Frame Movie Mode Question. When to use or not.

[Nathan Gifford]

After re-reading my manual on the XL-1, the manual suggests that frame movie mode increases resolution. Ok....

So my question is when does the group here use frame mode and when do they avoid it?

Now when I shoot in frame mode, I try to remember NOT to make fast pans. Does anyone have a tip for that?

What about shutter speeds...Other than controlling depth-of-field in bright conditions, is there an advantage to using other shutter speeds while in frame mode?

When shooting for the web frame mode seems to have some definite advantages, especially when there is lots of motion. Is there a disadvantage using frame mode when the target is either TV or cable?

Thanks,
Nathan Gifford

P.S. Chris this should be a good thread.

[Stephen van Vuuren]

When I first got my XL-1 two years ago, I used it all the time (since I had shot Super 8 and 16 before). But now I've learned how to do it in post without the resolution loss plus controlling what is frame and what is not.

I wish frame mode was true full frame 30 or 24 progressive mode, but alas, no.

Because of in frame mode one frame has two equal fields, you'll never worry about slow/fast motion flickering.

[Adam Wakely]

I do alot of wedding videos and I choose the frame mode for my photo shoots and reception dancing/party footages. It gives the visual of 'none reality'. I use the regular 'video look' for the ceremony and speeches. I wouldn't want to watch the news on tv in a 'film look' nor would I want to watch a movie in a 'video look'. Well that's what I'm used to to and alot of people I think.
The XL1's frame mode works well but you do have to pan a bit slower! I like using a small depth a field while in frame mode to give that real 'film' look as well.

Adam....

[Stephen van Vuuren]

Also, the Canon manual is misleading. Frame mode does not increase vertical resolution, it reduces it. Canon states that it increase vertical resolution by 1.5. What they actually mean is that resolution is 360 lines in frame mode vs the 240 lines captured by a field.

However, each frame is made of two field i.e. 480 line. So an interlaced frame is 480 (with interlace effects) and frame mode is 360. This is most noticable in still shots. Regular interlaced mode is noticably (to my eye) sharper.

That's why I started using it in post especially since I like to shoot 16:9 which invovled some resolution loss, though less than cropping 4:3 (see www.adamwilt.com) for more info on that issue.

[Michael Pappas]

Hello! What process is it that you use to emulate the frame mode?

Thanks

Michael Pappas

<<<-- Originally posted by stephenvv : Also, the Canon manual is misleading. Frame mode does not increase vertical resolution, it reduces it. Canon states that it increase vertical resolution by 1.5. What they actually mean is that resolution is 360 lines in frame mode vs the 240 lines captured by a field.

However, each frame is made of two field i.e. 480 line. So an interlaced frame is 480 (with interlace effects) and frame mode is 360. This is most noticable in still shots. Regular interlaced mode is noticably (to my eye) sharper.

That's why I started using it in post especially since I like to shoot 16:9 which invovled some resolution loss, though less than cropping 4:3 (see www.adamwilt.com) for more info on that issue. -->>>

[Stephen van Vuuren]

My frame mode recipe can be done in Premiere or After Effects (though results seem better and easier to work with in After Effects). Note that the basis of this is the method I got from the makers of the DV movie (The Last Broadcast) that they used for the final sequence in that movie.

Here's is my complete workflow including handling 16:9 with a Martox RT2000 system - adjust for your system as necessary:

1) Import 16:9 footage into 16:9 project in Premiere. Edit to taste. Export as is (16:9, 29.97 interlaced) i.e. no processing or rendering.

(2) Work with in widescreen square pixel comp (864X486). Import DV file from step one, interpret as widescreen DV pixels, separate lower field (Matrox Codec setting).

(3) I use RevisionFX Deinterlacer plug-in - great results and flexibility first using upper field, then using lower field. Then I make a new widescreen comp and put the upper deinterlaced comp at 50% over the lower at 100%. Gives a wonderful 30FPS progressive scan look without any resolution loss apparent to my eye or strange motion artifacts.

(4) I do color correction, FX etc. on the deinterlaced comp and then render out NTSC D1 (720X486) with a HQ stretch to 720X480, cropping 3 top and 3 bottom pixels using the Matrox DV RT codec, field render off as Matrox codec takes care of this. This results in a 720X480 file still 16:9 but all the AE square pixel stuff is right proportions etc.

If I'm outputting a DV 16:9 file or tape, I'm done. If I want 4:3 with letterbox, then I:

(5) Then I bring this file into a 720X480 4:3 Premiere Project. I do NOT use the "maintain aspect ratio" since it want to rerender the entire clip. What I do is use the Matrox RT2000 PIP at change setting to squeeze clip by 75% for letterbox. Since this is done real-time by CC Cube hardware chip, quality is to my eye perfect. And then I just export this to Matrox RT DV AVI again (no rendering required) but as a 720X480 4:3 DV file, complete with letterbox.

(6) This 4:3 DV AVI can be used for MPEG-2 or output straight to tape and quality is excellent. I generally do not deinterlaced lots of complex AE graphics as artifacts can merge.

[Michael Pappas]

Thanks! I have a similar setup for AE. But this appears to be better. What is the average render time for 1 minute of DV?

Michael



<<<-- Originally posted by stephenvv : My frame mode recipe can be done in Premiere or After Effects (though results seem better and easier to work with in After Effects). Note that the basis of this is the method I got from the makers of the DV movie (The Last Broadcast) that they used for the final sequence in that movie.

Here's is my complete workflow including handling 16:9 with a Martox RT2000 system - adjust for your system as necessary:

1) Import 16:9 footage into 16:9 project in Premiere. Edit to taste. Export as is (16:9, 29.97 interlaced) i.e. no processing or rendering.

(2) Work with in widescreen square pixel comp (864X486). Import DV file from step one, interpret as widescreen DV pixels, separate lower field (Matrox Codec setting).

(3) I use RevisionFX Deinterlacer plug-in - great results and flexibility first using upper field, then using lower field. Then I make a new widescreen comp and put the upper deinterlaced comp at 50% over the lower at 100%. Gives a wonderful 30FPS progressive scan look without any resolution loss apparent to my eye or strange motion artifacts.

(4) I do color correction, FX etc. on the deinterlaced comp and then render out NTSC D1 (720X486) with a HQ stretch to 720X480, cropping 3 top and 3 bottom pixels using the Matrox DV RT codec, field render off as Matrox codec takes care of this. This results in a 720X480 file still 16:9 but all the AE square pixel stuff is right proportions etc.

If I'm outputting a DV 16:9 file or tape, I'm done. If I want 4:3 with letterbox, then I:

(5) Then I bring this file into a 720X480 4:3 Premiere Project. I do NOT use the "maintain aspect ratio" since it want to rerender the entire clip. What I do is use the Matrox RT2000 PIP at change setting to squeeze clip by 75% for letterbox. Since this is done real-time by CC Cube hardware chip, quality is to my eye perfect. And then I just export this to Matrox RT DV AVI again (no rendering required) but as a 720X480 4:3 DV file, complete with letterbox.

(6) This 4:3 DV AVI can be used for MPEG-2 or output straight to tape and quality is excellent. I generally do not deinterlaced lots of complex AE graphics as artifacts can merge. -->>>

[Stephen van Vuuren]

21 minutes for one minute of DV for frame mode effect. Add full color correction, gamma, grain, other effects shoot it up to about 45 minutes.

But on my system, I can have two of these renders going at the same time in multiple instances of AE and still work in other apps (except Premiere if I want RT effects). However, it slow renders by about 50%.

But, in half an afternoon, I can render a full 30-minute project with color correction etc. Overnight will do a full feature length project.

Yeah... Fast pans, as well as dolly shots with a lot of verticle shapes (especially hot or white areas) in the shot will cause a nasty flicker in FMM.

For pans I developed a technique to more-or-less eliminate this.

Instead of panning along a straight horizontal, try panning on a VERY slight bell-shaped arc. Not too drastic an arc. Just enough to reduce the flicker on those verticle areas (sides of houses, wall corners, etc).

Chef

stephenvv writes that the frame movie mode DOES NOT in
fact increase vertical resolution. Actually, he says, the resolution is something like 360 lines (not even Standard Definition). Is this true?

How would that work in PAL? First of all, do I actually get a sort of 25p effect in frame movie mode (in PAL)? Second, what would the actual vertical resolution be (in PAL) in frame movie mode?

Thanks for your help!

[Rob Lohman]

Hi everyone...

I'd like to join the discussion..... I've read a couple of things that I find strange:

1. Why would the vertical resolution change at all?
In my opinion vertical resolution should not change
at all when you shoot in frame or interlaced mode....
The difference is that they store the information
differently.... While there seems to be a mixed opinion
as to whether the XL1(s) shoots in true progessive
or not (i've read articles that state true progressive
and others that say the XL1(s) combines fields to
get progressive) it should not matter for resolution...
this would be a very weird thing, since the amount of
information is *exactly* the same!

2. In your describtion of how to go from interlaced to
progressive in your NLE introduces some information
loss stephenvv, because you render back to a DV file....
Unless you go straight to MPEG2 you do not introduce
this (which is possible).... but if you simply cut and go
back to DV your introducing a second compression stage
which decreases the resolution and information in the
stream (basics of lossy compression)

Now... a question that I have after this discussion is
whether the XL1(s) is true progressive or not... I've
read some articles somewhere in the past stating that
if you want to do fast pans and stuff you beter put the
thing in frame mode so that it shoots progressive....
Cause fast pans are not nice to watch on interlaced
materials....

Okay... that's me for now....

[Stephen van Vuuren]

Hello Rob:

Per your points:

1. Vertical resolution because the Canon CCD in the XL1/GL1 is capturing the image differently in frame than in interlaced - Adam Wilt covered the tech details in an article in DV Mag (www.dv.com). Frame is progressive, just not full or true progressive capture. Because it uses a technical trick, it grabs 360 lines each second rather than 480, but it eliminate the interlacing between to the 2 240 line fields. A true progressive scan CCD would capture 480 lines each second (as some low-end Canon 1-chips do) and of course the Cinealta etc. with 24P. That's why Canon does not called it progressive and put a 30P sticker on the XL-1. The manual also states 1.5X increase in field vertical resolution - not 2X.

The bottom line is because the CCD is working differently, the information is not exactly the same; it's actually quite different. That's why some people love/hate the frame look. It's not the same image capture.

2. You absolutely correct that my process involves rendering, but the Matrox codec I use has been tested as one of the best for rerending with nearly identical results. Also note, the deinterlacer plugin I use loses data that it attempts to reconstruct through interpolation.

But also note I use the offset fields & double layering trick as well developed by the Last Broadcast guys in 1997 and used by others that helps restore this lost info by basically deinterlacing each pair of fields and blending them together after deinterlacing to restore lost resolution using upper/lower field dominance.

In my and other's testing, the results are sharper than frame mode with exactly the same motion effects. The one big problem is the postproduction rendering time.

So here's the hope the someone has an under 10K true progressive scan 3-chip DV camera in the works...

:)

[Chris Hurd]

Thank you, Stephen.

There's a very sharp guy at Canon USA, he's the senior technical representative in the video division, by the name of Simon Kerr. Earlier this year I asked him for a layman's simple explanation of Frame Movie mode, and he described it this way:

"Frame Movie mode produces the same results as Progressive Scan, but through a different route." In other words, even though the XL1 and GL1 do not have Progressive Scan CCD's, their Frame Movie mode will give you video that looks similar to progressive scan.

The end results of Progressive Scan and Frame Movie mode appear to have a lot in common. However, they each take different highways to get there. How's that analogy?

[Rob Lohman]

Chris,

That sounds nice but tells me not much more... I'd really
like to what the difference is between the routes? Since
I don't own a XL1(s) (yet :) i wondered the following: you
are talking about frame & progressive as if it is both
available on the machine? I thought it only had frame
mode (besides interlaced).... not progressive.... (although
you could argue that frame is progressive :) ..... maybe you
can clear this up? Perhaps someone who has good contact
with canon (you chris?) could clear up the 360 lines debate?
Cause I really like to know this before I buy the camera..
360 lines is just too low of a resolutionin my opinion...
especially if I'm going to chop pieces off due to converting
it to widescreen!

Great discussion!