CCDs Clocked at 24Hz or 48Hz?

[Steve Mullen]

Chris reports the interlace CCDs are clocked at 24Hz which would mean the FIELD RATE would be 12Hz!

That can't be correct.

Possibilities:

1) the CCDs are Interlace but of the type that can read-out an entire FRAME (1080-lines) as long as they are clocked at HALF their Interlace rate, or less. Since the interlace rates are 50Hz and 60Hz, such CCDs could output at 25Hz and 24Hz. If Canon is using this type of CCD they are getting 24p from the CCDs and could call it 24p. Additionally, the sensitivity would drop by 6dB in Frame Mode.

2) the 24Hz claim is wrong -- and the CCDs are clocked at 60Hz, 50Hz, or 48Hz. In this case the CCDs are outputing 24fps or 25fps interlace video. And, the sensitivity in Frame Mode doesn't go lower. And, Canon could not call it 24p -- which they don't. I've got to believe this is the way it works or Canon would have called it 24p.

BUT:

This leads to three critical questions:

1) Is Canon using just one field from the CCDs? This certainly seems to match Chris describing 24Hz video being sent to the DEVIC DV 2 chip. However, if this is true, in Frame Mode the effective vertical resolution is only 540-lines -- not at all what one needs for going to film!

2) Is Canon bringing interlace video into the DEVIC chip and deinterlacing it smartly to get 810-lines of static resolution but lesser V rez on moving objects? This, frankly seems the simplest approach to 24fps!

3) Is Canon using a of version of it's FRAME MODE to get vertical rez somewhere between 540- and 1080-lines? For example, during the read-out of one field -- the Red and Blue rows 1 and 3 are summed along with the sum of Green from rows 2 and 4 to obtain a luma signal.

By doing so -- they could get about 150% of 540-lines which would be 810-lines which is exactly what is obtained in Interlace mode. This would match Chris describing 24Hz video being sent to the DEVIC DV 2 chip. There would be no loss of V rez. And, no loss in sensitivity because each field uses Row Pair summation -- with one of the fields being dropped. << THIS IS MY BET

<< By the way, with 1440 H. CCD elements there is no need for offsetting Green by half a pixel to get extra H. rez. because HDV is limited to 1440 pixels. Is it confirmed they are using H. Green shift? >>

Lastly, is 2:3:2:3 and/or 2:3:3:2 pulldown used to get 60i?

Chis -- didn't anyone ask these kinds of questions? Without knowing these answers we have no real knowedge of 24fps and 25fps modes!

[Chris Hurd]

Hi Steve, this is going to sound like a lame answer, but it's the only one they gave me... they were working around the clock just to show these things at the Canon Expo. Now that that's over with, more information including a white paper on the Frame mode will be forthcoming. When they told me "clocked at 24Hz" they might have meant per field and didn't realize it... it was a rushed week. I asked specifically more than once about a loss of vertical resolution and was told no, definitely not. I'm looking forward to a detailed explanation myself.

[Thomas Smet]

Even if it was only 810 lines of resolution could any other 1080 1/3" chip camera with a 1/3" lens resolve any more detail than 810 anyways? Even the HVX200 may have an interpolated form of 1080p if it is made like the other Panasonic HD cameras. There may not be any more detail from the HVX200 for 1080p. Even if there was a slight bit amount more of detail how many would really tell between the two at that level of resolution?

The thing that has me worried about a frame movie mode type of method is what it would do to the chroma resolution. Any chroma detail that might be gained from uncompressed might be useless since the vertical chroma space might be interpolated.

Maybe somebody can answer this. How do interlaced cameras use the same chips to shoot still photos which are progressive? Maybe the XLH1 is doing the same thing but at much faster rate.

[David Newman]

Steve,

I'm certainly speculating the FRAME MODE possibility (I just blogged about that yet I didn't know that is what they called it.) This would achieve excellent results.

Regarding "is 2:3:2:3 and/or 2:3:3:2 pulldown used to get 60i?" In DV 2:3:3:2 was a necessary quality optimization for its progressive compression option -- allowing every frame of the 24p signal to be compressed on a frame boundary. However, HDV doesn't have this issue as the MPEG2 used in the HD2 1080i spec only uses field compression. A standard 2:3:2:3 pulldown will be just as efficient as 2:3:3:2, yet 2:3:2:3 is more pleasant on the eye for direct playback so that is what should be choosen. 24p extraction is easy from either.

[Steven White]

David,

Do you think Canon might alter the spec to support frame-based compression for the MPEG-2? If so, would this offer any significant quality advantage over interlaced compression?

I thought it was neat that if they implement the 2:3:2:3 without altering the spec, your HDLink software and AspectHD are already fully configured to handle it... and give progressive 23.976 fps Cineform files with native 23.976 fps timelines... how cool is that? ;)

-Steve

[David Newman]

Steve W,
Yes HDLink and Aspect HD's Premiere capture is ready for a 2:3:2:3 24p signal, so we expect the new Canon HDV camera to work will all CineForm products without any upgrades (which is good for us, as the are plenty of other new cameras to add support.) CineForm is currently the only NLE company to directly support pulldown on HDV -- I guess our work on CineFrame 24 will be very helpful for Canon.

Added progressive encoding would break the HDV spec, as a result none of the existing software would be able to edit it -- worse then HD100U's 24p mode which is within the spec, yet many NLE tools like Final Cut can't supporting it (yet.) If they where to try that, it would be cool, as there are some signicant compression advantages; plus CineForm would be again the first to support it. ;)

[Steven White]

CineForm is currently the only NLE company to directly support pulldown on HDV -- I guess our work on CineFrame 24 will be very helpful for Canon.

I'm still lovin' you guys for that BTW. It's a brilliant workflow enhancing tool. CF24 may have its flaws, but your implementation and support of it isn't one of 'em.

-Steve

[Steve Mullen]

Originally Posted by Chris Hurd

When they told me "clocked at 24Hz" they might have meant per field and didn't realize it... it was a rushed week. I asked specifically more than once about a loss of vertical resolution and was told no, definitely not. I'm looking forward to a detailed explanation myself.

We could both be right! Normally a CCD is clocked at the field rate and if only one field is wanted -- one is tossed out.

But Canon may truly clock the CCD once every 1/24th second getting only one field. The shutter-speed would naturally be 1/48th/second.

So they would get 24fps -- except it is 24 FIELDS per second from the CCDs being clocked at 24Hz.

Obviously this CANNOT give 1080-lines of V. rez! But interlaced video from any camera is only 75% of the CCD V. count -- or about 810-lines per frame.

Since Canon says they are doing interlaced scanning the most they can get from a field is about 405-lines. Therefore, there must be a loss in V. rez no matter what they claim.

If they use their Green shift Frame Mode technolgy they could get UP TO 150% greater V. rez, which would be 608-lines.

Canon can't have the resolution possible from progressive using interlace scanning unless they have defied the laws of physics.

Thus, there is -- like the CF24 -- a real difference between "24fps" and "24p."

Not criticising you at all Chris -- but it seems Canon may have tossed this together a day before IBC and we need to not believe everyting being claimed by them.

Remember the Sony PC where they had no idea -- or at least really wanted to avoid -- what CF24 was.

Right now, this seems to be a better CF24 -- no more.

Best Regards,
Steve Mullen

[Eric Brown]

Originally Posted by Steve Mullen

Right now, this seems to be a better CF24 -- no more.

Best Regards,
Steve Mullen

Let's hope not.

[Stephen van Vuuren]

Originally Posted by Steve Mullen

Right now, this seems to be a better CF24 -- no more.

Best Regards,
Steve Mullen

In this thread http://www.dvinfo.net/conf/showthread.php?t=51305 though the link is in german and translated, so take with a grain of salt that Canon is saying "no perceptible loss of resolution" which could mean it's not true 1080p but its not 540 like CF24 either - it it get more res than 720p in the HD100 and the lens is better quality than the stock on the HD100, they could have a nice niche to carve out.

The old tech in Frame Mode got around 360 lines (2/3 rez of DV) and if they improved this method, it seems the 810 number floated would be plenty good enough for quality 24p HD. Whatever they get over 720p, assuming it's artifact free is gravy.

Until the HD100 has better lens options, it seems the XL H1 has a market for those who need it's feature sets. Of course, it could also benefit by a speedy introduction of more lens options to become the leader in the sub $10k HD market.

[Boyd Ostroff]

Originally Posted by Stephen van Vuuren

around 360 lines (2/3 rez of DV)

360/480 = 3/4 :-)

[Stephen van Vuuren]

Originally Posted by Boyd Ostroff

360/480 = 3/4 :-)

Ooops! The big irony is I'm trying to go back to school to study some physic and calculus. I hope they allow calculators...

[Steve Mullen]

Originally Posted by Stephen van Vuuren

I hope they allow calculators...

INTERLACE NTSC CCD: 480-rows

ONE FIELD before Row Summation: 240-rows

ONE FIELD after Row Summation: 180-lines

NTSC INTERLACE FRAME: 360-lines

CANON FRAME MODE: 1.5 X 180 >>> 270-lines


PROGRESSIVE NTSC CCD: 480-rows

ONE FRAME: 480-lines

ONE FRAME after KELL FACTOR OF .85 >>> 408-lines

SOOOOOOOOOOOOOOO

NTSC PROGRESSIVE FRAME: 408-lines

NTSC INTERLACE FRAME: 360-lines

CANON FRAME MODE: 1.5 X 180 >>> 270-lines

NOTE: this is just luma -- I can't calculate what Canon's Green Shift will do to chroma rez especially before 4:2:0 encoding.

I'll be frank -- my HD100 just arrived today and I really can't see why anyone would spend MORE to get 24fps than an honest 24p. Especially, since you can't really use your old lenses.

One big advantage of the Canon -- if you mostly shoot video you get 108060i which I expect to look better than 72030p with Motion Filter engaged. But, if this is your situation, why not spend less and get a Z1 or even less and get an FX1! Then use DVFILM to convert 1080i60 to 24fps. Or, use AspectHD to convert CF24 to 24.

That's why the Canon just makes no sense to me at the price. And, of course, you've really got to like Canon's ergonomics.

Steve

[Steve Mullen]

Originally Posted by Steve Mullen

Since Canon says they are doing interlaced scanning the most they can get from a field is about 405-lines. Therefore, there must be a loss in V. rez no matter what they claim.

Canon can't have the resolution possible from progressive using interlace scanning unless they have defied the laws of physics.

OK -- there is one way. They could capture interlaced video at 24 FRAMES per second. There is no rule that says 24fps must have no interlace artifacts! Why do this? Because it would enable a good film look. For those going to film, they would use one of the several sophisticated software programs to deinterlace.

Steve

[Steve Mullen]

"Since Canon says they are doing interlaced scanning the most they can get from a field is about 405-lines. Therefore, there must be a loss in V. rez no matter what they claim.

If they use their Green shift Frame Mode technolgy they could get UP TO 150% greater V. rez, which would be only 608-lines."

ANOTHER POSSIBILITY: run the CCD in "frame mode" which turns off Row Pair summation yielding the full 540-lines. Now with Canon's Vertical Green shift technology they would get 1.5 X 540 >>>> 810-lines which is exactly the same as 60i mode.

So Canon could get 24fps with no loss in V. Rez.

But they would lose 6dB sensitivity in 24F and 30F modes.