Chris-
I would add that if a new camera arrives next year with a full size chip, aside from figuring out the lens situation, I would hope Canon would have adapters availible to use film prime lenses i.e. PL Mount. Wouldn't that be nice...

Jim Martin

ps- How's things in the Lone Star state? Still a little smokey here in LA.

And for a VIDEO camera, I'll reiterate my plea for NO rolling shutter artifacts. For artistic work, one can generally work around it. For some documentary work of fast moving objects/events, it is an unacceptable limitation. As I told the Canon reps at NAB last spring, other nice HD cameras already do rolling shutter. Don't be "me too." Advance the the technology and put out a larger sensor, very low light video camera withOUT rolling shutter and you won't be able to build enough of them.

Chris-
I would add that if a new camera arrives next year with a full size chip, aside from figuring out the lens situation, I would hope Canon would have adapters availible to use film prime lenses i.e. PL Mount. Wouldn't that be nice...

Jim Martin

ps- How's things in the Lone Star state? Still a little smokey here in LA.

So I am hoping this camera will actually be APS-C sized for this very reason!

I'd second the APS-C size sensor. And hopefully the Canon 7D is an indication they're moving in this direction. It's going to be a lot easier to build a decent, lightweight and affordable lens for this sensor size rather than a FF 35mm sensor. Focusing would be a lot easier too.

I like shallow DOF as much as the next person, but from a practical standpoint there is no way I want to attempt to focus in a run and gun scenario with a FF 35mm sensor. The indie filmmaker market isn't the only market out there. And personally, I think they're a little too obsessed with shallow DOF. For me APS-C or 4/3 is the ideal balance of DOF control, light sensitivity, lens compatibility and ability to focus.

Whether the sensor size is FF or APS-C, even a fast 10x lens will be a seemingly expensive proposition. I

While it doesn't have a servo zoom on it. The stock lens on the Panasonic GH1 seems like a reasonable compromise of quality and functionality. I would think it should be possible to make something like that for APS-C size. You could add a digital 2X extender assuming the chip would have some resolution to burn. So that would give you 28mm-560mm out of the box.

I hope Canon suprises us and develops a 3-chip APS-C sensor for the camera. A single chip, bayer filter camera seems so retro.

To me, three-chip is even more retro than single-chip... these days there's no significant difference between them in color accuracy like there used to be. A single-chip design can be larger and less expensive than a three-chip design.

Chris, that's true for 4:2:2 acquisition. But we really don't have a lot of examples comparing 3-chip 4:4:4 to Bayer 4:4:4 for the same size chip.

While it doesn't have a servo zoom on it. The stock lens on the Panasonic GH1 seems like a reasonable compromise of quality and functionality. I would think it should be possible to make something like that for APS-C size. You could add a digital 2X extender assuming the chip would have some resolution to burn. So that would give you 28mm-560mm out of the box.


Something that builds on this as a foundation?

Canon | 28-300mm f/3.5-5.6L IS USM Autofocus Lens | 9322A002AA (http://www.bhphotovideo.com/c/product/319784-USA/Canon_9322A002AA_28_300mm_f_3_5_5_6L_IS_USM.html)


I hope Canon suprises us and develops a 3-chip APS-C sensor for the camera. A single chip, bayer filter camera seems so retro.


Do you realize just how huge that total sensor block would be?

Ok it's Area 51, but I think this is going too much along the lines of "what I'd like to see" and not enough of the lines are "what's likely to happen? What would make business sense for Canon?"

And my gut feeling is that the next step is more likely to be something along the lines of the EX or the HPX301 - CMOS, 1920x1080 three chip, and 1/2" 0r 1/3". (And I hope 1/2") But we'll see.

In a way, it's the codec and media choice I see more interesting, and the more I think about 50Mbs MPEG2 on to Compact Flash, the cleverer a move it would be. Think about it. Cheap media, unlike SxS or P2. Better codec than the EX or the JVC cameras, giving 4:2:2. Better quality, and none of the editing problems of AVC-HD.

The other thing I would like to see are all three main formats (and the 60Hz equivalents): 1080i/25, 1080p/25, and 720p/50.

... Cheap media, unlike SxS or P2. Better codec than the EX or the JVC cameras, giving 4:2:2. Better quality, and none of the editing problems of AVC-HD.

The other thing I would like to see are all three main formats (and the 60Hz equivalents): 1080i/25, 1080p/25, and 720p/50.


#1 is reasonable if you assume Canon does NOT build on what it has done in the 5DM2 and 7D. Otherwise, you are out of luck.

#2 All those formats I believe are in the 7D. So the remaining missing piece is the video camera itself, and the appropriate lenses.

#1 is reasonable if you assume Canon does NOT build on what it has done in the 5DM2 and 7D. Otherwise, you are out of luck.
My assumption is that an "XL-H2" is more likely to build on an XL-H1 than a 5D2 - I regard the latter as a (very good) still camera with a video mode. In that respect, the move to higher than HDV quality would be answering what Sony have done with the EX and JVC have done with the HM700.

H264 would be one answer, but that is difficult to edit natively. It's no better quality than MPEG2 necessarily, just achieves a given level at a slightly lower bitrate. In a camera like this, MPEG2 at 35/50Mbs seems a good compromise - relatively easy to edit, high quality, but still a low enough rate to fit on SDHC/CF cards. Going to 50Mbs wouldn't just yield 4:2:2 - it would put clear marketing water between it and Sony and JVC at 35Mbs.
#2 All those formats I believe are in the 7D. So the remaining missing piece is the video camera itself, and the appropriate lenses.
Well, not quite - I don't think the 7D does 1080i/25, and I didn't think the XL-H1 did 720p/50. I'd see the ability to record 1080i/25. 1080p/25 and 720p/50 (as on the EX) as a very good thing.

Well, not quite - I don't think the 7D does 1080i/25, and I didn't think the XL-H1 did 720p/50. I'd see the ability to record 1080i/25. 1080p/25 and 720p/50 (as on the EX) as a very good thing.


You don't have to take my word for what the Canon 7D can do.
Have a look at this web page, click on the Features tab, and have a good read.
The format+framerates the 7D will do are right at the top.

Canon EOS 7D Digital SLR (http://www.usa.canon.com/consumer/controller?act=ModelInfoAct&fcategoryid=139&modelid=19356#ModelFeaturesAct)

Have a look at this web page, click on the Features tab, and have a good read.
Yes, and from that it's as I thought:
selectable frame rates: 1920 x 1080 (Full HD): 30p (29.97) / 24p (23.976) / 25p, ..........{plus other frame sizes....}
no 1080i/25.

At 1920x1080 you just get 30p, 24p, and 25p.

Considering video capable DSLRs, their rolling shutter when it comes to their CMOS sensor is absolutely horrid. Using live view on the D90 and the Canon, I was getting huge amounts of rolling shutter artifacts. When it's actually recording, this would never be able to be used for matchmoving...

If Panasonic's AVC-Intra is a sign of what's to come, the consortium that made HDV should make a professional tapeless format that isn't consumer centered like AVCHD is.

Once that codec is out, a XL-H2 could be possible.

Well my feeling is with this supposed XL H2, they will probably go with CF cards and an H.264 codec very much in line with what's coming from the 7D. Of course, it'll have HD-SDI, or hopefully 3G HD-SDI for 4:4:4.

What will make this camera amazing besides the tech being built for video in mind (a specific APS-C sensor that will try to minimize/eliminate CMOS cons?), is the fact that it'll be a camera which focuses on video first.

I would think it's form factor will be somewhere very close to what the XL series is now.

......... the consortium that made HDV should make a professional tapeless format that isn't consumer centered like AVCHD is.
Isn't that what 35Mbs or (even better) 50Mbs MPEG2 are? The latter (as XDCAM 422HD) has (like AVC-Intra 100) gaining full acceptance for future broadcast acquisition, according to recent EBU tests. Based on HDV (as MPEG2), but a much higher quality, and giving 4:2:2 in the 50Mbs version.

For a prosumer camera, 50Mbs means easily recordable onto fairly cheap memory cards.

Why bother with H264? It won't necessarily give any better quality than MPEG2, just give the same at a slightly lower bitrate. Against that is lack of support from NLEs, and a likely need to transcode before editing.

"During this fall, Canon will make one of its largest and most important launches in the company’s history and we hope you can attend. During a luncheon, you are invited to listen to CEO Jouko Tuouminen, Marketing presiden Monica Forsberg and trend analyst Magnus Lindkvist, talking about tomorrows trends and factors of success in the continually more digitalized everyday life."

Hmmm ... could this finally be the video camera we're waiting for? Or a 1DMark4?

Or a new product line like Vacuum Cleaners with Optical Image Stabilization?

:)

A Mark7 XLH2 vacuum with OIS? But which codec?

The latest gossip from IBC is that the September Announcement is regarding a 21MP video camera. Canon Europe have the Business Design Centre in Islington Booked for a exhibition in early October...

talking about tomorrows trends and factors of success in the continually more digitalized everyday life."


You need a 21MP video camera for your 'digitalized everyday life'?

You need a 21MP video camera for your 'digitalized everyday life'?

Ughh Yeeeeeeah ....

:)

Everybody at Canon in IBC was tight lipped about the Sep29 event but I got a hint that is about a video camera launch. On the other hand I cannot understand why they didn't launch it at IBC where it would have been more apropriate. Anyway if it is APS or bigger it must be Cmos because CCDs in small form cameras become very hot (thats why Sony put only a 1/2inch chip on EX1/3). If the 21Mp rumor holds then probably is FF. Ideally it will be FF 4/3 with the option for 16/9 APS

... Anyway if it is APS or bigger it must be Cmos because CCDs in small form cameras become very hot (thats why Sony put only a 1/2inch chip on EX1/3). If the 21Mp rumor holds then probably is FF. Ideally it will be FF 4/3 with the option for 16/9 APS

Emmanuel,

I'm not disputing that CCD is hotter than CMOS, but FYI the EX1/3's have 1/2" CMOS's not CCD's. BTW, Sony's F35 has one CCD, their F23 has three CCD's , Dalsa used a CCD, and I believe the Viper uses CCD's as well.

Maybe I wasn't clear. I meant that Sony decided to put a 1/2 CMOS instead of 1/2 CCD in a SMALL FORM camera like EX1/3 because such a small camera cannot deal with the heat emitted from the CCD's. SONY in the much bigger 1/2 3XX line of cameras, uses CCD. In the same manner all the cameras you mentioned (viper, dalsa, F23, F35), all are large cameras that can deal with CCD's heat.

I didn't make that out of my head. It was the response of a SONY tech in last year's IBC on why they used CMOS instead CCD in the EX1/3.

In a way, it's the codec and media choice I see more interesting, and the more I think about 50Mbs MPEG2 on to Compact Flash, the cleverer a move it would be. Think about it. Cheap media, unlike SxS or P2. Better codec than the EX or the JVC cameras, giving 4:2:2. Better quality, and none of the editing problems of AVC-HD.

That about sums it up for me except one other point, form factor. Chris related to this in the second post on the thread. I don't care what Canon calls the new offering but please, please let it be SIMILAR to the XL line in form factor. It needs to be a true shoulder mounted camera, not the form factor of the EX3. It needs to also have at least a 16x or 18x lens, 20x would be my vote. 10x and 12x lenses don't cut it in my line of work and are deal killers.

Since I believe this thread is about completely wild speculations here's my guess.

The XH-EF will be the next XL body style released from canon. It will in effect have the same sensor as the 7d size wise (aps-c) but I believe it will be less mega pixels (let's say 10) to give it good low light performance. I think it will have adjustable frame rates up to 60fps with the option to do 1080p instead of 720 (like the 7d) The stock lens will be an APS-C sized lens designed to be able to shoot wider as opposed to longer. There will not be a new series of lenses to go along with it, perhaps a second motorized zoom, either wider or longer but that is all.

The biggest improvement people will talk about is the viewfinder with something akin to the 5d-7d zacuto setup. the outputs will have an option for HDMI or HDSDI, hdmi being a grand cheaper. It will record to some media (CF, SXS, SDHC whatever) in a format that is use able and high quality but not readily inject-able into most editing systems. it will start at either 7500 or 1000 more then i can get if i sell all my current camera equipment, whichever is more. It willl be announced in late October early November and ship by the end of December.

Don't u understand. Canon sells the xl h1 now with a lens for 6k. If they come out with an xl h2 with a full frame body the price would be very very high. That is why the 7d and 5d provide such a value now. Another issue that arises is the problem of lenses, canon would probably have to make a new lens line. I don't see this coming, even the red scarlet has an advertised 2/3' sensor not s35 (7d) or full frame (5d). Now 2/3' broadcast cams cost 15k+. And alot of you are expecting 4:4:4 color space and what not. Maybe canon will make a new line of cams that have aps c sensors or 2/3 but its not going to be the same price the xl h1.

Isn't that what 35Mbs or (even better) 50Mbs MPEG2 are? The latter (as XDCAM 422HD) has (like AVC-Intra 100) gaining full acceptance for future broadcast acquisition, according to recent EBU tests. Based on HDV (as MPEG2), but a much higher quality, and giving 4:2:2 in the 50Mbs version.

For a prosumer camera, 50Mbs means easily recordable onto fairly cheap memory cards.

Why bother with H264? It won't necessarily give any better quality than MPEG2, just give the same at a slightly lower bitrate. Against that is lack of support from NLEs, and a likely need to transcode before editing.

I'm talking about a codec that can survive passes/generations of recompression, in the end the best looking even on heavily compressed HD cable. MPEG-2 is horrible when it comes to this so naturally, AVC-Intra is gaining ground. Just remember, not everyone has access to CineForm.

I'm talking about a codec that can survive passes/generations of recompression, ............
Can I refer you to this - http://www.dvinfo.net/forum/general-hd-720-1080-acquisition/137934-ebu-hd-acquisition-codec-tests.html - thread? And in particular the EBU document I refer to? Specifically referring to their main conclusion that : "All tested codecs have shown quasitransparent quality up to at least 4 to 5 multi-generations,.........
". That includes 50Mbs MPEG2 and AVC-Intra 100.
MPEG-2 is horrible when it comes to this so naturally, AVC-Intra is gaining ground.
Comparing codecs makes no sense without defining bitrates, and you're also ignoring the ADVANTAGES that MPEG2 has over other codecs. Practically, some codecs are better for some applications, others in other cases. MPEG2 quality may indeed be "horrible" for SD transmission at 2Mbs, but far from it at 50Mbs. Same with H264 - that's really horrible when used for mobile phone compression!

MPEG4 type codecs are actually based on MPEG2, they do all that MPEG2 does, but have the ability to use other features as well. If a coder was to use all the features, it may (theoretically) achieve comparable quality to an equivalent MPEG2 one at approx half the bitrate. Practically, at the moment, it's unlikely to achieve that quality at much less than about 75% (ball park figure only) of the MPEG2 figure for what we're talking about here. This all assumes equal GOP length etc.

Hence, yes, you could say that an H264 based codec at about 38 Mbs could give equivalent results to an MPEG2 version at 50Mbs. But is it worth it? The price to pay for the efficiency is a codec that is far, far more difficult to edit natively than MPEG2.

Go to AVC-Intra and it's the opposite argument. You need an even higher bitrate to give the equivalent quality to MPEG2 - hence the EBU equating it at 100Mbs to MPEG2 at 50Mbs.

Apply all that to this new hypothetical Canon camera and if you were the designer, you have to decide where you make the compromises. If I was in that position, then 50Mbs MPEG2 onto Compact Flash would make sense. The bitrate is low enough to fit onto even cheap CF easily, give relatively small file sizes, yet be high enough to be considered a "true" broadcast codec, as approved fully by the EBU.

H264? Yes, it would be even more efficient in bitrate terms, but my opinion would be that the editing issues wouldn't be worth the further bitrate saving. This is the conclusion that many manufacturers have come to - Sony, JVC, Ikegami, and Convergent Design to name four. If I were designing a consumer camera, or one for the absolute top end, it would be a different story.

Don't u understand. Canon sells the xl h1 now with a lens for 6k. If they come out with an xl h2 with a full frame body the price would be very very high. That is why the 7d and 5d provide such a value now. Another issue that arises is the problem of lenses, canon would probably have to make a new lens line. I don't see this coming, even the red scarlet has an advertised 2/3' sensor not s35 (7d) or full frame (5d). Now 2/3' broadcast cams cost 15k+. And alot of you are expecting 4:4:4 color space and what not. Maybe canon will make a new line of cams that have aps c sensors or 2/3 but its not going to be the same price the xl h1.

I don't think the camera will cost more then 10k, and I stand by my guess of 7500, which for a camera similar to this would be about right I think.

Canon manufactures thousands of different lenses for (my guess) just as many applications. Saying that they will hit a stumbling block in adding a zoom function to an already existing APS-C lens seems just silly to me. Additionally this won't be a "broadcast" camera. Canon is also significantly larger then RED, and to be honest not as interested in pushing he development of new technology, but if they can take the sensor they're using for the 7d and drop it into the old xl body and sell more by doing less, then that's a good model.

I think the 7d and 5dm2 are good value but they still have a long way to go before they offer the same features as the xl- body cameras.

I think the XL style body is a pain. I'd much rather see something resembling JVC's small shoulder mount cameras from the HD100 onwards or the HPX300. I'd prefer V Lock battery plates, but maybe that could be an option, similar to how it was added to the JVC's HD100 line. Proper form factor, a bit of weight and a decent OIS would be great for handheld.

In order to achieve a greater zoom range, they could use an over sampled sensor and perform a digital zoom without any loss of resolution. That could turn an 8x into a 16x for example. It would be a digital 2x extender.

APS-C is pretty large for video, I certainly think FF like the 5D would be too hard to keep in focus consistently. maybe 4/3" would be a good compromise between DOF and focus.

An excellent VF is a must! At least on par with the EX3.

Since I believe this thread is about completely wild speculations here's my guess.

The XH-EF will be the next XL body style released from canon. It will in effect have the same sensor as the 7d size wise (aps-c) but I believe it will be less mega pixels (let's say 10) to give it good low light performance. I think it will have adjustable frame rates up to 60fps with the option to do 1080p instead of 720 (like the 7d) The stock lens will be an APS-C sized lens designed to be able to shoot wider as opposed to longer. There will not be a new series of lenses to go along with it, perhaps a second motorized zoom, either wider or longer but that is all.

The biggest improvement people will talk about is the viewfinder with something akin to the 5d-7d zacuto setup. the outputs will have an option for HDMI or HDSDI, hdmi being a grand cheaper. It will record to some media (CF, SXS, SDHC whatever) in a format that is use able and high quality but not readily inject-able into most editing systems. it will start at either 7500 or 1000 more then i can get if i sell all my current camera equipment, whichever is more. It willl be announced in late October early November and ship by the end of December.

Nick-

I like the way you think....

Jim Martin

jim-

I hope canon likes the way i think too...

Mostly because I have a feeling I'll be in the market for a new camera in the next couple of months, but only if they release something like this. When they announced the xl-h1 and the h1s where and when were those announcements made?

I have a dream that one day canon will release a 35mm ef to xl adapter and I'll have iris control and a 35mm dof of all my ef lenses, and it will be awesome...

Be careful what you wish for :)

I believe Canon will announce the Canon EOS 1D Mark IV next week. And I do think there will be some special video features. What I really hope is that Canon will announce an IO box with the formfactor of the battery grip with two XLR inputs and a few controls that are typical on videocameras. We'll just have to wait and see.

jim-

I hope canon likes the way i think too...

Mostly because I have a feeling I'll be in the market for a new camera in the next couple of months, but only if they release something like this. When they announced the xl-h1 and the h1s where and when were those announcements made?

I think, and I'm just speculating here, there could be something in the 1st quarter of next year or at NAB.....again, it's just a guess. I'm also speculating that I'm going to win the lottery in the second quarter of next year and my screenplay gets picked up.

Jim Martin

Resurrecting an oldie (but a goodie).


Think: if the sensor size is 35mm full frame, as it is on the EOS 5D Mk.II, what kind of video lens would that require, and at what cost? Keeping in mind how large in circumference the front objective will have to be, and that the faster and wider the lens, and the higher the zoom ratio, the larger diameter that glass will require and therefore the more expensive it will be.


The EF lens does not have to be heavier or more expensive at all, for one very important reason: f-number scales with sensor size. Check out this other thread for the full explanation:

http://www.dvinfo.net/forum/canon-eos-7d-hd/465482-smaller-sensors-do-not-have-deeper-dof.html

The XL-H1s lens is 5.4-108mm. In order to have the same angle of view as the XL-H1s lens, an EF lens on a FF camera would need to be 32.5-650mm. Sounds pretty heavy, right? No. The reason is f-number. IN order to get the same DOF, light gathering power, and diffraction as f/1.6-3.4 on the XL-H1s, the EF lens only needs to be f/9.6-20.4. That's right: f/20! A 650mm f/3.4 would be quite heavy, but 650mm f/20 is very lightweight.

It may be surprising that f/20 on 35mm can really match f/3.4 on 1/3", but it's true. As explained in the other thread, it's easy to test for yourself. Compare the 7D and 5D2. If you shoot them both at f/11, then the 5D2 will have thinner DOF, less noise, and less diffraction. But what if you multiply the f-number by the crop factor? f/11 * 1.6 = f/17.6. Now compare the f/11 on 7D vs f/17.6 on the 5D2. Well, now the DOF is the same and diffraction is the same. But the 5D2 image will be too dark: increase the ISO, and they are the same. What may be surprising is that noise, too, is the same.

Of course, one of the big advantages of the XL-H1s is 3-chip. That gives it three times the light gathering capability of a single chip, without making DOF thinner. That would give it a low light advantage over a full frame lens set to the same DOF.


Also consider how expensive Canon's L-series of fast telephoto EF primes are. Then you'll get some indication of how the real limiting factor of an XL replacement isn't the sensor so much as it is the lens. The practicality of the lens requirements -- that it'll have to be motorized, will most likely include IS, will need a field of view at the wide end comparable to the mid-thirties in 35mm, and have some kind of usable zoom ratio of at least 5x or so, and not throw the camera cost off balance -- are the factors which will ultimately dictate the size of the sensor that Canon selects.

None of those are a problem if the f-number is scaled with the sensor size. Canon can make a 32-650mm f/10 - f/20 that has the same size as capability as 4.5-108mm f/1.6-3.4 on a 1/3" Bayer chip. Of course, it wont be any better: the low light performance will be just as dismal. f/1.6 on the 1/3" Bayer will have just as much noise as f/10 on the FF Bayer. If you want better low light performance, you have to have a f-number that is faster than f/10. The only way to do that is to make the lens heavier, or smaller zoom ratio.

The real problem is Marketing. Photographers are not aware that f-number scales with sensor size. They think f/1.6 on 1/3" is better in low light than f/11 on 35mm, when in fact they are the same for noise. If Canon really did come out with an f/11 - f/20 20X video zoom for 35mm, no one would buy it, even if it did allow them to get the exact same shots as their f/1.6-3.4 zooms. Even a f/8 or f/5.6 zoom would be a hard sell. But I would really like to see it happen, so I hope that Canon can somehow find a way to educate enough customers for it to sell well.

I agree with much of your reasoning, and especially about f-numbers, though feel it important to add that it assumes the same number of pixels on the chips under comparison. In that case, a 35mm sensor will have bigger pixels than a 1/3" one, hence each pixel need less light, hence a smaller aperture for the same performance.

Another way of putting it is to say that for a given technology, the low light performance is defined by the chip resolution and the diameter of the lens front element.

The only thing I wonder about is whether an f20 650mm lens really would be the similar in size/weight to an f3.4 108mm. True, we would be talking about the same size front element for all the reasons you explain, but would the greater focal length not inherently make the 650mm lens bigger and heavier?

I agree with much of your reasoning, and especially about f-numbers, though feel it important to add that it assumes the same number of pixels on the chips under comparison.


Actually, my position is that it doesn't assume that: 4 small pixels yield the same noise (when taken together) as one large pixel. For example, if you crop the 5D2 down to the same size as the 7D, you are left with an 8MP vs 18MP. The 5D2's pixels have over twice as much area. Yet the noise is the same. I could go into this topic in much more detail, but we really should save it for a thread that is specifically dedicated to pixel size.


The only thing I wonder about is whether an f20 650mm lens really would be the similar in size/weight to an f3.4 108mm.


It's hard to draw on specific examples because most photographic lenses are designed for phase detect autofocus which requires f/5.6 or faster. But there are a few more examples than the ones I already gave. Like how the Canon 400mm f/2.8 adapted for Four Thirds weighs about the same as 800mm f/5.6 on 5D2 (and both have the same angle of view and depth of field). You can compare size of Medium Format lenses with 35mm lenses of similar design to see they are about the same weight. Refractor telescopes of a similar design also tend to weigh the same as long as f-number scales with focal length.


True, we would be talking about the same size front element for all the reasons you explain, but would the greater focal length not inherently make the 650mm lens bigger and heavier?

It will certainly make it inherently bigger (longer), but I'm not so certain that it will make it heavier. I am not a lens designer, but the little I have read and seen seems to indicate that longer, slower lenses can weigh the same as shorter, faster ones.

Daniel,

I'm sure you would agree that it would border on absurd to make a large sensor camera w/ a very slow lens so it could mimic the performance of a smaller sensor camera. While it's possible to do, it makes little sense.

I do agree that w/ your statements about three chip designs. They don't get a lot of love around here it seems, but I think that's mostly b/c most of the acquisition is 4:2:2, so the extra chroma resolution (as compared to a Bayer pattern) gets lost. And people seem to forget that a mask allows only 1/3 of the light to reach any one photosite (I am unfairly assuming that there is an equal proportion of R, G and B in the composition).

Three chips are more complicated and heavier, but they do have the advantage of truly co-sited color. I also believe the chromatic aberration that is more prevalent in three chip designs has to do with the fact that the color resolution is finer than in masked designs. So how the diffraction pattern varies ever so slightly for different wavelengths of light becomes more apparent. On the other hand, I think the demosaicing process by it's very nature softens the resolution and blends the color info. thereby acting as a defacto CA filter. BTW, the HVX has a built-in electronic CA filter, which works quite well. So there are ways to remove CA in three chip cameras during acquisition.

Anyway, I went off on enough of tangent. I, like everyone else, am waiting to see Canon's latest offering.

Imagine the issues that having a sensor like that size will cause for event videographers or for anyone who need deep depth of field.


The 3-chip cameras will always have an advantage in deep DOF over single chip cameras of the same size because they have more light gathering capability. More light means they can stop down further and increase gain higher for the same amount of noise.

However, when you compare a large single-chip camera vs. a small single-chip camera, there is no advantage in deep DOF. That is because the large chip can stop down for DOF and increase ISO to get the same brightness, as discussed here:

http://www.dvinfo.net/forum/canon-eos-7d-hd/465482-smaller-sensors-do-not-have-deeper-dof.html

In my opinion, with the 7D shipping in less than a month and the 5D Mk. II having been out for nine months, the reason why we haven't yet seen a dedicated video camera equivalent from Canon is the issue of the video lens.

That's the least of their worries. Canon already has all the resources it needs to build a FF 4X video lens. All it needs is a market.

Personally, I think the reason why we haven't seen a FF Canon video camera is the same reason that the RED ONE weighs over 10 pounds: 9 MP at 24 FPS and takes a lot of hardware. How much more for 21 MP at 30 FPS? It's a data rate of 8,800 Mbps! Of course, the 5D2 side-steps the problem by just skipping 2 out of every three rows and binning 3 pixels. That gets it down to just 980 Mbps, which a single DIGIC IV can handle.

The result, of course, is the most terrible aliasing problems ever to disgrace the Canon brand name. I think the only reason Canon even let it out the door was because they thought quality didn't matter for PJ web videos. If they have higher standards for actual video cameras (and I would like to think that they do), then they'll delay the release of the camera until they can address the aliasing problem.

I hope Canon suprises us and develops a 3-chip APS-C sensor for the camera.


There is a good reason why no 3-chip camera is larger than 2/3": the lenses are far too expensive due to the lens design necessary, such as extremely deep back focus. (They're bad enough for 2/3". I'd hate to see how much it costs for an APS-C 3-chip lens.)

Furthermore, 3-chip has terrible color accuracy. The shape of the spectral response of color filter array sensors can be finely tuned at every wavelength, whereas dichroic beamsplitters always have color overlap crossover at exactly 50%, and it can't be changed because it's the very nature of a prism. Their delta-E is orders of magnitude worse than Bayer for certain colors.

A single chip, bayer filter camera seems so retro.

It doesn't give me that feeling at all. 3-chip will always have the advantage of gathering three times as much light without making DOF thinner (something no single-chip can do), so I think it will always have a place in low-light deep-DOF applications.

Thanks for the response, Peter.


I'm sure you would agree that it would border on absurd to make a large sensor camera w/ a very slow lens so it could mimic the performance of a smaller sensor camera. While it's possible to do, it makes little sense.


Not at all! Quite the contrary, in fact. First of all, it will only mimic the performance (noise) of smaller sensor cameras when in low light. In fact, it would be much worse if they were upgrading from 3-chip to 1-chip. But in ample light, users will get the full benefit of the larger sensor, including far less noise, more dynamic range, etc. In fact, they will not ever have to use their ND filters, a nice side benefit. Not to mention the flexibility of switching lenses. On 1/3", you can't switch your f/1.6 for an f/0.1 to get better low light. But on FF, you could switch your f/11 for f/1.4, gaining a massive low light and thin DOF capability.

Second, there is no free lunch. I'm responding to the idea that it's impossible to make FF35 lenses that match the 1/3" lenses. It is possible, but one has to accept that they will only be the same, not better. If one is not content with the current low light performance on the 20X 1/3" lens, then there is nothing to be done about it without accepting higher cost, more weight, or smaller zoom range.

Third, I think there will be a bit of lens design advantage, since the 3-chip lenses have more considerations (back focus, etc.) than a mirror-less (E.V.I.L.) 35mm lens. Furthermore, to achieve the same contrast in the final image, the FF lens only needs 50% MTF at (say) 20 lp/mm, compared to much higher spatial frequencies on 1/3", say, 120 lp/mm. Of course, the larger image circle balances this out somewhat, but one thing that doesn't balance out is manufacturing tolerances, which have to be much tighter for the same MTF.

Fourth, the only sense in which I think it is absurd is the marketing challenge. It's the same problem as other industries. In astronomy, many people think the magnification of the scope is the most important factor in the buying decision, so they think a 1000X refractor (50mm) is better than a 30X newtonian (250mm). In fact, the 30X newtonian is an order of magnitude better. In the same way, camera buyers are now fixated on f-number as the most important factor. They think f/1.6 on a small sensor will give them a better picture than f/8 on a large sensor. In fact, the f/8 provide a far better picture in low light.

Since 3-chip gives a 1.6 stop advantage over 1-chip, it would be poetic for the f-number to be 1.6 stops faster, so that the user could at least do the same in low light (though it will mean accepting thinner DOF). Ideally, the lens will be as fast as they can make it without sacrificing angle of view or zoom range, while still hitting a price that is commensurate for electronics in the rest of the camera. Maybe they could get it as fast as f/11 or even f/8.

But Daniel,

You are ignoring the fact that the main reason people want large sensors is for shallow DoF. Better low light performance and less noise isn't going to be enough if the DoF gets no shallower than you'd get from a 1/3 or 1/2 chip.

It's like building a Ferrari and limiting it to 85 MPH. Nice leather, beautiful paint, faster 0 to 60 than a Lexus, but if it can't sniff 185 MPH I don't want it. It doesn't matter that I might never floor the car for a 1/4 mile, I just want to know that if I do, I'll be going 120 MPH.

And in truth, shallow DoF has been the trend in cinematography now for quite some time. So a low F stop couples w/ a large sensor is more than just a vanity issue. The deep focus shots we see in "Citizen Kane" and "Raise the Red Lantern" are seldom used now.

Great post, Peter.

But Daniel,

You are ignoring the fact that the main reason people want large sensors is for shallow DoF.


That's only because large sensors, historically, have been prohibitively expensive. As they drop in price, people will use them for everything, not just shallow DOF.

For example, if a vidoegrapher goes to the video store, and the XL-H2 (1/3") is $8,000 and the XL-5D (36x24mm) is $8,000, and they both have a 20X zoom with the same DOF, it would be perfectly acceptable to choose the XL-5D, even if the videographer doesn't plan on using it with thin DOF. He'll get the huge advantages when in ample light (far better image quality than 1/3"). In fact, if Canon could make them both for the same price, they might as well not bother even making the XL-H2, since the XL-5D could do everything it can, and more. (The trick is convincing the customer that 650mm f/11 really is just as good as 108mm f/3.4.) Other customers can buy the exact same camera, but instead of a 20X f/11 zoom, get the 4X f/4 (say, 28-135). Still others can go with 3X f/2.8 zooms or f/1.4 primes.

I guess my point is that it's possible to "first, do no harm." If Canon came out with a new video camera that did not have a 32-650mm zoom, then the previous owners that depended on that functionality could not upgrade. If they kept 20X but added 10 pounds to the camera (by making it f/5.6), that too would hinder many buyers. But if they preserve the existing light-weight 20X functionality (and they can), then everyone can upgrade. Not that they will: just that they can.

...

Furthermore, 3-chip has terrible color accuracy. The shape of the spectral response of color filter array sensors can be finely tuned at every wavelength, whereas dichroic beamsplitters always have color overlap crossover at exactly 50%, and it can't be changed because it's the very nature of a prism. Their delta-E is orders of magnitude worse than Bayer for certain colors.

...

Wouldn't a combination of both color filters placed on the R, G and B sensors and the camera's image processing be able to compensate for this?

Essentially, mask each sensor so it doesn't receieve crosstalk from the the adjacent color. Then estimate how much color was lost to the filters by evaluating each sensor's border wavelength values. It sounds complicated and imprecise, but I would think it's easier and more accurate than demosaicing. Is this not done? Does it make sense?


P.S. I'm not convinced the filters would even be necessary, just the internal image processing "should" be enough.

Anyway, thanks very much all your input, Daniel. It's always a learning experience reading your posts ;).

Thanks, Peter.

Wouldn't a combination of both color filters placed on the R, G and B sensors and the camera's image processing be able to compensate for this?


Yes, I would guess that filters could re-shape the spectral response, but only at great cost to sensitivity, probably enough to undo the advantage of having three sensors in the first place. The reason is that filters can only increase relative sensitivity in one wavelength by reducing response in all other wavelengths.


Essentially, mask each sensor so it doesn't receieve crosstalk from the the adjacent color.


Actually, the problem with 3-chip is that they don't have *enough* crosstalk. In human vision, there is significant overlap in the green and red, so that there are many colors that look the same to us (we can't tell the difference), but look different to the 3-chip (because it doesn't have so much overlap).

In one sense, 3-chip goes beyond the capability of human vision; it's too precise to approximate our more relaxed perception.


P.S. I'm not convinced the filters would even be necessary, just the internal image processing "should" be enough.


Image processing would be able to achieve excellent color accuracy if every pixel stored a chart of how much of the light came from each wavelength. 5% from 550nm, 10% from 560nm, etc. But it doesn't: all the wavelengths are mixed together in the pixel.

You make some good, well thought out points, Daniel. I may not be 100% convinced, but...... :-)

In answer to one of Peter's points ("it would border on absurd to make a large sensor camera w/ a very slow lens so it could mimic the performance of a smaller sensor camera") then I suppose the answer is you get more options. Initially, I confess, I felt the same as Peter.

Either get a long zoom with small max aperture (to mimic small sensor performance) or get a fast short ratio zoom (or prime) if DOF or low light ability is of more importance.

Take a Canon EF-S 18-200 f/3.5-5.6. Motorize it. Problem solved ;)

Canon. You burst on the scene with the famous XL1. Everybody wanted one. Then as the other manufactures got going to play catchup the cry from your loyal diehard troups began chanting.

"Where's the XL2?" Out came the wishlists and the prophesy of how it will be. We waited with baited breath, couldnt sleep at night. Then, Drumroll, it's coming. Out poped a XL1s.
Umm we all said, "Looks a bit like the xl1" we thought. Canon assured it was a totally new camera. Only the paint was the same, but it's innards are entirely new. So we grabbed one, raved over it. Many said this is the xl2, but it's got a new name. We were still loyal Cannonites and hung in there with religious verver.

In the mean time we looked over our shoulders and saw Sony, and Panasonic and others leaping ahead and were breaking into HDV.

Were's the xl2, Surely as God made apples the xl2 hd version will be released at the next NAB. We waited, waited, and anticipated the new xl2 in hd. After a long gestation out poped another camera. They'd run out of red paint so used white. and gave birth to the xl2.
This what indie and ENG , pro's and wannabe pros were waiting for. A totally new camera Canon boomed again. The only thing that's not new is the microphone. We thought the only thing not new was it's SD. So we rushed out, bought one and found when we're in a media bunfight we were the only ones with a white camera and not hd capable.

So here we are again. The remenant, waiting for a revolutionary new super lens one inch chip daddy of em all canon hd camera. In reality so many I know have given up. The EX1 and EX3 have finally done it. All the loyal Canon users here are leaving or have left their Canon and embraced the EX1 and EX3, buying them in ones, twos and threes.

So I've had a xl1, xl1s and a xl2. I'm about to go to hd. I've been loyal to Canon for 10years, but it's time to stop waiting and look else where. I've been loyal to Canon all the way through. I even had a Super 8 Canon film camera once. It's nearly next year, I'll wait a bit longer, but not forever.

Canon's always slow, Canon's always late, but Canon always comes with high quality products. I think the XL-H1 or whatever it will be called will be a revolution again. You can't be market leader for 10 years in a row in this industry.

And the build quality of the EX1 and EX3 is so poor. Fisher price plastic microphone holders that break, paint that peels off... my XL-H1 still looks new. The EX1 and EX3 have a nicer chip and great image quality but their build quality makes me cry.