Jump to content


Photo

Canon 25F and JVC 720p


  • Please log in to reply
7 replies to this topic

#1 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11936 posts
  • Other

Posted 04 February 2007 - 01:39 PM

Hi,

At the moment, I have both a Canon XHG1 and a JVC GY-HD251, as well as the capacity to record them uncompressed, so we can answer the longstanding question about which might appear sharper - the true progressive mode of the JVC, or the higher-resolution but processed output available from the Canon. These following images are annotated either 1080p or 1080i (for the Canon in 25F and 50i modes respectively), or 720p24 or 720p60 for the JVC in either 23.976 or 60 frame progressive modes. There's not a lot of point in my putting JPEGs up for such a critical sharpness test, so I've just made links available to download the DPX. These DPX files were created directly from the recorded AVI data by Blackmagic's Framelink application. Test subjects are a Siemens star and an unusually cooperative model.

Both cameras were set up for a low-con, low-sharpening approach, so you're seeing as close to raw CCD data as is available. The XHG1 has a very similar CCD and optical setup to the XLH1, and although the lens is slightly different the results can be considered somewhat comparable.

doll_720p24.dpx
doll_720p60.dpx
siemens_720p24.dpx
siemens_720p60.dpx
doll_1080p.dpx
doll_1080i.dpx
siemens_1080p.dpx
siemens_1080i.dpx

The results? Well, I'm not very sure, to be honest. There is a visible resolution hit using the Canon 25F mode, as expected, and it's worth remembering that the chips are only 1440 wide anyway. I think it may still be slightly sharper than the JVC, but then the JVC is saddled with its' very mediocre Fuji 16x5.5 lens. The engineering compromises avoided by the servo lens on the canon (and on the XLH1 too) mean that it's very visibly a much, much better piece of glass - I yearn anew to try the JVC with better optics.

One big surprise is that the JVC is vastly noisier in 24-frame mode. This is a real eyebrow-raiser, as it seems that the camera doesn't become as much less sensitive as you'd expect when switching modes. This implies that the shorter-shutter-period 60p mode uses more gain than the 24p mode, which is counterintuitive given the higher noise level in the slower mode. It really is visibly much more noisy and I've no idea why. I'll be meeting the UK representatives of JVC on Tuesday and I'll certainly ask about it.

I still like the JVC pictures more, though. It's a nicer device, even though it is comprehensively outresolved by the Canon at least in 50i mode.

Phil
  • 0

#2 Thomas James

Thomas James
  • Basic Members
  • PipPipPipPip
  • 844 posts
  • Camera Operator

Posted 04 February 2007 - 10:49 PM

There is no way that the Canon outresolves the JVC. This thing about 720p being medium definition is just a myth. The fact is that 720p resolves higher than 1080i. This is because 720p gives you 60 full complete frames per second while 1080i can only give you 60 fields or 30 complete frames per second. Thus 720p has better temporal resolution. Don't be fooled when the naysayers tell you temporal resolution counts for nothing. The fact is that video is all about motion so it really is not fair to measure video resolution using static test charts. If static resolution charts are used 1080i will resolve higher but as soon as motion is introduced 720p will blow 1080i away.
  • 0

#3 Keith Mottram

Keith Mottram
  • Sustaining Members
  • 824 posts
  • Other

Posted 05 February 2007 - 08:01 AM

There is no way that the Canon outresolves the JVC. This thing about 720p being medium definition is just a myth. The fact is that 720p resolves higher than 1080i. This is because 720p gives you 60 full complete frames per second while 1080i can only give you 60 fields or 30 complete frames per second. Thus 720p has better temporal resolution. Don't be fooled when the naysayers tell you temporal resolution counts for nothing. The fact is that video is all about motion so it really is not fair to measure video resolution using static test charts. If static resolution charts are used 1080i will resolve higher but as soon as motion is introduced 720p will blow 1080i away.


um whatever... i have looked at the image and well my initial opinion is you're talking knackers. i mean i get the idea of your point, but having done a fair amount of work in uprezzing 1080i and 720p to 1080p with various results in shake i fail to see why a sharper image becomes softer than a less sharp image when moving. i suppose if you were to use no post production at all...
  • 0

#4 Stuart Brereton

Stuart Brereton
  • Basic Members
  • PipPipPipPip
  • 3057 posts
  • Cinematographer
  • Los Angeles

Posted 05 February 2007 - 08:18 AM

This thing about 720p being medium definition is just a myth.


The only person I have ever heard use the term 'Medium Definition' is you. 720p is part of the HD standard. I've never heard anyone, here or elsewhere, say otherwise.
  • 0

#5 Thomas Worth

Thomas Worth
  • Basic Members
  • PipPipPipPip
  • 372 posts
  • Director
  • Los Angeles

Posted 05 February 2007 - 08:34 AM

I still like the JVC pictures more, though. It's a nicer device, even though it is comprehensively outresolved by the Canon at least in 50i mode.

In order to get a fair comparison, you'd have to deinterlace the 50i frame, take one of the fields and then upscale it to 720. A locked-off shot of a resolution chart doesn't represent a real-world use of the camera (as I know you're well aware).

The increased bit depth afforded by the SDI capture is nice. I did some color tests with your DPX files and was delighted with the latitude. It's similar to shooting in RAW mode with a dSLR. Good stuff. A bit noisy, though.

Did you capture directly to DPX using the Blackmagic / Aja card? Did you ever figure out exactly what it was doing, i.e. copying the SDI stream directly to disk and only adding a DPX header?

This thing about 720p being medium definition is just a myth.

Raw picture data rate of 1080i/60 @ 10 bit 4:2:2: 148 MB/sec
Raw picture data rate of 720p/60 @ 10 bit 4:2:2: 132 MB/sec

The two share almost the same data rate. This would imply that they carry almost the same amount of information. The main difference between the two is the way they present the data.
  • 0

#6 Thomas James

Thomas James
  • Basic Members
  • PipPipPipPip
  • 844 posts
  • Camera Operator

Posted 05 February 2007 - 10:02 AM

Of course a sharper image can become less sharp when motion is introduced. We can have 720p at 60 frames per second or 1080p at 30 frames per second or even 1500p at 15 frames per second. Using static test charts 1500p15 would blow away 1080p30 but as soon as motion is introduced the 1500p picture quality would go to hell. It would either be a stroboscopic mess if the shutter speed is high or it would be blurry if the shutter speed is set low. 720p60 has the best temporal resolution yet many people insist that only 1080i or 1080p is Full HD. Yet for the fast action sports 720p is the highest resolution possible. Some consider the 1080p60 format the holy grail of high definition. Nevertheless 720p shot at 120 frames per second would result in some awesome slow motion playback for sporting events.
  • 0

#7 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11936 posts
  • Other

Posted 05 February 2007 - 12:00 PM

Hi,

Mr James - I think you're mixing things up rather. It's pretty obvious that a 1080-line image has more spatial resolution than a 720-line one, which is all we're talking about here. What's at issue is whether the deinterlaced Canon image sacrifices enough resolution that 720 is a better bet. I shot the 720 at 24fps anyway.

I've made up an image that should be a direct comparison, by scaling-up the 720p frame in Photoshop:

siemens_compare.cin

There's really very little in it. I think the 1080 is fractionally sharper, but that could still be down to lens issues. Certainly not enough in it to make me eschew the otherwise much nicer JVC.

In normal 50i mode, the Canon 1080 frame quite clearly and obviously outresolves the 720. Actually the canon produces what I'd call pretty damn convincing 1080, even up against very high end stuff - it's a crying shame it won't do true 25p.

All these frames were captured using a Blackmagic Multibridge Extreme with Premiere Pro 2.0 as the capture application recording Blackmagic codec 10-bit AVIs. These AVIs were then mounted with the Framelink application, which causes each file to appear as a new virtual storage device filled with DPX files. I then copied these DPX files to the web server. The split screen was made in Photoshop, with the only imaging operation being the upscale from 720 to 1080.

By the way, I'd consider the fairest way to view these would be to take them into Photoshop and apply a curves filter with a single point at coordinates 80,150. That's a fairly intense brighten without clobbering the highlights.

Phil
  • 0

#8 Phil Rhodes

Phil Rhodes
  • Sustaining Members
  • 11936 posts
  • Other

Posted 05 February 2007 - 12:48 PM

Hi again,

Here we have some comparative compression demos between the HDV and SDI. This is all 60p shot on the JVC GY-HD251, using a photo flash to synchronise the internal HDV tape recording and external hard disk recorder. The MPEG-2 encoder is, therefore, that of the JVC camera. I believe that it compresses to a 12-frame GOP at this frame rate.

I then captured the tape playback from the camera via its HD-SDI output and produced the split screen in Premiere, which is why they're TIFFs - Premiere won't export DPX (d'oh).

First a fairly straightforward example of a fast-moving subject - the rotating reels on a 1/4" tape deck. I'm being very, very unfair to the HDV here - it's 60 frames a second, and the I-frame of this GOP was (I'm pretty sure) the one that saw the photo flash, thus completely screwing up the codec's chances of making a decent run of the next eleven frames. Nevertheless, it's no worse than I've seen on other frames. I think what's happened here is that the yellow label on the reel has just come into view, representing a whole load of previously unseen macroblocks. The rotary motion of the reels is unfriendly to linear motion compensation, too.

reels_1.tif

I think this next example is a simple case of "insufficient bandwidth for desired image acutance," or to put it another way, 19Mbps isn't enough for 720p60, even when there's effectively no high frequency detail in the image.

mcu_two.tif

Of course if you have the unmitigated temerity to grade it in any way that increases contrast:

mcu_two_graded.tif

And not to drop any dead donkeys, since this isn't really a compression issue, but dropout is one thing that tape does and disks don't (at least, you'd better damn well hope they don't...). I was using totally generic miniDV tape which I've used thousands of hours of before without incident and the density is no higher for HDV, so I can only assume it's completely random chance.

reels_dropout.tif

Phil
  • 0


Tai Audio

Metropolis Post

Technodolly

The Slider

Media Blackout - Custom Cables and AKS

Paralinx LLC

CineLab

Glidecam

Opal

CineTape

Broadcast Solutions Inc

rebotnix Technologies

Ritter Battery

Gamma Ray Digital Inc

Wooden Camera

Rig Wheels Passport

Visual Products

Aerial Filmworks

Willys Widgets

FJS International, LLC

Abel Cine

Gamma Ray Digital Inc

Glidecam

FJS International, LLC

rebotnix Technologies

Rig Wheels Passport

Willys Widgets

Abel Cine

Opal

Ritter Battery

Technodolly

CineTape

Aerial Filmworks

CineLab

Metropolis Post

Broadcast Solutions Inc

Visual Products

Paralinx LLC

Tai Audio

Wooden Camera

Media Blackout - Custom Cables and AKS

The Slider