I think I know why Canon sensors lagged behind Sony's...

After reading through literally trillions and trillions of complaints over the last couple of years about the superiority on Sony's sensors (biggest imaging sensor manufacturer) I wondered why would that be?

As a semiconductor industry EE since the mid '90s, I've lived and breathed technological breakthroughs, manufacturing triumphs and marketting disasters. I feel like I know enough to have an understanding as to when something "doesn't feel right".

So, I've had this nagging feeling that something was amiss regarding Canon's sensor development. That it was progressing, but perhaps at a slower pace than it should and definitely at an even slower pace than the competition.

It felt like Canon, the technological and manufacturing behemoth that it certainly is, was "stuck".

Or was it?

Is it possible that the competition knew what was coming and they felt like they had to accelerate their development as fast as possible so as to claim as much market share before "the big event"?

What "big event"?

It looks like Canon is going to come out with a technology that will, in my estimate, create the "forever camera". OK, so in our day and age "forever" is probably between 5-10 years. Who of you has electronics that are that old? Put your hand down grampa!

The technology I'm expecting Canon to release soon (that will also render its own dSLR line "obsolete" or at least outdated) is AUTOISO per individual pixel.

I think this is the technology Canon focused the majority of its process/design development braintrust.

Now it's starting to make sense, doesn't it?
Releasing evolutionary updates of their excellent sensor line while working on the revolutionary sensor while releasing world-beater lenses to prepare the ground for it all.

Another thing that now makes sense was the Canon Body/Lens inequality of the last 3-4 years...
Why has the "Lens" side of Canon overwhelmingly released amazing lenses while the "Sensor" side of Canon has been so conservative and muted?

I guess the answer is the "Body/Sensor" side has been working on the solution that will make maximum use of the lenses...

So, what is the tl;dr version of this post?
If I were you, I'd sit tight and wait for the next generation of sensors from Canon that will leapfrog the competition.
That's what I'm going to do...
Wait for that FF with "infinite" DR, great high ISO handling, flippy touchscreen, wifi, gps etc. etc. etc...

The only thing left would be a decent photographer behind it to take advantage of it all...

Interesting concept.

While working in the yard today I listened to AM talk radio on a Nordmende Globetraveler Jr. purchased in 1963. I also happen to have my father's, purchased at the same time. German electronics of course...

It will be interesting to see what they do come out with over the next year or so. I hope you're right about a big breakthrough.

Auto ISO on a pixel level sounds like a techno-myth. If you have the ability to set ISO on a per pixel level you will need a metering system capable of metering on a per pixel level. Then the end result is that every pixel will be set to 18% gray rendering quite a boring picture.

andrikos wrote in post #15065496
After reading through literally trillions and trillions of complaints over the last couple of years ...

You understand there haven't been that many seconds elapsed in the last couple of years, right?

Let me ask you a question cause I've been curious about this. Canon, Nikon, Sony, and Pentax basically own the DSLR market in the USA, right? Am I right that only Canon develops and uses their own senors, and the rest use Sony sensors?

Why does Canon do this?

ScullenCrossBones wrote in post #15065632
You understand there haven't been that many seconds elapsed in the last couple of years, right?

The complainers are very fast.

gjl711 wrote in post #15065631
Auto ISO on a pixel level sounds like a techno-myth. If you have the ability to set ISO on a per pixel level you will need a metering system capable of metering on a per pixel level. Then the end result is that every pixel will be set to 18% gray rendering quite a boring picture.

I would agree with this...

Meters "look" at a complete scene and not the individual photons...

Fun speculation, but what's the evidence backing up such a wild rumor? Whenever someone goes off about the 70D or 3D or whathaveyou, they at least point to patents Canon's filed in recent years. I'd expect something as dramatic as what the OP is suggesting would leave a significant patent trail.

Sony has something of a financial advantage because of A) contracted market share and B) extensive electronics R&D dating back decades. Canon held a sensor advantage for a number of years, but has now been surpassed (modestly so, in relative terms). But Canon's not thrown in the towel. Rather than a spectacular game-changer, I expect Canon to close the gap by focusing on controlling high ISO noise and increased dynamic range. We're reaching the point where even a base level imaging sensor out strips what even professional photographers need (note I said "need" not "want').

In all honesty, I've been somewhat surprised that Canon hasn't launched a medium format DSLR system to leverage their in-house sensor production. I'd expect them to produce digital backs for other systems, if nothing else...

The Dark Knight wrote in post #15065634
Canon, Nikon, Sony, and Pentax basically own the DSLR market in the USA, right? Am I right that only Canon develops and uses their own senors, and the rest use Sony sensors?

Why does Canon do this?

Canon and Sony are massive companies, and they own their chip fabrication plants because they can ammortalize the costs across a wide product range (Canon and Sony make a lot more imaging products than just DSLRs).

The smaller companies cannot afford to spend billions on chip fabrication labs, and so have to buy their sensors from a third-party supplier. This means that individual sensors cost them more than it costs Canon and Sony to use their own tech (because they are obviously marking up the sensors at a profit when selling them to rival companies).

Basically, Canon can, Nikon can't. Nikon are a way smaller company than Canon.

See:
http://www.googlefight​.com …B&word1=canon&w​ord2=nikon

A new metering system would need to be developed that does "look" at the complete scene, and decides how to break up that scene, pixel by pixel, to increase ISO sensitivity in the potentially underexposed areas such as shadows or the foreground in a sunset scene. Why would this not be possible?

Essentially, metering evaluates the scene and chooses an exposure. Or meters a spot and chooses an exposure for that spot, which may not be the best exposure for the rest of the scene.

Why would it not be possible to develop a "multi spot metering" system which meters a grid or matrix of the scene (the more fine the grid, the more processing power that will be needed for this to work quickly) and sets the ISO in each grid/matrix cell to match the reading of that spot meter'.

Per pixel might sound crazy at first but it's not out of the question. If spot metering can work now, that concept can be applied to meter an entire scene and set the correct ISO values in different parts of the scene. It would probably make sense to have a user defined setting for a maximum ISO spread, such that your image is limited to say only 2 or 3 stops of ISO differential.

Interesting idea. Not sure if waiting a year for this is going to cut it, though - but you never know. For now I'll just get a 6D and keep using my GND filters

No reason why you couldn't record luminance values on a pixel by pixel basis. Doesn't the 1Dx already do that for metering (on 100K pixels)? Somewhere along the line you record how quickly each pixel goes from 0-256 or whatever and you have a record of relatively how "dark" that pixel is. The data stream would be absolutely massive but that is just a processing issue. Surely you could clip it to optimize the DR based on user settings (i.e. how much DR you want in the scene)

There isn't any need generally to have the DR of the image maximized all the time. But at least it would give you the option of expanding it without the need for multiple exposures.

I was sort of thinking about this a while back too. So much for making millions off of that patent.

jblaschke wrote in post #15065692
Fun speculation, but what's the evidence backing up such a wild rumor? Whenever someone goes off about the 70D or 3D or whathaveyou, they at least point to patents Canon's filed in recent years. I'd expect something as dramatic as what the OP is suggesting would leave a significant patent trail.

For your viewing pleasure:
Filed Dec 2009 (right after the introduction of 7D). Interesting timing, no?

http://appft1.uspto.go​v …l=50&s1=2010014​1792.PGNR.


It is dense and hard to read if not technical AND a patent lawyer, but the meat is there...

Copidosoma wrote in post #15065756
No reason why you couldn't record luminance values on a pixel by pixel basis. Doesn't the 1Dx already do that for metering (on 100K pixels)? Somewhere along the line you record how quickly each pixel goes from 0-256 or whatever and you have a record of relatively how "dark" that pixel is. The data stream would be absolutely massive but that is just a processing issue. Surely you could clip it to optimize the DR based on user settings (i.e. how much DR you want in the scene)

There isn't any need generally to have the DR of the image maximized all the time. But at least it would give you the option of expanding it without the need for multiple exposures.

I was sort of thinking about this a while back too. So much for making millions off of that patent.

I'm with you with the patent idea... I'm crying away my billions...

You say massive processing power...
Interestingly, the Digic 5+ (intro June 2012) is not a slouch...
Add a couple of them and you have unheard of processing power...

ScullenCrossBones wrote in post #15065632
You understand there haven't been that many seconds elapsed in the last couple of years, right?

Should I guide you to the definition of "hyperbole for effect"?