It's official: Canon XL2 announced

[Christopher Reynolds]

All I know is Im going to have to find someone who wants to buy a kidney, im still paying off my XL1s, along with my editing software suite, and my car!!!

[Greg Matty]

If the 16:9 footage looks at least as good as DVX-100 footage shot anamorphically, I think almost everyone will be satisfied. That may not result in a picture quality improvement, but it will eliminate the hassle of having to manipulate an anamorphic adapter. I guess we'll have to wait and see what the footage looks like.

Greg

[Nick Hiltgen]

<<As you know it
will always end up being 720x480 or 720x576.>>
Hey Rob maybe you should add that to the specs page, or create a seperate read only link that explains it, so you won't have to keep answering and explaining it. Also maybe include the info about how the 1.3x magnification works and what that means to the old xl1(s) user updating the camera body.

[Chris Hurd]

I mentioned both the 1.1x mag for 16:9 and the 1.3x mag for 4:3 in the second paragraph of my XL2 CCD Block Overview article. I suppose this needs to be worked into other areas of the site as well. I thought I had it in the XL2 FAQ; I'll have to re-check that page.

[Barry Green]

Originally posted by Greg Matty : If the 16:9 footage looks at least as good as DVX-100 footage shot anamorphically, I think almost everyone will be satisfied. That may not result in a picture quality improvement, but it will eliminate the hassle of having to manipulate an anamorphic adapter. I guess we'll have to wait and see what the footage looks like.

Absolutely. If it matches anamorphic DVX, I would be thrilled with it. The anamorphic adapter on the DVX can provide excellent results, but it takes work. If the XL2 matches DVX quality, gives real 16:9 at full res, and with none of the hassles, that's a mighty tempting combination!

Already I'm looking at it as a mini35 cam, since it's very difficult to get high-res 16:9 footage on a mini35/DVX. The built-in 16:9 on the XL2 could be just heavenly for mini35 footage.

[Greg Matty]

<<<-- Originally posted by Barry Green : Absolutely. If it matches anamorphic DVX, I would be thrilled with it. The anamorphic adapter on the DVX can provide excellent results, but it takes work. If the XL2 matches DVX quality, gives real 16:9 at full res, and with none of the hassles, that's a mighty tempting combination!

Already I'm looking at it as a mini35 cam, since it's very difficult to get high-res 16:9 footage on a mini35/DVX. The built-in 16:9 on the XL2 could be just heavenly for mini35 footage. -->>>

Barry,

Is mini35 a film format?

Greg

[Jay Gladwell]

Greg, mini35 is referring to an attachment (an expensive one, I might add) that goes on the front of the camera body and accepts prime lenses. You can read about it here:
http://www.zgc.com/zgc.nsf/c7a682995...2?OpenDocument

Hope this helps.

Jay

[Rob Lohman]

Nick: I'm working on something so we can link to one place that
describes it indeed. Thanks!

[hainline]

Has anyone seen the specs on the A/D converters used in the XL2?

Test shots to date, don't seem to include very wide lighting ranges.

The very anemic 9-bit units in the XL1 (I own one) were the greatest limit to shooting outdoors. Pray for cloud cover, large crew, scrim, and no wind ;-) One of the more sobering tests I ran (over 4 years ago) was shooting the same scene with an XL1, BetaSP, and an early Nikon D1. Tossing out the resolution advantage, the color and dynamic range advantage of the early (and quite faulty) Nikon D1 was exceptional, dusting both video cameras. 12 bit converters.

The reason one would want greater bit depth is that it enables greater dynamic range, or in film terms, greater lattitude. 12-bit is the current standard for Digital SLRs and newer video cameras like the SDX900. This is what let's one shoot in situations of greater contrast. (Think bright backlit windows.)

In the spirit of providing a little backround info (and no intention of diverting the thread). CCDs are analog devices. There is a voltage level for each pixel that varies depending on how many photons hit the cell (thus larger pixels are better in low light) for each frame's time slice. The pixels are serialized and clocked out, with voltage varying with illumination. This voltage typically adjusted a bit with the white balance circuitry.

Now this source has to be converted to a digital stream and further processed (frame building and pixel shifting in the XL1/s) This is where the A/D (analog to digital) converters are employed, one for each (R,G,B) chip. The analog voltage is converted into a 9 bit number. Then a custom ASIC crunches all the numbers and outputs what looks like, believe it or not, CCIR656 (8-bit uncompressed 4:2:2), (a little gem in hiding for the DYI crowd). This is then routed to another big ASIC that compresses this 20MByte/sec data stream down to the 3.5M/sec DV stream that is written to tape.

The DV data stream is an 8-bit standard. Just like the photo world's JPEG. However most camera's CCD can produce a wider range of values. This wider range of values is converted to 9,10,or 12 bits and then mapped down into an 8-bit range. However, a great deal of the dynamic range of the scene is maintained. This holds detail in the highlights among other things.

Now some of the dynamic range (lattitude) is a function of the CCD quality, but a good portion is a function of those old A/D converters. There is more to image quality than resolution.

As Bill Pryor said earlier: "It's all about lighting.". Better A/D converters get more of that light (range) down onto the tape.

The specs don't seem to mention this aspect.

Hopefully yours,

- Phil

[Aaron Koolen]

Thanks Phil, that's interesting and has made me ask a question I've been thinking about.

In in practical terms when talking about the dynamic range of a camera are we really talking about the level of banding that will be present in the final footage? I ask because to me, whether you have a 8bit, 12 bit or 128 bit a/d converter, your bright whites are always going to be the highest value and the blacks are going to be the lowest. So you will not "see" more into the dark, or the light. You will of course have more of a range in those light areas, and dark areas so you can physically see the detailed gradation. Am I correct in this, or does the CCD play more of a part than I think and so a high end camera would actually not blow out the whites like a cheapo one.

Aaron

[Jacques Mersereau]

<<<-- Originally posted by hainline : Has anyone seen the specs on the A/D converters used in the XL2?>>>

No, but you're right about a higher bit rate = better image. I too was
disappointed to hear the XL2 is only 8 bit. Doesn't Canon remember
the banding issues that are still common in 8 bit video? OTOH, I haven't
seen XL2 images and what they look like is what really matters.

<<<Test shots to date, don't seem to include very wide lighting ranges.>>>

I'm SHOCKED! ;)

<<<The very anemic 9-bit units in the XL1 (I own one) were the greatest limit to shooting outdoors. >>>

But if your up and shooting at dawn the video looks good :)

<<< Nikon D1 was exceptional, dusting both video cameras. 12 bit converters>>

Yup.

<<<Then a custom ASIC crunches all the numbers and outputs what looks like, believe it or not, CCIR656 (8-bit uncompressed 4:2:2), (a little gem in hiding for the DYI crowd).>>>

DYI crowd? Why in the HELL can't Canon port this output? Losers.
(they'll lose sales and we lose resolution!)
With some of the new video cards/boxes/decks, you could capture this signal
without a lot of trouble.

The other day, my boss had taken a 5 megapixel camera and did a time lapse
of our building's coffee shop using only existing light (glass wall).
He then made a crop action in photoshop and imported the wonderful wide screen
aspect ratio into quicktime. Man, how cool is having 3072x2048 pixels
to work with. Room to crop aspect ratio and pan and scan too.

The clarity, color and contrast handling combined with a 1:2.35
aspect ratio made amazing (albeit studdering) video.

[Bill Pryor]

Dynamic latitude is important, but without seeing the camera and doing some tests in real world lighting conditions, you really can't know all that much about it. Specs are important, but they don't tell everything. I'm going to hold off my judgement until I actually get to see some footage from the camera.

[hainline]

Bill, you're quite right regarding the real world footage. We are just nattering on about what might be possible, based on the limited info that has been available. I was just curious.

Certainly chips have moved on and Canon will used whatever made economic sense. I only get excited about this package because you can mount an Mini35 on it (in a reasonable fashion, as opposed to the DVX, which works well but is no fun to use)

>> DYI crowd? Why in the HELL can't Canon port this output?

Jacques is referring to my comment that in an XL1/s there is a buss that contains a CCIR656 (8-bit uncompressed 4:2:2) signal. There is a chip that converts this buss standard into SDI (you can't really plug a CCIR656 signal into anything we use). Circuit board real estate, power, and most expensive, another jack on the camera (BNC).

Like most companies, Canon looks at the number of people who might get excited about this idea (small hundreds, at best) and yawns. Direct to disk, uncompressed, in the field, is only of interest to poor startup filmakers and green screen affectionados. I have great empathy, but the numbers just are not there. At least Canon doesn't have a broadcast line of cameras to protect.

Back in the days *before* the P+S Technik PRO35, when the mini35 was just out, I was excited about direct to disk of this signal. The electronics are pretty easy, but the mechanical aspects could be a little daunting. The inside is all state of the art surface mount. Beautiful miniaturization, cute little sub-assemblies.

As for Jacques other comment:
>> I too was disappointed to hear the XL2 is only 8 bit

At little clarification is in order. The DV tape standard is 8 bit. So are many higher standards like DVCPRO50. So is JPEG. There is no problem representing an image with 256 levels of RGB or YCC, more accurately YCbCr. Capturing the image in a world that often has geater dynamic range is where the issues crop up. I'm sure the XL2 is better than 8 bit. The XL1, as I mentioned, was 9 bit.

Which brings me to Aaron's question:
>> so a high end camera would actually not blow out the whites like a cheapo one

Yes, more money does make a difference. It is both the CCD and the A/D converters that let one capture a high dynamic range. That bigger range (for example 12 bit) then maps down just fine into 8 bit. Typically without banding if properly exposured. Bigger CCDs help in low light, better A/D converters help with the highlights. A gross oversimplification, but basically valid. In the still world Fuji is working a sensor with two sizes of pixels, large ones for the shadows, and small ones for the highlights. This is hard stuff.

Jacques:
>> Doesn't Canon remember the banding issues that are still common in 8 bit video?

More of these problems crop up in heavy post manipulation in an 8-bit world. On the camera, a better A/D converter will catch the great variance of light. Then the onboard chips can then compress (a different kind than the DV compression) the signal down into the 8 bit range. Done successfully all the time.

There is also the desire to hang onto more of that dynamic range into the post process. In the STILL photo world, most pros are moving to shoot RAW files, where you have all 12 bits of data. Post-shutter white balance, push and pull a stop or two. In the high end video world you have the THOMSON VIPER Filmstream Camera where the data is 10-bit log.


One closing note on all these consumer tradeoffs. When you look at the schematics, you see how many chips and engineering (!!!) go into making the consumer ("one little red button") experience work. Feels like 80% of the cost. So much goes into auto-focus, tape control, compression, etc.

I too yearn for a $4k block camera, with 10-bit SDI out, progammable analog gamma, with a lense mount. Yes it could be easily built, but not for the very few would be interested. Funny thing about all this volume efficiency manufacturing.

So we go on discussing where a sweet spot might be that we can use.

I'll just wait for some challenging real-world lighting examples to show up.

Have fun. -Phil

[Charles Papert]

<<I only get excited about this package because you can mount an Mini35 on it (in a reasonable fashion, as opposed to the DVX, which works well but is no fun to use)>>

Care to elaborate on the no-fun issue?

[Andre De Clercq]

Dynamic range of a camcorder CCD is not determined by the A/D conversion bit depth. Only the grayscale resolution and thus the performance of knee processing, black stretching, matrix processing, WB precision... is determined by the bit depth.