Low Light Performance – Calculation examples for some popular cameras

[facts_please]

This is really a continuation of the previous thread ‘Comparing Low Light Performance of Different Cameras’, but as this is the interesting and useful part, I thought a new thread may be appropriate. The cameras under comparison are a Nikon D80, an Olympus E30, and a Finepix S5600.

Strictly speaking, the method used here predicts the ‘light gathering ability’ of the cameras, with the lenses that would typically be supplied. That is, with the aperture wide open, how much light (how many photons) is delivered to the sensor for a given shutter speed? As previously discussed previously, the signal-to-noise ratio is set primarily by the absolute number of detected photons.

In practice, smaller sensors with a higher proportion of dead area will perform slightly worse, because more photons are lost to the dead areas. Unfortunately I don’t have data comparing sensor performance which takes this and other second order effects into account, though as previously explained it should not be difficult to measure. However, I’ll give the smaller sensor in the S5600 the benefit of the doubt, and assume that the proportion of dead area is the same for all sensors, in which case the comparison essentially just gets down to comparing the total amount of light delivered to the sensor by the lens.


Nothing beats a real life example.
Let’s make a prediction for how much more light the Nikon D80 SLR can potentially collect and throw at the sensor, compared to the Finepix S5600, with an Olymus E30 thrown in as well just for fun.

For all lenses, the intensity of light at the sensor is inversely proportional to the square of the F-number. Therefore :-

The total light striking sensor is proportional to A/(FxF)

where
A is the sensor area, eg in mm^2
F is the lens F-number


Here are the specs and results for these cameras :-

Fuji Finepix S5600
Focal length 6.3 – 63 mm (38 to 380mm equivalent)
F3.2 - 3.5
Sensor area = 5.75x4.31 = 24.8 mm^2
5.2 megapixels
Amount of light on sensor = 24.8/(3.2x3.2) = 2.42 (at L=38mm)
Amount of light on sensor = 24.8/(3.5x3.5) = 2.02 (at L=380mm)


Nikon D80 (with standard lens)
Focal length 18 – 135mm (27 to 202.5mm equivalent
F3.5 – 5.6
Sensor area = 23.6x15.8 = 373 mm^2 (APS-C / Nikon DX size)
10.2 megapixels
Amount of light on sensor = 373/(3.5x3.5) = 30.4 (at L=27)
Amount of light on sensor = 373/(5.6*5.6) = 11.9 (at L=202)


Olympus E30 (with standard lens)
Focal length 14 – 54mm (28 to 108mm equivalent
F2.8 – 3.5
Sensor area = 17.3x13 = 225 mm^2 (4/3 size)
12.3 megapixels
Amount of light on sensor = 225/(2.8*2.8) = 28.7 (at L=28)
Amount of light on sensor = 225/(3.5*3.5) = 18.4 (at L=108)


OK. That was pretty easy, and the results are interesting. The calculated amount of light on the sensor is in arbitrary units, which is fine as it is only the comparison between cameras that interests us, not the absolute values. This comparison is with the lens apertures wide open.

Fortunately all three lenses have about the same 35mm equivalent focal length at full wide angle, making for a neat comparison. The two SLRs are roughly the same, with 30 and 29 light units striking the sensors. However, the point and shoot S5600 is not in the running, and can only muster 2.4 units of light, a whopping 12 times less than the SLR cameras! Let’s get a feel for what that means in the real world. For example, in a situation where the S5600 was flat out to get enough light, forcing the user to select a long exposure time of 1/10 second, the SLRs would do it in around 1/120 seconds. However, there is no avoiding the fact that the depth of field will be lower whenever that extra aperture is actually used.

In comparing the D80 with the E30, the D80 with this particular lens is overall a better performer. Over the Olympus zoom range of 28 to 108mm (35mm equivalent) the Nikon is slightly better, but there is not much in it. However, the Nikon has a much greater zoom range out to 202mm, making it a lot more useful, AND still maintaining a slight edge in low light performance. Impressive. However, the simple fact of the matter is that you get what you pay for, and what you are prepared to carry, and the Nikon performance comes at the expense of greater size and weight, and that’s how it always is and will always be.

The S5600 is very substantially worse because the lens is physically smaller, because the large zoom range (10:1) nearly always throws away light, and generally because it is a much cheaper lens.

OK guys. Was that simple, useful and real world? I don’t claim these predictions will be super accurate, but as far as I can see they should be more-or-less right, and good enough to be useful. A disclaimer is necessary though, as my expertise is more with science, and I know very little about photography. As always, comments from the gurus here are encouraged. I would be interested to check these predictions by measurement, but I only own one camera, and it is not even an SLR. I am considering the purchase of a new camera though, which is the reason why I took the trouble to think about this topic, so if and when I do I might take some measurements.

One last point. The interpretation is slightly more wooly when the number of megapixels is different. That is a topic in itself, but what it amounts to is that with post processing (or in camera noise reduction) the extra megapixels do not cause much more visible noise if the noise reduction is applied to the extent that all images contain about the same detail. In other words, it doesn’t greatly affect the above conclusions, and I don’t want to confuse matters by talking about it. In addition, larger sensors have superior dynamic range, and can provide lower noise images at low ISO than is possible with a smaller sensor even when there is light to burn. However, these are minor advantages for many of us, remembering that noise at low ISO is not usually a problem anyway. Of course, it goes without saying that there are many, many other factors to consider when choosing a camera.

If anyone would like to repeat the above calculation for their camera(s) of interest, could you post you calculation in this thread, preferably in the same format that I used that clearly shows the lens specs as well as the calculation, so we may all benefit?

Thanks.

[Davey]

Thanks for that. I find stuff like that interesting anyway but even more so than usual because it applies to me since my cam is a finepix 5600 and the most likely future upgrade is a d80 as it seems to suit most my needs for my next step up.

Interesting to see how much less light my lens gathers. I guessed it would be less but knowing by how much Even if the figures are not perfect due to other variables that can't be measured or accounted for (easily) they will be in the ball park I'd imagine. I guess it confirms what many would predict, and bigger or especially a higher quality lens on a small point and shoot wouldn't be wise since people looking for that kind of above average (average persons needs I mean) quality in a lens get an slr. Shelling out the pennies for finer optics in the form of a none removable lens is not likely for anyone really since it ties you to that camera and technology changes as well as needs hence I am planning to get d80 partly due to lens considerations and it's cheaper than a d200 but if I ever upgraded again to a pro body I could keep the lenses I shelled out on.

It would be interesting to compare other similar performers too, especially with high price differences. I know the price isn't just about sensor related points though but it would be interesting to see.

[Dave Humphries]

Thanks Colin,

Comparing against the S5600 I became concerned that I was lugging around a large chunk of glass that wasn't doing much in comparison with one of Fuji's tiny flat compacts, so here's the numbers for the 5x zoom, Z100, I'm not sure if the lens extends to zoom or if it internally zooms. Not that that matters.

Fuji Finepix Z100
Focal length 5.9 – 29.5 mm (36 to 180mm equivalent)
F3.8 - 4.8
Sensor area = 5.75x4.31 = 24.8 mm^2
8.0 megapixels
Amount of light on sensor = 24.8/(3.8x3.8) = 1.72 (at L=36mm)
Amount of light on sensor = 24.8/(4.8x4.8) = 1.07 (at L=180mm)

The figures are somewhat worse, so I am a bit relieved at that.

For completeness, my S6500;

Fuji Finepix S6500
Focal length 6.2 – 66.7 mm (28 to 300mm equivalent)
F2.8 - 4.9
Sensor area = 7.17x5.31 = 38.1 mm^2 (1/1.7")
6.0 megapixels
Amount of light on sensor = 38.1/(2.8x2.8) = 4.86 (at L=28mm)
Amount of light on sensor = 38.1/(4.9x4.9) = 1.59 (at L=300mm)

So, ignoring the top end because it's so disparate, that leaves me with about 1/6 the light reaching my sensor at 28mm as I'd get with the E-30 or D80 - does this mean noise will be 6 times less apparent with those cameras?

Thanks,

[facts_please]

Hi Dave,

Re Finepix 6500:
So, ignoring the top end because it's so disparate, that leaves me with about 1/6 the light reaching my sensor at 28mm as I'd get with the E-30 or D80 - does this mean noise will be 6 times less apparent with those cameras?

Most of the noise will be 'statistical', meaning it is due to the inherent uncertainty in detecting only a small number of photons. The signal to noise ratio scales as the square root of the number of photons detected, so 6 times more light should improve signal-to-noise by root(6), or approximately 2.5

However, you asked how much less apparent the noise will be, and that is subjective.

You can answer your question yourself with a simple experiment. Set up your camera to take a shot at high ISO such that the noise is quite bad. Then, increase the amount of light reaching the sensor by decreasing shutter speed by factor of 6, and lower ISO setting to retain same exposure. Compare the 2 shots visually, and that's your answer ....

Re your comparison with the Finepix Z100, it is heartning to see that the Z100 is worse, but to be fair you need to factor in the fact that the S5600 and S6500 have a much larger zoom range. There is a good reason why you don't find x10 zoom on ultra compact camera with small lenses .....

Cheers, Colin

[JonathanC]

Maybe its significant to refer to DOF of the various cameras listed in the original post?

What I'm trying to point out is that the Nikon example will not have the same DOF as the Fuji unless the Nikon is stopped down. Once stopped down to f/16 or so, the numbers get less damning of the Fuji, so its not quite so cut and dried as the raw data appears.

It boils down to the fact that the Fuji uses a much shorter focal length lens therefore it has greater DOF wide open than does the Nikon lens. If you need to get the DOF the same then you need to stop down the Nikon. If you don't need to stop down you definitely do better with the bigger sensor, the longer focal length and corresponding bigger lens aperture.

It appears to me that the "hole in the front of the camera" has a whole lot to do with the image quality and most of the IQ issues can be traced back to the total amount of light that you can capture.

Same goes for film cameras really, for better quality use a bigger format.

JC