UKC

Official release due tomorrow?

http://www.canonwatch.com/cw5-canon-eos-5ds-eos-5ds-r-ef-11-24mm-f4l-press-...

In reply to Douglas Griffin:

Had to happen as Canon didn't really have an answer to the Nikon D810. More interesting is the 11-24, that's crazy wide.

In reply to James Rushforth:


Not sure I'd have much (if any) use for one. The new camera would be nice, however...

In reply to Douglas Griffin: 50.6 mp - my pc struggles with 24 !

In reply to Brian:

Aye, you've got to wonder where this will end!

In reply to Douglas Griffin:

Still plenty of head room available. Cannon have old, cpu feature size. Lots more room can be found on the chip for a reduction in noise and power consumption etc.

"When it comes to manufacturing semiconductors, the “process generation” refers to how small the transistors are. The smaller the transistor, the less power it needs to do the same thing. That’s how our modern laptops with 14nm FinFET CPUs from Intel can have hours upon hours of battery life. The smaller the transistor though, the more precise your manufacturing has to be and the more leakage that can occur (which generates heat). When it comes to CMOS camera sensors, process technology matters just as much as it does with regular computer processors and heat equals noise.

Right now, Canon sensors including the EOS-1Dx are manufactured on a 500nm process. That’s the kind of process technology from the AMD 486DX4 era and Pentiums in the early 1990s. The Nikon D4 sensor uses a 250 nm process. That moves us up to the Pentium II or AMD K6 era. The Sony A7R/A7s sensors as well as the D800 sensor are built on an 180nm aluminum platform. Pentium III or Athlon XP."

http://www.slrlounge.com/samsung-nx1-vs-full-frame-cmos/

In reply to Douglas Griffin:

It will end when the consumer stop upgrading their kit when the gain is minimal. I had a Nikon D800. It was a great camera but you had to have top quality lenses to take advantage of the 36m pixel sensor. I eventually got fed up with lugging around heavy cameras and lenses so I sold the lot and bought the Fuji XT-1 as my main body and an XE-1 as a back up/ travel camera. I do a certain amount of studio work and even with the kit lenses am more than happy with the results.

In reply to Douglas Griffin:

So has canon's attempt at catch up almost certainly driven by the Nikon 810e payed off.

on the available evidnce thus far....no. The Sensor in the 5Ds has a dynamic range and signal/noise ratio which is average at best. what it will do very well is slow your machine down in process with massive files largely usefless with the vast majority posting to the net (at 72dpi)and not printing large format.

At more than 3K for the body....oh dear sadly Canon at first soundings appear to have come up way short with this one...such a shame for canon shooter.

In reply to Douglas Griffin:

Olympus have come up with a rather nifty heath-robertson approach to increasing the megapixels on their 16 mpx E-M5 using their sensor shift technology + electronic shutter. No use for high speed stuff, but for landscapes sounds very nifty:

"Another advantage of Olympus’ ultra-precise 5-axis stabilisation system is the potential to capture extremely high-resolution still images that exceed the normal capacity of its 16-Megapixel LIVE MOS sensor. By moving the sensor between each shot and merging eight single exposures into one, the E-M5 Mark II delivers a 40-Megapixel image with breathtaking levels of detail and radiance... snip over the top marketing ...."

In reply to Douglas Griffin:

I just noticed this (been on holiday ignoring the rest of the world), at last. You can thank me for the release as I bought a 5d3 about three weeks ago. Bang goes about 40% of its value, probably.

I look forward to see if it is as well rounded as the D810 or a flawed torch-bearer like the D800/E.

In reply to dread-i:


That's not entirely a bad thing you know. In fact I suspect it's a very good thing. Analog image sensors like Canon's CMOS sensors are very unlike digital CPUs.

When the pixel size on an image sensor is perhaps 6,000nm; what does moving from 500nm fabs to 14nm fabs get you? Too much leakage current, too much resistance in analog read outs, too much noise susceptibility, a whole new set of design rules to adapt and reinterpret for an application far outside those they were created for. Oh, and a massive increase in cost.

There is big room for improvement in the image sensors, not however from a process shrink but from replacing the Bayer (colour) filter array with something that conserves photons instead of throwing 66% of them away. Canon and many others have put patents out on various approaches (micro optic beamsplitter and mirror arrays, multi layer detectors, photonics structures etc) - that's the next big thing.

In reply to wintertree:


I know bugger all about this kind of thing, obviously, which is why I looked at this thread. But just a question, are CMOS sensors analogue? I thought that an analogue sensor (CCD) is one where there is a proportional relationship between light falling on the sensor and the electrons stored; between light and electric charge. But a CMOS sensor converts the analogue signal to digital in the sensor itself; rather than transferring the charge, it is detected and amplified by transistors at the column of a load of pixels. So each pixel has a charge to voltage conversion and the chip includes other stuff, amplifiers, noise correction... and the output is digital.

In reply to Mike Highbury:


Good question! The pixel sensing array is analog. There is a matrix of readout wires that is switched with transistors to convey the analog charge from detector pixels to analog to digital conversion electronics. Those ADCs and associated gumf can live on the detector chip itself or on a separate chip, as with CCDs. Conversion electronics are rarely replicated per pixel as this would increase cost, complexity and decrease the area available for light detection. Exceptions to this are research devices like SPAD arrays where digital logic must be replicated per-pixel by necessity.

I believe Canon implement their signal conversion off chip in their DIGIC processor chip(s).

To further clarify, with a CMOS sensor, charge is still accumulated (or depleted depending on your view) during an exposure, and then read out at the end of an exposure. a difference to a CCD is that the readout uses a separate switching matrix - a lot like a RAM chip - instead of transferring charge through adjacent pixels themselves. This gives a lot more flexibility in regioning, rolling shitter, differential exposure in a frame and removes the possibility of analog binning.