Question about 1440 x 1080
 

I'm totally lost about the Aspect Ratio and need some assistance. Can anyone explain, or direct me to proper thread which explains this?

When i shoot in HDV mode, my HVR-A1 shoots 1440 x 1080 (i believe), but the ratio of 1440:1080 is 4:3 instead of 16:9. But on the screen it appears as 16:9.

And when i shoot in Standard DV mode, and select 16:9, it comes out as 720 x 480, which is neither 16:9 or 4:3, but 3:2. I'm starting to think that # of pixels has nothing to do with aspect ratio.

One extra question, if I edit 1440 x 1080 footage and 1920 x 1080 footage on same timeline, would there be any problem? I'm thinking about getting a XH-A1 and use along with my HVR-A1.

This is funky, but in 1440x1080, you're not shooting square pixels. You're shooting rectangular ones that are 1.333 times wider than they are tall.

Essentially, you're shooting is a 16:9 image onto a 4:3 sensor; and the way this is done is because the camera lens and sensor are anamorphic - or, in layman's terms, "squished like a funhouse mirror."

Similarly, in Standard DV, 720x480 is 3:2, but if you squish it one way, so that the pixels are .9 times as wide as they are tall, you get a 4:3 image. If you squish it the other way, so that they're 1.2 times as wide as they are tall, you get a 16:9 image.

Thanks so much brian. It makes sense now.
So, if I shot 16:9 on SD miniDV, for the Setting on Final Cut, I should choose "DV-NTSC Anamorphic" instead of "DV-NTSC" or "DV-NTSC Firewire Basic"?

And, can 1440 x 1080 mix up with 1920 x 1080 on a same timeline?

The Canon XH-A1 shoots 1440x1080. There is no such thing as 1920x1080 HDV.

HDV camcorders don't use anamorphic lenses and sensors with 4:3 physical dimensions. They use sensors that are indeed 16:9 in shape.

If your NLE supports it, you can mix 1440x1080 and 1920x1080 footage on the same timeline. Works just fine with Edius.

No!
There's no anamorphic process at the lens/sensor stages! All the HDV cams ('pro' or not) have a 16:9 image from the lens to the sensor. If the sensor has a 4:3 shape, some top and bottom pixels are simply not used in HDV mode (and the full 4:3 surface is only usable for the 4:3 SD mode or the photo mode).

The conversion from the 16:9 ratio to the anamorphic 4:3 ratio is done at the electronic image processor stage (for HDV: 1440 horizontal pixels, each one representing 1.33 width of a 'displayable' pixel).

To the best of my knowledge (perhaps I'm mistaken) all HDV cameras use sensors that have a physical 16x9 shape. How the image gets to 1440x1080 from the sensor differs by camera. If I recall correctly, the Canon XH-A1 uses 16x9 shaped sensors with 1440x1080 pixels to begin with. I believe the HDR-FX1 uses 16x9 shaped sensors with 960x1080 pixels, and obviously there's some digital processing (pixel shifting) involved to generate a 1440x1080 image for recording. I'm not sure, but I think the HVR-A1 uses a single 16x9 shaped CMOS chip with 1920x1080 pixels (and downscales for recording).

I just looked it up. The HVR-A1 uses a 2.97 4x3 dimension CMOS sensor (cropping the image from the chip for HDV).

Actually, no conversion at all if the sensor has physically 1440x1080 pixels - it's just stored as is on DV tape. When playing back the tape, the data is read as is - the 1.333 pixel aspect ratio is inferred so that square pixel monitors can produce the proper scale to get a 16x9 display.

Of course - if your monitoring the output of the camera from an analog output - the conversion is done in camera.

To my knowledge the only cameras out there that use a sensor that is 1920x1080 is the SONY EX1 and EX3. All other cameras get to 1920x1080 through interpolation, stretching, or pixel shifting. All HDV output to 1440x1080 or smaller (think 720p.) This information is based on a video tutorial on the XDCAM EX format and can be viewed at the link below.

What's So Great about XDCAM EX? - on Vortex Media

(Note: I originally found this on the XDCAM thread here on DVi...)

Umm, I think some (all?) of the consumer AVCHD cameras use full 1920x1080 sensors. No?

It's just HDV that does 1440X1080.

Consumer AVCHD cameras use sensors with all sorts of different pixel counts. Canon has one out now that has close to a 9 megapixel sensor. Panasonic leans on pixel shifting with 3 chippers that have fairly low pixel counts. Most AVCHD camcorders that record 1920x1080, can also record 1440x1080.

Exactly.

Somewhat OT:

I wonder if a large pixel count bayer-sensor interpolating to a lower resolution beats a 3 chip with lower resolution?

The TM300 (and sister models) look promising. Previous Panny 3-chipper consumer AVCHD cams have been somewhat disappointing though.

Gonna be a while before I go for my next camera, current finances being what they are.

I think for me the issue in the furture is going to be low light sensitivity. I think, resolutio-wise, most of the upper-end consumer/pro-sumer cameras are doing fantastic. Exposre control, etc...

What I like about the TM300, from what I've read, is perhaps the best low light performance of any consumer cam thus far, combined with fairly robust manual control (especially compared to consumer cams from other major mfgs).

When the TM300 (and sister models) come down to well under $1k, I'll start seriously considering selling and replacing an HV20.

No. This statement is not true at all, and even though many of the newer cameras (consumer AVCHD) talk about megapixel chipsets those are packed into tiny chips that get very poor amounts of light compared to 1/3", 1/2", and 2/3" chips. Here is an article on the Videomaker .com website that pertains to this issue but for most cameras you will have to search model by model to find specific sensor pixel counts.

State of Imaging - on Videomaker.com -article 13054

The output may be 1920x1080 but the imaging sensor often is not. Here is a quick excerpt from the article with a few examples:

"...The Canon XL H1 has three 1440x1080-pixel sensors, the Sony HVR-Z1 has three 960x1080-pixel native sensors and the Panasonic AG-HVX200 has three 960x540-pixel sensors, yet they are all capable of reproducing 1080i (1920x1080-pixel) video. Each camcorder is capable of doing this by a measure of multiple technologies and algorithms that include pixel shifting and pixel interpolation..."

It should be noted that just because a camera records 1080 lines of video, that doesn't mean that the realistic resolution is that high. HD cameras vary considerably on realistic resolution (in the neighborhood of 600-800 lines is pretty common). It's pretty well known that 1080i out of an XL-H1 is sharper than 1080i out of an HVR-Z1U or HVX200, for example.

I may have gotten the technical details wrong, but I think that explains it.

I'm not as good with final cut, but I'd try DV-NTSC Anamorphic, or a pixel ratio of 1.2. Usually I just capture as 4:3 and use the Motion tab to stretch out the video to the full 16:9 when rendering.

1440x1080 can, indeed, mix up with 1920x1080 on the same timeline so long as they are the same framerate.

Robert, that is a great way to sum it up. I could not think of the words to say it so succinctly. The vid clip (I linked earlier) from Vortex Media is very interesting, and goes into good detail in comparing HD, HDV, and XDCAM but it is also over an hour long...

Additional thought... It is similar to taking a wallet sized image in photoshop and blowing it up to an 8x10. The results will vary depending on the subject you are shooting. Low detailed moderate contract images will probably print out fine but the interpolation algorithms will have a harder time making the enlargement of more detailed, higher contrast images, resulting in softening as the contrast lines get muddied...

Whilst AVC-HD may enable the recording of a 1920x1080 signal, virtually none of the cameras that use it are capable of doing justice to the recording format.

At the more expensive end, the HMC150, the sensors are 960x540 like the HVX200, so even after pixel shifting it's resolution is only about 1200x650. You're actually better off recording in 720p mode - it maintains the front end resolution, and does a better job of compression.

At the cheaper end, single sensors may indeed have many million of pixels, but simple numbers don't always tell the full story. Take account of de-Bayering and downconversion and you're likely to not have the reolution first hoped for, and a lot of nasty aliasing as well.

At least with 3 1440x1080 sensors, recording to HDV, you know where you stand.......

And that is a very misleading statement, if not downright untrue. None of them even make a 1920x1080 recording.

What is true is that those three examples of camera may well give recordings of pixel dimensions 1440x1080 (1280x1080 for the HVX200) but what the actual sharpness of the image is is totally unrelated. Just as Super8 film blown up to 35mm won't look as sharp as if it was originally shot on 35mm.

Pixel shift is generally reckoned to give a resolution enhancement of about 1.5x for luminance. So if used horizontally, can indeed give 1440x1080 from 960x1080 chips. But if used horizontally AND VERTICALLY (as with the HVX200), the improvement is spread between the two axes, so about sq rt 1.5X (about 1.2x) on each axis. Hence you can expect performance relating to about 1200x650 from the 960x540 chips referred to.

The three examples of camera given are an excerpt from the article linked in that post. I never said they make a 1920x1080 recording in fact if you read the thread you will see that was precisely my point. The output resolution in the NLE is 1920x1080, but the recording is smaller and the chips are even smaller than that. If it came off differently than I am sorry for the confusion but you and I were saying similar things...

We *are* talking AVCHD cameras here, aren't we. There's little reason for them not to - withing reason of course - depending on the class of camera and it's intended market/usuage. DOF is a issue - but then you'd be better off buying a canera with a removable lens.

And these are HDV camera's.

Even a camera like the EX1 (with three 1/2" 1920x1080 pixel sensors) doesn't resolve 1080 lines (but it is dang sharp). Hopefully, in the reasonably near future, we'll see a camera on par with the EX1, that records AVCHD (on cheap cards).

Any criticism wasn't directed at you, but rather the article, sorry.
But why should we want it? Use MxR adaptors and the EX can use exactly the same SDHC cards that you refer to for recording AVC-HD. My own experiences are that 35 Mbs MPEG2 is better quality than 21Mbs AVC-HD, whilst being far easier to edit. OK, up the AVC-HD bitrate and quality might achieve parity with the EX codec, but the effective need to transcode to edit remains. So why not just say stick with MPEG2 for a number of years, until processing power is more able to cope with AVC-HD.

The initial impetus behind AVC-HD was indeed given as "it's the only way we can record decent quality to cheap cards". Those people who so cleverly found out how to use SDHC cards reliably in an EX seem to have whipped the carpet out from under that argument......

The main reason I'd like to see an AVCHD camera, with comparable imaging to an EX1, is the considerably smaller size files needed to archive original footage. I'd also rather use recording media that the camera was designed to use, rather than hacking the hardware (that could easily be disabled by a firmware "upgrade" from Sony). It's not likely we will see such a camera real soon. By the time we do, editing AVCHD should be as smooth as editing MPEG-2 is today (probably smoother).

Octa-or-more-core processors, that are even more efficient per core than an i7, are not very far off really. We're also on the verge of seeing 64bit computing, as well as the power of massively parallel processing GPUs, finally being effectively unleashed for video editing and encoding purposes. More than an order of magnitude leap in performance seems pretty likely in the next 2-3 years.

AVCHD at 24mbps is certainly potentially capable of image quality performance that's at least on par with 35mbps MPEG-2. I was just looking at some footage out of a TM300, with a lot of motion, the other day. The image compression quality is pretty dang good (way better than HDV). Real-time AVCHD encoder chips are are coming along nicely (and will get better), while MPEG-2 encoding technology has pretty much peaked.

Gene- I quoted from the article linked and yes those are HDV which is where the 1440x1080 format comes from which is the topic of the OP, so i was trying to come back around.

David- Didn't take it personal, just hoping I didn't muddy the water with my statements.

I am with you David, on the XDCAM EX format. Lots of improvements over anything else out there at this time. From what I understand the new JVC XDCAM's come with SDHC slots already and the SxS slots are an add-on, though I am a fan of SONY it is an interesting development especially in tapeless acquisition.

I don't think MPEG-2 is going away anytime soon. For one, it's a broadcast standard that's going to be with us for many years.

Secondly, AVCHD is great, but not *that* great - it's needs to be a couple of orders of magnitude better than MPEG-2. In the consumer market - it makes sernse in some areas - but why bother in the professional market. The space savings is good, but not so much that you'd trade everything for it, and you're probably going to go for something easier to work with on ingest.

Of course, increased processing power allows us to do things we couldn't efficently do before - and we can expect AVCHD to be easier to work with. But with storage getting ever cheaper - why waste the cycles - which could be put to better use.

2 cents, etc....

I just don't see the file size as that much of a problem. I'll agree that very roughly 25Mbs AVC-HD may correspond to 35Mbs XDCAM-EX for quality, but that only equates to corresponding file sizes of 50MB v 70MB. Given the far greater difficulty of editing AVC-HD, the file size gain just doesn't seem worth it. If AVC-HD really had meant you could use SDHC, stick to MPEG2 and you can't, it would be a different story, but recording XDCAM-EX to SDHC has totally blown that argument.

And yes, the JVC cameras shoot 35Mbs MPEG2 to SDHC natively, as they are designed to do. I don't feel too worried about future disabling of EX cameras regarding SDHC use. Sony haven't said much about how they feel about the situation, but a UK magazine recently published an interview on the subject where they did comment publicly. They didn't endorse the use of SDHC cards as such, but the interview was as interesting as much for what they didn't say as what they did. Most people have come away with the impression that they are far from antagonistic to the use of SDHC cards. Even if they DID block the use of MxR adaptors by disabling the USB bus, there seems nothing to stop a future adaptor making use of the PCIExpress interface to feed a SDHC card. Quite simply, the genie's out of the bottle.

OK, never say never, and in years to come then maybe AVC-HD will be as easy to edit as MPEG2 is now, but maybe by then we'll be looking towards some even newer and better codec, maybe wavelet based. In the meantime, the bitrate savings of MPEG4 may be well worthwhile for transmission, but much less so for acquisition.

Cineform proved the advantages and commercial viability of a wavelet codec, at least for intraframe only compression, a few years ago now. I'm a little surprised at Panasonic, for introducing AVC Intra, well after that, as their codec of the future. I would think a Cineform like codec would be a vastly superior choice. At similar bitrates to AVC-I, the image quality is excellent (dang near lossless, for practical purposes) and it takes far less CPU power to encode (and decode). That should translate into encoding chips (for in-camera video compression) that use less power and generate less heat. Add to that faster performance on editing machines, and it sure sounds like a real winner all the way around, that might have moved Panasonic well to the front of the pack.

AVCHD may only be lighter than XDCAM-EX by about 30% (as far as file sizes), but that's not a small issue, in my mind. When I go tapeless I expect I will probably wind up generating a lot more footage than I do now, and I just can't bring myself to throw away any footage I shoot of my parents, kids, grandkids, or even my dog (who is perhaps the most entertaining member of the family!), as well as of friends, etc. I look forward to not having the expense of tapes, but I'm likely to wind up buying a heck of a lot of Blu-Ray disks.

Actually, most single-chip HD camcorders have full 1920x1080 sensors... pretty much every one that came after that first JVC. And just about every camcorder ever made has square pixels in the CCD, regardless of the storage format.

It's more complex with 3-chip cameras. In the olden days (SD days), you really needed three full frame sensors, since the main point of 3-chip was to give you full color pixels. But for HD, even subsampled color is brilliant, given the relative size of the pixels. Plus, you pay for full-frame sensors in light-gathering capability... the sensitivity of a sensor at any given technology level is based on the size of the sensor (plus any extra light that might be routed to it via a microlens). So there have been all kinds of arrangements, using half-frame-or-smaller sensors with offset, essentially using 3-chip more for light-gathering advantage than color advantage.

The Sony HVR-A1 has a 4:3 sensor that's about 1920x1440 or so... that yields the nominally 2.8Mpixel that's claimed for the sensor. When you're shooting stills (as if anyone does), you get the full field.. when you're shooting video, it does a 1920x1080 crop for 16:9. All pixels are active .. the "spares" are used in the digital image stabilization algorithms. Think about the 16:9 crop being a movable window, and the stabilization being an algorithm that decides just where to put that window in the sensor, based on montion detection.

When the image is processed for compression, the 1920 lines are downsampled to 1440. As with all formats short of high-end stuff, there's also color subsampling and other stuff going on here, too (4:2:0 in HDV mode, and perhaps additional image processing, depending the light and the mode). So the end result are pixels that are 1.333x wider than they are tall... not too weird for anyone who's been doing SD all these years.

And in fact, important enough that Blu-Ray officially supports 1440x1080 resolution video at 16:9, along with the usual 1920x1080 and 1280x720.

Actually, Gene is right, his statement is true. Most single-chip consumer AVCHD (and HDV) camcorders use full 1920x1080 sensors. Cases in point, the entire Canon VIXIA camcorder series. Hope this helps,

Chris ,can you point us to any source material that clarifies this? According to B&H's tutorial on the XDCAM format, the only camera's they have in the store with physical pixels that count 1920x1080 are the EX1 and EX3. Beyond that, i am having difficulty finding any information on physical pixel counts on the sensors of any Vixia cameras. While this may be a minor point, i still find it hard to believe that the vixia circa $600 cameras have sensors the same physical resolution as the EX1/EX3 and don't utilize the methods other HDV cameras use such as pixel stretching and pixel shifting. Could you give us any links to clarify this? Thanks.

You have to actually try before you can claim trouble. The first place I'd look (since that's where I looked) would be Canon's web site, if I want to know Vixia pixel counts. Go here: Canon U.S.A. Choose "camcorders", then "consumer", then find your Vixia of choice, and click on the "Specifications" page. Took me about a minute, including the time to read this post.

This would be telling... but hey. So I look up "Vixia HF S11", and I find these stats:
Image Sensor
1/2.6-inch CMOS, RGB Primary Color Filter
Total Pixels
Approx. 8.59 Megapixels
Effective Pixels
Video: Approx. 6.01 Megapixels (3264 x 1840)
Still Image: 16:9 Approx. 6.01 Megapixels (3264 x 1840)
4:3 Approx. 8.02 Megapixels (3264 x 2456)

So, in English, that means the 16:9 crop of the sensor yields a 3264 x 1840 image, which is downrezzed in software to 1920x1080 or 1440x1080, depending on your selected recording mode. I guess actual 1920x1080 sensors are "so last year"... well, the HF200 yields 2304 x 1296 from the sensor, while the HG's still deliver a true 1920x1080.



While this may be a minor point, i still find it hard to believe that the vixia circa $600 cameras have sensors the same physical resolution as the EX1/EX3 and don't utilize the methods other HDV cameras use such as pixel stretching and pixel shifting. Could you give us any links to clarify this? Thanks.[/QUOTE]

Dave i saw that too but those are effective resolution, which as I understand it is not the same thing as the physical resolution of the sensor.

For example using the cameras listed before, the physical resolution of the sensors in the Sony HVR-Z1 camera are 960x1080 but the effective resolution is 1440x1080 after pixel shifting. The Panasonic AG-HVX200 has three 964x540 pixel sensors that after pixel shifting and pixel stretching result in an effective resolution of 1440x1080. These are of course, HDV cameras, not AVCHD cameras.

I did actually try and all I can find are effective resolution stats. And so did you.

It's a case of whether you are talking about single or three chip 1920x1080.

In Chris's case, he does state single chip - I suspect that B&H are referring to three chip cameras, each chip being 1920x1080.

In the single chip case, half the pixels will be green, and a quarter each red and blue. A debayering process yields a Y,U,V raster out of that. But even if the output is 1920x1080, the debayering will inevitably mean that the luminance resolution will be nowhere as good as if the 1920x1080 chip was simply used in a monochrome fashion, or if three 1920x1080 chips were used.

To get the resolution equivalent to three 1920x1080 chips from a single sensor, that must have a pixel count greater than 1920x1080, and then downconverted. A reasonable rule of thumb would be about twice the pixel count of the output raster, or in this case around 4 megapixel.