Canon's 120MP Camera Sensor | Hacker News Reader

Canon's 120MP Camera Sensor

petapixel.com

203 pointsartsandsci140 comments

There's not much discussion about high ISO noise. I'd be interested to know how such a high pixel density affects that. If they could keep that desesity on a medium format sized sensor it would really be amazing and would really give large format 8x10 film a run for its money.

Hasselblad offer a 100MP 53 x 40mm sensor which they can then use to capture pseudo 400MP images. That's based on a multishot system where they can move the sensor by, I think, half a pixel by shaking the sensor. Only good for things like landscapes that don't move but still amazing.

(Also, off topic but I'm pretty sure the watch used there is an Omega Speedmaster Professional.

I'm not sure if impressed or ashamed that I can recognise it from its unbranded movement.)

Pentax has sensor shift in its current sub $1000 APS-C bodies (K70 and KP) and the full frame K-1. The shifting moves the sensor one pixel to allow four samples under the Bayer filter (red, blue, two greens). The cameras also ship without an anti-aliasing filter. Instead moire is removed by shifting the sensor one pixel to simulate the effect of a low pass filter. Of course, these cameras are rounding error on the price of the Hasselblad (before even thinking about lenses).

> Only good for things like landscapes that don't move but still amazing.

And also very good for high resolution document imaging, such as 300 DPI newspaper shots.

Here is a blurb about that sensor:

https://www.hasselblad.com/press/press-releases/hasselblad-i...

You're right about the watch, if you look at the bridge in the enlarged picture you can still make out the blurred letters.

What's the advantage over a scanner? Speed?

That and less handling, very important in the case of fragile documents.

https://picturae.com/en/

Very interesting company, and from what I know one of the first in the world to take delivery of that sensor.

The biggest advantage is being able to image things that aren't flat. It would be extremely useful for digitizing natural history collections (like the beetle in the article). Unfortunately, natural history museums, at least the smaller ones, like the one I am affiliated with, probably don't have the funding for a super high end camera like this and instead have to to get creative with flatbed scanners and light boxes in the mere $10,000 range. It's a shame, because the resolution and depth of field of the Hasselblad would be absolutely killer for, for example, herbarium specimens (pressed plants).

A flatbed scanner, anyway, spreads its resolution over its whole plate area. So if you want to scan something smaller than that full-size, you're getting a fraction of its resolution.

Want to flatbed scan a 2" tintype or even a 4x6 photo? You'd probably be better off using a camera instead, as you can fill the cameras imaging sensor with the subject. You can also control the lighting conditions better with a camera setup, so you don't get the obligatory flatbed scanner reflections.

That is not really how it works. Yes, a camera has a fixed array of dots and to extract the maximum amount of information from that sensor, you have to maximize the number of pixels covered by your image.

A scanner essentially just moves a single pixel over the source material. The resolution is limited not by how much of the flatbed you cover, but by how small that single pixel is and how precisely it can be moved. (And even the size of the sensor is fungible, if you can move it precisely you can oversample.)

Scanners have their downsides, no doubt, but raw resolution is their upside. The downside is the time it takes for a scanner to make an image -- that is why we carry around ultra-expensive silicon wafers for photography, and not a single light-sensitive element that moves around under the lens's image circle.

I think hat the parent is saying is that you can adjust the lens of the camera to have the subject better fit the whole frame. For a scanner you’d need a specific adaptor and there aren’t so many readily available.

You can image non-flat things

Oh, it definitely is a Speedy Pro - I believe (it is quite a few years since I last looked at this in detail) we’re looking at a c.861; my display back (c.1863) Speedmaster is very similar, but I believe there are a couple of minor differences (in addition to the added decoration, given the display back, turning the 861/1861 into an 863/1863) - so, my guess is 861, a somewhat modified Lemania, if memory serves.)

It does indeed appear to be an .861, likely circa 1969 given its serial number.

Sensor shifting is implemented in several camera bodies these days. Olympus has been doing it for a few years.

This has been a great tool for Micro 4/3 sensors to beat their sensor size constraints for higher-res imaging.

There have been rumors for a while of an Olympus implementation of this system that works fast enough for some moving subjects as well, which would be huge if true - you could do this handheld instead of a tripod.

This article mentions a way to manually do it without the need for stabilizing hardware: https://petapixel.com/2015/02/21/a-practical-guide-to-creati...

> It also has a nice side effect of putting less demand on the glass

I don't follow. How does a higher resolution capture -- however it's accomplished -- not demand more from the glass?

The Hasselblad sensor shift is amazing for photographing art or still life (i.e. cars etc) - I'm sure you can imagine how excellent this is for commercial photography.

Only for a smaller niche of commercial photography - the one that sells large fine art prints. You don't really need high resolution for billboards or posters - they are viewed from large distances. Magazines are not printed with that great quality. Even high quality photo books are more than fine with 20-30 megapixel shots.

A lot of cameras are sticking at 24MP because that's where a lot of lenses start to reach a weak point. Also, file size to utility becomes a consideration because most people expect cloud/sharing. I feel like 100+MP is going to be an enthusiast thing for a long time. It seems a lot like car speeds. Every car sold today can go over 100 miles per hour. Some people still care about how fast their car can go, but most don't really think about it or even know their car's top speed.

I have a couple of photography colleagues who use the 50MP Canon 5Ds. The detail they gather is astounding, more than my 21.1MP can grab even with a teleconvertor on the lens. So they can shoot with shorter focal lengths, which means less weight and better stabilisation.

They can then go into post-processing and "crop the hell outta it" and still end up with more detail than my image.

It is a cliché but high-density sensors really are a "game changer". They give so much potential to reframe the photo after shooting, selecting the most relevant area of the frame and still having enough detail to blow-up to prints.

Locked down on a good tripod with a remote shutter release, the very best prime 35mm lenses will resolve about 40Mpix worth of useful detail at their optimum aperture. The pro-level PJ zooms (24-70mm and 70-200mm f/2.8) will manage about 30Mpix on a good day. Handheld, all bets are off - you're at the mercy of image stabilisation and your own hands. In many circumstances, you'll have no choice but to use a sub-optimal ISO or aperture setting to get the shot, losing even more effective resolution.

High resolution sensors are very useful in some applications, but you'll often see little or no benefit in real-world use. All the major manufacturers offer both a >40Mpix and a ~24Mpix sensor option in their high-end cameras, because resolution isn't the be-all and end-all of a sensor.

The Canon Mark II 'big white' primes are said to be designed for 100MP. The smaller Mark II / III zooms are indeed in the 40 - 50 range. Bear in mind that the 24MP crop sensors have been pushing past the 50MP-full-frame-equivalent for several years and really show up any lens flaw.

The one example that stands out is the Canon 70-200 f4 IS zoom, which has an ageing early-digital-era reputation for being razor-sharp but never managed that even on my old 1Ds III and looks like mush on a 5Ds. Hence a Mark II is to be released this year.

Interestingly the even older non-IS f4 version was on the list of lenses approved for the 5Ds, which shows how difficult adding those extra optics makes lens design:

https://www.the-digital-picture.com/News/News-Post.aspx?News...

Top of the line tools really make a difference in photography. Back in 2007 when taking a photography class, you could see the difference in quality between a top of the line DSLR that some senior students had and point-and-shoot cameras without interchangeable lenses.

Even though the composition was good from both, the difference in quality really made you biased towards photos produced by the higher end gear. Depth of field, sharpness, noise, all those things.

And everything you mentioned as well - not having to take multiple shots, because you can always crop later, not having to get as close to the subject for people photography, so perhaps getting more shots.

> So they can shoot with shorter focal lengths, which means less weight and better stabilisation.

It doesn’t mean better stabilization once they crop. The weight isn’t a major factor in stabilization. The effective magnification is.

I think Dingaling means that, if you're holding the camera in your hand, you'll be able to hold it steadier if it weighs less.

Probably, but it's not really true. It's more comfortable to hold a lighter lens but the heavier lens has more inertia so resists micro-tremors more effectively. In the end it's pretty much a wash.

Exactly this. Being able to crop is why I use a 42MP camera.

I have a 5Ds. Once you have used one, you know that medium format is dead in the water as a category. That said, for most shooting scenarios they are definitely complete overkill, and lens quality becomes paramount.

A good quality 2x teleconverter will make your sensor the equivalent of 80mp in this context.

Ie with the same lens, a 20mp camera with 2x teleconverter will resolve more detail than a 50mp sensor.

But it's a losing race; adding a 2x teleconvertor halves the amount of light received, so I have to drop my shutter speed and / or raise the ISO to compensate. Both of which are disruptive to image detail.

Plus the distorting effect of the extra glass in the lightpath.

-Unless I am much mistaken, a 2x teleconverter will cost you two stops - meaning 1/4 of the light. (I’ve mostly used a rangefinder in later years; they do not lend themselves well to teleconverters. :-)

You lose almost no light.

You are changing the area of the sensor the part of the scene is spread over.

Its a triviality as a result of us expressing the F number as the ratio of the focal length and the aperture size.

The effective SNR of the sensor is increased due to more sensor being used for a given part of the scene, so the lower light level does not increase noise comparatively.

People often forget that 2x increase in one dimension equates to a 4x increase in area and number of pixels. I didn't expect that of this community though, which is how I'm interpreting the down voting here.

This is my profession so if the down voters know something I don't I'm very interested to hear where I've gone wrong.

Yeah. Teleconverters are true lens "stress" testers.

I tend to think this 120MP sensor is probably going to be sold commercially as a surveillance product first.

There's a lot of need for products like this in stadiums, arenas, large public areas etc.

That marketing video seemed to be highlighting specifically that. They even had the shot of the security woman at the very end.

See also WAAS [11] and ARGUS [22]

[11] https://foxtrotalpha.jalopnik.com/how-one-new-drone-tech-fin...

[22] https://www.youtube.com/watch?v=0p4BQ1XzwDg

The file size is insane though - imagine having to store hours of video from hundreds of 120MP cameras. Even the bandwidth required to continuously stream hundreds of those cameras to a central location would be enormous.

If it's for surveillance, you connect them to gigabit fiber.

Sony had an ad for a eye-of-sauron type application for a gigapixel imager array that could track someone across an entire city from one location. Surveillance will be the major consumer of >100MP sensors.

> I feel like 100+MP is going to be an enthusiast thing for a long time.

Or maybe not? Optical system design often allows you to trade one thing for another. So for instance, years ago, light field capture (each "pixel" is a little lens in front of several real pixels) was bottlenecked on image size. You're spending resolution to buy angle information. An order-of-magnitude or two of resolution. There's also exposure (HDR), and time (anti-blur), and so on. 100 MP sounds like a lot... until you start spending it in order-of-magnitude chunks to obtain other system features.

The lens quality barrier at 24MP isn't arbitrary; that's basically the equivalent digital resolution you get from 35mm film.

There are all kinds of films out there, including very very high resolution ones. Gigabitfilm is supposedly 720 lp/mm, which would give you

    36*24*(720*2)^2 = 1.79 gigapixels

(theoretically, of course)

Not true - obviously it slightly depends on the grain of the film but the theoretical resolution of high quality 35mm film is ~87MP

Full frame, obviously.

You'd need to be using Kodak Technical Pan at a rated speed of 12 to 16 (before filter factors, for continuous-tone images with smooth gradients) in order to get there, though. Kodachrome 25 (in the K14 process version) and Ektar 25 weren't terribly far behind, but neither had a great deal of latitude. And you'd also need something a little flatter than the typical film aperture and pressure plate to really use it. More typical 35mm films - those with usable sensitivities for general photography - wouldn't get you anywhere near there.

Sure, there’s tons of variation of course and 87 is the upper limit, but from what I’ve read it’s still the 40-50MP range before you hit the limits of a decent typical 35mm film.

Lens quality does start to become a factor a bit earlier than that too.

A lot of the comparisons I've read (and internally Leica's own tests) say otherwise.

https://petapixel.com/2014/12/18/comparing-image-quality-fil...

I mean... in that very article the 80MP camera looks worse than film (tho more sharpened, go go digital image processing!)

You appear to have provided counter evidence to your own position.

Take a look at the 35mm section, that's explicitly not the conclusion of the article.

"The results are very impressive and from our comparisons we think scanned color slide film has a digital equivalent resolution of between 12MP and 24MP depending on what aspect you look at (12MP for overall sharpness appearance, 14-16MP for luminosity resolution, and 24MP for color resolution). These figures are estimates based on a good Imacon or Drum scan and would be even higher for projected or enlarged images."

So you're basically reading that and going, well I think I'm right anyway.

Much higher brightness range on film too (potentially -- depends on the film of course).

To Canon's credit starting with the 35mm f1.4L II a few years ago, they've started releasing pro grade lenses with the optical performance to handle extreme resolution sensors.

Still though 24MP is more than enough for me.

The lens 'weak point' depends on the aperture you're using. Even relatively inexpensive lenses will resolve a hell of a lot of detail at f/8.

I do agree that it's an enthusiast thing, and will remain so for a long time, but it's not that hard to utilize that resolution.

I think most lenses, cheap or not, already lose resolution at f/8 due to diffraction.

Most of my lenses have resolving peaks somewhere between f/4 and f/5.6.

We need a better file format that JPG for sharing these large images.

JPEG is a compressed format; it intentionally sacrifices quality for smaller size. It still has its uses, but it should never be used where image quality is the primary concern. There are many widely-adopted alternatives that suit other purposes and have no data loss. One alternative, PNG, is supported everywhere that matters, but there are plenty of others.

We need lossy state of the art codecs, PNG sucks for large images. TIFF and EXR are also great. But I think we need a lossy large image format, probably based on AV1 and similar technology.

JPEG is still used because of ubiquity and because it always was good enough, we have over two decades of better replacements lined up if we need them.

We have jpeg2000 and dng.

> I feel like 100+MP is going to be an enthusiast thing for a long time. It seems a lot like car speeds.

Is there anything in photography analogous to speed limit? Because that's often a reason why folks don't go over say 80mph not to mention well over 100.

If tomorrow 100MP cameras appeared and were cheap enough I don't think people would think twice about getting them.

> Is there anything in photography analogous to speed limit?

Maybe price, weight and the size of the RAW files they generate? At some point it's not worth to invest more.

Yes, people would buy them, but they do require very disciplined shooting. That kind of resolution is very unforgiving of even slight movement or mis-focus.

That's true if you're using 35MM equivalent lenses. One must switch to medium format sizes.

Unless you're looking at the Mamiya or Hasselblad MF lenses, good 35mm lenses tend to out-resolve MF and definitely LF. Projection circle is only part of the effective resolution.

Only really cheap zoom lenses would max out at 24MP. Most lenses will have no problem with 100MP.

I'd say even 24MP is an enthusiast resolution. Most people are now fine with their smartphones, and the most they do is share their photos on social media sites in very low resolution.

Ever climbing megapixel is awesome and all, but I want ever increasing ISO performance. Low light is where the real innovation is. Especially in my line of business with food photography.

Yep. Pixels are great, but the big hole is in dynamic range. I'd love to get some high shutter speed shots of night time action in natural light, but I rarely have a use for creating prints the size of a drive-in movie screen.

The smaller the pixels get, the closer to log response dynamic range per unit area you're going to get.

That's simple because of statistics of photons registering at random locations. Happens even if readout noise or whatnot increases.

For those who doubt: imagine a sensor with so high resolution that each pixel is only being hit by just one photon or none at all. Such a sensor would have the ultimate dynamic range. No burned highlights at all. (Of course you might want to downsample the result before use. :))

Although you seem to be after quantum efficiency and low noise. You certainly wouldn't be getting either of those with the ultimate dynamic range extreme pixel count sensor.

I'm curious... what kind of action were you thinking of?

I'm not the OP, but nighttime street photography would be great, such as walking around NYC at night. Right now I am having to handhold at 1/10th second and 3200 ISO, which makes any moving subject a blur (in addition to the shake from my hands).

If you're doing still photos of food, surely you can just put more light on the subject or use a longer exposure.

With modern sensors, ISO is mainly limited by thermal noise. Astrophotographers often use cooled sensors to allow higher sensitivity, but it's tricky to cool it very much without moisture from the air condensing inside or on the lenses.

> If you're doing still photos of food, surely you can just put more light on the subject or use a longer exposure.

Longer exposure, yes, though that means having to set up a tripod for every shot which significantly increases the time it takes to get a shoot done.

Turning up the lighting is very difficult for food photography. Lights are hot and food doesn't last long under studio lights. (That's also why a lot of food photography has such shallow depth of field — a wider aperture lets in more light.)

>Turning up the lighting is very difficult for food photography. Lights are hot and food doesn't last long under studio lights. (That's also why a lot of food photography has such shallow depth of field — a wider aperture lets in more light.)

Sure, but that's only if you use non-LED continuous lighting. Alternatively, you could use flash and then not need a longer exposure. As long as you can balance the ambient with a low enough continuous light that it wont affect the food. Plenty of ways to skin the cat..

Flashes, and external lights are not really realistic in a busy restaurant environment. Majority of my shots are using natural light. I usually scope out places ahead of time, find the best seating by time of day, near windows.

I guess we all assumed still pictures of food in a studio instead of just asking you :)

Another option is to use a body that offers stabilization. Olympus has gotten their stabilization good enough that you can shoot handheld at pretty slow shutter speeds: https://blog.mingthein.com/2016/11/02/review-olympus-e-m1-ma...

I guess you haven't read this then :)

https://blog.mingthein.com/2016/08/19/stabilisation-is-good-...

He wrote that before he wrote the article I linked to. Also, this article mentions some deficiencies of stabilization in certain circumstances but nowhere does it claim that it's useless.

Oh, of course. I agree with you that it is not useless and I myself rely on it quite frequently. I guess I should have made my point clear. The relevant point of the blog post is that IS is not very effective at high resolution. I don't know if you've shot with a 5DSR or any other 50+ MP camera It is very tricky to get a super sharp image since the tiniest bit of movement gets picked up. I can only imagine it gets worse at 120MP..

Right, which is why I think Olympus has had success with it. They primarily (maybe only?) make micro 4/3 bodies, so the resolution is still relatively low (some of their bodies are at 16MP, others 20). I don't photograph food, but I'd think that a stabilized m4/3 body would be very useful for that purpose if one wants to avoid bringing a tripod.

Food photography is an art in its own right. Photographing real food is super hard, it wilts before your eyes. Many food pictures are of extremely expensive fakes! Hard to make the fake, but once you have it it is so much easier to work with than organic material in the studio.

Absolutely true about the fakes. A lot of times they use wax models of food.

I remember one example, most photographs of breakfast cereal (for advertisements or that appear on the box) use white school glue for the milk. Real milk is too thin and also ends up looking bluish in the photos.

Nice set of samples:

https://petapixel.com/2016/03/24/photos-show-secret-tricks-f...

Spot the non-food items :)

Astrophotographers also care about quantum efficiency and often shoot at higher ISOs. Astrophotographers use cooled sensors for a different reason. Long exposures build up heat, and heat causes noise, which is why you need to cool overclocked CPUs too.

Thermal noise is not relevant for food photography. Input referred read noise will dominate here 100% of the time, unless you're taking pictures of things lit so dark that you can't see them yourself.

Nikon, Sony, Canon, and Fuji have all been making large improvements in low light performance in the last 10 years with only modest improvements to increased resolution in comparison. Nikon for example released the D7500 with a 21mp sensor instead of 24mp and it has improved low light performance.

I still think this video from 2014 is pretty impressive:

https://www.youtube.com/watch?v=a1W-bPyYR0k

I wonder how latest models (a7 III would be a good candidate maybe?) fare in comparison?

The A7S is a very different camera. A high MP camera is rarely meant for low light photography. Traditionally its meant for studio/portrait/product photography work where the editor can crop into tiny details if required, and also the light is very carefully controlled/modified.

Yes, I'd like a 4-8MP sensor with great dynamic range and low noise. Since resolutions have gone beyond the point where most users notice, why aren't camera companies trading it off for other benefits?

It might be possible to use pixel binning to get a 30mp image at high iso that looks great, using the luminance of 4 pixels to derive 1 pixel.

ISO performance has already increased amazingly over the past two decades. I wonder how much more possible headroom it has.

There is a lot of headroom. What shows up in consumer cameras is never state of the art. I'm hoping BSI sensors will drive the noise down even further.

Why would you need better low light performance for food photography?

Busy, dimly lit restaurants, where I haven't gone ahead and explicitly setup shoots with owners. I'm not just going to walk in with a tripod at a random lunch. I promote food and cocktails for the Western, NY region. Mainly Rochester / Fingerlakes. So I'm out and about trying new places, ultimately deciding if I'm going to promote them before I even meet owners. I don't want preferential treatment, which usually comes when they know who you are. My followers expect that I'm giving them only the best food in the area, and not what someone has given me for free or done up extra special.

This is like a blog or what?

Instagram. If you're not on there promoting your food, then your restaurant is losing a ton of exposure.

If you're curious.

https://instagram.com/rocfoodies

I see. I was just thinking traditional studio style food photography with a food stylist, fake ice cubes, etc.

Nokia made two cameras, the Lumia 1020 and Lumia 808, that featured a 41MP camera. For the most part, the use case was pretty much to immediately downsample or use the high resolution as a way to create a synthetic high-quality digital zoom. The low-light quality was excellent as well.

The sensor was 8.80×6.60 mm (as opposed to the APS-H sensor here, at 29.22×20.20mm), which means 3x the sensor density. I've been waiting for a while for this "Pureview" technology to show up anywhere else; it's surprising to me that it has not.

I had a lumia 950 which had “only” a 20 MP sensor and the low light photography was stunning, better than what I saw with my own eyes in some cases. It was very slow though (even in daylight), which made it useless at action shots, so I guess other phones optimize for speed over image quality.

I did make some absolutely breathtaking landscape shots with that 950, and I’ve never made pictures that nice since switching to an iphone.

pic or it never happened :P

The 1020 still wipes the floor with any of the current smartphone cameras, though the new dual-camera phones are a very interesting development. You can stand to lose a lot of chroma resolution if you still have lumina resolution. Just try chroma noise reduction while retaining lumina noise in your favorite image editor.

>> The sensor was 8.80×6.60 mm

FWIW, this sensor is more commonly known as a 2/3" sensor, which was used in some recent Fujifilm compact cameras (X10-X30, XF1 etc) and is also commonly used for broadcast television cameras.

I had an X10/X20, and that smallish sensor could produce some really nice results in spite of its size.

I've been pining for a greyscale digital camera. I think maybe Leica had one but it seemed overly pricey (to the surprise of no, I suppose). Price aside though, having only the one I am aware of is slim pickings.

With no Bayer filter I expect four times the pixel density with no gain in noise (and in fact a slight improvement in speed since the lossiness of the filter is removed from the light path).

I guess I want the sharp, medium format portraits we saw from the 1960's — but digital.

I've considered building my own by either:

1) using a telescope CCD and building my own camera, or 2) try to remove the Bayer filter from the CCD.

I dislike both those options....

PhaseOne sell a 100MP IQ3 B&W digital back, which is basically exactly what you're after. The quality of this is brilliant - I've seen some really astonishing work out of it - but it might be out of price range a bit - you're looking at north of $50k for a working setup, so many people just rent them when needed.

There might be some B&W scanning backs that are a bit cheaper, but those don't really work too well if your subject is moving.

>I guess I want the sharp, medium format portraits we saw from the 1960's — but digital

The closest you're going to get to that now is the MF cameras from Fuji and Hasselblad. Of course, this will set you back a pretty bitcoin.

I have to say though, I just got a Nikon D850, and it's pretty nice--nearly the equal of my Hasselblad 503.

Phase One make a 100MP black and white digital back.

Slightly out of my price range.

https://www.phaseone.com/IQ3-100MP-Achromatic

Yeah, I can feed my Hasselblad a lot of Pan F for that kind of money.

Leica had the M8 Monochrom but yeah, Leicas are at least as much about the red logo on the camera as they are about actual good photography. Their lenses are quite good, but not unequaled.

Fuji has some cameras with a monochrome mode I think. And their sensor doesn't use a low-pass filter so it's sharper at the same MP count than some others.

https://en.wikipedia.org/wiki/Fujifilm_X-Trans_sensor

Leica never made a monochrome camera based on the M8, but both on the M9 and the M(240). The latter is current. A monochrome version of the M10 is espected to eventually arrive. Leica cameras might be pricy, but they are excellent photography tools. You might get higher res cameras from the competition, but for example the M is special with its small size and excellent lenses.

Thanks for the correction -- for a short time I was captivated by the idea of owning a Leica Monochrom when I first heard about them but could never part with the cash.

Yes they make excellent cameras -- particularly the film ones. With care they will last indefinitely. The digital Leicas had some problems at least early on. Maybe they are sorted out by now.

But for the money, you have to just want to own a Leica. The actual camera and optics are no longer unequaled (and for a lot less money) if they ever were.

You can certainly get cameras at a better price/performance level. So on a pure financial scale, they are difficult to justificat. But a M is unequaled by definition - it is the only digital rangefinder in production. Not to rave too much about rangefinder photography by itself, that means a very compact full-frame camera (roughly A7 size), but where it stands out is the tiny lenses. Even µFT struggles to match the M lens sizes. And at minimum for my back, lens size matters a lot. A M camera with 3 lenses fits into an absolutely tiny bag.

You could shoot film.

"Lomography" is cool enough now that even where I live BW film developers are back in business. And if it dies down I might just start doing it myself.

B&W is easy to develop yourself, too. And good 35mm cameras are sitting in closets everywhere. They can be had for next to nothing.

In case anyone else is interested in Bayer filters...

https://en.wikipedia.org/wiki/Bayer_filter

Leica still makes a monochrome model of their M cameras, the current one features a 24 mp cmos sensor based on the M(240). It is expected that there will be a monochrome camera based on the current M10 in about a year. They are excellent cameras with a stunning resolution, if you are into B/W photography.

you could pay someone else to do it https://www.monochromeimaging.com/

What would be useful on their site would be a comparison showing the same image shot with an unmodified sensor and then again with their filter-less sensor. If no one can tell the difference, is it worth it?

The problem of RAW conversion remains because every converter will still de-Bayer files from such a camera and thus degrade the IQ.

I don't know, seems like the conversion works well enough. You're always going to edit and change files in export anyways.

The site has a page explaining how to get around it.

You can get rawtherapee to not do it.

You could try a Sigma. I love my DP3 Merrill (keeping in mind the inherent limitations).

Yep, Foveon is one of the most underrated camera technologies. It's basically a medium format camera in a compact digital package. I fervently wish Sigma would make some progress on camera usability!

One example review, for readers who don't know about Sigma's awesome camera tech: https://www.photigy.com/sigma-foveon-sensor-review-dp3-merri...

I love how they choose 1080p to compare it to as though we should give a shit what 1080p stills look like on a 5D Mark IV.

Yeah, 1080p is less impressive when you call it 2MP.

This is going to be big for surveillance cams. Just put a fisheye lens on them and collect a hemisphere. There are already 4K fisheye surveillance cams. You do pan, tilt, and zoom later.

Next step is face popout in a GPU. Faces, license plates, and other items of interest can be saved at high resolution and the rest compressed to a lower level of detail.

Ah, I was waiting for the surveillance example in the video... they didn't disappoint and it was the last example.

Did I miss something? It was the only example! The video goes directly from

"Look at the high-res awesome awesomeness!" to

"Future Applications: Wide-area monitoring (a.k.a. video surveillance)"

EDIT: I find that depressing. Seriously? With tech like that, you can't at least come up with a more benign example? You really have to throw yourself at mayors and other hawks first?

You don't even need such a high-resolution sensor. At that level, you might as well have multiple cameras and stitch the results together.

And the military has already done it. See here:

https://en.wikipedia.org/wiki/Gorgon_Stare

and here:

https://en.wikipedia.org/wiki/ARGUS-IS

Sometimes the high-resolution sensor is used in place of a zoom lens (i.e., use "digital zoom" to 1:1). This way the camera's primary lens is simpler because it has a shorter focal length and there's less to break down. Can probably bring the costs down a little too.

Wikipedia seems to be confused here about the "unit cost" of Gorgon Stare (listed as $15M) and the "development cost" (also noted as $15M).

Or is Sierra Nevada Corp. really making the full cost in profit as of the second unit sold?!

Anyhow, those two examples are probably in a different price range, no?

Stadiums that serve alcohol use and need cameras like this. The Toronto Blue Jays have had issues with beer throwing idiots during post season games in 2015 and 2016. The incident in 2016 was particularly bad when a guy threw a full beer can at an Orioles outfielder during the wild card game.

Oh I never had any illusions of this not being used for surveillance. But that the manufacturer doesn't even try to frame it primarily as "look at the amazing photography you can make with this" is a sad sign of the times for me.

tips hat

> All pixel progressive reading of 9.4fps

> RAW photos shot by the sensor weigh in at 210MB each.

1974 MB/s. Finally an use-case for Intels fancy 3d-xpoint memory?

Knowing Canon it still only supports 1080p video...

The fact that this captures at 9.4 fps means comparing the image to 1080p kind of makes sense, but I wish they had a comparison to a more practical still resolution between 24-50 MP.

Hmmm... Technical merits aside, have you noticed that the only "future applications" they could come up with is surveillance?

> If you’d like one in the future, though, you should probably start stockpiling hard drives: RAW photos shot by the sensor weigh in at 210MB each.

I would be more worried about the amount of RAM you need to comfortably edit these photos.

Happy to hear that Canon is going medical.

OK but where is my SX70?

Confirmed for iPhone 2020

Reply on news.ycombinator.com