Why the hate for video capable DSLRs?

[Orangutan]

Any time during the design process you have a requirement that the device be able to do A and B (rather then just A or B) compromises will be made, designs become more complex, more hardware is required, and testing time increases.

This is the part that confuses me: I'm a former programmer (though not a trained EE) but I don't see A and B being different here.  To my mind, video is is basically 30fps of still images encoded in a single file, along with a bit of audio.  You might argue that the algorithms are different, and perhaps they are.  But the code must be written for consumer cameras, and the chips (presumably) use the same instruction set.  (I doubt the DigicN for PowerShot is different from the DigicN for EOS).  So the code is already written, and need only be incorporated into the EOS firmware.  Furthermore, the high-speed sensor reading necessary for video can only help still images.

In short, I'd ask you this: what, precisely, of the following chain of events is substantially different between still and video?  And which of that is not already a requirement of consumer cameras?

1. Expose sensor
2. Read data from sensor
3. Demosaic
4. Encode a frame
5. Save to media

Really, I don't get it: why is video not just a (nearly) free bonus?

[SPG]

I shoot in natural light. I demand that Canon remove the hotshoe!

Ok, seriously though...removing every feature that you personally don't use is not going to make your camera better. The reason is that all these features increase the number of people who buy the camera and increase the amount of money that can be spent on R&D to make the next camera even better.

I still haven't seen anyone show proof that video capabilities are harming the photo capability. Positing theories about R&D and engineering doesn't mean any more than the latest conspiracy theory. I want to see a video feature that has directly hurt photo performance.

[AvTvM]

In short, I'd ask you this: what, precisely, of the following chain of events is substantially different between still and video?  And which of that is not already a requirement of consumer cameras?

1. Expose sensor
2. Read data from sensor
3. Demosaic
4. Encode a frame
5. Save to media

Really, I don't get it: why is video not just a (nearly) free bonus?

I am no engineer.

But as I see it:

For video, a sensor with 2 megapixels (FullHD - 1920x1080) is all that is required. Videos are viewed on monitors or on screens/beamers - all of which offer - at best! - Full HD.
The sensor+readout+processor needs to be able to handle an ongoing massive stream of data without any interruption. Image quality of single frames however is a secondary requirement at best. 
The sensor needs to be of a type that can handle capture incoming light for "indefinite periods" of time. Cooling that thing is a major hardware issue.
These requirements preclude certain sensor types (e.g. CCD-FT) from being used in regular video-enabled cameras altogether. This narrows the choice of image capturing device and layout friom the start to a much narrower field than for "stills capture only".

For photography (stills) hat is needed/wanted?
Sensors with the highest possible resolution [currently 36 MP+ on 36x24mm "FF"sensor size], with the best possible S/R [translating into good DR, low hi-ISO noise, and all other goodies us photographers want). Image quality of each single frame/capture is paramount, speed is a secondary concern - 10 fps more than good enough for virtually anything. Exposure times are typically fractions of a second, typically max. 30 seconds, and only in very rae instances minutes. Cooling is much less of a concern than in video use ... and yes, live view has is blurring that requirement a bit. But bottom line: the whole data readout/processing pipeline needs to be geared to highest "single capture performance" vs. "streaming performance". 

These two sets of requirements are not 180 degrees opposite of each other, but there is a significant rift between them. Fulfilling both requirements necessitates an enormous amount of compromise on both ends .. for video and for stills capture, making the final product significantly more difficult to design, test, manufacture = more expensive, more prone to defects, less good in each of the 2 disciplines.   

"Video" in DSLRs of all things  with all the mechanical stuff (anything inside the mirror box) and a lot of the optical stuff (prism, viewfinder etc.) really *in the way of video* rather than complementary is an aberration in camera design. Looking at mirrorless cameras changes the pictures a lot, but for DSLRs its madness.

The sole reason HDSLRs got popular with videographers  is their relative pricing compared to the "traditionally outlandish prices" for (relatively) large-sensored videocameras. All of a sudden, a $ 2,000 body comes with a sensor that rivals old-school videocams at $ 100.000+ ... that is the appeal in HDSLRs. It is a boon for videographers for sure, but not for photography/stills capture!  HDSRLs are highly affordable to a large group of aspiring moving images people, and given the price they are more than illing to put up with all of the disadvantages HDSLRs bring to their work. Basically, all they use in a HDSLR is the sensor and the data processing pipeline. They (generally) don't want AF, but will rather add "follow-focus" contraptions and all sorts of bulky rigs around the poor little HDSLRs. They don't need an optical viewfinder (with heavy/bulky, costly glass prism). Basically, 50% of the cost of an HDSLR is "wasted" on them.  Or, put another way, the money put into AF, mirrorbox, prism, optical viewfinder should go into even more video-suitable sensors+data processing capabilities. No reason, why there should not be 2k video cams with an FF sensor and an EF/PL mount in front. Without AF system, no OVF, but top notch EVF. In one word: a videocam! Something like a Canon C300, but at $ 2,000 not at an insane cost of $ 16,000 $

At the same time stills shooters should get a 5D III with e.g. a 24 MP FF "stills-optimized" sensor [whatever type/design may be best for that], top-notch AF, top-notch optical viewfinder minus all video-related features for $ 2,000 or less. 

I guess that should answer  the question, whether video-capability is just "a little freebie" on a HDSLR. It is not. It is a product design aberration  from the very start!     

[Cetalis]

AvTvM, all you've just said is theoretical; it's certainly possible, but I have yet to see a defect in current DSLRs caused by the addition of video.

[AvTvM]

AvTvM, all you've just said is theoretical;

wrong. It's dead on, 100% real-life.

For all the reasons a number of people have detailed in this thread, digital cameras designed to capture both still images and video are a huge comrpomise from the very start. All of the points I've made are valid and very real.

[distant.star]

Thanks, Av.

This is a comprehensive explanation that makes a lot of sense to me.

In short, I'd ask you this: what, precisely, of the following chain of events is substantially different between still and video?  And which of that is not already a requirement of consumer cameras?

1. Expose sensor
2. Read data from sensor
3. Demosaic
4. Encode a frame
5. Save to media

Really, I don't get it: why is video not just a (nearly) free bonus?

I am no engineer.

But as I see it:

For video, a sensor with 2 megapixels (FullHD - 1920x1080) is all that is required. Videos are viewed on monitors or on screens/beamers - all of which offer - at best! - Full HD.
The sensor+readout+processor needs to be able to handle an ongoing massive stream of data without any interruption. Image quality of single frames however is a secondary requirement at best. 
The sensor needs to be of a type that can handle capture incoming light for "indefinite periods" of time. Cooling that thing is a major hardware issue.
These requirements preclude certain sensor types (e.g. CCD-FT) from being used in regular video-enabled cameras altogether. This narrows the choice of image capturing device and layout friom the start to a much narrower field than for "stills capture only".

For photography (stills) hat is needed/wanted?
Sensors with the highest possible resolution [currently 36 MP+ on 36x24mm "FF"sensor size], with the best possible S/R [translating into good DR, low hi-ISO noise, and all other goodies us photographers want). Image quality of each single frame/capture is paramount, speed is a secondary concern - 10 fps more than good enough for virtually anything. Exposure times are typically fractions of a second, typically max. 30 seconds, and only in very rae instances minutes. Cooling is much less of a concern than in video use ... and yes, live view has is blurring that requirement a bit. But bottom line: the whole data readout/processing pipeline needs to be geared to highest "single capture performance" vs. "streaming performance". 

These two sets of reuirements are not 180 degrees opposite of each other, but there is a significant rift betwenn them. Fulfilling both requirements necessitates an enormous amounts of compromises (on bioth ends .. video and stills capture!), making the final product significantly more diffcult to design, test, manufacture = more expensive, more prone to defects, less good in each of the 2 disciplines.   

"Video" in DSLRs of all things  with all the mechanical S___ (anything inside the mirror box) and a lot of the optical stuff (prism, viewfinder and assorted stuff) really *in the way of video* rather than complementary is an aberration in camera design. Looking at mirrorless cameras changes the pictures a lot, but for DSLRs ...

The only reason HDSLRs got popular with videographers  is their relative pricing compared to the "traditionally outlandish prices" for (realtively) large-sensored videocameras. a 2000 body will offer a sensor that can rival old-school videocams at 100.000. That is the appeal in HDSLRs. It is a boon for videographers for sure, but not for photography/stills capture! Because HDSRLs are highly affordable to a large group of aspiring moving images people, they have to and are of course willing to put up with all the disadvantages a HDSLRs bring to their work. Basically, all they use of a HDSLR is the sensor and the data processing pipeline. They (generally) don't want AF but will rather add "follow-focus" contraptions and all sorts of bulky rigs around the poor little HDSLRs. They don't need or want an optical viewfinder (with heavy/bulky, costly glass prism) - none of it! Basically, 50% of the cost of an HDSLR is "wasted" on them.  Or, put another way, the money put into AF, mirrorbox, prism, optical viewfinder should go into even more video-suitable sensors+data processing capabilities. No reason, why there should not be 2k video cams with an FF sensor and an EF/PL mount in front. Without AF system, no OVF, but top notch EVF. Yes, exactly - a videocam. Something like a C300, but at $ 2000 not at $ 16000 $.

At the same time there is no reason, why stills shooters should not get a 5D II with 24 MP FF "stuills-optimized" sensor [whatever type/design may be best for that], top-notch AF, top-notch optical viewfinder but minus all video capability in the entire processing pipeline for $ 2000. 

I guess that should answer  the question, whether video-capability is just "a little freebie" on a HDSLR. It is not. It is a product design aberration  from the very start!     

[Cetalis]

AvTvM, all you've just said is theoretical;

wrong. It's dead on, 100% real-life.

For all the reasons a number of people have detailed in this thread, digital cameras designed to capture both still images and video are a huge comrpomise from the very start. All of the points I've made are valid and very real.

If this is such a huge compromise, then surely there are glaring defects in the current batch of video capable slrs, i.e. 1Dmk4, 5DmkII and 7D. I have yet to notice any, hence my skepticism. I can understand how these are not ideal for video, but I have yet to see stills functionality compromised by the addition of video functionality.

[hollybush]

I have yet to see a defect in current DSLRs caused by the addition of video.

Wait a few weeks and I suspect you will, in the form of the Canon 1DX. Look at the position of the "movie" button on that. Maybe the testers' thumbs are shorter than mine, but it looks to be exactly where my thumb sits on the current 1 series bodies. That compromises steadiness and convenience for stills. If so, it's a catastrophic design error.

Putting faster and more CPUs in cameras to handle the video decreases battery life. Battery life is one of the biggest differences between good stills cameras and bad stills cameras. The 1D Mk IV has significantly worse battery life than the Mk III and the only extra thing the CPUs have to do is video.

[Orangutan]

Look at the position of the "movie" button on that. Maybe the testers' thumbs are shorter than mine, but it looks to be exactly where my thumb sits on the current 1 series bodies. That compromises steadiness and convenience for stills. If so, it's a catastrophic design error.

While that could be a legitimate complaint, it's not inherent to the inclusion of video tech; it's just a design error.  They could easily have put some stills-oriented button there instead.

Putting faster and more CPUs in cameras to handle the video decreases battery life...The 1D Mk IV has significantly worse battery life than the Mk III and the only extra thing the CPUs have to do is video.

This is completely specious reasoning:

• It's only burning the extra juice when it's processing video.  When processing stills it should be no worse than it was before.  Embedded CPU's like this don't burn a lot of juice just idling.
• The extra demands of video will push Canon to design more efficient processors and higher-capacity batteries.
• The demands of video will push canon to design faster processors, meaning better/faster in-camera JPEG (if that's your thing).

No, sorry: I've still yet to hear even a shred of cogent argument.  All I've heard so far is speculation, hand-waving and imagination.  If someone with real chip-design experience wants to chime in with an explanation of how video capability limits stills IQ at the sensel level I'd love to read all about it.  Note that even the Leica M9 (a stills-only camera, if ever there was one) scores worse on DxOMark than the 1D4 for low-light, DR and color depth.  Sorry folks: just not buying it.

[hollybush]

While that could be a legitimate complaint, it's not inherent to the inclusion of video tech; it's just a design error.  They could easily have put some stills-oriented button there instead.

A previous poster described in detail how the design process worked, and why it is more difficult to design a product to do two different things.

It's only burning the extra juice when it's processing video.  When processing stills it should be no worse than it was before.  Embedded CPU's like this don't burn a lot of juice just idling.

But it's not idling, and not halted. That CPU does things other than video. They had to use a bigger CPU, which has more gates and burns more power when it is doing those other things.  In some alternate design for the 1DX, there might have been a 4th processor dedicated to video that could be halted, but that isn't what we got.

The extra demands of video will push Canon to design more efficient processors and higher-capacity batteries.

I don't want a different, heavier, higher-capacity battery. I want to use my old ones. (Kudos to Canon, I can with the 1DX, but they are suspiciously silent on how many shots it wil last with an extra CPU on board, bigger than the old ones.) If there is new battery technology, I want smaller and lighter, not the same size with more capacity.

The demands of video will push canon to design faster processors, meaning better/faster in-camera JPEG (if that's your thing).

It's not, but I will admit it is for roughly the same people for whom video is. No raw video for them yet with current CPUs.

If someone with real chip-design experience wants to chime in with an explanation of how video capability limits stills IQ at the sensel level I'd love to read all about it.

Wasn't sensor heating and noise already mentioned?

Note that even the Leica M9 (a stills-only camera, if ever there was one) scores worse on DxOMark than the 1D4 for low-light, DR and color depth.  Sorry folks: just not buying it.

Strawman.

[Cetalis]

A previous poster described in detail how the design process worked, and why it is more difficult to design a product to do two different things.

It is harder to design something to do two things well, but it appears that video is being compromised, and not stills. In any event, I have yet to see anything compromised in favor of video.

But it's not idling, and not halted. That CPU does things other than video. They had to use a bigger CPU, which has more gates and burns more power when it is doing those other things.  In some alternate design for the 1DX, there might have been a 4th processor dedicated to video that could be halted, but that isn't what we got.

There's an entire CPU dedicated to AF, a non-video function. The 1DX was designed by Canon to cater to photojournalists and sports shooters (and yes, they have said that http://www.megapixel.co.il/english/archive/28040) and thus high processing power is a necessity, regardless of any video needs. Also, who's to say that the CPU hasn't gotten more power efficient with new advances in technology? Modern computer (x86/x86-64 at least) CPUs can scale back their frequencies if need be, and are more efficient than older ones.

I don't want a different, heavier, higher-capacity battery. I want to use my old ones. (Kudos to Canon, I can with the 1DX, but they are suspiciously silent on how many shots it wil last with an extra CPU on board, bigger than the old ones.) If there is new battery technology, I want smaller and lighter, not the same size with more capacity.

The 1DX is neither small or light; nor was it a design priority. Smaller and lighter would be an issue if it was a priority, yet right now it appears that the size of the reflex mirror and the relevant housing would prevent any change in battery size from making any substantial impact. As for lighter, I've yet to see any substantial changes in weight imparted by the battery, relative to the weight of the camera.

"The demands of video will push canon to design faster processors, meaning better/faster in-camera JPEG (if that's your thing)."
It's not, but I will admit it is for roughly the same people for whom video is. No raw video for them yet with current CPUs.

The entire electronics industry has always pushed for faster, more efficient, cheaper, and in general better processors; often achieving more than a few of these goals. Canon will push for better processors as long as its profitable.

Wasn't sensor heating and noise already mentioned?

Sensor heating, and the noise that comes with it, is only a problem during long exposures, live view and video. If one does none of those, it is not an issue. Also, every current digital camera suffers from the same issue, regardless of whether or not it has video. If anything, the video people complaining about it may mean that those of us who use live view and long exposures might benefit.



Also, if one looks at it from a video perspective, the current batch of DSLRs is optimized toward stills with video functionality added on. What I see are two lines of cameras: one optimized for stills, and one optimized for video (Cinema EOS), with both being able to capture stills and video, but being better at one, and I cannot see how relatively bad video functionality(which isn't even that bad) is something to complain about when it is essentially free, especially if one will not even use it.  If the 5Dmk3/X has no OVF, no AF, built in ND filters, a cooling fan, and a 4k sensor I'd agree that we need a stills optimized camera, but Canon is not going to do that.

[NormanBates]

But as I see it:

For video, a sensor with 2 megapixels (FullHD - 1920x1080) is all that is required. Videos are viewed on monitors or on screens/beamers - all of which offer - at best! - Full HD.
The sensor+readout+processor needs to be able to handle an ongoing massive stream of data without any interruption. Image quality of single frames however is a secondary requirement at best. 
The sensor needs to be of a type that can handle capture incoming light for "indefinite periods" of time. Cooling that thing is a major hardware issue.
These requirements preclude certain sensor types (e.g. CCD-FT) from being used in regular video-enabled cameras altogether. This narrows the choice of image capturing device and layout friom the start to a much narrower field than for "stills capture only".

I only read this part of that huge post

it shows you don't really know much about video, so there's no point reading all the rest

for a start, a 2 mpix sensor is not enough for 1080p video, as each photosite in the sensor captures only one color - ideally, for 1080p you want 8mpix, so you get full color for each pixel, as in the c300; if the sensor is 3:2, that's 10 mpix (3840x2560)

and if your final images will be 4K, you want a 18 mpix sensor (the RED EPIC is 5120x2700, make that 3:2 and it's 17.5 mpix)

so the resolution of the 1DX sensor is perfect for 4K footage - just make read-out times fast enough, and avoid overheating (that's a very fair point you made, but if you want high fps you have to take this into account too), and you're basically done

because all the arguments about the processor are utterly silly

processing video streams is NOT all that difficult

the digic-dv-III found on the $16K C300 can also be found on the $300 HF-R200, it's a dirt-cheap processor that does the job reasonably well (it can't do 10 bits, only 8-bits 4:2:2, but it's a 4 years old chip, the next iteration should do 10 bits)

THE WEATHER SEALED PLUGS IN THE 1DX COST A LOT MORE THAN A HIGH-END VIDEO PROCESSOR

[Orangutan]

Since Cetalis addressed much of this I'll try to keep this reply short.

A previous poster described in detail how the design process worked, and why it is more difficult to design a product to do two different things.

I saw little detail, and no references.  That long "explanation" seemed like handwaving to me -- all suggestion, no substance.  I saw little value in it.

They had to use a bigger CPU, which has more gates and burns more power when it is doing those other things.

Performance per-watt continues to go up for mobile devices.  As an example, consider netbooks, which have the performance of a Pentium 3, and use a trickle of power.  And yes, one of the strategies is to idle the processor, or to idle parts of it that aren't being used.  If that were not true, your camera would quickly run out of power just staying "awake."

I don't want a different, heavier, higher-capacity battery...If there is new battery technology, I want smaller and lighter, not the same size with more capacity.

This is legitimate, though some people do want higher capacity at same size/weight.

Wasn't sensor heating and noise already mentioned?

Mentioned?  yes.  Addressed well? no.  Sensors only heat when active.  If you're shooting stills then this is no problem for you.  So what's the problem?

Note that even the Leica M9...Strawman.

You did not specify how this is a strawman so I'll have to guess.  If you mean it's irrelevant because it's not a Canon then I disagree.  The premise of the anti-video crowd is that adding video features inherently undermines still photography, and that a camera that was designed, ground-up, as a stills-only camera would produce better-quality stills and a better still-shooter experience.  The Leica is a counter-example to the assertion that stills-only produces better IQ, even though it is a substantially larger sensor and was selected/designed for stills-only.  If there was something else you considered to be a strawman, please specify.

[AvTvM]

Mentioned?  yes.  Addressed well? no.  Sensors only heat when active.  If you're shooting stills then this is no problem for you.  So what's the problem?

The problem is: a stills+video camera needs to be able to withstand (e.g. heatwise) the requirements of   running full-bore video capture and processing for the duration of up to a full battery charge [or 2 of them when gripped] - or some artificial limit like 29 minutes (owing to some wretched customs regulations).
 
A stills-only camera would not have to take that into account from the very start. That alone would enable a significantly different layout of the system and its components, geared more towards delivering highest IQ in each single exposure, rather than in a continous stream. 

The premise of the anti-video crowd is that adding video features inherently undermines still photography, and that a camera that was designed, ground-up, as a stills-only camera would produce better-quality stills and a better still-shooter experience.  The Leica is a counter-example to the assertion that stills-only produces better IQ, even though it is a substantially larger sensor and was selected/designed for stills-only. 

Leica M9 IQ is sub-par not because it is a stills-only camera, but despite it ... its Kodak is a very old design (5+ years?), and at the time apparently the only sensor Leica could get their hands on. Leica has no sensor know-how whatsoever.  If Canon (or Sony) were to design a dedicated "stills-only" camera now (in 2012) with sensor+processing pipeline targeted at highest possible stills IQ it would be quite a different beast, that's for sure. 

[Orangutan]

The problem is: a stills+video camera needs to be able to withstand (e.g. heatwise)

....

A stills-only camera would not have to take that into account from the very start. That alone would enable a significantly different layout of the system and its components, geared more towards delivering highest IQ in each single exposure, rather than in a continous stream. 

Maybe.  Or maybe the heat-tolerance would lead to more efficient circuitry that results in better IQ.  I don't really know, but it would be interesting to ask a true expert in the field.  For now, this is open to speculation only.

Leica M9 IQ is sub-par not because it is a stills-only camera, but despite it...If Canon (or Sony) were to design a dedicated "stills-only" camera now (in 2012) with sensor+processing pipeline targeted at highest possible stills IQ it would be quite a different beast, that's for sure.

Again, this seems like speculation with more certainty than is supported by the hard facts available to us.