% of the recorded image. Fortunately, a variety of safe zone and aspect ratio guides help remind you to frame appropriately even though you're seeing only about 90% of what is being recorded.

I know, I know. Who cares about all that fluff? How's the picture? Too soon -- I need more time. I did manage to do a very quick frame-grab B&W rez comparison between the XL2 24p mode, the H1 60i, and H1 24F. The 24F was on a 60i timeline in PPro 1.5.1 because the limited version of the Cineform technologies licensed to that version of PPro did not have an option for a 24fps timeline. (I've downloaded the trial version of Aspect HD 3.3 but not yet installed it -- Cineform is working on a Canon 24F update to their products that's not yet released, but I'm hoping that Aspect HD 3.3 has 24fps presets that'll do for HDV use for now. Unless I hear that it's a bad idea, I'll try installing the current trial version in the next day or two). The 24F frame grab should still be a pretty fair comparison because this was a totally static test. Anyway, this was quick-n-dirty, but it is pretty clear that HD is for real. The 24F was close to the resolution of 60i, but 60i still won, as expected.

The pictures were all shot in 16:9 anamorphic, but the frame grabs are square pixel. Also, I created the rez chart by splitting a pdf of the chart into quarters, printing each part on legal size paper, and taping them together...you can see the alignment of the pieces of paper wasn't quite perfect. Just didn't want anyone to think it was some kind of split screen anomaly or something!

Hopefully tomorrow and Wednesday I'll be able to shoot a little real-world stuff. I also have a torture test in mind to look for GOP-induced motion artifacts. Some of this will depend on whether I can get true 24fps from the camera, through PPro, to exported file, which I haven't had time to verify.

I promise...I'll update with more info as I have more time, but sadly, I do have a job. So it may be a few days.

Glowing Text in Titles of PPro

By fiddling with the Shadow settings for text within the Titler applet, I was able to fairly easily create this glowing effect:

Recommend that you "Right Click>>Save As..." because the effect is subtle. At least for me, it doesn't show up well when viewed in MS IE6. Upon downloading and viewing in Photoshop CS, the effect is nice.

XL2 + 20x Lens Depth of Field -- A quick n dirty demo

This is a quick and incomplete look at the depth of field on the XL2 using the stock 20x lens, but should at least give folks a rough idea of what to expect. All frame grabs were shot in 16:9, 30p using manual settings: 1/60th at f4.0. This doesn't represent the absolute minimum depth of field that could be attained, but the lens can stay at f4.0 for the entire zoom range, so I choose that as "typical." They were saved as 720x480 square-pixel jpg's (so the frame grab aspect ratio is 1.5 rather than 16:9).

In each of the frames, the nearest crock, on the right, was used as the reference point for distance from the camera. The distant crock, in the left of each frame, was four feet more distant, and the mixer in the center background was eight feet beyond that (12 feet from the front, left crock). This is to give an approximation of a question asked on DVinfo: Can the depth of field be made shallow enough indoors to achieve a rack focus from a near talking head (4-5 feet away) to a distant talking head (4-6 feet more distant) while indoors, while keeping the background (7-10 feet behind the distant subject) out of focus?

As you can see, the XL2 doesn't break any laws of physics. The depth of field pales by comparison to 35mm film cameras. Yet, if you want to achieve a depth of field shallow enough to rack focus between two talking heads, it can be done. The combination of distance and telephoto will get you there, albeit with more care necessary to set up the perspective you'll need for the shot.

Is the horizontal resolution of the XL2's 16:9 inferior to 4:3?

There is little argument that the XL2 has an outstanding standard definition picture, whether shooting 16:9 or 4:3 aspect ratios. The question arose on the message boards about what's best to do if you want your final result to be a 4:3 file that is letterboxed to 16:9 (black bars above and below). That led to discussions about the 960x480 CCD area being down-sampled and whether or not that resulted in a loss of horizontal resolution despite the XL2's use of pixel shift technology and 12 bit DSP processing. So I set out to answer these two questions:

I created a 16:9 frame with blue painter's tape around a pair of laser-printed resolution charts taped to the wall (what a tolerant wife I have!). The width at the outer edge of the tape was exactly 720 times the width of a line on the resolution chart for either the fat end of the "4" line on the chart, or the skinny end of the "2" line. (Five black lines and four black lines at the fat end of the "1" line were 1+15/32" or 1.78125") should represent the theoretical limit of horizontal resolution for a 720 pixel-wide image. Oddly enough, my first attempt showed that the exact 16:9 rectangle (35.625"w x 20.10"h) was not quite as wide as the image the camera recorded in 16:9, so I widened my rectangle by 1/2 inch to make the aspect ratio match the camera's image. The rectangle as shown in the pictures here was 36.125"w x 20.10"h -- slightly more than 720 resolution chart lines wide. Considering that the correction was by eyeballing the capture window, the 1.4% increase in width (and therefore change in aspect ratio) was pretty close the 1.25% percentage by which 486 (D1 resolution) is greater than 480 (DV). I'm not sure how the camera and PPro handle rounding, etc, but that small discrepancy was probably not coincidental. Yeah, I should have changed the height, not width, of the frame. But I ain't doing it all over again!

Anyway, camera settings were 30p, 1/6oth, F6.2 with gain set to 0, resulting in 1 tick mark (1 stop) of overexposure according to the camera. Zebras set at 90% did not appear at all during the tests, though. Except for 2 ticks low on the sharpness, all menu presets were at default; I may be wrong, but I suspect a lot of serious shooters will turn down the sharpness a little to minimize halo effect. The wall I used as a backdrop is standard darkish beige and I didn't manually white balance, so I think that's probably part of the reason the white paper the rez chart was printed on looks a little greyish and didn't overexpose despite the camera setting for a stop hot. The 20x camera lens was 12 feet from the target, a little more than half way zoomed and both 16:9 and 4:3 were shot such that the sides of the blue painter's tape rectangle were right at the lateral edges of the frame. I'll also note that the XL2's viewfinder has a LOT of overscan -- much more than I think is necessary. It was enough to just make the 1.5" wide tape not visible on the viewfinder when PPro's capture window showed perfect edge-to-edge coverage.

Enough already! Here are the full 720x480 square-pixel frame grabs directly from PPro 1.5 timeline; the only processing was to convert them to max quality jpg from the original bmp. The first is the 16:9, then the 4:3, then the 16:9 footage as letterboxed into the 4:3 project. The final jpg is best-resolution snips of all three. (Remember, these are all square-pixel frame grabs; if you want, you can import tham into Photoshop CS and view them with the correct pixel aspect ratio).

You can judge for yourself but I must say, I checked my math several times and I'm amazed that the horizontal resolution actually slightly exceeds theoretical limits for 16:9! It is hard to declare a winner between 16:9 or 4:3, but similarly, it can't fairly be said that one is clearly inferior either. Undoubtedly some slop crept into the process somewhere, but it looks sweet! As far as the letterboxing question, seems a wash to me from an overall resolution standpoint. I would shoot 16:9 because it allows re-purposing of the footage later.

Screen Capture of Premiere 6.5 timeline

This screen capture shows my settings for an attempt to replicate a problem James is having with cropped or otherwise rendered areas changing from 60i to 30p. I am unable to replicate this.

24p Advanced Anomaly in Adobe Premiere Pro 1.5

Original 16:9 24pA files captured directly from the XL2 using PPro, or created by capture using Canon PC-DV software, play normally in Windows Media Player (WMP), despite wide reporting that WMP does not support 24pA. However, the same footage exported from PPro has grossly unacceptable rapid flickering between 16:9 and 4:3 (even though it is supposed to be 16:9 footage. I set out to understand the problem. I captured a short clip of the chronograph on my digital wristwatch using my Canon XL2 with 20x lens at 24p, 2:3:3:2. It doesn't really matter, but in case you're interested, the shutter speed was 1/48th. I imported the file into a 24p project in PPro (24fps timeline) and selected a 10 second segment (240 frame at 24fps) for export with both 24pA and then 30p NTSC settings. I added a standard countdown screen in the upper left corner using the PPro timeline prior to the exports.

The file resulting from Export>>Movie>>MS DV AVI using "24p Advanced" (really 23.976 in the video section), has every 5th frame erroneously in 4:3, making the clip unwatchable. When re-imported into PPro, this AVI clip played normally in the 24fps timeline.

When the same 10 second segment is exported from Export>>Movie>>MS DV AVI using "30p NTSC" it plays normally in both WMP and in the PPro timeline after being re-imported.

Both exported AVI files are actually 300 frames long at 30p, preserving the 2:3:3:2 architecture of the original 24pA file -- still exactly 10 seconds, as you can see from the clips. Here's what I speculate is going on:

Based on the every-fifth-frame aspect ratio problem, I believe (but don't KNOW) that the "24p" exported file plays incorrectly in WMP because WMP can't interpret every fifth frame properly (the "repeat" frames that are not read by a 24pA-aware application) and the aspect ratio flag is misread as 4:3 or missed altogether and defaulted to 4:3. On the other hand, the original camera output and the PPro export using 30p NTSC are both apparently properly flagged as regular 30fps NTSC video (so WMP can read them) and just happen to have a field/frame arrangement of 2:3:3:2. (But if that is true, I don't know how PPro properly recognizes 24pA footage without fail).

Although WMP can't read these 24pA AVI files properly, PPro itself does handle them just fine. As mentioned above, I re-imported the 10 second files back into the 24p timeline. I then exported each of them to a separate mpg file from Export>>Adobe Media Encoder using MPEG2 at 720p, 23.976 fps. The end result, as far as I can tell, is two identical 24p files...the fact that one file had been a 24pA AVI and the other had been 30p didn't seem to matter one bit. Take a look:

Although I would sure like to understand the technicalities of how the 24pA and 30p exported files differ, there is no problem working with them within PPro. The only limitation is that 24pA exports can't really be played on WMP at present -- and the Big Question is what makes a 24pA camera capture play fine in WMP, but not the same footage exported from PPro?

IF ANYONE FULLY UNDERSTANDS THE ANSWER TO THIS ISSUE, PLEASE EITHER EMAIL ME OR POST ON:

Conversion artifact during pass-through capture of Hi8 video

I did a small test to verify that an analog source, in this case Hi8 footage played on my old Sony camcorder, could be captured directly to hard disk as a DV AVI file using the XL2's A-D converter as a pass-through. As described on pp 99-100 of the XL2 manual, no tape is required or desired in the XL2 for this process. Yup, it works. Realize, though, that since there is no DV timecode on the Hi8 (or other analog source...VHS, etc), you cannot set IN and OUT points; the data stream flow only IN to the computer. So for this, only real-time capture is possible, at least in Premiere Pro 1.5.

ALSO, as I'm sure we have all seen in many video files that started out analog, there were thin black lines at the left and right margins, and a narrow band of distortion at the very bottom of the image in the AVI file.

Here are two frame grabs in jpg format. The first demonstrates the black lines and bottom distortion, all of which I presume is due to the technical differences between analog and digital signals. The second picture is the same video frame after applying PPro's Crop effect (2% left, 1.5% right, no top, 1.5% bottom) to get rid of the distortions. I exported them as square-pixel 1280x960 bmp files from a PPro timeline, then saved them to jpg files in Photoshop CS. I doubled the output resolution from PPro simply to minimize the impact of any jpg compression artifacts.

There's also interlace artifact, but I didn't concern myself with it for this exercise. For what it's worth, a little sharpening (usually from 10-15) -- in addition to whatever specific tweaks a clip may need -- makes these old Hi8 tapes look fairly decent...still not XL2 footage, but still pleasant to view.

A Puzzle: Did interlace artifact creep into a 24p project?

Here are three screenshots (with permission) from Matthew Cherry's first short posted on DVinfo.net. He reports that he shot this with a Canon XL2 using 24p with 3:2 and edited in a 29.97 timeline on Final Cut Pro. This was then presumably exported to a 30fps Quicktime *.mp4 file.

Where there was motion in the frame, such as the movement of a coat being hung up or the arm swinging in the third picture, it appears that there are "jaggies" from interlace artifact. Please read the thread at

Side-by Side Comparison of the XL2 and GL2

This is just an impromptu comparison of these two cameras in 30p mode (technically the GL2 is "frame mode") using 16:9 NTSC. The settings for both cameras were 30p, Tv=60, no custom presets, and audio set on Automatic Gain Control. The two cameras were set up literally side-by-side about 6 inches apart, approximately 12 feet away from the subject and run simultaneously. A little parallax is, of course, evident when comparing the images from the two cameras, but since they were A and B roll right next to each other, the effect should be negligible for comparisons of the final images.

853x960 bitmap

Other than adding titles, the video is just as captured into PremierePro 1.5. PPro's Adobe Media Encoder was set to NTSC 720p to minimize any visible artifacting in the miniDV (720x480, 1.2) clips. The still image was created by exporting a 720x480 bitmap of the first frame from each camera, upsampling to 1706x960 (2hx2w times the widescreen 853-480), and copy-pasting an 853x480 (half height, half width) selection of pixels from each image into the final bitmap. The XL2 appears on top, the GL2 on bottom.

I think the results are unquestionable: the XL2 image is so obviously better that even if I did inadvertently do something in the processing to give it some advantage, it couldn't possibly change the winner of this contest!

GL2 Narrow-Wide Comparison: The Banana Spider

This big gal took up residence outside our breakfast nook in the summer of 2003. Our daughter Olivia promptly named the arachnid -- what else -- Charlotte and forbade any anti-spider behavior. Even the exterminator steered a wide berth during our quarterly service.

So what good could I make of this monster? With all the angled body parts, fine "hair" and webbing, Charlotte was a good model while I compared the resolution of different modes of my Canon GL2 camcorder. Here are "best case" and "worst case" frame grabs:

Both were taken with a WD-58H adapter and no filters. Once the video was on the timeline in Premiere Pro, I exported one frame to *.bmp using square pixels. In PhotoShop, I then saved as *.jpg at maximum quality factor (12).

My opinion, for whatever it's worth: yes, there is a definite reduction in resolution in the worst case of 16:9 frame mode but not too much. There's no doubt that the GL2 is now my "B Roll" camera for anything 16:9. However, the 4:3 is still nothing to scoff at...if I do happen to shoot in 4:3 for some reason, especially in a run-n-gun situation, it may be my choice even over the much larger XL2.