|
ISO Lies: When ISO 100 is really 72
Your camera's ISO settings may not be accurate
By Mason Resnick
February 2, 2009
My suspicions began a few years ago, after I had purchased a Canon 20D. My pictures were coming out just a tad dark, even when I used manual metering and went with the “sunny 16” exposure rule (shutter speed is the same as the ISO at f/16 on a sunny day). The histograms, which indicate proper exposure, seemed to be shifted ever so slightly to the left, indicating underexposure. Something was not quite right.
Mystery solved
Then, this past November, DxOMark.com went live, and my suspicions were confirmed. Each of DxOMark's objective lab tests of over 50 digital cameras, which are now freely available online, includes the camera's actual, measured (as opposed to the manufacturer's claimed) ISO sensitivity. It turns out my camera's measured ISO sensitivity was off by about a third of a stop at all ISOs!
Check the charts: Here are two extreme examples of two cameras' ISO sensitivity results. The closer the circle is to the line, the closer the measured ISO is to the manufacturer's ISO. In the top example, the measured ISO is very accurate, one of the best results. In the bottom chart, measured ISO is off by nearly a full stop at all ISO settings, one of the worst performances. (Charts courtesy DxOMark.com.)
As I studied DxOMark's test results I was surprised to learn that my camera turned in a relatively good performance. As I looked at other camera test results, it became clear that measured sensitivity at the sensor is a bit lower (sometimes, higher) than the specs indicate. In some cases, the difference between the claimed and measured ISOs are as much as a half stop or more, but a third to a fifth-stop difference is typical. Only a handful of cameras consistently come within a few ISO points of the manufacturer's stated ISO in DxOMark's tests.
Cost, age don't matter
The other big surprise was that there wasn't much correlation between the cost of the camera and the accuracy of the measured ISO accuracy. There are starter DSLRs with better than average accuracy and pro cameras that showed only average results, as well as starter cameras that did poorly and pro models that did well. Camera age made no difference, either: Cameras produced in 2004 in some cases fared better than those introduced in 2008. Some did worse. There was no pattern.
Should you be alarmed? Nah.
Now for a reality check: While this information may be surprising, it's not really a big deal, because it is so easy to overcome. In the real world (as opposed to the lab) most cameras' variation in measured ISO from the manufacturer's ISO will have a minimal affect on the image. You only really start to see it if the gap is 1/3 of a stop or larger, and you can resolve this in-camera by adjusting your exposure compensation.
After the fact, if you shoot RAW and then run the image through Adobe Raw converter, the software may automatically fix the exposure. If not, you can simply adjust the exposure while in your conversion software. When working with JPEGs, simply refer to the histogram, and adjust exposure accordingly. (It's good to do that anyway.)
The difference between the official ISO and the measured one is usually fairly consistent when changing ISOs. If you find a third-stop difference at ISO 100 (ISO 80), you'll probably also have a third-stop difference at ISO 800 (that would be around ISO 640). Some models, however, are not consistent. That's why I've listed each ISO performance below, by camera.
A bit of historic (read: film) perspective
If you think this is a compelling argument to return to film, sorry. That's not a valid excuse. Those treasured mechanical cameras, when it came to exposure accuracy, were generally worse. While the actual shutter speeds of a camera fresh out of the factory may have been fairly accurate (within 1/3-1/2 a stop), with time and use the coils and springs that run the shutter would loosten and shutter speeds would slow down. That's why semi-annual cleaning, lubrication and general tune-ups at the shop were recommended.
Film speed could vary slightly from batch to batch (but the difference was usually very slight, especially with slide film). For print film, this difference was hardly noticeable because of the variations inherent in the print process.
What if your camera wasn't tested?
As of this writing, DxOMark has tested 54 cameras--most of them DSLRs. Since they will only test cameras that capture RAW images, most compact cameras will not be tested. A quick way to test your camera's exposure accuracy is to photograph a blue sky using manual metering, if your camera has that option. Check the histogram and if it is slightly too the right, compensate using exposure compensation by 1/3 stop and check again. When you get a nice even curve that begins at one edge and ends at the other, your exposure is accurate. Set your camera's EV preferences to reflect your results. This is a down-and-dirty way of doing it but for snapshooters, it should be enough.
When ordering prints, many labs (including AdoramaPix) give you the option to let the person running the printer to make adjustments for the best possible print. This is a good idea if you order prints direct from the camera or memory card with no alterations and want to make sure any wacky ISO sensitivity difference doesn't show up on the print. An experienced printer can compensate for slight exposure variances. At AdoramaPix, this service is free.
So, how did your camera do?
Here's a quick look at how all the cameras DxOMark tested performed when tested for ISO sensitivity, in alphabetical order by company. How did your camera do? Scroll down to find out!
Canon
G9
Measured ISO sensitivity consistently 1/3-1/2 stop higher than Manufacturer spec. ISO 100 was measured at 139, ISO 400 was measured at 551, etc. The lowest manufacturer ISO (78) was measured at 98.
G10
Measured ISO sensitivity stayed consistently within 1/3-1/4 stop faster than the manufactuer ISO from ISOs 100-1600. For example, ISO 100 was really 117. The exception was the lowest manufacturer ISO (78), which was very close at a measured ISO 83.
EOS 5D Mark II
Measured ISO sensitivity was consistently 1/4-1/3 stop less than manufactuer ISO. ISO 100 was measured at 73, ISO 400 was measured at 285, etc. The difference was slightly greater by ISO 6400.
EOS 50D
Measured ISO sensitivity (93) was very close to manufactuer ISO at ISO 100, and stayed roughly within 1/4 stop less than the manufacturer ISO through 1600. The difference increased slightly, to 1/3 stop, at ISOs 3200-12800.
EOS 1000D (Rebel Xs)
Tested ISO sensitivity (95) was close to manufacturer ISO of 100, and stayed consistently with 1/4 stop of the manufacturer ISO at all speeds.
EOS 450D (Rebel XSi)
Tested ISO sensitivity was consistently around 1/3 stop less than the tested ISO at all speeds. At ISO 100 the measured ISO was 86. At ISO 200 it was 151, at ISO 400 was 303, etc.
EOS 1Ds Mark III
Tested ISO sensitivity was consistently around 1/2-1/3 stop less than the manufacturer ISO at all speeds. At ISO 100, the measured ISO was 73, at ISO 200 was 144, at ISO 400 it was 285 etc.
EOS 40D
The measured ISO sensitivity was consistently approximately 1/3 stop less than the manufacturer ISO. At ISO 100, the measured ISO was 87. At ISO 200 it was 145, at ISO 400 it was 294, etc.
EOS 1D Mark III
Measured ISO sensitivity was consistently around 1/3 stop less than the manufacturer ISO at all speeds.
EOS 400D (Rebel XTi)
Measured ISO sensitivity was very close to manufacturer ISO at most speeds, especially ISO 100, which tested as ISO 93. ISO 200 was the widest gap (measured: 167) but in most cases the measured sensitivity was approximately 1/4 stop less than the manufacturer ISO.
EOS 30D
Measured ISO sensitivity was close to manufacturer ISO at most speeds—around 1/3-1/4 stop less. ISO 100 tested as ISO 103, an insignificant difference. ISO 200 was measured at 169, ISO 400 was 363, etc.
EOS 1D Mark II N
From ISO 100-800, measured ISO was approximately 1/4 stop less than the manufacturer ISO. However, the gap widened a bit at ISO 1600 (Measured: 1276) and 3200 (Measured: 2282)
EOS 5D
Measured ISO was slightly lower than, but very close to the manufacturer ISO at all speeds. ISO 100 was measured at 92, ISO 200 was 181, ISO 400 was 357, ISO 800 was 737 and so on.
EOS 350D (Rebel Xt)
Measured ISO was very close to the manufacturer ISO at all speeds, no more than 1/5-stop less.
EOS 1Ds Mark II
Measured ISO was close to the manufacturer ISO at almost all speeds, with an approximately 1/3-stop less from ISO 100-1600. ISO 100 was measured at ISO 84, ISO 200 was 163, ISO 400 was 334, and ISO 800 was 666.
EOS 20D
Measured ISO was close to the manufacturer ISO at all speeds, with an approximately 1/3-1/4 stop loss. The measured ISO at 100 was 87, at 200, it was 177, at 400, it was 344, etc.
EOS 1D Mark II
Measured ISO was extremely close to the manufactuer ISO, with approximately 1/4 stop loss at all speeds except ISO 3200, at which the measured ISO was 2542. ISO 100 was measured at ISO 93, ISO 400 was measured at 360, etc—an insignificant difference.
EOS 300D (Digital Rebel)
The measured ISO for the Digital Rebel was slightly higher than the manufacturer ISO. The difference is just under 1/3 stop. At ISO 100, the measured ISO was 116, at ISO 200, it was 240, etc.
EOS 10D
As with the Digital Rebel, the EOS10D's measured ISO was consistently slightly higher than the manufacturer ISO. The Measured ISO at 100 and 1600 were closest to the manufacturer settings, registering ISO 113 and 1695, respectively. At 200-800, the difference was approximately 1/4 stop.
EOS 1Ds
The earliest of the Canon DSLRs tested by DxO had the most accurate measured ISOs. At ISO 100, the measured ISO was 99. At 200, it was 199. At 400, it was 393, and at 800 it was 794—all visually insignificant differences. At the top speed, ISO 1233, the measured ISO was a less accurate 1520.
Fujifilm
FinePix S5 Pro
Measured ISO was extremely close to the manufacturer ISO at all speeds, with a difference of a few points underexposure. At ISO 100, measured ISO is 95. At 200, it is 191 and remains consistent right up through the top ISO of 3200 (measured: 3160).
FinePix S3 Pro
Measured ISO for the S3 was the same as the manufacturer ISO for ISO's 100-400, and extremely close at ISO 800 (measured: 821). One of the best performances of all models tested.
Konica Minolta
Maxxum/Dynax 5D
Before the company left the camera business, Konica Minolta produced a pair of innovative DSLRs, but ISO accuracy was not the strong suit of the second model, the 5D. Measured ISO was as much as 1/2-2/3 stop higher than the listed one, although accuracy improved at the highest speeds. At ISO 100, measured ISO was 178. At ISO 200, it was 328. At ISO 400, it was 654. At ISO 800, it was 1113. At ISO 1600, it was 2021, but at ISO 3200, it was extremely accurate at ISO 3180.
Maxxum/Dynax 7D
As with the 5D, the 7D's exposure accuracy varied greatly. At ISO 100, the measured ISO was 173. At ISO 200, it was 294. At ISO 400, it was 576. At ISO 800, it was 1010. At ISO 1600, it was accurate with an measured ISO of 1595, but was slightly lower at ISO 3200 (measured: 2798).
Leica
Leica M8
Leica's measured ISOs were remarkably close to the manufacturer ISOs. The Manufacturer ISOs are non-standard but at all speeds sensitivity was within a few points of the measured ISO. At ISO 160, the measured ISO is 151. At ISO 320, it is 304. At ISO 640 it is 619. At ISO 1250 the measured ISO is 1256, and at ISO 2500 it is 2536.
Nikon
D90
Measured ISO was consistently approximately 1/2 to 1/3 of a stop less than the manufacturer ISO. At manufacturer ISO 200, the measured ISO was 145. At ISO 400, it was 292 and at ISO 800 it was 589, etc.
Coolpix P6000
Nikon's flagship compact camera's ISO was consistently 1/3 to 1/2 stop higher than the manufacturer ISO, except at the highest ISO, where it was around 1/10 stop higher. At ISO 64 the measured ISO was 94. At ISO 100, it was 145. At 200, it was measured at 268, etc.
D700
The D700's measured ISO was consistently around 1/3 stop less than the manufacturer ISO. At ISO 200, the measured ISO was 162. At ISO 400, it was 327 and similar gaps were detected until ISO 12800, which was measured at 8436, and ISO 25600, which was measured at 14085, which is still impressively sensitive.
D60
Measured ISO was extremely close to the manufacturer ISO at all speeds. At ISO 100, the measured ISO was 93. At ISO 200, it was 191. The gap widens slightly at ISO 3200, which measured as ISO 2733.
D300
Measured ISO was approximately half a stop less than the manufacturer ISO. At ISO 200, the measured ISO was 145. At ISO 400, it was 288. At ISO 800, it was 571, etc.
D3
Measured ISO was consistently around 1/3 of a stop below the manufacturer ISOs. At ISO 200, the measured ISO was 161. At ISO 400, it was 326. At ISO 800, it was 635, etc. The gap widened at ISO 25600, when the ISO was 16134—still an impressively high sensitivity.
D3x
The D3x's lowest manufacturer ISO, 50, measured at ISO 78--identical to the measured ISO at the manufacturer's ISO 100! From 200 through 1600, measured ISO was consistently around 1/3-1/2 stop lower than the manufacturer's ISO. At ISO 200, it was 170. At 400, it was 337. At 800, it was 674, and at 1600 it was 1369. The gap widens sligthly for ISO 3200 (actual: 2504) and 6400 (5015).
D40x
Measured ISO was consistently very close to the manufacturer's ISO, staying within 1/4 of a stop underexposure. At ISO 100, measured ISO was 94. At ISO 200, it was 189. At ISO 400, it was 374, and at ISO 800, it was 736, etc.
D40
Measured ISO was consistently around 1/4-1/3 of a stop less than the manufacturer's ISO at all ISO settings. At ISO 200, it was 178. At ISO 400, it was 360, and at ISO 800 it was 705, etc.
D80
Measured ISO was consistently well within 1/5 stop below the manufacturer's ISO at all settings except for ISO 3200, which was around 1/3 stop less at an measured ISO 2425. At ISO 100, measured ISO was 96. At ISO 200, measured ISO was 191, and at ISO 400, measured ISO was 388, etc.
D2Xs
Measured ISO was remarkably close to the manufacturer's ISO at all speeds, staying within a few points in either direction, depending on the speed. At ISO 100, the measured ISO was 104. At ISO 200, it was 209. At ISO 400, it was 419. At ISO 800, it was 811. At ISO 1600 it was 1594, and at ISO 3200 it was 2981.
D200
Measured ISO was remarkably close to the manufacturer's ISO at all speeds, staying within 1/5 stop less than the manufacturer's ISO. At ISO 100, measured ISO was 96. At ISO 200, it was 195. At ISO 400, it was 388. At ISO 800, it was 759. At ISO 1600 it was 1494, and at ISO 3200 it was 2680.
D70s
Measured ISO was consistently within 1/5 stop faster than the manufacturer ISO at all speeds except at ISO 1600, where it was very close with an measured ISO of 1661. At ISO 200, the measured ISO was 216. At ISO 400, it was 436. At ISO 800, it was 851.
D50
Measured ISO was consistently 1/4-1/5 stop slower than the manufacturer ISO. At ISO 200, the measured ISO was 187. At ISO 400, it was 364. At ISO 800, it was 718, and at ISO 1600, it was 1467.
D2x
Measured ISO was consistently well within 1/10 of a stop in either direction of the manufacturer ISO, with a minimal effect on measured exposure. At ISO 100, the measured ISO was 104. At ISO 200, it was 209. At ISO 400, it was 419. At ISO 800, it was 811. At ISO 1600, it was 1594 and at ISO 3200 it was 2981.
D70
At all speeds, measured ISO was consistently less than 1/5 stop above the manufacturer ISO. At ISO 200, measured ISO was 216. At 400, it was 436. At ISO 800, it was 851 and at ISO 1600 it was 1661.
D2H
Measured ISO varied from as much as 1/3 stop more than the manufacturer ISO from ISO 200-1600, to about 1/5 stop less at ISO 3200 and 6400. At ISO 200, the measured ISO was 238. At ISO 400, it was 488. At ISO 800, it was 989, and at ISO 1600, it was 1926.
Olympus
E520
At Manufacturer ISO 100, the measured ISO was 129, approximately 1/3 of a stop more than the official ISO. However, From ISO 200 (measured ISO 211) through 1600, the difference narrowed down to about 1/5 of a stop.
E3
Overall, measured ISO is very close to the manufacturer's ISO, although it is slightly faster at ISO 100 (measured ISO: 118) and 200 (227). By ISO 100 (measured ISO: 412) it is within a few points of the manufacturer's ISO.
Panasonic
Lumix DMC LX3
At all ISOs, the LX3's measured ISOs were nearly a full stop less than the manufacturer ISO. ISO 100's measured ISO was 60, and at 200 it was 122. By ISO 800 it was 458 and ISO 1600 was measured at only 901.
DMC FZ28
Overall, the FZ28's measured ISOs were within 1/5 stop above the manufacturer's ISO, a relatively small gap. ISO 100 was measured as 117, ISO 200 was 232, ISO 400 was 460, ISO 800 was 885, and ISO 1600 was 1735.
DMC L10
At all ISOs, the measured ISO was a few points above or below the manufacturer ISO, an excellent performance. At ISO 100, measured ISO was 106. At ISO 200, it was 212. At ISO 400, it was 409. At ISO 800, it was 785, and at ISO 1600 it was 1544.
Pentax
K20D
Measured ISO was consistently around 1/3-1/4 of a stop slower than the manufacturer ISO. At ISO 100, the measured ISO was 87. At 200, it was 171. At 400, it was 339, etc.
K200D
Measured ISO was consistently very close to the manufacturer ISO at all speeds. At ISO 100, the measured ISO was 95. At ISO 200, it was 189 and at ISO 400 it was 371, etc.
K10D
Measured ISO was consistently within approximately 1/4 of a stop below the manufacturer ISO at all speeds. At ISO 100, measured ISO was 90. At ISO 200, it was 181, and at ISO 400 it was 352, etc.
Samsung
GX20
Measured ISO was consistently within 1/3-1/4 of a stop below the manufacturer ISO. At ISO 100, it was measured at ISO 87. At ISO 200, it was 173. At ISO 400, it was 339. At ISO 800, it was ISO 667, and so on.
Sony
Alpha 900
At ISO 100, the measured ISO (119) was slightly higher than the manufacturer ISO, but from ISO 200 on, the measured ISO was consistently approximately 1/3 to 1/2 of a stop less than the manufacturer ISO. At ISO 200, the measured ISO was 151. At ISO 400, it was 303. At 800, it was 615. At 1600 it was 1264, etc.
Alpha 350
Measured ISO was consistently within around 1/4 stop less than the manufacturer ISO. At ISO 100, the measured ISO was 87. At ISO 200, it was 181. At IS0 400, it was 366. At ISO 800, it was 727, and so on.
Alpha 100
Measured ISO was consistently around 1/4 stop more than the manufacturer ISO. At ISO 100, the measured ISO was 121. At ISO 200, it was 238. At ISO 400, it was 464. At ISO 800 it was 927, and at ISO 1600, it was 1804.
(A note to math geeks: I am not a math geek, but I'm married to one (MIT, class of 1982), and I ran the above numbers by her before publishing this article. I frequently mention above that measured ISOs may be off by fractions of a stop. While I could have run the algorithmic calculations for the exact numbers (which my wife patiently demonstrated to me), I wanted to strike a balance between precision and meaningful information. And so, when I say that a measured ISO was around 1/4 of a stop more than the manufacturer's ISO, I am rounding things off. Any more precision, as it relates to the effect on the exposure of most photographs, would be meaningless.)
Our thanks to DxOMark.com for allowing us to use and interpret their data.
Mason Resnick is the Editor of the Adorama AIRC Learning Center.
© 2009 Adorama Camera, Inc.
|
|
|
 |
 |
|
|
 |
| |
 |
 |
 |
| |
|
Supplies
Canon EOS-5D Mark II Digital SLR Camera Body, 21.1 Megapixels with 3.0" TFT LCD - USA Warranty
Canon EOS 50D Digital SLR Camera Body Kit, 15.1 Megapixels with USB Interface & 3" LCD, Black
Canon Digital Rebel XSi SLR Camera Body Kit, 12.2 Megapixels, Black with EF-S 18-55mm f/3.5-5.6 Image Stabilizer Lens
Canon EOS-1Ds MARK-III Digital SLR Camera with 21.1 Megapixel, 1.5x - 10x Zoom and 3" LCD Screen
Canon EOS-1Ds Mark II Digital SLR Camera Kit, 16.7 Megapixels - Refurbished By Canon U.S.A.
Nikon D90 12.3 Megapixel Digital SLR Camera Body with DX Format CMOS Sensor
Nikon Coolpix P6000 Digital Camera, 13.5 Megapixel, 4x Optical Zoom, 4x Digital Zoom, 2.7" LCD Display, Black
Nikon D700 Digital SLR Camera, 12.1 Megapixel, 0.72x Zoom, FX Format, CMOS Image Sensor
Sony Alpha DSLR-A900 Digital SLR Camera Body with 24.6 Megapixel & 3.0" LCD Screen, HDMI Output
|
|
|
 |
|
 |
|
| |
| |
|
Entertainment Center
Expert tips
