Quick question concerning low-light photography

[Abitconfused]

If we assume that, generally, a full-frame sensor functions more efficiently in low light than a half-frame sensor. Is this due to the larger surface area of the full-frame sensor or larger photosites or both?

[Dave Humphries]

I may be wrong Ed, but being pragmatic, I'd say it's the larger photosites that result in less noise that allows the manufacturers to offer what seems to be a 'more efficient' sensor (by dint of how much less they need to 'amplify' the signal at base ISO).

IIRC the effect of the larger surface area of the sensor and the bigger lens required for it mean that is largely 'cancelled out'.

No doubt I may learn something from others answering.

Dave

[Shadowman]

I may be wrong Ed, but being pragmatic, I'd say it's the larger photosites that result in less noise that allows the manufacturers to offer what seems to be a 'more efficient' sensor (by dint of how much less they need to 'amplify' the signal at base ISO).

IIRC the effect of the larger surface area of the sensor and the bigger lens required for it mean that is largely 'cancelled out'.

No doubt I may learn something from others answering.

Dave

I've also heard that the anti-aliasing filter or lack of it also helps the full frame sensor produce higher resolution images, most cameras have the filter to reduce distortions.

[rpcrowe]

If we assume that, generally, a full-frame sensor functions more efficiently in low light than a half-frame sensor. Is this due to the larger surface area of the full-frame sensor or larger photosites or both?

You may state "generally" the full frame sensor is more efficient in low light.

However, that can be a function of the individual sensor. I am thinking that the crop sensor of my 7D2 is more efficient all around than the full frame sensor of my 6D2. When working in low light, the dynamic range of the sensor also has a bearing on the "efficiency".

[Manfred M]

I've also heard that the anti-aliasing filter or lack of it also helps the full frame sensor produce higher resolution images, most cameras have the filter to reduce distortions.

Yes and no. More modern cameras have "weaker" AA filters than older models. These do not reduce distortion but are there to reduce the risk of Moiré patterns occurring.

[Manfred M]

A simple analogy is to put two measuring cups out beside each other in the rain. Use a standard 4-cup (1 liter) and and a standard 1-cup (250 ml) cup. Go out and measure the total amount of water in each. The 4-cup measure will have more water because it has a much larger diameter opening than the 1-cup measure.

Replace the rain with photons and the two cup sizes with larger and smaller sensors. Larger diameter sensors will measure a greater number of photons.

[DanK]

All of this assumes that the sensors are otherwise similar in quality.

It's really not FF/crop that matters, but rather pixel density: to fit more pixels in, you have to make them smaller. Some of the most expensive current full frame cameras have pixel densities similar to that of older crop cameras. They take advantage of advances in sensor technology to compensate for small photosite size.

[dem]

Larger sensor area means better ISO performance, larger photosites mean better dynamic range.

[DanK]

Larger sensor area means better ISO performance, larger photosites mean better dynamic range.

Can you explain? I can't see why the size of the sensor, rather than the photosites, would matter.

[Manfred M]

Larger sensor area means better ISO performance, larger photosites mean better dynamic range.

Sorry that does not make any sense. Larger sensor means higher resolution versus a similar photosite size on a smaller sensor. There are significantly more photosites (pixels) in a larger sensor (assuming as Dan has pointed out, we are writing about the same generation of sensors).

Dynamic range, signal-to-noise ratio, colour bit depth are all related to photosite performance. Any ISO that is higher than the "base" ISO of the sensor comes from amplifying the signal.

[xpatUSA]

I've also heard that the anti-aliasing filter or lack of it also helps the full frame sensor produce higher resolution images

John, this might give a better understanding of AA vs no AA:

http://www.falklumo.com/lumolabs/art...0AAFilter.html

Just read the first few pages ...

most cameras have the filter to reduce distortions.

Not sure that's right, is there a link? My cameras use lens profiles for that purpose.

[Shadowman]

John, this might give a better understanding of AA vs no AA:

http://www.falklumo.com/lumolabs/art...0AAFilter.html

Just read the first few pages ...



Not sure that's right, is there a link? My cameras use lens profiles for that purpose.

Ted,

Thanks for the link, my comments on the filter were based on past information I've read on lowlight performances of cameras and the comment about distortion can be found here. In my quest for better lowlight performance, less noise; most information commented on sensor size and also the anti aliasing filter, cameras without the AA were thought to be better performers. I've had good results with the D750 which has the AA filter and is FF, wondered why the D500 which is crop format was considered a better lowlight performer, assumed it was the lack of the AA filter. Had a chance to play with the D500 and it is good at high ISO settings but like all systems there is a limit to how high you can go, my preference regarding ISO settings is to only go as high as I have to.

http://www.digitalcamera-hq.com/arti...liasing-filter

[dem]

Dynamic range, signal-to-noise ratio, colour bit depth are all related to photosite performance.

Sure. I think we all agree that the parameters that are defined for a single photosite are not affected by the size of the sensor.

Can you explain? I can't see why the size of the sensor, rather than the photosites, would matter.

When talking about ISO performance, people often say that APS-C is a stop better than m43 and full frame is a stop better than APS-C. To me this means that if you are happy shooting full frame at ISO 6400, you will get similar noise levels in your prints at ISO 3200 on the APS-C format and at ISO 1600 on the m43 regardless the pixel count.

50Mpx full frame Canon 5ds and 16Mpx m43 Olympus M10 mk ii were both made in 2015 and have similar pixel density:

http://snapsort.com/compare/Canon-EO...E-M10-II/specs

If we produce an A4 print from shots taken at ISO 6400 with each camera, the full frame image is going to look much cleaner. Yes, at a pixel level, 100% magnification on the computer screen, the noise in both shots will look quite similar. However, the full frame shot image has got 3 times as many pixels to fit on the same A4 print!

If we look at "Low light performance" figures on snapsort.com for some Nikon cameras:

APS-C:
===============================
D3200 24Mpx Mar 2012 1,131 ISO
D5200 24Mpx Oct 2012 1,284 ISO
D5300 24Mpx Oct 2013 1,338 ISO
D7100 24Mpx Feb 2013 1,256 ISO
D7200 24Mpx Mar 2015 1,333 ISO

Full frame:
===============================
D800 36.2Mpx Dec 2011 2,979 ISO
D600 24.2Mpx May 2012 2,980 ISO
D610 24.2Mpx Oct 2013 2,925 ISO
D810 36.2Mpx Jun 2014 2,979 ISO
D750 24.3Mpx Sep 2014 2,956 ISO

we'll see that regardless the megapixel count the full frame format is at least twice as good as the APS-C in terms of ISO noise, e.g.

http://snapsort.com/compare/Nikon-D7...kon-D750/specs

[Manfred M]

When talking about ISO performance, people often say that APS-C is a stop better than m43 and full frame is a stop better than APS-C. To me this means that if you are happy shooting full frame at ISO 6400, you will get similar noise levels in your prints at ISO 3200 on the APS-C format and at ISO 1600 on the m43 regardless the pixel count.

Now I understand your mean by "ISO Performance" and that was not clear to me from your original post.

You seem to be looking purely at the "digital noise" side of sensor performance and here I would agree with you. The issue is tied to the level of magnification one needs to upsample or downsample the images from various sizes of sensors and how that affects the final look. All things being equal; fewer pixels with smaller sensor size means more the noise will appear more grainy and hence be more noticeable than with a larger sensor. Heavily crop a larger sensor image and this advantage disappears.

[xpatUSA]

Ted,

Thanks for the link, my comments on the filter were based on past information I've read on lowlight performances of cameras and the comment about distortion can be found here.

http://www.digitalcamera-hq.com/arti...liasing-filter

Thanks for your link, John. I see there that they use the term "distortion" to mean moiré - which is known by a few of us simply as a form of aliasing, not "distortion".

Loose terminology strikes again ...

[xpatUSA]

I may be wrong Ed, but being pragmatic, I'd say it's the larger photosites that result in less noise that allows the manufacturers to offer what seems to be a 'more efficient' sensor (by dint of how much less they need to 'amplify' the signal at base ISO).

Dave

Dave, the standard definitions in ISO 12232 of "base" ISO do not include "amplification of the signal".

They are dependent on the exposure level in lux-sec that saturates the sensor (Hsat).

For example 'ISO Speed' is defined by ISO = 78/Hsat.

There's lot more to it than that - but perhaps too technical for this forum.

[shreds]

Dave, the standard definitions in ISO 12232 of "base" ISO do not include "amplification of the signal".

…….There's lot more to it than that - but perhaps too technical for this forum.

Ted,

As a very long time contributor and participant of this forum, I would beg to differ and find that even the most complex of discussions are useful and interesting to those of us lurking in the shadows.

It is what sets CiC apart from the rest in many respects….very few technical subjects are taboo, and I would rather read and learn about matters that are on the periphery of my knowledge (we all can always learn….it never stops), that gives a better understanding.

Sean set the CiC forum up ten years ago as a 'learning' site and that is not just for beginners, so if you can enlighten us further, I for one would be delighted to read on. Technical or otherwise.

(Which sets another thought alight….ten years of CiC ought to be celebrated in some way…..or have I missed that thread. I had better set it off by thanking Sean and everyone involved for creating such a superb resource).

[xpatUSA]

Originally Posted by xpatUSA Dave, the standard definitions in ISO 12232 of "base" ISO do not include "amplification of the signal".

…….There's lot more to it than that - but perhaps too technical for this forum.

Ted,

As a very long time contributor and participant of this forum, I would beg to differ and find that even the most complex of discussions are useful and interesting to those of us lurking in the shadows.

Sean set the CiC forum up ten years ago as a 'learning' site and that is not just for beginners, so if you can enlighten us further, I for one would be delighted to read on. Technical or otherwise.

I'll try. By basing measured ISO on sensor saturation, similarly to ye olde film days, the amount of amplification (gain betwixt sensor and the ADC) goes away as a factor UNLESS that gain is enough to saturate the the ADC instead of the sensor 'wells'.

One way to determine the point of saturation of a sensor is to shoot a gray card on it's white side until the raw histogram starts a 'brick wall' on the right. Then calculate the exposure for that saturation from:

I = πTvL_obj cos⁴(T) / 4N ² = π * 0.9 * 0.98 * 0.9783L_obj / 4N ² = 0.6777L_obj / N ²

http://www.imatest.com/docs/sensitivity_ei/

where I is luminous intensity in cd/m2

From that we can calculate lux - and from the exposure time get the exposure in lux-sec that saturated the sensor a.k.a Hsat.

Note that photo-cell size doesn't come into it either because, all other things being equal, lux is an area-based measurement.

In theory, the camera metering should always give an output of 'mid gray' for an average gray scene (the gray card, either side). In practice, that rarely happens, mostly it's a bit over - no idea why.

A 'base ISO" shot should come, in sRGB or aRGB, at about 100/255.

A "SOS ISO" shot should come, in sRGB or aRGB, at about 118/255.

[Manfred M]

"SOS ISO"

I'm not familiar with that term Ted. What is "SOS ISO"?

[xpatUSA]

I'm not familiar with that term Ted. What is "SOS ISO"?

Sorry, Manfred.

It stands for "Standard Output Sensitivity" first introduced by the Japanese CIPA:

http://www.cipa.jp/std/documents/e/DC-004_EN.pdf

It was incorporated into the ISO Standard 12232 as of 2006.

I understand from Doug Kerr that some later Canons use it instead of saturation-based.