FF low-light advantage vs. crop?

jblaschke wrote in post #9470175
Yeah, New Braunfels is quite a photogenic town. I work in San Marcos, which ain't half bad, either!

Austinite here,,, I'll be visiting Schlitterbahn at least once this summer though and maybe float the river...

District_History_Fan wrote in post #9470151
The 50D offers some pretty dang good high ISO performance. The 5D2 is better, probably by a couple of stops.

As a previous 50D owner, I wouldn't say it offered "pretty dang good high ISO performance". Everybody has a different idea of whats good though. For me anything over 800 was not really usable for print, etc...

xarqi wrote in post #9470210
Sensor size is totally irrelevant. Photosite (loosely termed "pixel") size matters.

Precisely and unequivocally, 100% just plain wrong.

Sensor size is the only physical metric that really counts - thanks to how technology has advanced, pixel size is essentially irrelevant.

In noise terms, my 30D is worse than my 40D is worse than my 7D - and this isn't because of NR (I don't convert in DPP, so in-camera NR is ignored), it's because more/smaller pixels does not increase noise across the sensor: it might mean more noise per pixel, but the sensor is getting the same amount of light whether the light is shared between 10 mp or 18 mp.

Factor in improvements in sensor design such as gapless microlenses that are closer to the photodiodes; cooler-running sensors; more sensitive photodiodes; reduction in the size of the underlying/support circuitry, and you get better noise performance at the same time as more and smaller photosites.

"Full frame" sensors have a noise advantage because they're bigger and more light hits them.

Want something to read? Read this. I particularly recommend the section titled Why aren't denser sensors noisier at higher sensitivity?

GSeries1 wrote in post #9474717
Austinite here,,, I'll be visiting Schlitterbahn at least once this summer though and maybe float the river...

Schlitterbahn is best experienced either early or late in the season. If you must go at the height of summer, mid-week. Otherwise you'll spend most of your time in line. The Comal River is great, and the Guadalupe seems to have good flow right now. During the week is the best time for those--they're just a logjam of people otherwise. Burn one of those vacation days, it'll be worth it!

Thanks for all the sensor input, guys. You've given me lots of reading material!

Some of the quietest and best low light cams aren't FF. FF has nothing to do with low light performance and may even diminish it...

The nuts and bolts of it: http://www.clarkvision​.com/articles/index.ht​ml#part_4

blackhawk wrote in post #9475344
FF has nothing to do with low light performance

Sorry, but it just does, and Mr Clark unequivocally agrees.

On this page he says (with my emphasis):

No ambiguity there. Other things being equal, bigger sensors are cleaner sensors - and FF is as big as it gets for DLRs right now.

Keith R wrote in post #9476451
Sorry, but it just does, and Mr Clark unequivocally agrees.

On this page he says (with my emphasis):

No ambiguity there. Other things being equal, bigger sensors are cleaner sensors - and FF is as big as it gets for DLRs right now.

No that's not what he's saying as both the 1DMK-2 and the MK-3 are out performing the FF's in the lower ISO ranges, having a higher dynamite range in these ISO settings.
The 1D MK-4 has not weighed in yet either...
The models like the 5D II and the D-3 that have better low light sensitivity are sacrificing dynamic range at lower ISO settings to do so.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

Pixel Pitch Size not sensor size factors in heavily by limiting contrast due to the effects of diffraction.

HOSTED PHOTO
please log in to view hosted photos in full size.

OK, try this on for size.

1) Larger photosites are inherently less noisy, therefore photosite size is a significant contributor to noise performance.

2) For the purposes of generating RAW data for one photosite, the total number of photosites in the sensor is unimportant. There is no cross-talk, either of signal or of noise.

3) The only noise present in the raw data for the entire sensor is the noise present in the data for each photosite.

4) Sensor size does not affect the noise in the data.

That is the logic that lead me to state that photosite size is important in determining noise, whereas sensor size is not.

So...
5) This noise is what I think is being called "per pixel noise", but there are claims of a different sort. If you accept the logic above, then this other type of noise can only derive ultimately from pixel noise as modified during post-processing.

6) There is noisy data, the magnitude of the noise depending on photosite size, but not sensor size. If, during processing, this data is interpolated, or averaged in any way then random noise will tend to cancel.

7) To reach the same print size, data from a larger sensor will need a greater degree of interpolation since the required magnification is less. Therefore, in the post-processing of a larger format image, there will necessarily be a greater degree of noise cancellation than for a smaller format. The noise remaining after processing is the "per frame" noise.

Where does that leave us?
The ultimate source of noise is the photosite, and larger photosites are less noisy.
Post-processing of noisy data can cancel some noise, the degree of cancellation depending on sensor size.

Are larger sensors inherently less noisy?
No. However, they allow a greater degree of noise cancellation during post-processing.

Discuss. (4 points)

Hotrodguru wrote in post #9474746
As a previous 50D owner, I wouldn't say it offered "pretty dang good high ISO performance". Everybody has a different idea of whats good though. For me anything over 800 was not really usable for print, etc...

What size were you printing?

I have a macro I printed at 8x10 that was heavily cropped, and it looks great to my eyes. I imagine 1600 would be quite useable, and 3200 as well for smaller prints or if I did some NR processing on it.

xarqi wrote in post #9470210
Sensor size is totally irrelevant. Photosite (loosely termed "pixel") size matters.

so is it by coincidence that larger sensors have larger photosites?

ed rader

ed rader wrote in post #9478582
so is it by coincidence that larger sensors have larger photosites?

ed rader

Coincidence of design, yes. Nothing about sensor size dictates photosite size, nor vice versa.

That said, since both larger sensors and larger photosites offer advantages, it does seem sensible to use the extra space available on a larger sensor to accommodate larger, rather than more photosites.

White papers for the 1D MK-4:

http://www.usa.canon.c​om …-1D%20Mark%20IV%20WP1.p​df

The pixel size is 5.7 um for the MK-4
7.2 for the MK-3
8.2 for MK-2

Interesting trade offs.

jblaschke wrote in post #9470115
I know the sensor design and internal processing affects this as well, but I wanted to see if there was an established rule for the sensor size itself.

Yes, there is an established rule. Multiply the f-number by the crop factor to get equal noise. Since the crop factor is 1.6X between the FF and APS-C, then f/2.8 ISO 160 on APS-C has the same noise as f/4.5 ISO 400 on FF. To directly convert the crop factor to the number of stops, use this formula:

log_2(crop_factor) * 2

Now, only accounts for sensor size. The performance characteristics of the sensor and electronics will have an effect as well. In some cases, the performance per area is the same, such as in the case of the 7D and 5D2, so the 5D2 has the 1.3 stop advantage.

This effect of sensor size is very easy to demonstrate through experimentation. Please see the following comparisons:

Noise scales with DOF for equal technology

Here are two example crops from APS-C and FF areas:

70mm f/4 ISO 640 on APS-C

111mm f/6.3 ISO 1600 on FF

And here they are again embedded directly:

70mm f/4 ISO 640 on APS-C: