Workprinter and High-Definition

[mattias]

So in order to sample film lines (or any kind of detail) that are not strictly vertical or horisontal it takes a lot more pixels per mm than lines per mm on film.

well, if you call a factor of the square root of 2 a lot. or is my geometry off?

the last time i did this test i saw around 600 lines iirc. search the archives for details. the opinions of people who discuss numbers without ever having shot a test chart should be taken with a grain of salt. theory is interesting for sure but there are so many variables that testing is the only way.

/matt

[christoph]

If avarage super8 image has 600-700 lines (which I don't doubt) , it takes a lot more resolution to transfer it to video.

yeah, you definitely need some oversampling.. but you should distinguish between scanning sensor and storage format. for example:
- the cinealta has a 1920x1080px sensor, stores it on a 1440x1080px codec, and gets about 1200x950 lines resolution.
- the canon HDV has a 1440x1080 sensor, stores it in a 1440x1080 codec, and resolves about 800 lines
- the sony HDV has a 960x1080 sensor stores it on a 1440x1080 codec, and resolves about 700x750 lines.

so you dont need twice the amount of pixels because todays image sensors use pretty clever algorithms
if we're strictly speaking resolution, i think 720p HD would hold up the full super8 resolution, but since we're using consumer grade cameras with average lenses, interpolated sensors, and compressed codecs etc, starting off with 1080i is maybe not a bad idea.

so, a very high quality HDV workprinter transfer could look like this:
- capture single frames at 1080i (1440x1080 pixels with 1.33 aspect), (if you really want to max it, use the analog/HD-SDI output and an uncompressed capture card)
- color correct
- decide what's going to be your high def master (and ask yourself, how do you play back HD anyway?).. then resize and crop to that size (i'd probably use a 960x720 pixel quicktime file with photo jpeg codec, or a 720p24 DVCPRO HD file).
- downconvert to PAL/NTSC, apply pulldown etc.

oh well, sorry for all the numbers.. obviously if you start to operate at this levels you should have a fairly good understanding of batch processing workflows. ;)
++ christoph ++

[christoph]

Obviously the topic is open to interpretation.
[snipped a lot of calculattions]
If we assume Christoph's 4000 lines of resolution for 35mm, the equivalent Super8 frame would then be .254 x 4000 = 1020 lines of resolution, fully in the HDCAM ballpark.

sigh, i knew that would happen ;)

first, i was giving ballpark figures here
second, if you take my vertical numbers (550 for super8, and 2500 for 35mm) you see that they match up pretty well - which is were i was probably wrong because of....
third, 35mm prime lenses are of much higher quality than most/all super8 lenses.

all calculations aside, if you really think that super8 has the same resolution as HDCAM i'm probably wasting my breath:
http://www.chicagohd.com/images/frame-grab-02.tif
http://www.chicagohd.com/images/frame-grab-04.tif
http://www.chicagohd.com/images/frame-grab-01.tif

++ christoph ++

[mattias]

all calculations aside, if you really think that super8 has the same resolution as HDCAM

there seems to be two kinds of people. those who *think* it does and those who *know* it doesn't. :-)

/matt

[Dave Anderson]

so, a very high quality HDV workprinter transfer could look like this:
- capture single frames at 1080i (1440x1080 pixels with 1.33 aspect), (if you really want to max it, use the analog/HD-SDI output and an uncompressed capture card)

I think the question was asked if Dodcap is capable of capturing anything more than 720x480? Does anyone know?

[BK]

All I can say is that I managed (with some difficulty) to align my Sony HDV camera with one of our CineMate units and the difference in detail as seen on an HD monitor was quite staggering, compared to SD. Frankly, I was surprised at the difference. It was not a perfect fit due to the rather short zoom

Well Roger, guess you have to saw off that plastic Carl Zeiss lens and use a "real" lens on your Z1.

http://www.eidomedia.com/hdv/

Bill

[tlatosmd]

SUPER 8 motion picture film has about 1315 lines of resolution! Over TWICE the amount of digital video!

To say that 720 pixels will yield 360 lines of resolution is a starting
point

This type of back-of-the-envelope calculations gives us a required pixel
count of 500 to 1500 per mm on film, as opposed to the 200 suggested above.

That's what I like to hear!

And I suppose this all refers to potential of the camera original film. Whether you reach this upper limit is also determined by lens, etc.

the best you can do is to set up a resolution target, point your camera at it, film it and .. (gasp) ... count.

One shouldn't confuse these lines with pixel resolution, though. Whether it's two, four, or five pixels per line is the question they have tried to solve over at Davideo's link. Even if you'd just take two, 1,000 lines (or more, as quoted on Andreas's site where Davideo also linked us to) would equal 2,000 pixels.

Anyway, why count, and not just print the number of lines on the chart? As soon as it's too many, it's only grey matter.

yeah, you definitely need some oversampling.. but you should distinguish between scanning sensor and storage format. for example:
- the cinealta has a 1920x1080px sensor, stores it on a 1440x1080px codec, and gets about 1200x950 lines resolution.
- the canon HDV has a 1440x1080 sensor, stores it in a 1440x1080 codec, and resolves about 800 lines
- the sony HDV has a 960x1080 sensor stores it on a 1440x1080 codec, and resolves about 700x750 lines.

so you dont need twice the amount of pixels because todays image sensors use pretty clever algorithms

They're clever because they downsize two times before playback, sacrificing detail? Even on top of that, they digitally compress it.

all calculations aside, if you really think that super8 has the same resolution as HDCAM i'm probably wasting my breath:
http://www.chicagohd.com/images/frame-grab-02.tif
http://www.chicagohd.com/images/frame-grab-04.tif
http://www.chicagohd.com/images/frame-grab-01.tif

When I look at those pics at 100 percent zoom, they look like bad JPEG compression, with halos and similar dirty looking patterns. Of course they might look niftier when looking at them at about 50% zoom level which is when they fill a 1024*768 monitor.

[christoph]

That's what I like to hear! :)

that's excactly the problem: you'd like to hear something, and you try to ignore anything that would tell you something different.

One shouldn't confuse these lines with pixel resolution, though. Whether it's two, four, or five pixels per line is the question they have tried to solve over at Davideo's link. Even if you'd just take two, 1,000 lines (or more, as quoted on Andreas's site where Davideo also linked us to) would equal 2,000 pixels.

you really shouldnt believe anything you read on the web ;)
the first link is from a production newsgroup (although scott dorsey and robert morein are technically very knownledgable) and the second one is just plain wrong (other than on the cost issues).
but nobody argues that resolution doesn't take into account grain pattern etc, it's just ONE way to get something measurable.

Anyway, why count, and not just print the number of lines on the chart?

if you look at a resolution chart you'll find that's what they actually did.. i was just trying to be funny.

They're clever because they downsize two times before playback, sacrificing detail? Even on top of that, they digitally compress it.

what are you talking about? as discussed before, you'd actually need 2000 pixels on the chip to resolve 1000 lines.. the cinealta does it with 1080! i'd call that pretty clever. the digital compression is not a chip problem, but a codec problem, and if you ever edit HD you'll see why they do that.

When I look at those pics at 100 percent zoom, they look like bad JPEG compression, with halos and similar dirty looking patterns.

HDCAM is 1440x1080 with 3:1:1 and additional compression, you could bypass that if you use the HD-SDI out.. but what you're seeing are rather unfortunate lighting and/or heavy post production. it was the first samples i came across, but feel free to search for some more, such as:
http://hd24.com/images/HDSDI%20toys.tif
http://hd24.com/images/HDCAM%20toys.tif

i'm not saying that HD looks better than film, but we were talking about resolution, which is easily measurable.. take the best super8 camera you can find, point it at a resolution chart, use vision2 50D film, scan it on a 8000dpi film scanner and see how many lines you get.

++ christoph ++

[S8 Booster]

s8 gives you guys the option to shoot anything you like on film from experimental to pro - you are the most limiting factor anyway so what is the fuzz?

resolution?

id like to quote Rolls Royce from back in the 60s. when they were asked how many kWs (ok, Hps then for you nostalgy nerds) their V8? motors produced their answer was always "sufficiently" with a posh nose up attitude.

point if you havent got it: s8 has more than adequate resolution to allow you guys to realize any film dreams you may ever have and generally more resolution than most "filmmakers`" capacities will ever need.

there are umpteen variables between s8 film - film and hd originated digital stuff that comes into play by comparisation. each and every one of them can shuffle the results on either side. btw; from audio world we know that an oversampling of 2:1 is the minimum to obtain anything like "true" response for reproduction. bet film sampling isnt way off. more is better always but not allways required.

if you like film and want to learn about filmmmaking on film shoot film.

if you want the cleanest, sharperst and stone dead images despite technical artifacts as a bonus - available for reasonable bucks - get some hd originating stuff and get bored to death.

a true filmmaker can do independant of format.

christophs` v2 50d suggestion for s8 will be as close as you get in s8 and more than adequate to reveal real life s8 performance.

s8 done right looks hell great. stick with it.

my twinola DuNots

s8hôôt

[Konton]

Cinecap doesn't do HD. I asked the programmer. But when my Panasonic HVX-200 shows up . . . . well I'll do a little testing before I sell my Workprinter.

[MovieStuff]

Cinecap doesn't do HD. I asked the programmer.

It doesn't do HD using HD compression but it does do uncompressed of most anything you feed it. DTS is capturing in HD uncompressed using a butt-load of hard drives in a Raid configuration. This is a specialty set up that they have been working on and not something the average WorkPrinter user would likely pursue.

Roger

[MovieStuff]

All I can say is that I managed (with some difficulty) to align my Sony HDV camera with one of our CineMate units and the difference in detail as seen on an HD monitor was quite staggering, compared to SD. Frankly, I was surprised at the difference. It was not a perfect fit due to the rather short zoom

Well Roger, guess you have to saw off that plastic Carl Zeiss lens and use a "real" lens on your Z1.

http://www.eidomedia.com/hdv/

Bill

Hah! That's interesting. I have been giving some thought to offering a mod service for DVX100 owners so they can put M42 lenses on their camera. Personally, I like the Zeiss lens on my Sony HDVZR1U. Sharp as a tack, in my opinion. I see no real reason to change it.

Roger

[ekta-clone]

So in order to sample film lines (or any kind of detail) that are not strictly vertical or horisontal it takes a lot more pixels per mm than lines per mm on film.

well, if you call a factor of the square root of 2 a lot. or is my geometry off?

the last time i did this test i saw around 600 lines iirc. search the archives for details. the opinions of people who discuss numbers without ever having shot a test chart should be taken with a grain of salt. theory is interesting for sure but there are so many variables that testing is the only way.

/matt

You saw 600 lines on the film. Tell me how can you sample those 600 lines with 600 pixels if the lines are diagonal?

[mattias]

You saw 600 lines on the film. Tell me how can you sample those 600 lines with 600 pixels if the lines are diagonal?

why on earth would i do that? to prove what? i have obviously never made any such claims so i can only assume it's you who have some kind of point to make? if so i suggest you make it and leave me out of it. thank you.

/matt

[mattias]

not that i expect anyone to care since facts don't seem to interest anyone but those who already have them, but ccd pixels are not necessarily aligned in rows and columns and their number don't necessarily match the resolution of the format it's read out as.

sampling ccd data digitally is a very similar problem to scanning film. as a matter of fact the main reason you never get the full res theoretically possible in a video camera is anti-aliasing, a function designed exactly in order to make lines in all directions resolve the same.

what makes people think that political apples and oranges debates would somehow resolve (sic!) this complex matter better than a really simple test chart? on a side note it's funy that those who have never shot a test chart are also those who have the least understanding of the theory, and vice versa. ;-)

/matt