[kadajawi]

ISO noise is not dependent of number of pixels, but of the size of every pixel, between other factors. But in the last years, even that some small progress was made, the size is still the most important factor. Because of this, if the pixel size is bigger, for a better ISO, the number of it on a certain surface is smaller. So, ISO noise is dependent of the number of pixels, if we count the same sensor size.

Almost. Last year Samsung has started making a 28 MP APS-C sensor that has the same pixel size as a normal 20 MP APS-C sensor. The tech that went into it only has a big effect on high pixel counts... i.e. if you'd apply it to a 16 MP sensor the benefits wouldn't be as big as they are with a 28 MP sensor. Otherwise I agree, the bigger the pixels the better the low ISO performance, but pixel size is not necessarily related to the MP.

[ElJamoquio]

Lets keep it simple. it is possible. that it does. (pixel density)
Otherwise i would not see any reason (except framerate and increased picture quality thanks to sensor pixels sized 6.94µm) that Canons top of the notch bodies "only" feature 18MP.
Despite that i am pretty sure, that we wont reach 10fps with 36MP. And with 36 MP you are as far away from endless raw shooting as we are from world peace now.

JMHO

Pixel size doesn't affect noise in any actually-discernable way.

Some people think it does because they looked at images at 1:1, which effectively 'zooms' the larger MP images closer to the image.


Canon's sensor division is behind the times and has been so for at least the last 7 years.... and contrary to that last point are now introducing a 50 MP camera IIRC (which I may not be).

Valid reasons for smaller MP is smaller file size and the associated faster storage + potentially quicker frame rates.

---------- Post added 05-04-15 at 10:18 PM ----------

Almost. Last year Samsung has started making a 28 MP APS-C sensor that has the same pixel size as a normal 20 MP APS-C sensor. The tech that went into it only has a big effect on high pixel counts... i.e. if you'd apply it to a 16 MP sensor the benefits wouldn't be as big as they are with a 28 MP sensor. Otherwise I agree, the bigger the pixels the better the low ISO performance, but pixel size is not necessarily related to the MP.

They're measuring something unconventionally if they say they have more pixels of the same size on the same size sensor.

Could be they're talking about back illumination or something but if so the microlenses already effectively accomplish a similar thing.

[patarok]

I hardly doubt that Canon is far behind...
If all that is true and pixel density doesnt affect noise and bigger pixels only have influence on iq at LOW ISO??? levels..., than please tell us, why: on compact cameras with smaller sensors there is more noise on hi iso pictures than on low iso pictures... ? . I also think if we go further with this theme we should move to some other thread.

[JimmyDranox]

A simple comparison will show the truth.

Sony A7s, A7 and A7r. Same technology, same manufacturer, same sensor size, all launched in several month. But different in pixel size.

[kadajawi]

Pixel size doesn't affect noise in any actually-discernable way.

Some people think it does because they looked at images at 1:1, which effectively 'zooms' the larger MP images closer to the image.


Canon's sensor division is behind the times and has been so for at least the last 7 years.... and contrary to that last point are now introducing a 50 MP camera IIRC (which I may not be).

Valid reasons for smaller MP is smaller file size and the associated faster storage + potentially quicker frame rates.

---------- Post added 05-04-15 at 10:18 PM ----------



They're measuring something unconventionally if they say they have more pixels of the same size on the same size sensor.

Could be they're talking about back illumination or something but if so the microlenses already effectively accomplish a similar thing.

It is backside illuminated, yes. And yeah, the microlenses should do it, buuuut... somehow it is a better sensor than what the competition has to offer.

If you even out the resolution, then yes, ISO performance isn't related to megapixels... to a certain extend. Microlenses can only do so much, and at some point pixels are going to be smaller. Dynamic range should be a different matter though. Fewer MP generally means bigger buckets.

[clackers]

ISO noise is, for the intent at hand, completely independent of the number of pixels.

When you output the image the noise will be lower due to better averaging.

[Gimbal]

Dynamic range should be a different matter though. Fewer MP generally means bigger buckets.

Yes, as in a wider bucket, but that doesn’t mean that the bucket is deeper. So when the photon-rain starts it will fill up just as quickly as a smaller bucket with the same depth.

[patarok]

A simple comparison will show the truth.

Sony A7s, A7 and A7r. Same technology, same manufacturer, same sensor size, all launched in several month. But different in pixel size.

What truth? What is it about? Have i missed something?

[JimmyDranox]

ISO noise level. Take a look at comparometers, and see.

[Ian Stuart Forsyth]

A simple comparison will show the truth.

Sony A7s, A7 and A7r. Same technology, same manufacturer, same sensor size, all launched in several month. But different in pixel size.

And the truth is that the A7s uses a sensor technology that is not found in any other sony sensor and is a generation more advanced than the any previous sony design, thus its a design function that gives the high iso ability not the size of the pixel. The first implementation of the design was in a small Aptina sensor with high density pixel count with much smaller pixels than what is found in the A7s and it to showed the same advantages as the much larger pixel in the A7s

From the work done by Jack Hogan, JimKasson and Bobn2 the A7”s uses a switchable conversion gain that we first seen in Aptina sensors and marketed as DRPix. At around iso1600 we can see the a7’s does something abnormal to previous sony sensors.

“Conversion gain is pretty much the reciprocal of the capacitance of the parts which connect the collected to the read transistor gate. What it controls is the 'gain' from e- to output voltage (since Q = CV). This controls both input referred read noise (assuming that read noise is a voltage noise, it will be translated to photoelecrons by the capacitance - i.e 1/CG.). It also controls the FWC, sine in a CMOS design it is the swing of the output transistor that is the limit. Thus small capacitance gives low read noise and low FWC. High capacitance gives high read noise and high FWC. The switchable conversion gain technique (patented by Aptina and branded DRPix) has a capacitor in the pixel which can be switched in at low ISOs and switched out at high ISOs.'

So what it looks like they are able to do with the A7’s below iso 1600 is to implement a High capacitance which gives a higher read noise but with a larger FWC, Above 1600 they switch to a smaller capacitance which in turn gives a lower read noise but also with a lower FWC.

There was a good paper on how Aptina the first that I know was to implement this and the original article I sniped this from but I cannot find it at this time.
So as you can see its a design that gives the a7s the iso advantage not the size of the pixel
(From memory) Eric Fossum also gave us some insight to this new design but at this time I am unable to find it.

[JimmyDranox]

I know about that Aptina technology.

So, your conclusion is that this technology, and only this, is the factor which affects ISO noise? Nothing else? If you think so, it's OK for me. Even if I don't agree.

No more comments from me from now on. I'm not in a mood to debate anymore. Forget what I have said.

[Rondec]

I know about that Aptina technology.

So, your conclusion is that this technology, and only this, is the factor which affects ISO noise? Nothing else? If you think so, it's OK for me. Even if I don't agree.

No more comments from me from now on. I'm not in a mood to debate anymore. Forget what I have said.

I think the question is why the A7s sensor performs better (normalized) than the sensors in the A7 and A7r, which have similar SNR curves. The A7s is not better at lower isos, but at high iso, it is amazingly able to hold on to dynamic range. It does seem to have different tech in it than the other full frame sensors out there, although it all seems like magic to me.

[Thomas Cantwell]

How about an enormous viewfinder, like .96x magnification.

[richandfleur]

How about an enormous viewfinder

I think that's actually pretty critical.


The viewfinder is the primary difference nowadays between DSLR's and Mirrorless cameras, now that mirrorless has phase detection autofocus on the sensor.


Manual focus is greatly aided by focus peaking and digital zoom, so for DSLR's to maintain their perceived benefit, it needs to really push and maximise this benefit.


Again, nothing here that is specifically FF based, but having said that FF should allow for a larger viewfinder, to match the good old days of 35mm film viewfinders.
Whilst it should be awesome and all, I've personally never been that blown away by the APS-C optical viewfinder, despite Pentax being one of the best of the bunch in this area.




What could be 'different' about Pentax FF?
Decent video would please me.
Real Sensor Shift based stabilisation during video, mixed with focus peaking and maybe digital zoom. Main thing is the focussing and stabilisation aids need to work during video capture, not beforehand only as presently.

[kadajawi]

I think that's actually pretty critical.


The viewfinder is the primary difference nowadays between DSLR's and Mirrorless cameras, now that mirrorless has phase detection autofocus on the sensor.


Manual focus is greatly aided by focus peaking and digital zoom, so for DSLR's to maintain their perceived benefit, it needs to really push and maximise this benefit.


Again, nothing here that is specifically FF based, but having said that FF should allow for a larger viewfinder, to match the good old days of 35mm film viewfinders.
Whilst it should be awesome and all, I've personally never been that blown away by the APS-C optical viewfinder, despite Pentax being one of the best of the bunch in this area.




What could be 'different' about Pentax FF?
Decent video would please me.
Real Sensor Shift based stabilisation during video, mixed with focus peaking and maybe digital zoom. Main thing is the focussing and stabilisation aids need to work during video capture, not beforehand only as presently.

Yup. I agree. And higher bitrates, maybe older codecs like MJPEG at really high bitrates for when every ounce of image quality matters, and file size doesn't. Clean HDMI out at 1080p at least, possibly 4K. Internal 4K recording wouldn't hurt, but not necessary. The digital zoom should be a 1:1 crop from the sensor (to avoid aliasing), and it should be possible to enable it while recording. Hit a button and you're in there. Maybe consult Magic Lantern, hire them.

As for the viewfinder I think the way for a truly great viewfinder is using that tech that Ricoh has patented... a mirror that can change translucency. Quite a lot of light is lost by the need to redirect some light to the AF sensors... but what if the light is only redirected/let through the mirror when focusing? What if otherwise 100% of the light goes to the OVF? The result would be a spectacular OVF that is like in the 35mm film days before the invention of AF. This would also make a ton of sense in an APS-C camera, even more so IMHO (an APS-C camera that can almost compete with the competitors FF offerings?!).

OVFs need to improve, cause EVFs surely do. The A6000 one is really big, and looks pretty good. Resolution could be higher though. The OM-D EM-10 viewfinder is a bit smaller (but still ok), and even rougher, but the lag... I don't see a lag. Still have to see the NX-1, which should beat all of these, with even less lag and a better image hopefully. At one point they will overtake the OVF.