Is 720p30 really progressive?

[Steve Mullen]

Today I decided to pass on exposure issues and look at shutter sppeds. JVC claims 1/30th is correct while I have believed 1/60th gets you a speed equal to film with a 180 degree shutter.

Using NYC traffic and walkers I found the following:

NO MATTER WHAT THE SHUTTER-SPEED -- WITH MOTION -- THERE ARE ALWAYS 2 IMAGES CAPTURED PER FRAME. The only effect speed has is that at 1/30th the 2 images are blurred and so less distinct. At higher speeds they become very clear. I always see a taxi moving as two displaced images locked together.

I believe the 2 images are time-displaced 1/60th apart. How can this be IF THE CAMERA IS SHOOTING PROGRESSIVELY?

I don't think it is!

At p30, I think the shutter is still activating every every 1/60th S. Perhaps the consumer CCD can't run at 30fps. It remains clocked at 1/60th S. So every moving image is captured twice per frame. (It is the same temporal artifact we find in interlaced video.)

At the end of 1/30th second, the two exposures are treated as one frame and recorded. (I suspect the two 720-line exposures are added to increase light sensitivity.)

To check this, I switched to SD. Now the 2 images are gone. One exposure per frame. Altering shutter-speed now does as expected. Slower > more blur. Faster > more strobe.

Switching back to moving objects in HD makes ones eyes hurt.

IF I'M RIGHT -- then 1/30th is the ONLY speed that is acceptable -- which may explain why JVC is so strongly recommending it. Moreover, it is critical to lock 1/30th with SA.

I'll leave it to others to comment on "is the increase in resolution -- which there is -- enable the JVC to be called an HD camcorder."

I HOPE I'm wrong. Others need to confirm. But I remember Ken's answer to questions about the strobing: "it's in the camera's design" he said.

Ken, care to comment?

[Paul Mogg]

Steve, the look of the video when locked to 1/30th is certainly a 30p look as opposed to 60i if that is what you are questioning. I do believe it is doing progressive 30p. I think you need to ask Ken (as I did before) to provide a breakdown of exactly what this camera is programmed to do with the shutter/iris under various lighting conditions and in the various shooting modes. The engineers who designed it must certainly know, and it would certainly save a whole lot of trial and error on our parts. I did not get a satisfactory answer when I asked for this, though I didn't follow up on it more.

[Joseph George]

Steve, from what you described, it certainly appears that the camera is not HD.

To start with, HD starts with 1280 effective horizontal pixels; the No. of effective horizontal pixels onb this camera is exactly between that and SD. So should it be called semi-HD?

720p Vericam has 3 CCDs that have pixels of all 3 colors precisely aligned. This is 1 CCD camera with less effective pixels that the format itself, and the pixels of the different colors are naturally not aligned. So there is a shift between the pixels, which means color picture degradation, although not serious.

There is no reason for the CDD to be incapable to shoot at 30p, if it is capable of shooting at 60p. I think that the best explanation for the 2 images is that the camera shoots at 60i.

This, the inability to lock the exposure, etc. -- it is all ridiculous. Plus the camera has unbalanced mic input, to which you plug in the XLR box -- and this changes it from consumer to pro!

So this camera actually has 720/60i CCD signal, and the images are acceptable on stills, because on stills even NTSC looks progressive. If the image moves, it is either terrible at higher speeds, or blured at 30p. That is the jittery image that everyone described in the past.

Congratulations to Matsushita on making another camcorder, to sell it by leaving a lot of facts, and promoting it as something it is not, and at the same time making it useless for any pro work, so their more expensive camera market stays protected.

I don't know if you guys know, but the DVX does not have gain up in progressive, so the Lux rating becomes 24; it does not auto focus in progressive. It does not lip-sync, while a lot older PD150 lip syncs perfectly. Depending on the mode, the sound comes in up to 3-4 fields ahead of the picture. The built-in mike is basically unusable, because of noisy motors. It just goes on and on. A lot less expensive PDX10 uses B/W viewfinder and special pixel arrangement on LCD, so their horizontal resolution is 500-850 lines. The DVX is at about 360.

I personally think that JVC owes us a lot of explanations, as to why is this thing called HD, why is it pushed as a pro camcorder, and what is this thing good for.

I personally think that you should wait for a Canon or Sony MPEG2 HD camcorder.

As to JVC -- you did bring out the world's first HD still camcorder. Congratulations! if anything moves, it is no longer HD. Does this thing take stills? If it does it is the world's most expensive 1 megapixel still camera.

I understand that the optical image stabilization is not very usable either.

[Steve Mullen]

<<<-- Originally posted by Paul Mogg : Steve, the look of the video when locked to 1/30th is certainly a 30p look as opposed to 60i if that is what you are questioning. -->>>

I think I/we have been assuming the look is 30p.

I'm very used to 30p from the DVX100. 30p is "an even stutter" because the motion is being captured 1/30th S apart.

Judder looks like "stutter with a beat."

The JVC captures 2 "images" at regular intervals -- 1/30th S apart.

If there is NO motion there is only 1 "image." But as motion speed increases, two images appear. As motion speed increases the displacement of the two images increases. There are 2 buses going by. Each window is 2 windows. Each door is two doors.

I think Ken gave you a true answer about the camera's design. But failed to really explain the implications of his answer.

Sort of like answering "an iPod battery will last 2 years" without adding "it can't be replaced." :)

[Heath McKnight]

Does this explain the ghosting effect I see when people walk around? Maybe I should take the camera back...HELP!

heath

[Steve Mullen]

<<<-- Originally posted by Heath McKnight : Does this explain the ghosting effect I see when people walk around? -->>>

Ghosting is a PERFECT description -- except both images are of equal intensity.

I'm going to use what I know and try some general shooting using locked 1/30th.

I've been shooting traffic going by my camera. I'd normally pan WITH a taxi, not let it zoom by.

[Heath McKnight]

On my XL-1, I always had, in total manual control, the effect of non-blurring, like in stop-motion animation, was how I felt moving images looked to me. If that's the case with the HD10, then that's cool. And what a lot of 28 DAYS LATER looked like in action sequences or the windmills shot.

Be honest, Steve, did I just blow $3200?

heath

[Joe Russ]

uhm, 1/30th is a poor shutter speed and any ghosting you see is because the shutter is open too long to take a clear image of anything moving. but, its an interesting idea you raise. ill do some tests myself, but i personally dont believe its 2 images sampled at 1/60th. the reason i believe the hd is only 30p (and not 2 60i/p samples combined) is because the proccesor cant handle that resolution at 60 frames a second. the pictures im getting dont appear to be showing signs like interlaced (normal dv) cameras either.

interesting never the less

[David Newman]

Steve,

I do not observe this effect, but I will need to recheck with some 1/30th shutter footage when I get to the office. Even if there are two "exposures" per frame, this would have no impact on whether the CCD is progressive (resulting in the video being progressive.) The might be be a good sign the camera is heading for 60p. Anyway, the double shutter per frame effect can be seen on another progressive scan camera -- the orginal Canon Optura. The output is definitely progressive in that old Canon and the new JVC cameras. There is nothing to worry about here.

[Steve Mullen]

Got feedback from JVC. They see the same thing we do. But they say the two images are 1/30th S apart, not 1/60th S.

They claim it isn't a capture issue. It is spectulated that it is a visual phenomena. It's in our heads. :)

Somehow when each frame is repeated twice -- to go from 30p to 60p -- our eye sees any moving object in two locations. The faster the motion, the further apart the apparent images.

Proof, they say, is to look at each frame one by one. And, they are right!!! Each frame is perfectly clear. Of course, the higher the shutter-speed the clearer. But, there is no double image in one frame. Which is why still captures don't show any problem.

There is one other possibility. It's not in our heads. Possibly the camcorder's upconverter is blending 2 frames to "smooth" the rate conversion. (It's the same upconverted to 1080i or down to 480p or 480i.)

Of course the double image is still there, only the explanation has changed!

[Frank Granovski]

I don't have any of these described problems with my very old GR-DVL9500 progressive scan cams. Maybe there's something wrong with it?

[David Newman]

Steve,

Here is proof that the camera is doing nothing weird and is working perfectly. This footage was taken at 1/30th shutter. The camera was spinning on the len's axis to give a streaking effect. If you observe the stage lights you can see that there is no strobing, and all the motion in defined (and there isn't two images per frame.)

Frame1
http://www.cineform.com/images/frameSM0.jpg

Frame2 - 1/30th of second later.
http://www.cineform.com/images/frameSM1.jpg

Frame 1 & Frame2 mixed to show the prefect 30p capture.
http://www.cineform.com/images/frameMIX2.jpg

P.S. These images have been scaled to 480x270 in GIMP from the original 1280x720. I used GIMP to show the overed frames. If anyone needs full size samples please email me.

[Steve Mullen]

<<<-- Originally posted by David Newman : Steve, Here is proof that the camera is doing nothing weird and is working perfectly. -->>>

You have confirmed what JVC claims and I agree with (I've revised my post).

But, we still all see double images during normal playback. Either these are from our eyes or from the camera's rate converter.

And they still cause viewing fatigue because this is not how either video or film looks.

Lowering the shutter-speed to 1/30 simply creates so much blur the two images obscure each other. That really lowers detail on moving objects! I still believe 1/30th is too slow for motion. 1/60th is correct, but the strobing is very annoying.

[Joe Russ]

1/60th is fine (for normal motion) and converts very well to 24p (important to me) with no motion stutter.

[Michael Hyun]

if this is merely a camcorder playback issue, then theoretically the playback should be fine when other HD playback devices are produced.

this is an interesting idea....

i have been selecting my shooting strategies based on the output of the camcorder rather than what the HD source video might actually look like.

if this is the case, it is possible that some of the shooting situations i have been ruling out might actually work out down the line.