Mjpg 24p 1280x720

[Ken Ross]

I saw that Ian. The thing that's too bad (at least from my perspective) about the GH1, is that there is far more detail in the GH1's 1080 24p video than its 720p output, but motion stutter makes it unaccpetible to me. When you view a static scene at both outputs, it drives you nuts to see how much potential there is from this camera.

It's the same issue I've had with any cam I've owned that does 24p. To my eyes I'll never understand why some people like this motion issue. I understand the desire to replicate the 'look of film', but if this is what the movies looked like, I'd never go out to see them.

[Steve Mullen]

Originally Posted by Ian G. Thompson

Well, in terms of detail, this test conducted by Jack Daniel Stanley shows that the GH-1 has lots of detail in 720p...even more than the HPX170 @ 1080p. I thought this was a very good test,

Ian, when I shot all three codecs at PMA and viewed the results -- to me, M-JPEG had far more clarity -- which I thought was really sad since I wanted p60. At PMA the prototype shot true 1080p24 with no pulldown. So when Pana went with adding 2-3 pulldown I could see no advantage -- and still don't.

Now about the no-B frame explanation and the supposed "overload" caused by 1080p24. FullHD has 2 million pixels that need to be compressed 24 times per second. Let's round this up to 30 times per second. 720p60 is 1 million pixels that must be compressed 60 times times per second. Both are the same load. So, in fact, 1080p24 is fewer pixels to compress per second -- so it places slightly LESS load on the encoder.

However when 2-3 pulldown is added, "split" (judder) frames are created that are inherently difficult to compressed. (Harder than interlace. And interlace is harder to compress than progressive because under motion each field has the same objects in different locations.) So this GREATER load might well balance-out the fewer pixels per second making both 720p60 and 1080p24/60i about the same "load."

Therefore, both should encode to equal bit-rates. But from what I've read, 720p60 gets 2X the data-rate of 1080p24/60i. That crazy! And, wrong!

The only way I see how this could be if someone designed the encoder based upon a huge logic error: "OK -- 24fps is half of 60fps so for equal quality, 60fps must get the full 17mbps and 24fps only needs 8mbps.

At CES, I reported the GH1 video looked horrible. Even on slow pans tree leaves just turned into mush. I wondered how no one at Pana noticed how bad the video looked.

It's possible that the encoder is an off-the-shelf LSI chip that has no modifiable firmware. Japanese companies typical build an initial quanity -- like 20K units -- and will sell them without any changes until they gone. Then they phase in a new part.

PS: Without mentioning names, I was blown away when -- on a whim -- I compared my Casio EX-F1 to a 3-chip camcorder that sells for many many times more money. My F1 provided far more clarity!

I have to wonder if the lens on a still camera that shoots 6MP to 25MP images -- doesn't inherently have a really high MTF at relatively low video resolutions. So I suspect that "clarity" -- which looks different that "video" resolution -- may well be better with still cameras. Moreover, these camera's chips are already well beyond Red's 4K!

[Steve Mullen]

Originally Posted by Ken Ross

I understand the desire to replicate the 'look of film', but if this is what the movies looked like, I'd never go out to see them.

BRAVO, Ken!

[Ken Ross]

Originally Posted by Steve Mullen

herefore, both should encode to equal bit-rates. But from what I've read, 720p60 gets 2X the data-rate of 1080p24/60i. That crazy! And, wrong!

Steve, are you sure about that? I was under the impression that both 720p & 1080p (using AVCHD) were getting the same bitrate on the GH1, but I could be wrong.

As far as the small still cam looking better than the pricier 3-chip videocam, I think there are other factors. One is color accuracy and it seems the digital still cams do better on average than the typical camcorder.

The other issue is the size of the chips. Generally a good digital still camera will have a larger chip (even if it's 1 vs. 3). The larger chip generally gives the camera a larger dynamic range than a typical consumer/prosumer vidoecamera.

I can see this for example in comparing my GH1 to my Z5. As good as the Z5 is, the GH1 has a greater ability to capture a larger tonal range and seems to come closer to capturing a scene as you saw it. On the other hand, since I'm limited to 720p on the GH1 and the Z5 can produce smooth 1440X1080 (even though it only captures about 900 lines of horizontal resolution), I still prefer the overall 'look' of the Z5 for video. Of course other issues such as the Z5's far superior autofocus, power zoom and a host of other image adjustment parameters, make the Z5 a superior tool for capturing video.

Now if the GH1 were able to produce smooth 1080p motion and had a better autofocus, all bets would be off. ;)

[Ian G. Thompson]

Originally Posted by Ken Ross

Steve, are you sure about that? I was under the impression that both 720p & 1080p (using AVCHD) were getting the same bitrate on the GH1, but I could be wrong.

This is true...both those modes are @ 17Mbps. I think what Steve might have meant was the GH-1's MJPEG mode which is actually 720/30p @ 30Mbps. It actually looks decent and is much easier to CC.

Originally Posted by Steve Mullen

At PMA the prototype shot true 1080p24 with no pulldown. So when Pana went with adding 2-3 pulldown I could see no advantage -- and still don't.

I agree.

But as far as the 720/60p getting twice the data rate….it does not. It’s the same as the 24p. And also the mud is a non-issue in 720/60p. Do you think the interlacing (or pull-down) is contributing to the 1080/24p mud? You barely see any in 60p at all. It only happens during some high motion with high details with 24p. This is typically where interlaced artifacts show up in other cams footage that was not de-interlaced correctly. I understand the issue of lacking B-frames but then again the Canon 5D ll also lacks B-frames....but is set to a higher data rate.

[Ken Ross]

I don't think this 'mud' issue is well understood. On another site someone did a test and commented that he say 'mud' in stationary objects, but not moving objects! I saw exactly the same thing.

In a shot from my window, grass that was stationary looked muddy, yet moving tree limbs and leaves were perfect. So I think something else is going on here.

[Steve Mullen]

Originally Posted by Ken Ross

I don't think this 'mud' issue is well understood. On another site someone did a test and commented that he say 'mud' in stationary objects, but not moving objects! I saw exactly the same thing.

In a shot from my window, grass that was stationary looked muddy, yet moving tree limbs and leaves were perfect. So I think something else is going on here.

I posted "But from what I've read, 720p60 gets 2X the data-rate of 1080p24/60i." because when I skimmed thru the 1000 posts are remember several posts saying the saw 108024 running -- even motion -- only about 8Mbps while 720p60 was at 17Mbps.

If those posts were wrong and both are at 17Mbps AND the mud is NOT only on motion -- then it look like the "issue" is encoding 2M pixels verses 1M pixels.

Since the first step of encoding is moving a kernal over the image -- if the encoder CPU is speed limited -- it may use a less "fine" grain search which would cause both a loss of fine detail AND poor motion handling. In which case there is no firmware fix. They'll have to wait for a 2X or 4X faster encoder CPU. Of course, that will consume far more power and generate far more heat.

[Wacharapong Chiowanich]

Steve, I live in a PAL area where ALL the flat panel TVs (LCDs and plasmas) sold here are 50Hz or a multiple of this, 100Hz-200Hz for higher-end models. Will the 1080/24p output from the GH1 or from other sources such as Blu-ray discs (Hollywood movies encoded in 24p) for that matter look worse on these screens or the same as it does on the 60Hz-multiple screens in the US? Unlike 24p-to-60i/p, I find the math of converting 24p to either 50i or 50p hard to fathom.

Wacharapong

[Steve Mullen]

Originally Posted by Wacharapong Chiowanich

Steve, I live in a PAL area where ALL the flat panel TVs (LCDs and plasmas) sold here are 50Hz or a multiple of this, 100Hz-200Hz for higher-end models.

In 2008 there was only one LCD hdtv that could display more than 600-lines of V rez. It used LED backlights that could be turned-off briefly to replicate the erase cycle in plasmas.

There are two ways of getting 100Hz from 50Hz: repeat each frame twice or interpolate a frame between. The latter is done by a Motion Compensated Interpolation which SMOOTHS motion. It is often called "dejuddering." This works fine for 50p and 50i -- although interplolation artifacts can be nasty. 200Hz most likely has to use Motion Compensated Interpolation.

There is a nice side-effect of Motion Compensated Interpolation on this one LCD -- the interpolator actually generates ALL the displayed iamges and these images have 1000-lines of measured vertical resolution. Turn-off the interpolator and resolution drops in half. (This is why you want a plasma or DLP and not an LCD. However, if you do buy an LCD it must be illuminated by LEDs -- otherwise rez drops to about 330-lines.)

=======

The issues get more complex with both 1080p24/60i and 1080p24. The latter ONLY comes in via HDMI from a BD player that has 1080p24 output and the hdtv must accept 1080p24. Since each frame has 1080-lines, one would hope the LCD would repeat each frame 5 times and display 120fps that would measure 1000-lines. Unfortunately, this critical resolution test was not performed!

With 1080p24/60i the reviews all note that turning the interpolator OFF let's film look more like "film" but the downside of a 60Hz display is a loss of about half the vertical resolution. Given these coments, it's reasonable to assume that the interpolator does "dejudder" film. But, one needs the interpolator to get 1000-lines of V rez.

==========

Frankly, I don't think it's surprising that reviewers devote a paragraph to resolution tests performed on 60i WHERE THE INTERPOLATOR DOESN'T HURT MOTION. This supports the manufacturers marketing of 120Hz and 240Hz.

[Wacharapong Chiowanich]

Now, I think I understand the options for 24p materials on 60Hz-multiple LCD TVs but what about 24p on 50Hz-multiple (PAL) LCD TVs? How could an LCD TV interpolate 24p output from a Blu-ray player into 50/100/200Hz display? Does something more with respect to spatial or temporal resolution have to give?

[Steve Mullen]

Originally Posted by Wacharapong Chiowanich

Now, I think I understand the options for 24p materials on 60Hz-multiple LCD TVs but what about 24p on 50Hz-multiple (PAL) LCD TVs? How could an LCD TV interpolate 24p output from a Blu-ray player into 50/100/200Hz display? Does something more with respect to spatial or temporal resolution have to give?

1) Repeat frames and run at an even multiple of 24 such as 72Hz (the best -- only done by the Kuro) or 120Hz.

2) Add 2-3 pulldown to 24p to get 60p than repeat each frame. (Bad)

3) Use an interpolator to create the necessary frames (Bad).

Fundamentally there is only one monitor -- the Panasonic Kuro -- that for sure does 24p with a refresh rate of under 100Hz (needed to keep film looking like film), doesn't use an interpolator which causes artifacts and "smooths" motion, and has 1000-lines of V rez. Plus has a great black level.

Since they are no longer made -- you'll have to look at 60Hz DLP. Or, we need tests run on the new LED-based LCDs runing at 120Hz. (Anything higher does nothing positive!)

PS 1: to clarify -- we want the "true" 24fps "sampling" judder but we want to eliminate the "false" 2-3 pulldown judder needed to record 24p in 60i.

PS 2: you may not feel 100Hz refresh itself causes film to look like video. I do becaue I come from a 60Hz world. However, as I said in my first post, I do NOT know if the changed "look" comes from repeating every frame twice or it comes from interpolating the tween frames.

Hard for a visitor to tell HOW the LCDs they see work. Maybe you can shed some light. Do all 100Hz look the same to you? Do some make 25fps film look like video? How do they work?

[Steve Mullen]

Originally Posted by Wacharapong Chiowanich

Now, I think I understand the options for 24p materials on 60Hz-multiple LCD TVs but what about 24p on 50Hz-multiple (PAL) LCD TVs? How could an LCD TV interpolate 24p output from a Blu-ray player into 50/100/200Hz display? Does something more with respect to spatial or temporal resolution have to give?

I just realized you asked about 24p in 50Hz. Your HDTVs also run at multiples of 60Hz -- which is why you also get to see NTSC.