[jsherman999]

.

The last thread got contentious and was closed (as I knew it would be,) so let's try to keep it civil! Leave the snark at the door and try to write and answer in good faith, meaning no Gish-Gallop or face-saving - stick to the facts and your true assumptions. And keep an open mind in case your past assumptions might be in need of a re-eval!

Let's start slowly.

Show of hands:

How many people think the micro four thirds (smaller than aps-c) Olympus OMD E-M5 likely produces as noisy images as the FF Canon 5DIII, or 5DS?

Does the K5 or Nikon D7000 likely produce the same noise in output as the D800?

(all images displayed on screen or print at same size, say in relatively low light at ISO6400 - one isn't blown up more than the other, and taken at the same framing, exposure.)

---

Last edited by jsherman999; 03-15-2016 at 08:52 AM.

[Gimbal]

But, isn’t ”all images displayed on screen or print at same size - one isn't blown up more than the other” impossible, unless they all have the same resolution?

[jsherman999]

But, isn’t ”all images displayed on screen or print at same size - one isn't blown up more than the other” impossible, unless they all have the same resolution?

No.... Same display dimensions. When you display a 16MP shot in your preview pane on your screen, is it presented smaller to you than say a 24MP or 36MP shot?

Or when you print that 16MP shot at 8x10, and then print a 24MP shot at 8x10 - this is what's meant by 'same display size'.

[Gimbal]

Right, but if you display a 16Mp and a 36Mp at the same size they have to be enlarged or shrunk differently. At least if they are displayed on a computer screen since the screen has a fixed resolution which both pictures have to match.

For instance a 4k monitor is about 8Mp, so a 36Mp picture has to be shrunk a lot more than a 16Mp picture. The pictures have to be individually shrunk/enlarged to end up with the same size.

[jsherman999]

Right, but if you display a 16Mp and a 36Mp at the same size they have to be enlarged or shrunk differently. At least if they are displayed on a computer screen since the screen has a fixed resolution which both pictures have to match.

For instance a 4k monitor is about 8Mp, so a 36Mp picture has to be shrunk a lot more than a 16Mp picture. The pictures have to be individually shrunk/enlarged to end up with the same size.

That is very true, and hits one one of the points that will be touched on in this thread as it progresses - the effect of normalization of images, either between different number of MP, different size sensors, or both. In other words, the difference between viewing a 100% crop/swatch of an image and the whole image at display size.

To keep on track, let's move through the first question first - which would you expect to show more noise in the same-dimension (whole) image, the m/43 OMD or the FF Canon?

[mecrox]

That is very true, and hits one one of the points that will be touched on in this thread as it progresses - the effect of normalization of images, either between different number of MP, different size sensors, or both. In other words, the difference between viewing a 100% crop/swatch of an image and the whole image at display size.

To keep on track, let's move through the first question first - which would you expect to show more noise in the same-dimension (whole) image, the m/43 OMD or the FF Canon?

At base ISO (or nearly) why should either show any noise? This surely only becomes an issue as one moves up the ISO scale or starts to blow stuff up hugely or push things too far recovering shadows, etc. Everything has to start from a properly exposed image and, at a guess, most images aren't. In addition, some sensors or processors seem to handle noise better than others.

So I would guess, if exposed properly then at base neither (they would both have no noise worth the name) and at 6400 the smaller sensor.

[jsherman999]

At base ISO (or nearly) why should either show any noise? This surely only becomes an issue as one moves up the ISO scale or starts to blow stuff up hugely or push things too far recovering shadows, etc. Everything has to start from a properly exposed image and, at a guess, most images aren't. In addition, some sensors or processors seem to handle noise better than others.

So I would guess, if exposed properly then at base neither (they would both have no noise worth the name) and at 6400 the smaller sensor.

You're right, let's take a typical low-light shot at say ISO 6400 (edited OP to reflect) otherwise the noise profile becomes academic and not as visible.

Which would you expect to show more noise, or would they be about the same? Imagine two preview panes on your 27'' display next to each other, or two 11x16 prints, and again, same framing and exposure (f/stop and shutter speed.) (EDIT: I see you answered, thanks, so the question is for other participants.)

(all we're doing here is putting the 'reality check' phase of the thread first as opposed to burying it under other posts - it's OK at this phase to just say what you expect, m4/3 vs. FF, FF newer-gen - which one should show less image noise. We can and will move from expectations to representations of data and even images.)



.

---

Last edited by jsherman999; 03-15-2016 at 09:16 AM.

[mecrox]

You're right, let's take a typical low-light shot at say ISO 6400 (edited OP to reflect) otherwise the noise profile becomes academic and not as visible.

Which would you expect to show more noise, or would they be about the same? Imagine two preview panes on your 27'' display next to each other, or two 11x16 prints, and again, same framing and exposure (f/stop and shutter speed.)

.

I would expect similar noise on-screen but less noise in print with the FF sensor? My understanding is that the issue isn't what things look like at 100 per cent but after reduction/rescaling?

[LensBeginner]

I'd say noise at 100% would depend on pixel size and sensor tech generation, ie on pixel count, sensor size and sensor tech.

At same print size a larger and newer sensor (eg FF CMOS as opposed to 1/2.3 CCD, just to single out two extremes") should show less noise, but noise itself would only visible from a certain ISO upwards (that however depends on print size, viewing distance etc...)

Open mind and all that

[jsherman999]

I'd say noise at 100% would depend on pixel size and sensor tech generation

Yes

, ie on pixel count,

Yes

sensor size

No. (not 'noise at 100%' - that has nothing to do with sensor size)

and sensor tech.

Yes.

At same print size a larger and newer sensor (eg FF CMOS as opposed to 1/2.3 CCD, just to single out two extremes") should show less noise, but noise itself would only visible from a certain ISO upwards (that however depends on print size, viewing distance etc...)

Of course, but to keep things less extreme and to stick with our comparson, how about two CMOS sensors, on m43 and one FF, FF equal or newer-gen?

Open mind and all that

Thanks, it's all I can ask

---------- Post added 03-15-16 at 11:03 AM ----------

I would expect similar noise on-screen but less noise in print with the FF sensor? My understanding is that the issue isn't what things look like at 100 per cent but after reduction/rescaling?

Yes, but keep in mind 'on screen' can be same display size as well, it doesn't depend on printing (original question isn't asking about 100%, yet.) So, at same display size either on screen or in print, which would you expect to have more noise?

[LensBeginner]

Yes



Yes



No. (not 'noise at 100%' - that has nothing to do with sensor size)


*snip*

I meant as a byproduct of those two quantities which allow you to calculate the one I listed before... it's also independent of pixel count if two sensor are of different size.
sensor size / pixel count = pixel size
So yes, we are saying the same thing

To the point: I'd guess more noise m43 (but not much, and only visible from a certain ISO upwards), almost indistinguishable result between FF "old" and "new" CMOS sensors.

[Rondec]

The struggle with equivalence is real. One major issue with it is that it requires a similar generation Sony sensor. Pop in a Toshiba or Canon sensor and things don't work as well. Try to compare a K10 to a K5 and everything is totally weird. But hey, it is a starting place.

The magical thing happens when you downsize higher megapixel images to a normalized end result. This, I think is what folks struggle with the most. Why turning a 36 megapixel image into an 8 megapixel image should give an extra EV of dynamic range is beyond me. I understand why it decreases noise, but the D810 at base iso has EV 13.6 dynamic range - screen and 14.6 EV print. It feels like the dynamic range is what it is. On the other hand, a K5 has more dynamic range than a K3, regardless of sampling and this could be a reason that someone would choose a K5 II over a K3, if landscape is their main interest.

Oh well, I think I understand it well enough. As I have said many times before, the thing that aggravates me the most is that equivalence proponents use it to show why full frame is the best format -- even better than medium format -- for all purposes.

[jsherman999]

The struggle with equivalence is real. One major issue with it is that it requires a similar generation Sony sensor. Pop in a Toshiba or Canon sensor and things don't work as well. Try to compare a K10 to a K5 and everything is totally weird. But hey, it is a starting place.

Yes, and well-put. This is why unless someone has the sensor data available and looks it up or already owns both cameras, you can't 'know' one sensor will perform better than another, even if one is bigger than the other. That's why this phase of the question is about expectations - which would you expect to be less noisy at the same display sizes (not 100%,) m43 or FF, with the FF slightly newer.

The magical thing happens when you downsize higher megapixel images to a normalized end result.

We're getting to that A side question: a higher-MP image taken from similar sensor gen and same sensor size, would you expect the image noise to be similar after downsizing? And: what if it was same sensor gen, same pixel density, but a larger sensor (same framing/exposure,) how would the images compare after normalizing display size?

...

Oh well, I think I understand it well enough. As I have said many times before, the thing that aggravates me the most is that equivalence proponents use it to show why full frame is the best format -- even better than medium format -- for all purposes.

Depends on what anyone ever means by 'better', and then when someone says 'for all purposes' they usually really mean 'for their purposes.'

And that goes both ways You get proponents of any format making those claims, which is really about IQ preferences and use cases, and usually, cost.

But if I could get back to it - what would be your expectation for the original question, with those specific bodies?

.

[mohb]

what would be your expectation for the original question, with those specific bodies?

you need to get out more

[Rondec]

Yes, and well-put. This is why unless someone has the sensor data available and looks it up or already owns both cameras, you can't 'know' one sensor will perform better than another, even if one is bigger than the other. That's why this phase of the question is about expectations - which would you expect to be less noisy at the same display sizes (not 100%,) m43 or FF, with the FF slightly newer.




We're getting to that A side question: a higher-MP image taken from similar sensor gen and same sensor size, would you expect the image noise to be similar after downsizing? And: what if it was same sensor gen, same pixel density, but a larger sensor (same framing/exposure,) how would the images compare after normalizing display size?



Depends on what anyone ever means by 'better', and then when someone says 'for all purposes' they usually really mean 'for their purposes.'

And that goes both ways You get proponents of any format making those claims, which is really about IQ preferences and use cases, and usually, cost.

But if I could get back to it - what would be your expectation for the original question, with those specific bodies?

.

In general, newer sensors and larger sensors tend to perform better. This is not always the case, the K10 did better at low iso than the K7 did and about the same at high iso. The A7s sensor will outperform all of these cameras in high iso situations, but not in low iso situations. As before mentioned, Canon sensors tend to do more poorly than Sony sensors -- not with regard to SNR, but with regard to dynamic range in the lower iso ranges.

I think that all of these features are relatively unimportant when it comes to making a good image. Images catch the eye when they contain the right light, subject, and composition and particularly for web size images, there are few situations where you can truly say that the size of the sensor made or broke a particular image. I have seen some images where noise was detrimental to the final image, but I have seen plenty too where shallow depth of field harmed the final result.

Good photographers work with the tools they have at hand and figure out ways to make them shine. They seldom compare camera specs, but often compare images, as those are the goal of photography, not SNR, dynamic range, or shallow depth of field.

(thought this black and white image might fit a little with your style, Jay)