Erkki Halkka

Yep - other than the sound stuff, the only important omission (IMO) is black stretch, which helps FX1 a bit in contrasty situations.

It will look better, if you scale to an uncompressed SD format - how big the diffrence is depends on the footage. Wether it's worth of the extra step or not - i.e. for bluescreen work it's DEFINITELY worth it, for something like a documentary (not much post) it's maybe not. You can do the conversion in the camera if you wish, but you'll end up with a regular DV file, and there's probably no big difference compared to shooting DV right away.

We mastered at 720p in-house, but the stuff that went to TV channels for broadcast was SD-PAL. We could have worked directly in SD resolution for the post (as we did with our next vid - which is still unreleased), but the band wanted to add a 720p HD wmv file to their DVD. The reason for using 720 instead of 1080 as the resolution of choice was twofold: first of all, we were able to slightly zoom in the images without making it look too different from the rest of the footage, second of all, 720 has half less pixels to render and store etc. than 1080. The video wasn't originally intended to be shown in 1080 HD, we made the blowup just for HDFEST. A friend of mine saw it in LA screening, and said that some of the shots did look a little soft, but most were surprisingly OK even in the big screen - would have been cool to see it like that ;-) Edit: added a little to the above sentence After being scaled to SD resolution, HDV has better color sampling than i.e. DigiBeta - it's basicly equal to 4:4:4 for NTSC, and 4:3,75:3,75 for PAL. There are some (very) minor artifacts visible after scaling, which can be filtered out to the extent of being nonexistant. With bluescreen, i wouldn't really suggest using HDV on anything higher than SD - but with greenscreen, one can get decent results even at full 1080 HD resolution, if everything goes more or less perfectly all the way from the shoot to post. When scaled to 720 HD in post, there's already more leeway - the effective color sampling should be close to 4:2:2.

Uh, HDR-FX1.

Hi, and sorry for the delay with my reply - This was an all greenscreen project, the backdrops are all 3D. We shot with HDR-Z1, without any adapters or filters. The additional DOF was done in post, by blurring the backgrounds (and, in some cases, selectively blurring the FG too).

As far as i know, HDV is 4:2:0. I've got a web page on the differences of HD formats too - here: http://www.kolumbus.fi/erkki.halkka/HDform...HD_formats.html

I could send you audio test files with 8kHz and 44kHz sampling, where you probably couldn't hear the difference (say, a bass track without much treble in it). That doesn't mean there are no differences in other cases. As said many times before, the difference doesn't show much in all pictures. On yours, the red logo on the hat is the best candidate for seeing color sampling errors - and it in fact does look a tiny bit mushier on the 4:1:1 image. If you zoom into the logo and A/B compare the images, you can see the difference. But it's very, very small in this case. Try making red text over black background, at 4:4:4, 4:2:2 and 4:1:1. Or look at the examples i posted earlier, to see the color sampling differences better. That red hat artifacting shot i posted is a direct .png screen grab from PAL MiniDV footage i shot on vacation, with the enlarged portion done in photoshop. It's actually 4:2:0 - but 4:1:1 should look rather similar. All the color sampling examples with text were made by creating a title in SonyVegas, saving that to various formats (Uncomp, DV, DVCPro50, HDV) and then taking a direct .png screen grab of those. The heading of the page is "Steppy Edges" - more or less exactly the same thing as the "blocky colors" we've been talking about... He's trying to get rid of it by filtering - i use more or less the same methods if i need to reduce color sampling errors when doing keying work, or mastering in higer sampling formats with DV/HDV footage. Instead of just blurring the color channels, i prefer median filtering - it gives slightly sharper edges. As said before, on most raw footage, the difference isn't big enough to be worried about - if it was, low color sampling wouldn't be used. The problem raises it's ugly head with original footage IN SOME CASES, and quite a lot more often, if the footage is manipulated a lot (color correcting individual colors, keying etc.). I too have a page about keying compressed / low color sampling video, have a looksee: http://www.kolumbus.fi/erkki.halkka/HDVKey...and_keying.html

Even if the above was true (if you weren't wrong about how color spaces work making your point 100% moot), you'd still be wrong. See the above gradient. The top portion of it uses 64 shades (same as 7 bit), the bottom portion uses 128 shades (8 bit). Unless you have a very bad monitor, you should see banding only in the top half of the image, because there's not enough shades to show the gradient properly. If your monitor is bad, both show banding.

I don't think it's necessarily about the higher than 22K frequencies - it's about the high but hearable frequencies (near 20kHz) having too few samples for each wave length, causing distortion. 4:4:4 - 4:2:2 - 4:1:1 is analogue to 96, 44 and 22 kHz audio sampling... you can't always hear the difference between 96 kHz and 22 kHz sound, but sometimes you do. It totally depends on the source. As far as mp3's go - 128 Kbit sounds usully rather OK. But in SOME cases... That said, i've destroyed my hearing by playing in rock bands in the 80's, i couldn't hear the difference at all - hasn't stopped me from mixing stuff for video though, i just need a pair of good ears to come and check the mixes for me in case of nasty errors in the high end ;-)

Charlie, you've totally misunderstood this 4:4:4 / 4:1:1 color space thing. 4:1:1 has exactly the same amount of color SHADES asd 4:4:4 - the amount of available colors depends on bit depth, not chroma sampling. Lower chroma sampling reduces SPATIAL chroma resolution, giving blocky images. In 4:1:1, color is saved at only 1/4 resolution - or in other words, color information is saved only for each 4th pixel. It's not much different than sampling audio: sampling resolution tells how many samples are taken (i.e. 48 000 samples of audio / second - every pixel of luminosity channel [4] and every 2nd pixel of both color channels [2:2] in 4:2:2 color sampling), bit depth tells how many diffrent levels those samples can be at (256 colors / channel in 24 bit RGB images or 256 audio levels at 8 bit audio). I.e 4:2:2 could just as well be marked as 1:1/2:1/2. If you have i.e. 100*500 pixel image, you'll get 4:4:4 1000*500 luminosity resolution 1000*500 color resolution 4:2:2 1000*500 luminosity resolution 500*500 color resolution 4:1:1 1000*500 luminosity resolution 500*250 color resolution Please, have a look at this image again, it shows the artifacts well. http://eki.pp.fi/temp/Eki/ColorSampling/Madeira_411.png So does this: http://eki.pp.fi/temp/MiniDV_uncomp_from_HDV.png

...it might also be interesting to know how much experience with HDV the participants have. So, how many of us here has shot HDV? Done post on it? Worked on HDV to film project(s)? Seen HDV to 35mm print footage in a theater? "Yes" to all accounts here ;-)

I wouldn't feel comfortable operating due to lack of experience, but if someone else operated the camera for me, and/or offered a bit of backup when making technical decisions, i'd be happy to give it a try.

Shoot 35mm film as DOP: No. Direct stuff shot on 35mm film: Yes. Done post production / color timing with telecined film: Yes. Done post production / color timing with 2K film scans: Yes. Been in telecine: Yes. In one of the 10 big post houses: Not in U.S.A. - Yes in Finland.

I use it a lot, when i need to give "film look" finish to video or CGI. The beauty of it is that it uses actual film grain. The grain looks real simply because it IS real. I have a small collection of digitized film grain, or more accurately, black footage from film transfers, i.e. the beginning of an archive film, or a telecine session. I simply put these on top of video in additive or screen transfer mode mode. It works pretty well. I control the amount of grain by color correcting the grain layer - more contrast = more grain. Shooting a gray card so that it's exposed to 50% luminosity, and compositing that in overlay mode would be theorethically slightly better method. In my "black" method, the grain always lightens the image. With "gray" method, some of the grain lightens, some darkens the image - dunno how much that matters in real life ,-) I was just referring to technical quality... that has been out of the reach untill now. That said, there's places outside Hollywood... usually they too work in the same way though ;-)

That pretty much sums it up, i think ;-)

The price savings are really relevant only if the film is done in true guerilla-style. You can shoot with HDV, edit and color correct your work in PC workstation. I'd say total cost of buying equipment to do a somewhat serious HDV production would be around 15-20K, lights and other stuff like dollies not included.. That would include the camera, a tripod, good mic or two, a workstation, enough HD space and all necessary software to make more or less all you can imagine - including compositing, color correction, audio post, 3D and other SFX. Add in 10K for random rental costs, travel etc. This place http://www.dvfilm.com/faq.htm says they'll do the 35mm film transfer starting from 20K. That would be a total of something around 50K for the technical budget... Now, if you were able to get some good people to work with you on "future profit sharing" ideology... you get my drift - full length feature with under 50K budget - and if it sells, the next one would be only 30K, as you already own the equipment ;-) I'm pretty sure i'm not the first one thinking this route - it might very well mean we get to see some movies that wouldn't have been made at all otherwise. And, who knows, some of them might be really good.

I dunno - depending on the scene, even that reverse angle thing might work. What i was tinking of though, was shooting one whole (easy) scene, i.e. something outdoors in an overcast day without the sky showing (you get my drift) and insert that scene in the film. I'm still pretty sure it could be really hard to notice that, especially if you weren't deliberately looking for it. As far as making a whole feature in HDV goes... i'm pretty sure we'll soon see something that was done this way. And it will look petty OK for an indie. Not as great as the best, but decent. Edit: And no, i don't think 35mm film and HDV are "interchangeable" so that you'd get the same result regardless of which you used... that'd be a bit silly ;-)

Please reaload, the .png was broken... there's three shots, varicam, hdcam and 16mm. I thought you were talking about recognizing the film vs video skin tone - the image size shouldn't affetct that... LCD screen might, but if the difference is too small to see on an LCD screen, i'd say it's not a real issue. If that's the case, doing pretty much anything on the set has so much bigger effect on the picture than the difference of formats, that it makes the argument moot. If you have the bandwidth, have a look at that comparision video - it has a load of examples of the differences, some favoring film, some favoring video.

Okay - So, do you still think it would have been impossible to insert some HDV shots within a 35mm movie, without you noticing it?? Meaning, you can positively ALWAYS see the difference between film and HDV?? Because that's what this discussion started with ;-) Actually, if you wanna make film look like video, the easiest way is to shoot 50fps, F11 or more, telecine it with best light, then speed it up by 200% to get 50i cadence. You could fool pretty much anyone just by doing this. I've actually tried it - it works!

Sorry, the HDV clips were NOT from the video. The video is available as HDCAM, VARICAM and 16mm. Quite nicely selected scenes, showing different situations, interior, exterior, day, night, flames, closeups, wide shots, fast motion etc. Anyway, interesting comparison, even though it's very big file, that "Szenen aller Formate im AB-Vergleich" is a good thing to watch - it has split screens and enlargements of the various formats etc. It also has some 35mm shots as well as DigiBeta. For those who don't want do download half a gigabyte of video, i took the liberty of taking a screen grab of the video, with each of the formats showing... tell me which is film? If i didn't know, there's no way i could tell the difference. Okay, i thought you did claim you had NEVER seen a SINGLE SHOT that was taken in video, that could be mistaken for film. I recall you said: "You claim that skintones shot on video can be made to look like those on film, but I have never and I mean NEVER seen that. Hell, I even recognize whether a film went through a DI or not by looking at the skintones."

I trust you - but at least in digital domain, the color is simply math: it's just numbers. To make one color to turn into another, you can simply add and substract etc., ending up in a perfect match. As far as recognizing digital intermediates goes - there's always some loss when moving from one format to another, and especially so when making analogue-digital-analogue transforms. Something gets lost, and you can see it. What exactly it is, could probably be measured by comparing two positives, one of which has gone through DI and the other which has not. Naturally, this should be done using the same original negative - it should even be possible to use this test to modify the DI process so that it's more transparent. I was trying to find some A/B tests with film and video (to test that skin color matching) - and ran into this: http://www.hd-channel.com/index.html?/programme.html They have the same pop video shot in Super 16 and various HD formats (including HDCAM and HDV). I'm downloading the (big) windows media files as i type...verrry interrresting.

Okay, i do get your point - but i got to disagree with you a bit. I still say that film's higher tonal range isn't the thing here. Skin tones fall within a certain range of hue, saturation and luminosity, when you watch them from film, in a theater.. As far as i know, this range can represented in 8-bit YUV, or 8-bit RGB without problems, as well as displayed by RGB devices, easily. The fillm's larger tonal range is an advantage only in very saturated deep colors (cyan especially, i've heard). If you took a color densitometer with you to the theatre, took readings from the skin, the readings wouldn't be outside computer screen's range. But now to the part where i agree with you - colors *do* look different in the computer monitor. This doesn't really have as much to do with the abilities of the monitor itself, but rather with the surroundings - you watch films in a darkened theater, the monitor in dimly lit (at best) room.

Maybe this is the essence of it: Video cameras are made to produce "mathematically correct" imagery, trying to get as linear reproduction of incoming light as possible. Technically speaking, this is good, but it doesn't always look cool, so visually oriented may consider it a fault. Film, on the other hand, is not as linear / neutral. The midrange has more contrast than lows and highs. To the technically oriented, this is a fault, but it happens to look visually pleasing. So, in practice, we tend to adjust the linear video to resemble nonlinear film, to get that cool look. I don't see why it would be impossible to make a video camera, that has a "film stock" setting built in - it could be a simple lookup table, based on measurements of actual real life film behaviour. When the sensor gives color X, the in-camera software would transform it into color Y. What Y actually is, would depend on the selected film stock. Or, in my case, i could adjust it to low contrast, but color wise linear output. Or whatever else, using some kind of "lookup table maker" software. It's just math, not even too complicated as such - we need to start calling the manufacturers, maybe they'll implement this to their next models ;-)

I don't think the 3:1:1 vs 4:4:4 sampling has much to do with this - it only affects the spatial color resolution. 8-bit linear vs. 10-bit log might have much bigger effect on the process - maybe log color space simply works better with the color correctors. But i think the real reason is the camera itself - like different film stocks, different video cameras do have their own look, and maybe "the F900 look" simply isn't what you're after... some other model might give you instantly what you want.

This really should be a matter of timing the footage correctly - not an color depth issue. Flesh tones usually are are in the mid range of luminosity, and of quite low saturation. This shouldn't be a problem area in either of the formats. In VERY, VERY general sense, film tends to make skin pink and video tends to make it yellowish. The technical difference is small, easily within the range of adjustments, both ways. Did i mention this is a VERY, VERY broad generalization? Could you point me to a film frame grab that has excellent skintone in your opinion - i'll try to find a similarly lit video clip and color correct it to match... might be a fun test?? *** I stand corrected on the film printers - it seems more or less all my info on film techniques is rather outdated ;-)