Question for Jim about Line-Skipping

[Gunleik Groven]

Maybe Panasonic doesn't care.... or maybe they are trying to protect their $60,000 2/3" 3-CCD 1080p P2 Varicams? Think about it.

They might, and have every good reason to do it from an economical perspective, as these cameras still sell very well and are very good at what they do.

I find it hard to compare them with an AF-100 as to useability, though.

Potential image-quality is (un-)fortunately seldom the only parameter involved when choosing a camera. Flexibility in a given context, conformity to image-workflows, reliability and ease of use on-set are much more important factors when a camera is chosen for a budget production.

 

[Gavin Greenwalt]

I'm never a fan of conspiracy theories when they result in the conspirators losing money. I would imagine the volume of prosumer cameras that they could push over someone like Sony would dwarf their entire Varicam sales division if they wanted to raise the price to offer it. But I bet they've done the market research and determined that there aren't enough people that'll pay the extra. Look at the scarlet reaction to a modest price increase for HDRx.

The real question is why doesn't the NEX VG10 offer anything other 60i? Because they don't care. It can't possibly be to protect another product (they have no other product). It can't possibly be because of cost (It can't cost that much more to offer another framerate if well designed). They just don't see the point.

 

[Gunleik Groven]

Yup. That sounds like total ignorance and a pre-designed upgrade path...

 

[Dan Hudgins]

Motion OR still

Motion OR still

Canon and the other HDSLR makers should offer a JUST motion camera model, and that should use Quad-HD rather than de-Bayer with Bin or Skip.

It would solve most if not all of the image issues having to do with the sensor and OLPF+IR/UV.

When you have Quad-HD or 4K sensor you can get 1920p or 2K without interpolation, you get four pixel groups, one red pixel, one blue pixel, and you mix the two green pixels, but you do not need to interpolate.

The red and blue are out of register by 1/2 1920p pixel, but that is not a major issue since the CA from the lens would be several times that off center.

Using 4K or Quad-HD sensor lets you get fast conversion without large CPU power, I use that in my "workprint" de-Bayer and its about 200x faster than doing a "full de-Bayer" with area filters. You get more noise without the de-Noise from area filters, but if the sensor noise is low enough to start with you can get maybe a usable image that has fewer processing artifacts that way.

In the HDSLR, they are using a high-resolution sensor made for still images, to shoot motion of lower resolution, and the OLPF is not heavy enough for the lower resolution images. If people are purchasing the cameras for motion ONLY use, why don't they make a motion only version like 7D-M or something?

 

[Tom Lowe]

I'm never a fan of conspiracy theories when they result in the conspirators losing money. I would imagine the volume of prosumer cameras that they could push over someone like Sony would dwarf their entire Varicam sales division if they wanted to raise the price to offer it. But I bet they've done the market research and determined that there aren't enough people that'll pay the extra. Look at the scarlet reaction to a modest price increase for HDRx™.

The real question is why doesn't the NEX VG10 offer anything other 60i? Because they don't care. It can't possibly be to protect another product (they have no other product). It can't possibly be because of cost (It can't cost that much more to offer another framerate if well designed). They just don't see the point.

Sony has $250k F35 cameras to protect, and F23s.

Gavin, you are saying it's not a conspiracy because it would lose them money. But you are assuming these companies have the slightest clue what they are doing. They don't. Canon stumbled into the 5D2 blindly, because the AP requested 30p video for their photographers. Two years later, Canon has doubled their corporate profits. Don't assume Panasonic is smart enough to know that $10K prosumer camera could outperform their entire "pro" line.

 

[Gavin Greenwalt]

The good news is that Cell Phone cameras are going to start eating into the handicam market in the next year or two. They're already 720p and not... terrible.

Cell Phones will be "Good enough" for soccer games and birthday parties. I suspect the entire handicam segment of the market is going to wake up in the pretty near future and decide quality is the only way to distinguish themselves from the 'free' camera in everyone's smartphone.

 

[Jim Retar]

Gavin, you are saying it's not a conspiracy because it would lose them money. But you are assuming these companies have the slightest clue what they are doing. They don't. Canon stumbled into the 5D2 blindly, because the AP requested 30p video for their photographers. Two years later, Canon has doubled their corporate profits. Don't assume Panasonic is smart enough to know that $10K prosumer camera could outperform their entire "pro" line.

for a company that doesn't know what it is doing.... doubling their profits sounds pretty good.

I would think you, more than most on here, would remember that these cameras are fantastic still image DSLRs and Canon etc concentrate on that aspect. The fact that they are even listening to video requests is a miracle.

Who knows what digic 5 will bring... Canon keep quiet about their tech, maybe they have figured out how to resize and not pixel bin/line skip. time will tell.

Regarding the holding back of the low end market... It is worth remembering the market for film/tv is tiny in comparison to stills... so why should they pursue it in the same manner as a billionaire with a passion for film and a near bottomless fund? RED is fantasy land (great for us) but to expect other hugely successful companies to follow suit is crazy... they have to limit their low end products or they go bust.

 

[Matt W.]

lineskipping wouldn't affect sensitivity.

You're right; I phrased it wrong. Lineskipping would leave sensitivity the same but increase noise levels by three stops. Obviously neither binning nor line skipping is ideal. But binning isn't that bad other than aliasing, which the AF100 seems to fix (at the expense of losing still picture functionality).

Anyhow, once the 7D doesn't bin or the Red is under $5,000 I'll be happy and take out a loan (or save up for a few years) and buy one. I don't think I'll ever have the opportunity to work with a more expensive camera than $5,000, unfortunately, unless I go to film school and can borrow equipment. And I don't have the background to get in. So...price matters, too.

So for me...7D is good enough and at the right price. Obviously if you can afford a better camera, there are MUCH better cameras, but I'm not as sure as everyone else Canon can make a "true" cinema camera at a prosumer price point. My next purchase will probably be a 7DII or whatever comes out--for price alone. Hopefully I'll be able to rent a red some day, though (I have used them in the past, when I had more money).

 

[M Most]

Gavin, you are saying it's not a conspiracy because it would lose them money. But you are assuming these companies have the slightest clue what they are doing. They don't. Canon stumbled into the 5D2 blindly, because the AP requested 30p video for their photographers. Two years later, Canon has doubled their corporate profits. Don't assume Panasonic is smart enough to know that $10K prosumer camera could outperform their entire "pro" line.

I could turn that around and say don't assume that you're smarter than multi-billion dollar corporations just because you think you understand their inner workings, market analyses, target audiences, and technical capabilities based on your own personal needs, prejudices, and opinions, when in fact you have no information regarding any of those things, and likely don't represent a major part of their intended audience in the first place. I could also point out that Jim Jannard has said that he didn't expect to sell anywhere near the numbers of units of Red Ones as he has in fact sold, but I don't hear you saying he wasn't smart enough to predict what would happen.

Products finding markets that were to some degree unintentional is pretty common with smartly designed devices (do you really think Steve Jobs and Jonathan Ive designed the iPad to be a movie slate?), and it often has nothing to do with the intelligence or lack of it of the company building the product.

 

[David Mullen ASC]

These companies know their current markets pretty well, it's reaching into new markets that is more of a risk and harder to predict.

I think that Panasonic has plenty of bean counters working for them that do nothing but crunch numbers all day. And there is also probably a bit of the "the market that wants to only pay $5000 for a camera will probably accept 'x' features even if compromised, and if the camera is successful, we can see about a $10,000 camera with 'y' features and a $20,000 with 'z' features."

Obviously if they crammed more quality into a cheaper camera, it would sell well -- that's only logical. I'm sure if BMW dropped their price per car to $20,000 it would sell well. I'm sure that iPads sold at $100 a pop would sell well. The real question is whether it would make a profit for Panasonic, and by profit I mean not just that model's sales but how they affected the sales of their other cameras as well.

You could see this in conspiratorial terms, or simply in business terms -- Panasonic has a certain "legacy" line-up of cameras to deal with, plus there is a whole professional infrastructure and long-term customer base to factor into their decisions. They can't just "start from scratch". Now does that make it a conspiracy? An act of stupidity? Or is it a burden on their part that they have to manage around? They aren't Red, with only one model on the market and 5,000 or so customers. So you can't expect them to make decisions as if they were Red.

What I see is a lot of compromising going on by Panasonic, "compromising" both as a negative term (new products that have questionable limitations built into them) and when the term means finding some sort of middle way between two extreme choices. They have a whole line-up of cameras that generate profit for their companies -- obviously so, otherwise they'd be out of business. Now they have this whole new technology, larger single sensors, different optics needed, different recording and post infrastructure perhaps needed, all of which may or may not play well with their main customers of professional broadcast equipment, the people with the deep pockets. So what's the solution -- tell their customers of the past thirty years "goodbye!" because Panasonic is junking their whole 3-CCD broadcast ENG/EFP line-up and instead going into something different, maybe something involving RAW recording and new post processing techniques? Or do they try to fit this new technology carefully into an existing line-up of cameras? I think this is where the "compromising" comes in, both for ill and for good.

I'd also try and think for a moment like the typical broadcast shooter, maybe someone shooting for the Golf Channel or something, or a doc for the History Channel, or the local news, etc. Do you really want to shoot a golf tournament with DSLR-style equipment? Large sensors, low depth of field, lack of ENG zoom optic choices? Do you want to deal with RAW recording?

I think we must accept that there is still a big market for "traditional" broadcast video equipment, so given that, given that Panasonic is going to keep providing gear for that market, then how do they also introduce this other way of shooting, this more cinematic larger-sensor technology? How do they introduce it in a way that it doesn't just throw all their product lines into mass confusion? This is what I mean by the "legacy" issue. We don't run Panasonic so we don't have to worry about that. We can play fantasy games here where we just toss out our dream cameras on the market and make a lot of money and everyone goes home happy. I just don't think it is as simple as that for corporations like Panasonic or Sony. They don't have that sort of freedom.

 

[Graeme Nattress]

The 7D/5D don't line skip; they bin pixels together, like with sRAW. The aliasing is roughly the same with this technique as it would be with line skipping (2MP sampling from a high megapixel chip) but sensitivity isn't sacrificed. If these cameras really used line skipping, they'd be a lot slower for video than for stills--on the order of three stops.

Even the Red can't deliver 1080p bayer in real time (for monitoring OR recording). The Genesis had to use an alternative sensor design (stripes) to allow real-time recording. The ALEXA has a very low-res chip. Debayering full res in real time is hard and the AF100 probably doesn't do it.

But claiming that the AF100 line skips? That's unsubstantiated. Chances are it bins pixels and uses an appropriately stronger anti-aliasing filter, so all it's sacrificing is resolution, which doesn't really matter with output to a 1080p codec.

This "line skipping" talk is largely FUD. You just have to look at this site to see Red's good at marketing and that's all this is. Not that RAW isn't obviously better, but there's nothing inherently terrible about what the AF100 is likely doing.

That cannot be so because sRaw, although aliasy, is aliasy evenly in the horizontal and vertical direction, whereas the image we see in video is more strongly aliased in the vertical direction.

The line skipping is not FUD. It's the only explanation for the images we see on the Canons and binning does not produce the artifacts we see on the images from the Canons. You can try that yourself by taking some raw bayer data and writing a line skipping demosaic and comparing it to a binning demosaic.

The RED One can indeed output fully demosaiced and downsampled 1080p on playback. The Epic is obviously doing that in playback and record.

Until I get an AF100 on a zone plate we won't know what it's doing, but I can hardly imagine it doing a full demosaic and downsample. Time will tell.

Graeme

 

[Tom Lowe]

Tom... I think the problem is your 1st assumption. I'd be very surprised if their sensors have a clock speed and A/Ds capable of read/resetting all pixels at 24fps. I don't know for sure... but if the sensor was fast enough, why would they not deliver it's full capability?

Jim

Does this mean the sensor is the toughest nut to crack at the moment?

Let's put RAW aside for the moment, and assume Panasonic or Sony wanted to "do it right" and fully downsample the images coming off an APS-C or 4/3 sized Bayer sensor.

I was not really able to understand exactly how much computing power we would be talking about for a camera to demosaic on the fly and downconvert 13MP+ Bayer to 1920x1080 RGB, and then compress it in some way. I don't think even the highest-end Macs and PCs can do it? (Leaving aside the Rocket.)

Jim often scolds these other companies for line-skipping, so my purpose on this thread was to understand exactly what it would take, in terms of adding cost to a camera, for them to do it properly?

Sorry if I muddied up the thread with talk of Panasonic and Sony potentially withholding this type of technology to protect their outdated, ultra-expensive "cinema" cameras... but I do think those considerations play into this.

 

[Jeff Kilgroe]

Canon sRAW is binned. Canon DSLR video is horizontally binned and vertically line-skipped. Not only is that what the aliasing indicates, but it's a necessity brought on by the sensor itself. Just as the RED One must window-down to operate a smaller section of the sensor at increased frame rates, the current crop of DSLRs must do the same -- operate less of the sensor to maintain 24, 30, etc.. FPS. So, the options would be to record video at a smaller center window of the sensor, or to utilize the full area of the sensor and then skip lines to keep the sensor from spontaneously combusting.

We also have to consider the internal bandwidth and processing abilities of these smaller cameras. Even if the sensor were able to cycle in full at 24+FPS, the amount of onboard compute power needed would increase several times over in order to bin the video just in the same manner as the sRAW algorithm does it. Lines recorded seem to be Bayer-pattern line pairs, so you have one G,B line paired with the R,G line above or below it, then it skips two pairs / 4 lines and records the next pair.

This is not FUD, it's just the way it is.

As for the AF100 -- Panasonic takes a similar approach with their DSLR cameras' video modes, if not identical. They are doing the same in the AF100. This time around they have a much stronger and more specialized OLPF, which goes a lot farther to mask the aliasing artifacts and it seems to work quite well. However, in the various clips floating about the net now, none of them are free of aliasing artifacts. There are lots of people claiming they are, but unfortunately it's not so. Look at the 300x still of the red car that Sanjin posted. First look around the car's mirror. Follow the lines in the hood and around the headlights. Ugh-ly.

IMO, the Sony NEX-VG10 is a better buy than the AF100. It has better sensitivity, nearly 2 stops greater DR, APS-C/S35 sensor size, equally adaptable lens mount and it's less than half the price, including an 18-200 zoom. It does lack SDI and XLR ports, but honestly, who cares? Anyone who cares about those features will be recording to an external recorder anyway and the camera delivers "uncompressed" via HDMI. The onboard mic is equal, possibly superior, to the AF100 mic. Or I guess you could buy an external recorder with audio input or just run a nice audio recorder too. You can buy a NEX-VG10 plus a Sound Devices 702 and still come out cheaper than the AF100.

I'm really perplexed as to why the AF100 is so attractive. Other than the adaptable lens mount and the larger sensor size, it has little to offer. For overall image quality and a better onboard codec, I would take an EX3 any day, and it, with the 14x Fujinon lens, is about the same price as an AF100 body w/ a decent external recorder.

 

[Matt W.]

Interesting... The AF100 is so new it may just be hype. You're probably right about the Canon doing something strange. Otherwise they'd be more forthright about it. I don't know. IMO, the low-light performance is too good for it to be line skipping only, though. Something along the lines of the hybrid system you mentioned above makes sense.

I have to assume real time debayering for like 20MP takes a shit ton of processing power. I think the maximum capture rates for JPEG on the 5D, 1D IV, D3, etc. are exactly where the technical limitations of those cameras are (in terms of sensor speed and debayer speed) because Nikon and Canon are competing pretty hard for spots/photojournalism, where FPS matters. I don't think those cameras are crippled at all in terms of 8fps JPEG or whatever. That's the state of the art. Just look at how slow medium format digital cameras are.

My question is why not make a 3k chip, like Arri does, for a mass-market cinema camera? I guess the expense of custom fabbing one is too high when a dSLR sensor can just be cannibalized. Canon's next generation of full frame cameras are going to be way in excess of 30MP. They aren't going to be able to real time bayer that any time soon. It seems like you need a lot of horsepower to make a true hybrid camera and Canon doesn't think the market will support a pure "cinema camera" when 7Ds are flying off the shelf despite their flaws...so the only answer is the 7D or something like it, at least right now. Who knows. By the time I can afford one whatever's available will be awesome so I'm just glad of that.

 

[Tom Lowe]

Well many of us here have been saying for years that a dream "prosumer" camera would be:

Large sensor
3K RAW or 1080pRGB
High-ISO performance
60fps
Interchangeable lenses

Do that for ten grand, without line-skipping, and you will sell truckloads of cameras.

 

[Jannard]

Well many of us here have been saying for years that a dream "prosumer" camera would be:

Large sensor
3K RAW or 1080pRGB
High-ISO performance
60fps
Interchangeable lenses

Do that for ten grand, without line-skipping, and you will sell truckloads of cameras.

Sounds like Scarlet S35.

Jim

 

[Gunleik Groven]

It does, but it is 5k, isn't it?

G

 

[Sanjin Jukic]

Sounds like Scarlet S35.

Jim

Great news!!!!

Jim, could you give us more info about Scarlet S35 estimated shipping time????

Or it could be something like a simple answer: Summer 2011???!!!

 

[Auguste Andre]

It does, but it is 5k, isn't it?

G

A lot of us here are expecting 5k, not less.

 

[Hrvoje Simic]

The real question is why doesn't the NEX VG10 offer anything other 60i? Because they don't care. It can't possibly be to protect another product (they have no other product). It can't possibly be because of cost (It can't cost that much more to offer another framerate if well designed). They just don't see the point.

Yup. That sounds like total ignorance and a pre-designed upgrade path...

"Not caring" and "total ignorance" lead to VX1000, VX2000/PD150, FX1/Z1 and EX1, all of which were pioneering, serious steps in camera history and became standard tools.

NEX cameras are transitional cameras in a new consumer market segment.