Erkki Halkka

Synthetic film grain doesn't - real does. Shoot a gray card with the film stock of your choice, digitize it, composite on top of video - voilá!! Edit: Fixed a typo

You can always add grain in post production, if you wish.

Color sampling is just one part of the HDV vs. Film equation. I only meant that in some shots, the color sampling errors are not visible. Don't get me wrong, it's not perfect, but i think it's better than you suppose. You shouldn't overexpose video. At all, if possible. You should expose for highlights, and adjust gamma to get correct exposure in midtones and darks. The aim is to make sure none of the color channels gets oversaturated in areas where you need to keep information (i.e. a bare lightbulb is usually ok to be blown out). Of course, there are occasions where this is not enough to capture the full dynamic range of the scene (your flame scene was probably one of these), but it *should* get you quite a bit of the way there. This adjustment is best done in-camera to as big a degree as possible, before compression / bit depth reduction. You can push up the gamma like this untill you feel the darks are getting too noisy or you see visible color banding. If the goal is to match film, you can adjust untill the amoint of noise in shadows matches the film's grain amount. You'll then just need to add grain only to the midtones and highlings, matching the film. See this example (i happened to have a MiniDV here at home right now) From this example, you can see a few things: First of all video cameras, at default settings, have a rather linear gamma. This means, simply adusting levels in post is more or less exactly the same as adjusting exposure, or gain in camera. The top right image is overexposed video (Orig2), the bottom right video is underexposed video (Orig1) adjusted with simply cranking up levels. The result is almost exactly the same - only big difference is that the adjusted image has more noise. The image on the bottom left was also color timed from the underexposed top left video clip - but this time by adjusting the gamma curve. As you can see, the shadow areas look rather similar to the overexposed clip, but the highlights are NOT blown out. As far as the color range goes, it's quite rare to run into subjects that fall outside the hue/saturation range of video cameras. More or less all things in nature fall within the range, including human skin. It's just a matter of color timing them properly. The biggest difference to film is just how far things can be pushed. If you don't need to push far, you can get very similar results - the differences start to show in extreme situations. This is especially true when digital intermediate is used - even though the imagery is high bitdepth most of the way, film printers tend to be 8 bit / channel (the same as i.e. MiniDV), AFAIK.

To me, as a person who spends more than half of the time doing post production, "cleaner" is always a good thing. "Flat" and "neutral" is also a good thing. Why?? Simply because you can always degrade the techical quality, by adding grain/noise, increasing contrast, blurring etc. Removing grain, de-blurring, or trying to rescue details from crushed blacks is much harder... Simply put, neutral, grainless, low contrast image can be manipulated to look like pretty much anything - but it's an one way street. Think of the law of entropia ;-)

Well, this thread has evolved to something much more... not sure if "more" is always better ;-) But back to Charlie's claim, I don't think anyone disagrees, that IN SOME CASES it's very hard or impossible to tell 4:4:4 from 4:1:1. That's part of my reasoning why HDV could pass as 35mm in theaters, IN SOME CASES. The thing is, in other cases, the difference is obvious. The result CAN be fixed to some extent in post, but it's not a miracle or something, you don't magically get that missing information from thin air - you can just make the errors harder to notice. Edit: fixed the above sentence, it was missing half of the words ,-) Here's a grab from MiniDV vid i shot on holiday at Madeira (Nice place, BTW). In most areas of the frame, low color sampling produces rather acceptable results, it would probably be hard to see the difference to 4:4:4 at 100% size. But then there's some tourists that have bought matching bright red caps... http://eki.pp.fi/temp/Eki/ColorSampling/Madeira_411.png It doesn't take Superman vision to see the errors. I've been trying to get to shoot some side by side shots with video and film lately. But, shooting film for SD is getting really rare here (i've been involved in maybe 2-3 film originated projects in the last year or so, compared to dozens a year just a few years back). And on those 2-3, i've only participated in post. We did do some side by side tests earlier, 35mm VS DVCPro50, i'll try to find something from those - but the thing is, i don't work at that company anymore. Even in those tests, the video portion of the shoot was an ugly duckling... as the shoots are usually a bit hectic anyway ;-) Me to talent: "Hey, don't run off just yet, i'll take one more shot" - then i take the vidcam, handheld, and shoot something for a few seconds, with the settings that happened to be in the camera. "Thank you!". Not exactly a fair comparision.

I added one more - HDV_Chart_cineframe_off_det08_focusmagic3.jpg. It's cineframe off, detail at 8, sharpened with alternative method of deconvolution. Nice, i must say... The problem is, i haven't found deconvolution plugin for i.e. After Effects - i did this with a demo of "Focus Magic", with a setting of 3 in "focus". It's a standalone piece of software, works only with one jpg image at a time. Results are good, but processing footage manually frame by frame... no thanks ;-)

...and to show how the printed test chart would look like with a "perfect" HDV camera, i scanned the printed test chart 300dpi, color matched, scaled and rotated it to match. There's now two new images called "HDV_Chart_Graphic_HDV_scanned.png" in the folders. This is more or less the best any camera could have done with this test chart. Edit: I must say, the HDV camera does VERY well here!!

This is something i'd really like to know the answer to: On this camera, does a detail/sharpen setting of 0 mean "this is the unprocessed image straight from the CCD's", or does it mean "Here's the image, blurred a bit". If it's the first one, they have a really good sharpening algorithm in the camera, better than most other cameras i've seen, better than unsharp mask -type sharpening i.e. in photoshop. If it's the second one - i'd really like to know what setting gives the unprocessed image... Anyway, in a perfect world, cameras would give supersharp results without any electronic sharpening... with HDV, we're not there yet. The 720P variant of HDV is 19 mbits/s, i recall some Panasonic models use this. 1080i variant, which is what i.e. Sony HVR-Z1 uses, is 25mbits.

Me too - i don't own a camera either. Only one i ever had was... Super 8 ;-) I've been thinking about getting my own HVR-Z1. I'd probably get the price back in year or two in rentals. But, it's not exactly a cam you wanna take with you on a holiday. So, i've been checking out that new small one chip Sony. It seems like an excellent "home-cam" - but lacks some crucial manual settings, like real control over f-stop, gain and shutter speed. So, i ended up buying nuthin' ,-) *** I did some test chart shooting the last time i had HVR-Z1. As i only wanted to compare some settings for the next day's shoot, i didn't pay much (read: any) attention in framing the test chart correctly, or lighting it for that matter. Anyway, i guess the stuff might be interesting to you dudes too... http://eki.pp.fi/temp/Eki/HDV_resolution/ There's two folders, one for full size HD and another for PAL D1 widescreen, square pixels. Both contain same images, i different resolutions. Here's what's there: 1. HDV_Chart_Graphic_HD_original.png This is an image, where i took the original PDF of the test chart, and color matched, scaled and rotated it so that it more or less overlaps the chart i shot iwith HDV. This is the "perfect" goal to aim at. The chart i shot was printed to A4 paper from the same PDF, with a regular inkjet. The test chart print itself wasn't of high enough resolution to resolve the finest lines in the chart. Some thin lines were also a bit thicker printed than in pdf. 2. HDV_Chart_Graphic_HDV.png This is the same as above, saved to .mt2 HDV stream, then stretched back to 1920*1080. This is the best HDV as a format can do - compare it to the previous one to see how much you lose because of the FORMAT. 3. HDV_Chart_cineframe_off_det00.png This is HVR-Z1 in 50i mode, with sharpening at zero. Rather soft to my taste... 4. HDV_Chart_cineframe_off_det08.png The same, but sharpening at "8". This is the setting where the balance between visible sharpening artifacts and overall sharpness is in best balance, IMO. I use settings between 5 and 8, depending on case. 5. HDV_Chart_cineframe_off_det08_WARPSHARP.png The same as above, but sharpened with "warp sharp" - instead of enhancing edge contrast, pixels are subtly warped towards edge areas. My number one sharpening method of choice nowdwys - i process most of my HDV like this. 6. HDV_Chart_cineframe_off_det15.png Camera's sharpening at full, setting of "15". Sharp, yes, but i have an allergy for edge ringing ,-) 7-10. **** Same as 3-6, but with "CineFrame 25" mode on in the camera. The result is softer, but i'd say it's usable in many cases, especially if scaled down to SD resolution. Overall, HVR-Z1 doesn't really do that bad - it's better than i anticipated, when i first had the camera in my hands... as said, very good for what it is.

Sorry, you totally lost me there - both use 25mbits / second. Unless you mean 720P variant, which does use lower data rate...

Here's an image - without heads again to protect the innocent - On the left, HDCAM footage, on the right, HDV, both scaled to 1920*1080, then cropped. HDCAM was keyed ontop. HDV is a bit softer, but i used much less sharpening than the HDCAM DOP. In my eyes, HDV actually looks better in this case - with some sharpen filtering and levels adjustments, they'd look really similar, IMO.

Probably not. I guess the difference should be the same as between top of the line SD camera and a MiniDV. I don't actually know what camera / lens combo was used on the HDCAM projects i've done post with, but i certainly hope none of those was a high end model. My HDV material was roughly in the same ballpark with those. The HDCAM footage i've worked with looked about the same as Betacam, or i.e. Dvcpro footage looked like a few years back, when looking at it 1:1 size, i mean, the same as compositing an HD frame from four Betacam frames. HDV footage looks like compositing that HD frame from four MiniDV frames. As said, different, but in the same ballpark. Top of the line HDCAM's *should* look like compositing the HD frame from four top of the line DigiBeta frames... uh, i hope this didn't sound too insane ;-)

You are mixing up visual information (how the actual images look) and datarate. They are two different things. Here's uncompressed way to describe ten pixels of black: "black, black, black, black, black, black, black, black, black, black" Here's a compressed way to say it: "10 times black" After decompression, both give the same picture, even though the later only used about 1/4th of the data. Well, i took it as sarcastic - My first language is Finnish, sometimes subtleties get lost. At first, i was about to be sarcastic on my answer too, and just say "Habbit??"... but ended up typing an essay ;-) Well, most of my stuff is commercials etc. material that goes the way of the dodo just a few weeks after they're done - i didn't think about archiving at all really. At that, film probably shines - both on resolution and color fidelity. Of course, if a digital intermediate is also saved, one can make safety copies of it without any generation loss for all eternity ;-) I've shot a few documentaries, which will probably run for years to come - they were shot at standard def video, and archived on both DigiBeta and data on a removable hard drives. Way back when, in the early 90's i also shot one drama series... haven't heard of it since ;-)

Yep, i'm pretty sure. The following is mostly from reading "reliable sources", but i've filled in some blanks myself... correct me if i'm wrong - that's happened before. Ahhem... The color information is kept intact untill compressing to HDV mpeg. Sony's HVR-Z1, the HDV camera i'm most familiar with, has three 960 x 1080 CCD's, which should give something very close to full 1920*1080 luma resolution with pixel shift, theoretically. The green pixels are offsetted by half a pixel from blue and red - the luma signal is calculated from these all. In practice, the optics are probably to blame for the fact that the real life resolution is lower - just as with most standard definition video cameras... but it's still much, much better than SD. The chroma information, AFAIK, is 960*1080 at this stage. You could call it 4:2:2 "in camera". From these, the data is reduced (scaled) to 1440*1080 luma signal / 720*540 chroma signals, then compressed. As PAL resolution is 720*576, the actual color sampling of downscaled HDV seems to be 4:3,75:3,75. In NTSC, one should get full 4:4:4. My real life tests suggest that this holds true - i like examples - wanna see one?? On the left side, HDV footage downscaled to uncompressed PAL D1 video. All color channels are clean, more or less perfect. Better than with i.e. DigiBeta. On the right side, there's yet another example of MiniDV's color sampling problems you can't tell the difference looking at the full color image, but looking at the individual channels (or trying to do chromakey)... you see it ;-)

Of course, digital cinema has some clear advantages over film too - like being able to monitor the shoot at full quality in real time, checking "dailies" at full quality on location, possibility to even make a full quality edit of the scene at location, ability to shoot an hour in a row, no telecine needed for digital intermediate, no generation loss whatsoever if projected digitally etc. ...it's a different beast. In many ways not as good as film, but in other ways - better. Dunno - it's a matter of economics. As soon as the projector prices drop under certain treshold, the change will probably be very fast. Film prints are expensive - moving data around is cheap.

Origin outputs 16 bit HDRI (high dynamic range images), as far as i know. Here's a bit of info about HDRI in general: http://en.wikipedia.org/wiki/HDRI There's two standard dynamic range 16 bit TIFF stills at their site, available for download. These are only 2K resolution, probably scaled down to get the size manageable - the zips are 12Mb each. These have already been "exposed" from the raw HDR originals - so color correcting these gives poorer results than color correcting the originals, AFAIK. http://www.dalsa.com/dc/documents/documents.asp However, i did a bit of gamma adjustment to the images just for fun, both darkening and lightening the pics - because of the image being 16 bit, the result of my rather extreme adjustments weren't that bad... there's some noise in the lightened shadows, not sure if that'd be present if the adjustment was done using the original, raw file. These are 8 bit png's, so they're much smaller downloads... http://eki.pp.fi/temp/Eki/Dalsa/

Here's what they say at Dalsa site: "Origin's exposure latitude is comparable to the best film stocks and offers at least 12 stops of linear response for an astonishing range in the whites as well as the blacks. In one shot Origin can handle both the naked flame of a candle and the delicate, nuanced shadows on candlelit faces. It can handle the full glare of the sun reflected from a window and still resolve the subtleties of the shadows below. When you light for profound effects, you'll appreciate Origin's profound advantages."

Well, the hot trend with clients continues to be film. But as they're paying the bills, many of them (well, most here in Finland nowdays) make do with something that's "close enough" ;-) Actually, one big reason to use film for SD i forgot to mention earlier, is that clients want "the film look". Actually shooting film is a sure way to get it. Also, shooting film has a bit of "glamour" or "magic" attached to it, that has nothing to do with actual imagery. It makes clients feel "this is big time" - and pay the "big time" bills happily...

No, no, and no. That is simply untrue. HDV has the same amount of DATA, but about four times more resolution, both in luminance and chroma (4.5 in NTSC, 3,75 in PAL). Because HDV uses the (much mocked) long-gop mpg compression, the resulting imagery is MUCH better than it would have been at the same data rate using intra-frame compression. The HDV compression is similar to best quality DVD compression, just with 4 times more resolution and 4 times more data. DVD's use about 1/4 of the data rate compared to MiniDV. The visual quality is roughly on par, because of better compression. If what you said about HDV was true, DVD's should have just 1/4 of the visual information compared to MiniDV. The downside of long-gop compression is, as you mentioned, the possibility of getting bad artifacts when there's a lot of motion in big areas of the frame. I personally have only shot one shot where this became a real problem (a greenscreen shot with fast moving, transparent plastic bag). We managed to fix the shot by applying additional motion blur in the post, but it was a close call... Partly because it's a habbit, partly because of better color correction abilities in telecine, partly because of bigger resolution. Why is that bigger resolution necessary?? First of all, most video cameras simply don't resolve even the full standard definition resolution in practice. Without electronical / digital sharpening, the images look soft. This happens also in pro formats, like DigiBeta and DVCPro50 etc. Only recently we've started to get cameras that actually produce sharp images without additional sharpening. Second, with bigger original resolution, you have the ability to re-frame / enlarge the pictures in post, without sacrifying quality. This is a big bonus, especially in SFX work. Also, the video shooting formats use 4:2:2 color sampling at best. This becomes an issue when i.e. trying to create perfect bluescreen composites. The reason why i've moved to HDV from "pro" SD video formats is the same. Because every pixel in the final SD master is combined from four original pixels, i get effectively 4:4:4 color sampling. The image is as sharp as one can get at SD, without additional sharpening. I can reframe the shots to some extent without losing visual quality, also because of the 4X oversampling. Believe or not, also color correction works better with HDV than with regular SD footage - again because of the oversampling. Now i can shoot slightly underexposed, to keep the highlights from clipping a bit better, and adjust gamma to brighten the image. This causes the noise levels to pick up, but because each pixel is averaged from four, i can do much bigger adjustments before the noise starts to show in the SD master. The fact that i can digitize and edit everything on my home workstation "as is" is a bonus - i do have an SDI card on the machine, but i don't own a DigiBeta or DVCPro50 deck, so untill HDV i had to rent those if i wanted to digitize at home. Or go to an editing facility to do the work... So, using HDV is definitely not a "trend" issue for me, it's a big positive step both in quality AND in ease of production. **** Here's some example images of color spaces in action, if people want to see the differences with their own eyes: http://eki.pp.fi/temp/Eki/ColorSampling/ http://eki.pp.fi/temp/Eki/ColorSampling/Co...xamples_444.png Saved to uncompressed 4:4:4 avi, then exported to .png. This is exatly the same as original text graphic. http://eki.pp.fi/temp/Eki/ColorSampling/Co..._420_HDV_D1.png This is the graphic upscaled and saved to HDV .m2t stream, then downscaled to SD resolution, then png'd. Even though HDV is 4:2:0, the effective color sampling at SD resolution is about 4:4:4. http://eki.pp.fi/temp/Eki/ColorSampling/Co...22_DVCPRO50.png This is the same graphic saved as DVCPro50 codec 4:2:2 avi, then png'd. If you look closely, you see a bit of color sampling artifacts here. http://eki.pp.fi/temp/Eki/ColorSampling/Co...s_420_PALDV.png This is the graphic saved as PAL MiniDV avi, 4:2:0, the red and blue text look rather horrendous. The result is similar to HDV's color sampling at full resolution. http://eki.pp.fi/temp/Eki/ColorSampling/Co..._411_NTSCDV.png This is the graphic saved as NTSC MiniDV avi, 4:1:1, the red and blue text look rather horrendous again.

Yep, i've seen the article - a good one. The reason why i said dalsa surpasses 35mm wasn't so much the resolution (which does look visually just slightly better than 4K film scan to me), but it's lack of grain and higher exposure latitude (correct me if i'm wrong, but isn't film negative somewhere around 9-10 stops at best?). That's why i said it's "approaching 65mm"...as it's being somewhere between the two qualitywise. There actually are 4K projectors already - not in common use though... Yep. It's a bit clumsy right now - but in 5-10 years... probably editing 4K video shot with my cell phone, on my cell phone ;-) I hope i haven't sounded militant - i wasn't aiming at that at all. Believe it or not, i'm currently involved in post production of a job that's done entirely in 4K, and love it... i do like quality too. The stuff that's shown in the theaters will probably evolve to something higher quality than now - not because the artists demand more quality, but rather because HDTV quality home theaters are getting too close... something must be done to keep the audience. Anyway, to me HDV is kind of a revolution - much in the same way as desktop publishing was in early 90's, and digital audio studios and CGI on home machines were a bit later - or MiniDv was in standard def video world. It's something that will allow Joe Average to get technically "close enough" of big time, with cheap off the shelf equipment - the current generation of it is kind of at the edge, the next gen will probably be over it already. So, the difference won't be who has the big bucks, but rather who's good. And, as i mentioned earlier, i was pleasantly surprised at the quality when watching it on 35mm print. All this said - in my daily work, which is SD for television, HDV has replaced DigiBeta and DVCPro 50 as the shooting format of choice. Not because it's cheaper (that's a non-issue with commercials etc. - i'm not the one paying the bills), but rather because it simply looks better.

Don't get me wrong - it didn't look "bad". It looked rather good, just as good as any other 35mm prints i've seen, and the theater i saw it at was about the newest in town so it should supposedly be technically good. As other have pointed out, there's NO additional generations due to subtitling - that was my mistake. It's just that the release print always is inferior quality compared to the original - and there's nothing that can be done about it really. I must disagree with you to some extent, the reduced quality of actual prints DOES work as an equalizer - not as much as scaling down to standard def video does though. There probably is a treshold where additional "original quality" ceases to affect the output in any discernable way. For example, i really doubt anyone can see the difference between 4K digital intermediate and 2K digital intermediate, after scaled down to standard definition TV. In both cases there's plenty of oversampling there. For film release prints, i suppose the difference can be seen. But difference between 4K and 8K probably would be lost. If you remember my original argument, it was that in that theater, some shots done with HDV, cut within the film, would probably have passed unnoticed by the audience. *** I totally understand what you're saying about rather trying to make 35mm look like 65mm... to me, making MiniDv look like Digibeta is just as cool. I simply love to try to stretch each medium as far as it can go. You know, to turn turds to diamonds ;-)

Well, *I* was the colorist, and i do take at least half the blame on highlight clipping, on these examples. I deliberately modified the gamma of the video with the notorious "S-curve", increasing contrast in midtones, darkening blacks and lightening whites - that was the desired look. Also, not all video cameras are equal in this respect. There's been a lot of progress from 5-10 years ago.

Oh, you can see it too. Shoot a car's red taillights at night, or anything else that has bright red or blue details against black. At 4:4:4 they look OK, at 4:1:1, you'll clearly see very blocky colors. 4:1:1 will fool the eye most of the time, but not always. Another thing is keying, as mentioned. There's a world of difference there. I should know, i'm just working on a 200+ shot blue/greenscreen production, shot with HDV. Luckily, this will be mastered on D1 PAL, so after scaling HDV down we have more or less 4:4:4 color, and the keys look quite good. 4:1:1 (or 4:1:0) stuff can be processed to look decent at full rez (i.e. by median filtering the color channels), but that only works if you keep the stuff at 4:4:4, or at least 4:2:2 from there on. As far as that 4K camera goes, i was talking about Dalsa Origin: http://www.dalsa.com/dc/origin/origin.asp 4K resolution, 12 stops of latitude, directly uses cine lenses. This one should already be better than 35mm film negative is technically, AFAIK, approachingf 65mm film.

Mmm.. that's pretty much it. But... Actually, on these examples, these all are partly something that i could be blamed for as colorist / post dude, rather than a fault of the medium. The originals were rather low contrast, there was plenty of detail in the shadows, and a bit more also in highlights. It's just that that high contrast look is/was the trend of the day - so it was done to look this way in color correction, deliberately. Same goes mostly with the sharpening - i always shoot with very low detail settings, and sharpen in post. On Video monitor, that edge enhancement is less visible - the stuff just looks crispy. Since the day these were made, the arrival of HD/HDV and also better sharpening algorithms has luckily reduced the need for unsharp mask type sharpening for video. Which is good. Very good. One cool thing about those 35mm shots in the examples was that i got two transfer passes on DigiBeta - there was a special "highlight" pass that was underxposed otherwise. I then used hand-drawn masks to blend between the two. I must admit, that's something that would have been nearly impossible to get with video (without altering lihting). I guess that depends on taste - i happen to like the "high-tech plastic quality" - film grain etc. is something i consider an "effect", that can be added in post if desired - probably because i've mostly worked on stuff that originates on video, and ends up on broadcast TV.

Cool - what was the telltale sign that made you think "video"?? BTW, None of the video stuff was HDV (HDV didn't exist as a format when these were made a few years back), it was all originally interlaced D1, shot with DVCPro50, so it wasn't actually dumbed down at all...