Cinematography Mailing List - CML

Progressive Imaging

First, thanks to Michael Bravin for his gracious offering. As a result of our off-line conversation, I agreed to attempt a non invasive, clearer delineation of my thinking on progressive imaging, so here it is. Because of the liberal mixing of imaging, transport, and recording terminology here on the CML which has muddied the imaging only analytical issues, I have avoided comment, but I have observed over much time that not all progressive is the same even just at the imaging stage, and that is, frankly, the only area my postings primarily refer to. Since my observations have been made over the past 5-7 years through the genesis and implementation of the Digital Cinema/HD/24P era, and predate the notion, so often suggested herein, that they are based on my qualitative (some accuse commercial) biases, I hope no one is offended by any such reader perceived, but here goes.

I have been viewing progressive images in the process of comprehensive visual and lab testing of multiple formats and frame rates for quite a while, well before the advent of 24Psf. First, one of the challenges with shooting at frame rates other than 60Hz rates has been and is monitoring. Even now most monitors either require a 60Hz input or one that has had a 3:2 pulldown or a 1.5:1 in the case of 48Hz modes. So when viewed on a monitor, 24 and 30 frame progressive imaged formats almost always exhibit field/frame latency associated with pulldown cadences that cause a phenomenon we call "judder" or "flicker". Judder, well understood and employed by filmmakers, and misunderstood and rejected by broadcasters, probably killed the 480P30 (LDK 2000) cameras from Philips, as the resultant 2:1 pulldown (created by cloning each frame in a DSP frame buffer) was used as a 60Hz interlace emulation for interlace system interoperability, but the judder or latency drove broadcasters crazy whenever horizontal motion was present, even though the frames created were truly progressive from a Frame Transfer sensor and could, in fact be "deinterlaced" to produce a 30Hz (frame) 480 Progressive image with no dicemination artifacting; in other words, a truly progressive image. In the case of the 480P30 the quasi interlaced frame was actually composed of a frame "interlaced" with a clone of that frame, created in a frame buffer, so the resultant progressive frame would appear to be interlaced for purposes of interlace system equipment recognition and use with a notable vertical resolution increase obvious to the naked eye. Because the 24Psf concept emerged so closely on the heels of the frame paired 480P30 experiment, some of us believe it was that specific 480P30 progressive implementation, known as paired frame progressive, that germinated the 24 psf concept.

Cinematographers have long dealt with judder by controlling the speed and nature of horizontal motion in their images, but those new to film and HD notice it, evidenced by the many, many postings on the CML questioning its causes and solutions. It is judder itself that led me to question whether psf is actually progressive imaging or, possibly, field paired progressive due to the heretofore fielding only nature of the FIT sensor. From the beginning, if compared to 24 frame film (obviously viewed after processing with 3:2 pulldown added) or other 24P imaging devices, the judder characteristic of 24PsF has been a little more quirky (non-linear) and exaggerated, especially at higher panning rates.

This has been a source of questioning here on CML quite often.

No one has presented an answer, and I am not either, except to suggest that, even at 1/48 (180 degree) electronic shutter, 24PsF displays a different, more noticeable judder component. This has led me to believe, since judder is a direct function of image latency, that that latency could be and may be the result of the use of a frame buffer in camera, suggesting the possibility that the FIT sensor might actually be operating in a fielding mode with a field buffer to produce a 2 field coherent progressive frame...

(...continued in Progressive, Part 2…)


(...Continued from Progressive, Part 1...)

It has occurred to me that if, in the 24 Psf imaging process, there was, first, a 2:1 buffering process for conversion from 24 field/frame image to 48 to which was added a 1.5:1 pulldown for 60Hz display enablement, all of that could explain the quirky judder characteristic of the 24Psf cameras. Would that be interlacing? Not in the sense that such a "composite" progressive frame would necessarily contain the temporal discontinuities of a fielded, interlaced frame, but it would be very much akin in nature to the frame paired progressive framing employed in the earlier LDK 2000. Whether the Sony FIT sensor is an actual progressive or a fielding sensor with buffering, I will not suggest that I know for sure, and even if it is, it is, perhaps, only a minor engineering discussion; only Sony knows for sure anyway. But the visual evidence leads in that direction. Would that buffering be a bad thing? Maybe not, but both Panasonic and Thomson do produce 24P cameras with a genuine progressive output that has virtually the same visual pulldown artifacting as 3:2 film for television, without that (possible) imaging frame buffer. And for HD/video release it does create some interesting and unique judder exaggerations in scenes with higher panning rates. But if it is true, I wonder why, except for the transport/recording justification of HDCam recorders, it was done, without comment or explanation from the manufacturer since it is obviously possible to do otherwise based on observation of 24P cameras from the big three. If not I continue to seek the true source of the quirky judder.

All of that relates, of course, to HD for video and television release only, and, of course, at film out, judder should be nonexistent, so a non issue for film applications, leaving only qualitative and marketing issues to differentiate between the options and determine production methodology in that latter medium.

HD and Digital Imaging Services

>All of that relates, of course, to HD for video and television release only, >and, of course, at film out, judder should be nonexistent

If, as you say, judder is nonexistent at film out (and therefore also nonexistent in the computer data frames used to create the film out) then the camera (and the data path to film out) should be eliminated as the source of the judder.

On the other hand, the video viewing system, which at least in the case of Psf relies on interlaced rather than progressive scanning, has a number of variables (including phosphors) which could cause problems. Another possible cause could be the 3:2 pulldown or other circuitry used in the output stages.

Noel Sterrett
Baytech Cinema

I haven't seen a Psf film out, so can't judge whether there is any latency in it.

HD and Digital Imaging Services

George Palmer wrote :

>I haven't seen a Psf film out, so can't judge whether there is any latency >in it.

I haven't either, but if it's going to film out why wouldn't there be judder? It's 24fps. What am I missing?

Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614

>I haven't either, but if it's going to film out why wouldn't there be judder? >It's 24fps. What am I missing?

Wade :

Yes, I agree that there should be judder as even with improved projection, any 24fps media should exhibit judder, I just haven't seen a 24Psf film out to be able to judge its judder characteristic compared to 24fps film or other 24fps HD filmed out. My original comment that judder is not an issue on film out is relative to the accepted norms for judder in 24fps media when shot with the proper framing and panning constraints. Whether Psf contributes to exaggerated judder in that mode, I simply have not observed.

HD and Digital Imaging Services

George C. Palmer wrote :

>I haven't seen a Psf film out, so can't judge whether there is any latency >in it.

Go See "Once upon a Time in Mexico".

Sean B. Fairburn

© copyright CML all rights reserved