Super 16 Not HD Resolution

>The places I've been calling though have surprisingly told me that if I'm >expecting HD resolution out of super16, that I'm going to be a bit >disappointed.

Jason,

Well you know I ran some material at IBC projected digitally from HDCam, even the Kodak member of the panel thought that it was 35mm originated.

It was Super 16 and looked very different to the HD originated material that was shown.

I was showing it before and after grading, making the point that in these situations Super 16 was the cheapest way to go because of the shoot time saved because of the latitude I had in the grade that I wouldn't have had form a HD master, whatever format I used.

And in terms of resolution, well it depends on what you call resolution, if you're talking about enhanced pictures that give an impression of sharpness but contain little information then HD will win every time.

Ultimately it has to be a question of what do you want it to look like?

That's the only question that matters.

Once you decide that then all you have to do is dig through the bullshit, from both sides.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based
www.cinematography.net

---

Jason,

I second Geoff's thoughts that Super16, HD, and 35mm all look different when projected. A lot also depends on what you are shooting - for example, shooting Monument Valley Landscapes on Super 16 or HD will be noticeably different than 35mm or best choice 65mm. If on the other hand, you are shooting actors in interiors - the differences are minimized. Resolution isn't everything - if it was - more DP´s would be throwing away a lot of their diffusion filters and silk stockings.

I have seen lots of projects projected on multiplex size screens. The ones that succeed work because the look aids the story. In the end, there's only one way to know for sure what will work for you - shoot some tests. Everything else is assumption, someone else's.

Robert Goodman
filmmaker/author
Philadelphia, PA

---

>The rental houses have also been telling me (those that also offer HD >equipment, so I'm not sure how much of this is innocently biased)

That phrase you posted right there says it all...

John Babl

---

Geoff Boyle wrote :

>if you're talking about enhanced pictures that give an impression of >sharpness but contain little information then HD will win every time.

Geoff,

On the Viper/F900/5218 exposure tests that you conducted, and have the DPX files on the website, I noticed that the resolution among the three was about the same, although it did seem like the F900/Viper was resolving the tighter circle-patterns on the Arri sharpness chart a bit better than the 35mm film (I know that the telecine might have been out-of-focus, but it didn't look like it was by much if it was, so I'll basically call it a tie among the three).

Now that was 35mm, what happens with Super16? That's what makes we wonder if these DI guys in NY/LA are correct, because if the 35mm had some problems resolving the entirety of the sharpness rings, but was on even par with the Viper/F900 (the Viper was the best of the three), then how can Super16 have more information that would bring it on parity with the 35mm (which was already about equal to the HD cameras)?

The HD footage might be enhanced footage, but it's still resolving information nonetheless.

Again, I'm just wondering because I've heard it taughted by Kodak that Super16 is HD res (which I kind of wondered about, but accepted because I figured they knew best), but now I'm hearing from the people doing these hi-res scans of super16 that if it's HD resolution I want, super16 might not deliver the useable resolution I'm imagining it would.

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

Jason Rodriguez wrote :

>The rental houses have also been telling me (those that also offer HD >equipment, so I'm not sure how much of this is innocently biased) that >shooting on HD will give me better digitally projected blow-ups than >super16

You're using words like "good" and "better" without defining what "good" and "better" mean to you. To video people, lack of grain, extreme depth of field, and enhanced midrange sharpness constitute "better" when compared to film. To film proponents, controllable grain, additional exposure latitude particularly in the highlight areas, the ability to shoot at any frame rate, smaller camera size, and more controllable depth of field (less so in 16mm, of course) make that medium "better."

With video, you'll get additional running time without reloads, no grain, enhanced sharpness (not real additional resolution, of course), more extreme (and therefore less controllable) depth of field, limited exposure latitude (especially in the highlights) and less manipulative capability in post due to the limited color palette, 4:2:2 sampling, and compression applied at the time of original recording (although the last item, compression, is much less of a factor than many here make it out to be). With film, in your case 16mm, you'll get film grain, a bit more control of depth of field, much more exposure latitude in the highlights, a bit more natural contrast in most cases, more manipulative capabilities in post (both in color timing and the ability to reframe with minimal loss), a more "filmic" color palette and look (because, well, it's film) and, if you need or want it, a more stable and compatible archival element. You need to figure out which of these characteristics are important to you based on the script and your own vision for it.

Perhaps most important is the "Robert Rodriguez Factor." By this I mean that if you don't have a lot of experience as a cinematographer, particularly film experience, you're probably better off shooting video for the simple reason that you can see the image as you're lighting and shooting - thereby saving you from yourself. I believe the reality is that if you have or are an experienced cameraman, you'll ultimately get more satisfying and interesting results shooting film, but if you aren't or don't have an experienced cameraman, you'll be better off sticking with something like video that will involve less "guesswork" on your part.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles

---

Robert Goodman wrote :

>you are shooting actors in interiors - the differences are minimized

Yes, it will mostly be actors/interiors with a little exterior work, but not too much.

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

>So where and how do you plan to project?

>It will be (as far as I can see right now), digitally projected, with a screen >probably no larger than a medium-sized Cineplex screen

--- Sounds like something that WMV9-HD could handle...what does the rest of the group think?

Jeffery Haas
freelance editor, camera operator
Dallas, Texas

---

Jason Rodriguez wrote :

>The places I've been calling though have surprisingly told me that if I'm >expecting HD resolution out of super16, that I'm going to be a bit >disappointed.

Who are you talking to? Places that have a vested interest in HD over film? S16 has many virtues.

If this project will have to suffer extreme digital compression (like the Discovery Channel's HD material) then there are potential disadvantages to S16 (grain, being different in every frame, doesn't compress nicely) -- which is why they have told people not to use S16 for Discovery productions -- much to the chagrin of many wildlife cinematographers, who really dislike the video alternatives they are being forced to use.

But for theatrical projection, on film or DLP (and why not just do a 35mm filmout? -- you can show it everywhere, not just in a few theatres) S16 is a fine format.

I've seen some ok looking HD on screen, but there's always something muddy about it -- I think people get enamoured with all the shadow detail they can get with these cameras, and light in ways that don't translate to the big screen. Altman's recent "The Company" is a good example -- parts of it looked fine, parts of it looked murky in ways that I don't think were intentional.

I think part of this is due to the fact that the current generation of laser recorders favor dupe stocks, which are designed to reproduce a film image with, essentially, unity gamma. There's absolutely no personality to these stocks (a good thing when duplicating film, a terrible thing when transferring digital images to film IMHO).

The new 4K film recorder we are introducing after NAB is designed to use camera stocks, with all their wonderful artefacts and texture (i.e. grain) and runs at up to 8 fps, meaning, if one wanted to have 35mm dailies from a digital shoot, or cut film, it would be easily done. The higher speed of the recorder also means that the price of a feature filmout could drop to under $15K -- which would make it far more likely for people to release on film rather than digitally.

I still think that viable, widespread digital projection is many years away, as there's still no impetus -- financially -- for theatres to go down this path. After all, today's $20,000 35mm projector will last for 20 years -- but today’s $100K digital projector will be replaces with something far better in three years. $1000 per year or $33K per year -- tough call? I don't think so.

Remember too that for large releases, most of the prints made for the US market are cleaned, treated with scratch-removal machines, repaired, and used for the non-US release. So the prints are going to be made regardless, until all theatres sport digital projection gear.

Note also that it's very likely that digital films will be far more susceptible to bootlegging, even if protected by software keys. There are many clever hackers out there with nothing better to do.

Jeff "trying to be informational" Kreines
www.kinetta.com

---

Michael Most wrote :

>You're using words like "good" and "better" without defining what "good" >and "better" mean to you. To video people, lack of grain, extreme depth >of field, and enhanced midrange sharpness constitute "better" when >compared to film.

Excellent point, Mike! Great post.

Jeff "likes good alternative points of view" Kreines

---

Jeffery J. Haas wrote :

> ----Sounds like something that WMV9-HD could handle... what does the >rest of the group think?

Why compress it?

I don't see the logic unless there's a need to get it to fit into a pipeline or storage medium that can't accommodate it otherwise.

Jeff "compression isn't inherently desirable" Kreines

---

>Why compress it?

>I don't see the logic unless there's a need to get it to fit into a pipeline or >storage medium that can't accommodate it otherwise.

-- I'll have to dig up the original post (so dang many of them, not that that's a bad thing!) I was under the impression that he had mentioned WMV9-HD somewhere...so I "connected the dots".

It would not be my first choice, as Geoff handily mentioned some others, like HDCAM, which would most likely do a lot better. I also realize that lots of venues aren’t digitally equipped, and I was shocked to learn that digital projectors don’t have as much longevity. I assisted in the install of a JVC DLA-QX1G recently and I was impressed with its performance but HDCAM was used during the initial setup. I don’t know what the expected product life of the unit is, but the price tag was pretty hefty.

Jeffery Haas
freelance editor, camera operator, installer
Dallas, Texas

---

Not long after Vision 2 '18 came out I shot a couple of small tests. One extreme test where I pushed it two stops and had the shadows at 2 footcandles. Of course grain came into play, the blacks became a dark shade of grey, but the image was very satisfactory. Now we have Vision 2 '29 with even less contrast, so HD will have to give up its claim on superior shadow detail.

Pushing it two stops essentially under candle light, you can still get a 2K scan at 16bit log, which far from the most advanced HD format.

Tenolian Bell
Cinematographer NY

---

Hi Jason,

Geoff Boyle wrote :

>if you're talking about enhanced pictures that give an impression of >sharpness but contain little information then HD will win every time.

Jason Rodriguez wrote :

>it did seem like the F900/Viper was resolving the tighter circle-patterns >on the Arri sharpness chart a bit better than the 35mm film (I know that >the telecine might have been out-of-focus, but it didn't look like it was by >much if it was, so I'll basically call it a tie among the three).

The Kodak transfer was significantly out of focus, the Fuji was sharp.

No detail correction, aperture correction, was added to any of the film transfers, this makes a significant difference to apparent sharpness.

Have a look at this link.

It won't be a permanent one.

/sharpness.htm

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based
www.cinematography.net

---

Geoff Boyle wrote:

> /sharpness.htm

CRT's are not good much beyond about 10MHz ... so, the full impact of the test may be significantly more dramatic than shown.

Martin Euredjian
eCinema Systems, Inc.
www.ecinemasys.com

---

Great stuff - this is CML at its best for me.

It was an interesting trade-off to me - the HD looked softened, but harmonious and the film showed more structure, but at the same time it introduced a weird (2nd order?) artefacts.

Dale Launer
Writer / Filmmaker
Santa Monica

---

Please bear in mind that these files are copies of copies.

They may well have undergone compression at some point.

Even so, they have undergone the same compression /changes and the differences are still marked.

If you look at the full res version of the first image then I think you'll get the clearest idea of the differences, the others are just there for the hell of it.

I have more of the highlight latitude type examples if you're
interested, I didn't upload them because they are done with an old version of the HDCam, V2 as against V3, mind the film scans are from '79 so maybe it would be fair to show V2 HDCam......

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

Wow,

I've been wanting to see one of those ISO charts shot with a HD/Film camera for a long time now!

I'm really surprised by the performance of the HD camera. I've seen these charts a ton since a well-known digital camera site I know uses these charts to measure the resolution of manufacturer's digital cameras. Anyways, 850 LPH (lines per picture height)? That's horrible!!! The Nikon D1, which was a digital still camera that has LESS pixel resolution that the F900 does around 1200 LPH! I'm not sure if it's aspect ratio differences (3:2 versus 16:9), but they should be the same resolution. Is this the results of pre-filtering in the HDCAM tape format down to 1440x810 (I think it is)? If so, then that should be about right, 800 LPH for a pre-filtered signal.

BTW, what was also interesting is that the 500T film is performing on par with the current crop of six megapixel digital cameras, like the Canon 10D, Nikon D100, etc., which I find funny since Kodak's been claiming that film has 12+ megapixels, yet digital cameras at six megapixels are matching it's resolution on the ISO charts in LPH.

Well, anyways, that was very enlightening. Thanks again,

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

>BTW, what was also interesting is that the 500T film is performing on >par with the current crop of six megapixel digital cameras, ... which I find >funny since Kodak's been claiming that film has 12+ megapixels,

Jason,

Any chance you're comparing 500T in S-16mm motion picture [vertical running] format whilst Kodak is comparing 35mm horizontal format STILL film [24x36mm image area]?

Cheers,

Clive Woodward,
Perth, Western Australia.

---

Hey,

BTW, Geoff, what resolution were the film scans made at?

I'm thinking that 1800 LPH is a good ways beyond the Nyquest frequency response possible with a Spirit 2K transfer.

If the film was scanned it at 4K, and it's being compared to HD, well . . .

Either way, it's very enlightening to see what could be possible with film, even though we don't have the infrastructure right now to go with 4K scans for an entire film.

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

>If the film was scanned it at 4K, and it's being compared to HD, well . . .

I'm sure Geoff would be happy to redo the tests and scan the HD at 4K too.

Jessica Gallant
Los Angeles based Director of Photography
West Coast Systems Administrator, Cinematography Mailing List
https://cinematography.net/

---

Jessica Gallant wrote :

>I'm sure Geoff would be happy to redo the tests and scan the HD at 4K >too.

No, no,

Please don't get me wrong,

I'm just wondering how the film so easily outclassed the HD, and my only conclusion can be that the film was scanned at a higher resolution, because isn't the camera inside a Spirit basically an HD camera (souped up of course, but still I believe the native resolution of the Spirit (not the Spirit 4K) is 1920x1080, please correct me if I'm wrong)? If it is, then how can one HD camera have almost twice the resolution of the other in LPH?

I was just trying to sort out this comparison in light of the work flows that we have available to us right now. But again, it is amazing to see what is in store, although I was a bit surprised that even scanned a higher resolutions, 500T only has the same absolute resolving power as a six-megapixel Digicam (Canon 10D, Nikon D100) at 1600-1800 LPH; so you don't need a 12 megapixel Digicam to match film's resolution like Kodak's been saying, at least not yet-although I do realize that a 4K scan at academy is 12 megapixels. But it looks like a 3K Digicam should compete very favourably.

In other words, the 4K Dalsa should be quite interesting...

Jason Rodriguez

---

>No, no, Please don't get me wrong...

Sorry. I was just being silly.

Jessica Gallant
Los Angeles based Director of Photography
West Coast Systems Administrator, Cinematography Mailing List

---

Just to confirm, is that 35mm or Super-16 500T being compared to HDCAM? The charts don't say.

David Mullen ASC
Cinematographer / L.A

---

Jason Rodriguez wrote :

>If it is, then how can one HD camera have almost twice the resolution of >the other in LPH?

Good point that the Spirit's basically got an HD camera inside it.

Perhaps part of the answer for higher resolution is that it also has simple, high quality optics specifically designed to focus only on a film plane a few inches away, with a diffuse light source, and the film has already consolidated the contrast and color space somewhat (as opposed to designing a lens that must be f/1.4, can focus through a long range, not flare with bright sources, not breathe, etc).

Additionally, the data collected can go straight to disk, uncompressed. Part of the problem we see with HD is the image data we must throw out in order to store it on a format such as HDCam.

Also, as you probably know, if you use a decent film scanner you get higher quality than you would from a Spirit - but not real-time transfer rate.

Mark Doering-Powell
LA based DP

---

WOW I've seen some BS Tests before and this is obviously just another one of them.

I cant even clearly make out the number "6" as in 600 lines on the HD rez chart.

I don't know what the person shooting the test was doing, but I seem to be getting much greater mileage out of my resolution when I shoot HD than this test shows.

As I have said before "Believe nothing until you test it for yourself"

I'm Glad that Geoff will not be keeping this up for long...it's just incorrect.

B. Sean Fairburn
Director of Photography
Castaic Ca

---

Mark Doering-Powell wrote :

>Good point that the Spirit's basically got an HD camera inside it

And if it does, then it's impossible, re: Nyquist, to get than many LPH (1800) in a 2K scan at 16x9. Again, if I'm wrong, please correct me in where my error lies.

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

B.Sean Fairburn writes :

>I'm Glad that Geoff will not be keeping this up for long ... it's just >incorrect.

I would suggest that it's more correct than some of the comparisons I've seen.......

It's 35mm and a 4K scan, of course.

As for the resolution visible of the HD, I did say that it had ALL been compressed and copied several times.

I also pointed out that this affected the film just as much as the HD.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

Woodward, Clive wrote :

>Any chance you're comparing 500T in S-16mm motion picture [vertical >running] format whilst Kodak is comparing 35mm horizontal format >STILL film [24x36mm image area]

According to Geoff, that was 35mm with a 4K scan, and Kodak has repeatedly stated in their panels that you'll need a 12 megapixel digital camera to compete with their 35mm stocks-this scan obviously shows that's not the case, since 6 megapixel (3K x 2K) digital cameras can do this many LPH's.

So while the actual resolution of the image is not the same (12 vs. 6 megapixels), the actual resolving power (1800 LPH) is.

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

Jason Rodriguez wrote:

>According to Geoff, that was 35mm with a 4K scan, and Kodak has >repeatedly stated in their panels that you'll need a 12 megapixel digital >camera to compete with their 35mm stocks-this scan obviously shows >that's not the case, since 6 megapixel (3K x 2K) digital cameras can do >this many LPH's.

6 megapixel digital cameras don't shoot at 24fps.

Just as you don't compare motion picture film results with large format stills, you can't compare 35mm motion picture images to "large format" digital stills. Personally, I think a lot of this "testing" is a bit presumptuous anyway.

Geoff 's tests show a certain result. Sean says he has tests that show a different result. One can set up tests any way one wants and perform them until the results they thought they'd get are there. That kind of thing is done all the time. Just ask Sony. Or Panasonic. Or Kodak.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles

---

Geoff Boyle wrote :

>As for the resolution visible of the HD, I did say that it had ALL been >compressed and copied several times.

Forgive me if I missed a post (I was travelling), but, it might be useful to post a description of the exact process by which these images were generated. My comment about monitors is only relevant to scans being done on an HD Telecine system. If your monitor can't deliver top MTF there is no way you can focus optimally. Same with a camera, of course.

If the HD process is camera to HDCAM to film recorder and then 4K scan a few questions arise:

What exactly does "compressed and copied several times" mean?
What equipment?
Any processing in between?
How many times is "several times"?
What was used to go from HDCAM frames to film?
Was there any processing (like scaling) in making data for a scanner?
If so, what software was used for scaling/pre-processing?

I'm not a fan of HDCAM, but those pictures look terrible. Shot and treated optimally it can do better than that, of course. And, I think everyone understands that "optimally" might also mean NOT using HDCAM for recording.

Martin Euredjian
eCinema Systems, Inc.

---

As Martin Suggested

What Lens was used on the HD CAM and where was the detail set

What was used to do the Back Focus and what was the path through the "Compression Gauntlet" from original through to what we see.

What I say on the HDCAM version was at best Out of focus.

Not an Accurate comparison

Sean Fairburn
LA HD DP

---

Michael Most wrote :

> 6 megapixel digital cameras don't shoot at 24fps.

Not this year.

Next year should see 4K cameras capable of 60 fps (progressive), if certain component manufacturers deliver what they promise.

Jeff "not being vague, just can't say more" Kreines

---

Sean Fairburn writes :

>What Lens was used on the HD CAM and where was the detail set. >What was used to do the Back Focus and what was the path through >the "Compression Gauntlet" from original through to what we see.

>Not an Accurate comparison

/sharpness.htm

These images are the ones produced in a test by Roger Morton et al, Kodak (Rochester) and published in the SMPTE journal Feb/Mar 2002, pp85-96. If you refer to this article, it gives some background on how the images were made. (These exact images are in the paper).

5 HD cameras were rented - "3 - 24P, one 30P and one 30i" (60i?) Lenses were primes from Canon, FUJI and Panavision "Cinematographers shot both scenes and targets, video engineers attended to the 24P cameras"…

>
"The field of view of each camera was aligned to marks on each target".

The outputs were put into progressive format, cropped to 1.85:1 and "Kodak software was used to transform the data into a global color space, analyse the resulting data and review the resulting images". The cameras were set to zero sharpening and the one with the most boosted frequency response was selected for the sharpness and resolution tests.

Film shot was 5279 and 5274, full aperture (for the reported results.) It was scanned at 4096 x 3072 and cropped to 4096 x 2214 resolution, at 10 bits RGB. Kodak software was used to grain reduce, sharpen and transform the cineon image into a common printing space.

Matt Cowan

---

>
Thanks for the info Matt,

I just find it kind of weird that Kodak can't seem to get those 1080i/p cameras to 800 LPPH. If you go over to Scott Billup's site where he's shot a bunch of cameras with the DCS CamAlign chart, you'll see that they can easily resolve that amount of LPPH.

For instance check this one out :

http://www.pixelmonger.com/hd_assets/hdl_ajhdc20.jpg

That's the Panasonic AJHDC20, a 1080i camera.

The Varicam is here :

http://www.pixelmonger.com/hd_assets/cam27V.jpg

While the Varicam is just about making it to 800 LPPH (surprisingly in the corners it's having no problems, but seems to have a little bit of trouble at the top and bottom of the frame), the 1080i camera is having no problems whatsoever. Kodak's chart is placing the resolving LPPH at around 750 lines, which seems a little low.

And of course this doesn't tell us how much resolution is coming out of a Spirit telecine or even a 2K scanner, which is how many films right now are being scanned at in the DI process.

But either way, this was quite enlightening, as I was always wondering where that 12 megapixel number from Kodak that always gets thrown around came from, and this seems to be where it might have originated.

Jason Rodriguez
Post Production Artist
Virginia Beach, VA

---

>
Jason Rodriguez wrote :

>I just find it kind of weird that Kodak can't seem to get those 1080i/p >cameras to 800 LPPH. If you go over to Scott Billup's site where he's >shot a bunch of cameras with the DCS CamAlign chart, you'll see that >they can easily resolve that amount of LPPH.

I find the artefacts in the Kodak test much more annoying than the lack of resolution.

Interestingly, the images that Jason referred to on Scott's site were both made from Panasonic cameras. I remember at the Tech Retreat a couple of years ago, Panasonic displayed a waveform monitor showing the difference in spectral response among the HDCam, DVCPro HD and D5 HD. Their point was that Sony purposely allows aliasing in order to make their pictures appear sharper while Panasonic filters properly and provides fewer artefacts like those shown in the Kodak test. Different strokes...

All that said, Scott has an F900 image :

http://www.pixelmonger.com/hd_assets/hd_f900.jpg

which still looks a lot cleaner than the one in question. It could be that Scott's was not recorded with HDCam compression and the Kodak image was. Of course, most HDCam footage is recorded on standard HDCam recorders so you have to decide which image is appropriate for your analysis.

As Mike pointed out, you can make tests show whatever you like.

cheers,

Charles R. Caillouet
Vision Unlimited/LA

---

>
The cameras were set to zero sharpening and the one with the most boosted frequency response was selected for the sharpness and resolution tests.

Film shot was 5279 and 5274,Kodak software was used to grain reduce, sharpen and transform the cineon image into a common printing space.

I find it interesting that on the HDCAM the sharpness was set to "0" yet the Film was grain reduced and "Sharpened" Not a Fare comparison.

How about you project your test in either Film or HD originated from both and lets see a one light on the film and transfer the HD over straight then project them in Film. OR One light transfer the Film to HD and project both HD origination and Film on HD and do a side by side there.

Funny also that early opponents of HD complained of too much Resolution, and Definition and the images were too Sharp. Since when have we been scanning Digital intermediate out at 4K full aperture (for the reported results.)

>It was scanned at 4096 x 3072 and cropped to 4096 x 2214 resolution, >at 10 bits RGB.

I thing Poster did this on Stuart Little 2 but other than that I would say that this is not common practice as its just not economical.

Since 90% of Kodak is Digital what are they defending??
When does their HD camera come out?

You won't hear about it at NAB... this year.

B. Sean Fairburn
LA HD DP

---

>
> According to Geoff, that was 35mm with a 4K scan,

35mm motion picture, that is after all, what these lists are about.

A still frame would be twice the size.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
Michael Most writes :

>One can set up tests any way one wants and perform them until the >results they thought they'd get are there. That kind of thing is done all >the time. Just ask Sony. Or Panasonic. Or Kodak.

Exactly!

The only way is to test for yourself.

I do a lot of my own testing and post the results here, like the Film/HD tests and the lens tests, I always describe how they were done so people can try/match my methodology.

I said that the sharpness tests were not my own but I do think that they match my own experience/tests.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
>I just find it kind of weird that Kodak can't seem to get those 1080i/p >cameras to 800 LPPH.

The same way that I find it weird that HD people can't seem to get decent pictures out of 16mm.

As I said before, do your own tests, if you don't know enough to do your own tests then get a film/DI person to shoot the film tests and a HD person to shoot the HD tests and compare the results.

That way you get people trying to get the best out of their respective formats.

OK, so Sean, send me some files that make HDCam look better!

B.Sean Fairburn writes :

>I find it interesting that on the HDCAM the sharpness was set to "0" yet >the Film was grain reduced and "Sharpened" Not a Fare comparison.

It had labelled quite clearly the pictures that are sharpened and those that aren't.

What more do you want?

If they'd wound up the detail on the HDCam you'd be complaining about that, "it's be better if they hadn't wound the detail up" or "I have better detail settings"

Isn't it better that they show both with and without correction?

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
Martin Euredjian writes :

>What exactly does "compressed and copied several times" mean?

It means that I have had these files for quite a while and have transferred them from machine to machine etc.

The first comparison, which to me is the most valid, was originally in a format that would not display on a web site so I converted it to the highest resolution jpg. that I could.

Once again, ALL the image went through the same process so the comparison is still valid.

Cheers

Geoff Boyle FBKS
Director of Photography

---

>
Geoff Boyle wrote:

>What exactly does "compressed and copied several times" mean?

>It means that I have had these files for quite a while and have >transferred them from machine to machine etc.

I thought you where describing the image manipulation that produced these files originally.

Look, it's not rocket science. An imager with 1920 horizontal pixels will, at most, reproduce 960 black/white transitions. It's not going to get better than that, no matter how hard you work at it and how you setup the camera. We also know that film, when shot and exposed properly, can blow away most consumer-grade (non gov./mil.) digital imaging solutions. None of this is a revelation of any sort.

Again, maybe I lost a couple of messages. What's they point being made?

Surely this isn't another film vs. HD thread? Remember when we had to come-up with creative spellings for "film"?

Martin Euredjian
eCinema Systems, Inc.

---

>
Martin Euredjian writes :

>Again, maybe I lost a couple of messages. What's they point being >made?

It was a response to somebody asking if they were being misinformed when they were told, by HD rental companies, that HD was sharper than film.

I've also said several times.

You test.

You use what looks best for you.

Whatever it is.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
>An imager with 1920 horizontal pixels will, at most, reproduce 960 >black/white transitions.

Since Sony downsamples the image to 1440 lines prior to compression, wouldn't HDCAM be limited to 720 black/white transitions? Or am I missing something?

Jessica Gallant
Los Angeles based Director of Photography
West Coast Systems Administrator, Cinematography Mailing List

---

>
>OK, so Sean, send me some files that make HDCam look better!

Be Happy To

Michael...May I borrow your new 28 mm Zeiss DigiPrime?

B. Sean Fairburn
LA HD DP

---

>
>Be Happy To

Excellent!

Anyone else ?

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

Jason Rodriguez wrote :

>"If you go over to Scott Billup's site where he's shot a bunch of cameras >with the DCS CamAlign chart, you'll see that they can easily resolve that >amount of LPPH."

Jason :

First "LPPH" means "Lines Per Picture Height" To measure "LPPH" you must have lines running in the horizontal direction as does the chart used by Scott.

NOTHING in the images that Geoff posted have any lines running in a horizontal direction. Circular lines do NOT count.

The full paper is here :

http://www.kodak.com/US/en/corp/researchDevelopment/productFeatures/

>dCinema.shtml

The way this imaging test is conducted is that the printing on the chart is scaled or designed so that on a film or video frame that is framed over the WHOLE chart you determine how many of the alternate white and black transitions are detectable. This is determined for every size and aspect ratio chart.

When displaying the image it is OK to display or scan and display the image while enlarging the portion of the image of interest. If you Zoom in on the HORIZONTAL line pairs (Alternate white and black lines) you will find that the alternate white and black horizontal lines do not display EQUAL levels of White and Black levels. This means that there is aliasing. NOT TRUE resolution.

I will try and find some more tutorial information to post this week. Keep asking questions but DO NOT jump to conclusions based on incorrect or incomplete information.

Regards,

Bill Hogan

---

>
>I find it interesting that on the HDCAM the sharpness was set to "0" yet >the Film was grain reduced and "Sharpened" Not a Fare comparison.

First, I think Sean means a "0" Detail when he refers to sharpness in the HDCAM. But even when Detail is at "0", there is ALWAYS Sharpness in that camera. Aperture (which is, in reality, pre Detail enhancement) is always ON in that (and most other current video/HD/24p/digital cameras). So much sharpness that it can actually be measured as aliased resolution at frequency response levels higher than the stated resolution of most CCD cameras. I find it interesting that in 9 1/2 years of doing side-by-side camera tests (shoot-outs for customer purchase evaluations) as an employee of Philips/BTS, I could NEVER get competitor camera manufacturer reps (camera demo engineers) to follow completely what were then the standard resolution testing guidelines for such testing : Gamma, Detail, Matrix, and Aperture OFF/LINEAR.

Everyone would turn off Gamma and Detail OFF or LINEAR (as appropriate)l, but, even when it was possible to turn it off, our camera would always be the only one with Aperture OFF. Nowadays, it is virtually impossible to turn Aperture off, as the function is no longer controlled by a physical switch or jumper as it was in analog and early digital cameras. It is now controlled in software at a level usually not enabled outside the manufacturing facility, where it is available only for sensor/processing optimisation.

This discussion thread is precisely the reason why that is and always has been so important, however, to understand why this is so relevant and yet we never truly measure the "unsharpened" output of any video/HD/24p camera in the modern era, and never will again. In every camera-shootout I have done, even when it was possible, aperture was ON in competitors cameras even when the client specified that it be OFF. And, though just to avoid litigation, one of the most respected network lab tests I was involved with, knowingly allowed the aperture to remain ON in its (customary) camera-of-choice while demanding that mine be OFF.

Just shows that somebody always gets what they want out of any test, but you also might never get a truly unbiased or even accurate test performed in any test. Ask any inmate in any prison; somebody's guilty in their cell block, it's just not them.

GEORGE C. PALMER
HDPIX, INC.
www.hdpix.com

---

>
>First, I think Sean means a "0" Detail when he refers to sharpness in >the HDCAM. But even when Detail is at "0", there is ALWAYS Sharpness >in that camera.

George is right...as anyone that has worked with the Sony HDCAM knows that when you turn detail "ON " and set the numbers to "0" when the scale is +99 to -99 then "0" is in the middle

"0" is to much for my taste but as Bill mentioned more is not always helpful.

And remember the Post about the F900/3 detail being turned 180s where +99 was the minimal amount of edge enhancement for black and Freq./3 has flipped this to -99 is minimal edge enhancement for black and Freq.

So when I set Detail??? I set the master detail to -69 (easy to remember) and I set Black Limiter to whatever will give me the smallest edge I set the Frequency to whatever will give me the smallest edge. And since no one will tell you how to set this you must test it for yourself.

This is where the testing comes in and dialling in the multiple parameters that affect the frequency response and detail.

The Smart Guys (you know who you are) know how to set this but they are often forced by those that don't know what they are doing to set differently.

And like Salt in the soup or Cayenne pepper in the Gumbo, "Detail" is often added to taste.

I have often mentioned to my students that when you shoot any test pay close attention to everything possible because you never know who will be looking at your test and what conclusions they will draw. As in this case with a dynamic range test being used for Detail comparison.

B. Sean Fairburn
LA HD DP

---

>
>(Set Detail) "ON " and set the numbers to "0" when the scale is +99 to ->99 then "0" is in the middle (even though) "0" is to much for my >taste...(so) I set the master detail to -69"

And that mitigates the Detail issue, but even with the Detail at "-99", even with Detail OFF in the menu, there is still Aperture created Sharpness and why it may be valid to allow grain reduction on film in these Sharpness comparisons. That is also the reason we never actually measure the true "spec" resolution of a video/HD/24p/digital camera, and why we can't actually compare one to the other because Aperture correction parameters are different and specific to each camera manufacturer, and even adjustable in every camera in factory level software adjustments. So Detail is not Sharpness. Sharpness is always there even when Detail is OFF.

GEORGE C. PALMER
HDPIX, INC.

---

>
Bear in mind when considering originating on S16mm that any potential broadcast will be seriously flawed. If digital projection is the only final use. then S16 - HD will look fine -- I have done many excellent transfers on the Spirit to D5.

The problem comes when HD is compressed for transmission for broadcast. Discovery first raised this when they prohibited S16 originated material on Discovery Theatre, much to the dismay of many film producers. They have put together a tape that shows the path from transfer, through HD master to the final 18 mps compressed output.

Using a pixel analysis program, they can isolate compression artefacts, and the proof is plain to see. There is a sea of moving pixels in non-movement areas of the frame. This is the same rationale, btw, for not allowing DV originated material. Upconverted SP and DBeta do not have these artefacts. Either the compression technology will have to improve or people will have to stick to HD origination if it is to be broadcast on cable or satellite.

Rod Paul

---

>
>Either the compression technology will have to improve or people will >have to stick to HD origination if it is to be broadcast on cable or >satellite.

---Which begs the question :

"What is the state of current compression technology today and who is at the forefront of this technology?"

Jeffery Haas
Freelance Editor, Camera Operator
Dallas, Texas

---

>
Rod Paul wrote :

>The problem comes when HD is compressed for transmission for >broadcast. ... There is a sea of moving pixels in non-movement areas of >the frame. ....Either the compression technology will have to improve or >people will have to stick to HD origination if it is to be broadcast on >cable or satellite.

There is theory (and tests) and there is reality. The reality is that Gilmore Girls and One Tree Hill are both aired on the WB network every week in HD, are both originated on S16, and exhibit none of these disastrous problems. So much for testing.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles

---

>
Jessica Gallant wrote :

>An imager with 1920 horizontal pixels will, at most, reproduce 960 >black/white transitions.

>wouldn't HDCAM be limited to 720 black/white transitions?

I wrote "imager". Once you start adding processing and compression all bets are off.

For example : What is the maximum attainable MTF in HDCAM when there's motion in the frame?

It's a constant data rate compressor, so, something must give.

It'd be interesting to see a test where you have a resolution pattern in the centre 1/15th of the screen (that's 128 pixels horizontally) and lots of motion everywhere else.

Martin Euredjian
eCinema Systems, Inc.

---

>
Martin writes :

>An imager with 1920 horizontal pixels will, at most, reproduce 960 >black/white transitions.

Please explain. I think I know, but me and assumptions don't have a great history together.

Brent Reynolds
DP / Film maker
Tampa, FL

---

>
>An imager with 1920 horizontal pixels will, at most, reproduce 960 >black/white transitions.

Also, note that it is more than likely that this isn't even possible due to optical pre-filtering. I probably should have said "under ideal laboratory conditions" rather than "at most".

Brent Reynolds wrote :

>An imager with 1920 horizontal pixels will, at most, reproduce 960 >black/white transitions.

> Please explain.

The imager (the chip) has 1920 pixels horizontally. If you devise a pattern and apparatus (notice I didn't say camera) that illuminates every other pixel with "white" while keeping adjacent pixels dark ... you have 960 black/white transitions across a line.

Martin Euredjian
eCinema Systems, Inc.

---

>
Martin Euredjian wrote :

>An imager with 1920 horizontal pixels will, at most, reproduce 960 >black/white transitions.

>Also, note that it is more than likely that this isn't even possible due to >optical pre-filtering.

What does optical pre-filtering accomplish? Do all 3-chip prism cameras do this? Even pre-digital format cameras like BVP-5/BVV-5 combo's?

How does 720P-Varicam compare to HDCam F900/3's projected off of filmout prints and direct digital projection? I'm just curious.

Tom McDonnell
DP/Operator
New Orleans, La

---

>
Rod Paul writes :

>Using a pixel analysis program, they can isolate compression artefacts, >and the proof is plain to see. There is a sea of moving pixels in non->movement areas of the frame.

So the compression schemes are now dictating origination format.

That is akin to being ordered to use a lens with poor resolution, because it looks like it has more depth of field.

Steven Gladstone
Cinematographer
Gladstone Films
Brooklyn, N.Y. U.S.A.
East Coast List administrator - Cinematography Mailing list

---

>
Michael Most wrote :

>There is theory (and tests) and there is reality. The reality is that Gilmore >Girls and One Tree Hill are both aired on the WB network every week in >HD, are both originated on S16, and exhibit none of these disastrous >problems. So much for testing.

Good point.

On watching HBO last night via digital cable I noticed that the usual awful digital artefacts (really bad posterization in the blacks, dissolves that looked nothing like dissolves, etc.) were really noticeable on The Sopranos, but on Curb Your Enthusiasm the only artifact that was disturbing was some jaggies. Wonder if the lower inherent res of the video-originated Curb is the reason?

Digital cable is really awful.

Jeff "HBO's new offices are in Artifact City" Kreines

---

>
Rod Paul writes :

>Either the compression technology will have to improve or people will >have to stick to HD origination if it is to be broadcast

I find the whole Discovery issue one of frighteningly flawed thinking.

The approach seems to be, we have a crap compression system so you have to give us pictures with as little information in them as possible so we can cope.

Call me old fashioned but I thought the idea was to transmit the best pictures you could, not drag everything down to the lowest possible common denominator.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
Reynolds, Brent wrote :

>Please explain. I think I know, but me and assumptions don't have a >great history together.

If you are photographing a chart with alternating black and white lines, and position that chart so each line is one pixel wide, you'd be able to resolve 960 black lines against white. No more than that, as the alternate pixel is required for white.

Note that, in reality, as others have said, this is a theoretical number, as many cameras use optical anti-aliasing filters (which blur the image slightly) etc.

Jeff "prefers real test subjects to charts" Kreines

---

>
Martin Euredjian wrote :

>It'd be interesting to see a test where you have a resolution pattern in >the centre 1/15th of the screen (that's 128 pixels horizontally) and lots of >motion everywhere else.

Yes, that would be interesting!

Jeff "would pay-per-view for that one" Kreines

---

>
>I wrote "imager". Once you start adding processing and compression >all bets are off. For example: What is the maximum attainable MTF in >HDCAM when there's motion in the frame? It's a constant data rate >compressor, so, something must give.

Thank you very much for the information!

Jessica Gallant
Los Angeles based Director of Photography
West Coast Systems Administrator, Cinematography Mailing List

---

>
Steven Gladstone wrote:

>So the compression schemes are now dictating origination format.

Yes, the Discovery Channel and BBC instance being a good (!) case in point.

Jeff "always uncompressed" Kreines

---

>
>Either the compression technology will have to improve or people will >have to stick to HD origination if it is to be broadcast

There are folks who lurk on this and other email forums for who this could be a mantra. In the rush to adopt formats and compression standards, first the ATSC, then SMPTE accepted MPEG-2 as the "official" (fixed) transmission compression standard. Unfortunately, MPEG-2 also became the core of many other compression stages in the HD acquisition process. MPEG-2 is, at best mediocre at efficiently compressing the many elements and combinations of picture content and picture motion, since it has very little, if any, active adaptivity. Coupled with the additional distribution channel selected satellite compression system, this can be very destructive to the received image, from the point of view of end-to-end artifact production. As a result, extreme combinations of complex picture content (i.e. high contrast, the sheer number of visual elements, and high motion rates) can "break" the compression algorithms in use all to easily. Hence the "swarms of moving pixels" (not really pixels, but actually artifacted picture elements) in some received satellite signals.

There are theoretical and practical solutions all of which are related to the application of "intelligent" layered compression techniques which have been proposed over and over at SMPTE working group meetings for inclusion into or substitution of the MPEG-2 or the (amended) MPEG-4 family of compression standards. Gary Demos, who sometimes lurks here, knows this subject intimately and his comments (to the extent he can politically or legally comment) might be very enlightening. So the only way for the compression technology to improve is to convince SMPTE, our FCC, and our satellite program originating companies to first include a more advanced, intelligent, adaptive compression system in their official compression tables, then to convince the end users (by regulatory coercion if necessary) to implement said improved compression methodology. A bit counter-intuitive to the free market system, but maybe somebody has to actually stand up and be the "father" for a change.

GEORGE C. PALMER
HDPIX, INC.

---

>
> Jeff "prefers real test subjects to charts" Kreines.

When I was in the hi-fi biz back in the '70's - it was widely accepted that solid-state amps were better than the old tube amps they replaced.

The solid-state amps measured much better in distortion tests.

Then some audiophiles pulled their old amps out of the garage and hooked them and for the hell of it - and lo and behold - they sounded BETTER, despite the fact they had much higher "distortion".

Well, the distortion they were measuring - THD (total harmonic distortion), and IMD (intermodulation distortion) had thresholds to human hearing around .5-1.0 percent. Anything below that could not be heard. But solid state amps could be made to have .05% of these two distortions, and even lower - by adding feedback. But, if one actually designed an amp to have low feedback, it sounded better - it clearly sounded better. Yet it had higher distortion.

Well, there are probably hundreds of kinds of audible distortions, we just don't what they are or how to measure them. This feedback concept actually gave rise to a new kind of distortion - call TID (transient intermodulation distortion). And it encouraged salesmen to rub two pieces of paper together and says "That's what specs sound like".

Now, testing HD vs anything is interesting - because you can actually SEE it. But seeing it on test charts is helpful, but one could (and should) safely argue that movies about test charts don't do well. It would be better to have a series of stock items that could be shot, and standardised lighting set-ups (Some Dare Call It Science!) - and shoot them. Do the same with motion.

If it was standardized -( like a test chart) - we could see real-life examples - and post them. The test - if conducted in Geoff's studio in England, or in Timbuktu - would be identical.

It would be nice to have real live human being in these tests - but impossible.

Dale Launer
writer.filmmaker
Santa Monica

---

>
Dale Launer writes:

>It would be nice to have real live human being in these tests - but >impossible.

I see a market for cloned test subjects. Kind of a Sony clony of your owny.

Brian "Sorry, it's been a tough Monday" Heller

---

>
Jeff writes :

>you'd be able to resolve 960 black lines against white.

Thanks, that's what I thought we were talking about, but wasn't sure, though in retrospect, it seems pretty obvious now. I have been trying to keep up with the posts while working and missed something. So, with that in mind, let me ask another obvious question in regard to the following:

>as many cameras use optical anti-aliasing filters (which blur the image >slightly) etc.

From what I remember of aliasing from my simulation field engineering days - anti-aliasing helps transition from color to color by choosing intermediate values for pixels that are where colors converge. This is in a software sense and I'm sure an oversimplification.

Let's say I am in the ballpark - I then am guessing that the optical anti-aliasing filters reside on the chip and are over each cell (pixel) which diffuses light thereby achieving what
I was referred to above - yes?

Comment and correct

Brent Reynolds
DP / film maker
Tampa FL

---

>
>I see a market for cloned test subjects. Kind of a Sony clony of your >owny.

Well, I was thinking more along the lines of getting Cindy Crawford to offer up her genetic map - it certainly makes an attractive argument for human cloning.

Dale Launer
Writer / Filmmaker
Santa Monica

---

>
Brent :

Not exactly; in prism based focal planes (HD Cameras) the anti-aliasing filter is usually a single filter or coating applied somewhere (depending on the optical block and camera manufacturer) prior to or even on the (light) entrance port to the prism. It usually consists of a filter with appropriately scaled and numbered horizontal and, sometimes, vertical etchings which provide enough pixel level edge diffusion to prevent too much high frequency picture information to the sensors on the optical block (prism). Combined with a very small Green CCD sensor physical offset (from an exact "bore-sighted" registered overlay) this minimizes aliasing (chromatic and monochrome resolution aliasing) in a multi sensor array, and also, does introduce an element of limiting resolution.

GEORGE C. PALMER
HDPIX, INC.

---

>
I would like to ask a question about all of this video vs. film.

Is it not the definition of a cinematographer or Dp to use what ever medium to create the final look?

It used to be in the days of old... 16, 35, or 70mm

Now that video has become close to 16. I'm not saying that it is yet, but obviously many people think so, that the evolution of this has become a reality and many are doing so.

OK, lets live with this because it has happened and will only become more of a debate, but the reality is, film is still better in many ways, the other reality is cost factor! that's where everyone is having a problem. You can shoot video cheaper and you do not have to be a Dp or cinematographer (as a background) but a camera person, if even that!. Do not get me wrong there are many videographers that are great DoP's but there are many videographers that cannot light, use lenses, depth of field, depth of focus, pick the proper stocks, know how to transfer (video or film) that make the debate more difficult! I see that this will never change and will go on forever, but the fact is, If you are a professional Dp /cinematographer you will be able to use any medium available. One of the problems here is ... video is not film.... YET, Yes in time it will become the same I'm sure. Just because of the technology advances. we can put a man on the moon and not make a video camera that looks like film. It will happen and we will all have to live with this and adapt with the change!!!!

So again the question is are you a professional Dp / cinematographer?

Can you use the medium requested to create a piece of art?

I was once told by an older Dp that a professional cinematographer can re-create any scene anywhere by knowledge of lighting, proper use of lenses, depth of field, color temp, knowledge of stocks, timing or film transfer. I'm sure I'm missing something but what I'm getting at is the director sees a scene and he wants that in his film.A professional Dp will be able to re-create that same scene with his or her experience and knowledge pretty much the same way. Lets not get off the track and say some great shot where the sun is going down in Alaska, I'm talking about a studio setting or location that any pro could re-create.

I think we are all here for the same reason... to create the best we can in a visual medium.

So what is the problem?

Many of us older DoP's and possibly younger DoP's have the experience of film. It's definitely harder to use and you must know what you are doing or you will end up with crap! in video you get what you see, so if you have 100 units on the scope it looks great! not necessarily art or even close to a film look but it does look like 100 units. Since video is so much cheaper and every one has a camera and can edit system at home, they all have a way to make a movie or commercial. as a director or producer its great because you will learn from this. I will say as a Dp you will probably still learn from this but anyone wanting to become a pro Dp should learn to shoot and light film style. It is an ART and it is somewhat falling to the way side of film making.

But my point is still...

As a professional Dp / cinematographer should you be able to use the medium requested?

Kevin Cable
dir/Dp
South Florida

---

>
I just got a note from Gary Demos; he says he'd be glad to give us a few paragraphs with his take on compression.
Anybody interested?

GEORGE C. PALMER
HDPIX, INC.

---

>
>I just got a note from Gary Demos; he says he'd be glad to give us a few >paragraphs with his take on compression. Anybody interested?

Gary's opinion would be welcome and valued. His knowledge is vast and his experience is deep.

Dave Stump ASC
Vfx Sup/DP
on location
Nassau Bahamas

---

>
George C. Palmer wrote :

>I just got a note from Gary Demos; he says he'd be glad to give us a few >paragraphs with his take on compression. Anybody interested

Go for it.

Mark Smith

---

>
>I just got a note from Gary Demos; he says he'd be glad to give us a few >paragraphs with his take on compression. Anybody interested

Sure. And while we're waiting, here's mine :

For any given data rate, there are formats (e.g., 1920x1080 24P 4:2:2) that can be supported "uncompressed", and higher formats (more pixels, more color, or a combination of both) that can be supported with "compression". The question will always be whether or not the uncompressed lower format "is/looks" better than the higher format after compression. The answer will likely turn on the state-of-the-art of the compression algorithm at that point in time, but the basic uncompressed-lower format/compressed-higher format trade-off comparison won't change for quite a while.

For those who are really interested in answers, it is probably best to discuss/analyse formats and compression separately. But then it might not be so interesting.

Noel Sterrett
Baytech Cinema
www.baytechcinema.com

---

>
Geoff Boyle writes :

>Call me old fashioned but I thought the idea was to transmit the best >pictures you could, not drag everything down to the lowest possible >common denominator.

Geoff Boyle, you are an antiquated old fossil. The idea is to make as much money in as short a period of time as possible. This constant obsessing about image quality is getting to be a real distraction.

Brian Heller
IA 600 DP

---

>
Noel :

I agree. The subject of formats and compression should be discussed separately. You won't find me mixing them unless they are found together, as in videotape recording or other such application in which the two should never have been mixed to start with. That's why we are all so excited about uncompressed recording processes such as yours.

GEORGE C. PALMER
HDPIX, INC.

---

>
Noel Sterrett wrote :

>For any given data rate, there are formats (e.g., 1920x1080 24P 4:2:2) >that can be supported "uncompressed", and higher formats (more >pixels, more color, or a combination of both) that can be supported with >"compression".

But there are ways of recording higher data rates, so formats with greater resolution and color bit depth need not be compressed. Yes, it takes greater bandwidth and more color space, but sometimes people will gladly trade bandwidth and storage space for increased resolution or bit depth.

For example, one could use four CineRAMs together to record a 4K 4:4:4 signal, if the camera provided the proper split outputs.

It might not be cheap, but it's certainly possible.

Jeff "had dinner with Noel at IBC" Kreines

---

>
The following is Part 1 of some comments from Gary Demos on compression...

George,

Feel free to pass along the following comments on compression to the CML mailing list :

Certainly MPEG-2 compression has become very widely used in delivering standard-definition and high-definition digital video. MPEG-2 was completed around 1994, and is thus about ten years old. Even at the time, I gave demonstrations at SMPTE and elsewhere that MPEG-2 could be substantially improved. I demonstrated this improvement across the entire range of compression ratios, from very low ratios, and very high quality, to very high ratios and marginal quality. I even gave demonstrations to the FCC chairman and to some of the FCC commissioners. However, in early 1996 the FCC ratified the ACATS US DTV standard for over-the-air broadcast. A couple of years later, DVD and DirecTV adopted MPEG-2, as did digital cable.

The original MPEG-2 momentum was therefore strong enough to carry it into the dominant position for delivering digital video. However, with each year that followed the original creation of MPEG-2 in 1994, a broader number of people began to be aware of the weaknesses inherent in MPEG-2. A counter-swell campaign of MPEG-2 supporters promoted the idea the MPEG-2 improvements could be made entirely in the encoding system that would substantially repair many of the inherent MPEG-2 flaws. I argued that such encoder-only improvements would have limited benefit, and I believe that it is now widely recognized that the majority of flaws within MPEG-2 are fundamental to the MPEG-2 coding standard (thus cannot be repaired by encoder improvements), which was always my position (although some significant improvements were implemented in encoders). I also believe that the value of layered coding (both spatial and temporal) has been more broadly recognized.

Now, here we are in 2004, and the question has finally been raised about what should be done with MPEG-2, now that it has been widely recognized as being flawed (mainly due to its having been very early in the evolution of compression technology). I believe that the most critical issue to consider is whether any new candidate coding technology can provide for high quality, high compression ratio, or span the entire range. Many new coding systems, such as MPEG-AVC/H264, were designed using sub-500kbps bit rates, and thus were honed on high compression ratio as their primary goal. Those who argue that such coding systems can also provide high quality are mostly speculating, in my opinion, since the digital processing within the engines of these codecs do not contain appropriate design features for high quality.

At the other end of the spectrum, many high quality coding systems do not provide very much compression. Further, such high quality systems, such as HD-CAM and DVC-PRO-100HD, often reduce the horizontal luminance and chroma information, thereby limiting the image resolution to below the source resolution (and below 1920x1080 or 1280x720). Many of these high quality systems, including these two examples, do not use "inter-frame" (between frame motion vectors) coding. Thus the frames are coded as "intra-frames" (each frame stand alone). This limits the potential compression ratio, but simplifies the compression system by eliminated the processing needed to match regions within different frames.

I believe that any next-generation compression coding system should aim to span the range of compression and quality. I believe I have demonstrated that this is feasible, and I continue to find this to be possible with continued improvements. However, high quality is not possible unless the underlying coding system is inherently designed, in all its aspects, for high quality. Previous extensions, such as increasing the bit rate of a profile of MPEG-2 and allowing the underlying 8-bit coding system to be extended to 10-bits, do not yield a high quality coding system, but rather result in mediocre performance. I believe that such an extension is under consideration for MPEG-AVC/H264, but I would expect similar weak performance at the high quality end...

End of Part 1.

Part 2 of comments from Gary Demos on compression...

Further, I believe that the correct way to evaluate the performance of a compression system is a central issue. I was intrigued by T.I.'s DLP demonstrations of film butterflied with the DLP projection, since the image could be compared in motion at the seam. This butterfly technique has been quite helpful in scrutinizing image compression quality in motion, and is much more revealing that "A after B" or "A above B" comparisons. I have long felt that SNR and PSNR are inadequate for evaluating coding quality. There are many widely known examples of better compression at lower PSNR and visa-versa. I believe that a much more thorough instrumentation of compression quality is possible, and I have been working on the outline of how this can work. The basic ingredients are the use of difference histograms for each of red green and blue (emphasizing outliers) and the partitioning of such histograms and signal-error measurements into brightness bands to more closely match source noise characteristics, and perceptual distinction characteristics.

Twenty years ago, using Kodachrome-25 film on my Digital Film Printer system, I discovered that wide-dynamic range image presentation (such as motion picture theatres) needed at least 11-bits to be imperceptibly coded. This two-decade-old result has recently be reverified by Barten (in the late 1990's), and this past year by Tom Maier of Kodak working together with the USC Entertainment Technology Center and the Digital Cinema Initiative, as well as additionally by Charles Fenimore of NIST. Tom Maier and Charles Fenimore presented their results at the Nov 2003 SMPTE conference in NYC.

Although home presentation does not yet rival movie theatres in dynamic range, the goal of coding technology should be to be able to achieve better than an 11-bit noise floor (greater than approx. 70dbPSNR by the existing flawed measurement standards). If such can be achieved, then the fundamental coding technology has a sound underlying engine. Further, with the right codec architecture, such a codec should be able to be set to high compression ratios, at quality rivalling any other coding technology at all such ratios.

That's my view of where compression should be heading.

Gary Demos

GEORGE C. PALMER
HDPIX, INC.

---

>
>If it was standardized -( like a test chart) - we could see real-life >examples - and post them. The test - if conducted in Geoff's studio in >England, or in Timbuktu - would be identical.

That was exactly what I tried to do with my original film out tests.

To do tests that could be duplicated anywhere, and challenged if that's how people felt.

It's vital that we have some form of comparison and that this comparison should be transparent.

I used various levels of exposure of MacBeth charts.

I also asked people to send me their files if they disagreed, as some did, with my results.

To date we have seen NO other files..........

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
>I just got a note from Gary Demos; he says he'd be glad to give us a few >paragraphs with his take on compression. Anybody interested?

Yes.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

>
That was exactly what I tried to do with my original film out tests.

To do tests that could be duplicated anywhere, and challenged if that's how people felt.

It's vital that we have some form of comparison and that this comparison should be transparent.

I used various levels of exposure of MacBeth charts.

I also asked people to send me their files if they disagreed, as some did, with my results.

To date we have seen NO other files..........

Apparently everyone wants YOU to do it.

Perhaps we should start a thread of easily obtainable items (like a can of Coca Cola, and...frozen kidney pie - no just kidding) and ground rules for the motion capture.

Dale Launer
Writer/Filmmaker
Santa Monica

---

>
Tom McDonnell wrote :

>What does optical pre-filtering accomplish? Do all 3-chip prism >cameras do this? Even pre-digital format cameras like BVP-5/BVV-5 >combo's? ...

Yes, all CCD-Cameras including old BVP-5 using optical low pass filter in front to reduce aliasing. But resolution and characteristic of this filters are in accordance with the amount of horizontal an vertical pixels on CCD-surface. So filters of SD-Cameras has to be different to filters of HD-Cameras.

Joerg Friedrich
Camera-/Camcorder-Training
srt, Germany
www.25psf.de

---

>
Brian Heller writes :

>The idea is to make as much money in as short a period of time as >possible.

This is the sad truth. I have been lucky enough to have had the occasion to pitch film ideas on several occasions, and here's the breakdown of those conversations:

10% Concept and story...
90% What's it gonna cost, who's name is attached, how much money will it make and when?

In only one of the pitches did format come up in the sense of quality as opposed to strictly cost.

But, you guys already knew this...

Brent Reynolds
DP / Film maker
Tampa FL

---

>
>Discovery Channel Engineering dictum: There is a sea of moving pixels >in non-movement areas of the frame (on S-16 originated material). This >is the same rationale, btw, for not allowing DV originated material

Forgive me if I am missing something, but why would compression artefacts in the uplinked image, be more evident in non-changing frame areas when the HDCam (or D5) master is taken from the S-16 transfer instead of from a HD camera tape?

Peter Pilafian
DP, LA and Wyoming
www.hpix.com

---

>
Peter Pilafian wrote :

>Forgive me if I am missing something, but why would compression >artefacts in the uplinked image, be more evident in non-changing frame >areas when the HDCam (or D5) master is taken from the S-16 transfer >instead of from a HD camera tape?

I think the deal boils down to the fact that since a film (filum) image is made up of a random sea of ever changing pixels, which causes the lame assed compression scheme they use for delivery to choke and actually degrades the image further.

HD, OTOH, since it is comprised of a horizontal x vertical pixel array which is fixed, presents less problem to the compression algorithm because there is no frame to frame movement caused by the moving sea of random grain structure.

Or something like that...

Mark Smith

---

>
Geoff,

The Resolution chart was shot today.

F-900/3 Zeiss DigiPrime Lens 28 mm

Detail Off

Detail On at -99

Detail On at 0

Detail On at +99

Full Frame 23.98 HDCAM shooting DSC Labs Chart
(Junior)

Michael Bravin will be doing the Still Frame Grabs then cutting out the resolution fans to give them to you to Post.

Thanks for the opportunity to provide another point of reference.

B. Sean Fairburn
Director of Photography
Role Model Productions LLC
Castaic Ca 91384

---

>
B.Sean Fairburn

>Thanks for the opportunity to provide another point of reference.

Thank you for doing them Sean.

Anyone fancy doing the same with Vision 2 100 ISO of the same chart and sending me the frame grabs.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based

---

Copyright © CML. All rights reserved.