View RSS Feed

ReggieB

The Truth About ISO: Four Thirds beats APS-C

Rating: 9 votes, 5.00 average.
The Four-Thirds format is much derided outside the ranks of its users. In applying for a photography gig recently, I was told that only “professional equipment” could be used, when I pressed for information, I was told that this meant Canon or Nikon. All others need not apply. I have a feeling there are a few users Leica, Hasselblad, Mamiya, among others that would take issue with this. But as a Four-Thirds user, I do too. I am going to do two posts on the alleged faults of Four-Thirds. The first (high-ISO noise) is an utter myth. The second (Dynamic Range) is less often mentioned, and it is time to put a stop to that, as it is the true weakness of the system. But that will be reserved for my next post, today, it is time to dispell that first notion.

It is not just the CaNikon faithful who espouse the utter falsehood that there is a disadvantage to the Four-Thirds sensor, relative to APS-C sized sensors, in the high ISO department. Popular Photography (who is not alone) perpetuates this myth in its reviews. When reviewing the Panasonic GH1, under the “what's not [hot]” (ie the cons) they list “Noisy at higher ISOs.” Under the “What's Hot” (Pros) of the Canon EOS-7D? “High-res images with low noise at high ISOs.” Let's evaluate that, shall we? DxO Labs offers a wonderful tool which analyzes the technical ability of digital camera sensors, and you can get a handy little graph of the Signal Noise Ratio (SNR) of various cameras. Below is a SNR graph which shows the Canon 7D, the Nikon D300s, and the Panasonic GH1:



It takes a 25% change in the SNR to register a (barely) perceptible change in the amount of noise in an image. There simply is not a noticable difference in the performance of these two cameras in terms of noise. Popular Photography also describes images at ISO 1600 produced by the E-3 to be “Unacceptable” in terms of noise, but I submit to you the following image, produced by the E-3 at ISO 1600, to show just how acceptable images can be (note, there is zero noise reduction applied, the following was a RAW file converted in Gimp).



Popular Photography is not the only publication espousing this myth, and they are not entirely at fault. There is one major culprit behind this (and a possible second), which has led much of the photography world to dismiss Four-Thirds cameras as being inferior at high ISOs. The first is relative gain at the same “ISO” in Four-Thirds vs Canon and Nikon cameras, and the possible second (which I will not get in to in depth, since there is simply no way to know for sure) is noise reduction applied in Canon and Nikon's JPEGs, even with noise reduction turned off in camera.

Unlike the second, the first reason is scientifically provable. There is no such thing as ISO in the world of digital, it simply doesn't exist. What OEMs must do is turn up the gain on the sensors in their cameras and estimate their sensitivities relative to old film ISOs, since that is a number that photographers are used to. DxO labs also estimates what they believe to be equivalent sensitivities, then they compare stated specs against what they believe to be “correct.” Again, this is only their own estimation, it is difficult to know who is right. Still, the DxO Mark is useful in that it can give you relative sensitivities, that is, even if DxO Labs and Olympus are wrong as to what is correct (I will show that they are very close, as are Canon and Nikon to each other), it still gives you valuable information since it can tell you that, no matter who is correct, when I set my E-3 to ISO 1600, the sensitivity of my camera is higher than when you set your D300 to ISO 1600. The following charts display how Panasonic, Nikon, and Canon all set the sensitivity on their sensors, the base line is what DxO labs consider to be “correct.”



Again, this is not to bash Canon or Nikon, they may actually be more correct, as these are all estimates. So, when you set the ISO on the GH1 to 1600, DxO estimates that it is at the equivalent to ISO 2154, whereas the Canon 7D is at 1223, and the D300s is at 1114. Olympus tends to be right on the mark based on DxO's standards. The E-3, for example, is at 804, based on DxO's estimates, when set to ISO 800, and 1584 at ISO 1600. This goes a long way to explain why DPReview, Pop Photo, et. al., do tests that show JPEGs out of Four-Thirds cameras that are significantly noisier at, say, ISO 1600 than their competition, while graphs of the SNR at DXO show much smaller differences. Our Four-Thirds sensors are set to a higher sensitivity at the “same setting.”

To compare more fairly the GH1 to its competitors, these testers would have to set the GH1 to ISO 800 (where it is at 1126) and its competitors to ISO 1600, since that is actually higher than the sensitivity on the D300 at ISO 1600 (1114), and very close to the 7D at ISO 1600 (1223). This is the case throughout the ISO range, when setting the GH1 to ISO 1600 vs ISO 3200 on the D300s and 7D, the GH1 is at 2154 vs 2129 for the D300s and 2278 for the 7D. These testers claim that the GH1 only goes to ISO 3200, something that many likely see as a disadvantage, since the 7D and D300 go higher. But the D300 only goes to ISO 6400, and at that setting, its sensitivity is very close to the GH1's at ISO 3200.

What all of this means is that, with the same aperture and shutter speed settings on the D300s the GH1's ISO setting would be a full stop lower to get the same exposure. And about that when comparing the 7D. That is a staggering difference in sensitivity. What you find in a graph of the SNR of the E-3 vs D300s vs 60D, is that the E-3 is at a one stop disadvantage (that is being generous, and it is more like 1/2 stop to the D300s and 2/3 to the 7D) in SNR when comparing measured ISO as opposed to manufacturer stated ISO. Now, compare the Zuiko lens lineup to Nikkor and Canon L lines, and what you find is that the equivalent pro-level (35-100, 90-250, etc) lenses are a full stop faster wide open, and by most accounts much sharper wide open. To the point that I nearly always shoot my 35-100 at f/2.0, the same with the Leica D 25mm f/1.4, I use it wide open often. Against many Canon L and Nikkor lenses I have more than a full stop of usable advantage (with more depth of field at equivalent field of views because of the crop factor of the sensor). In the sub-full frame market, I assert that Four-Thirds is superior to APS-C sensors in terms of high-ISO noise. My next update will get to the real drawbacks to the sensors, and the issues that Panasonic should be working on: dynamic range and color depth.

Submit "The Truth About ISO: Four Thirds beats APS-C" to Digg Submit "The Truth About ISO: Four Thirds beats APS-C" to del.icio.us Submit "The Truth About ISO: Four Thirds beats APS-C" to StumbleUpon Submit "The Truth About ISO: Four Thirds beats APS-C" to Google

Updated 24th October 2010 at 01:56 AM by ReggieB (Typos and formatting)

Tags: None Add / Edit Tags
Categories
Uncategorized

Comments

  1. John Perriment's Avatar
    What an excellent first blog entry, Reggie, the first of many I hope. You make some very salient points and a convincing arguement for the quality of Four Thirds sensors vs APC-S size. Perhaps we cannot really blame the journos at Popular Photography amongst others, they have been caught in the hype and are not necessarily as good as we are at using our cameras correctly. You mention the likelyhood that some noise reduction is probably applied to Canikon JPEGs even when the noise filter is turned off and I believe there is a suspicion that as far as NR goes some camera's raw files are not as raw as they might be. I'd rather have the detail and sort out the noise to my own liking anyday. It will be interesting to see how the E-5 compares in DxO tests. Meanwhile I look forward to your next blog entry and some more interesting facts.
  2. ReggieB's Avatar
    Thank you, John. I have enjoyed your Blog entries, as well. That is a real compliment coming from you!
  3. Ross's Avatar
    Thanks Reggie,
    That opens up some more understanding. It would be interesting to see a real photo with similar lens set according to those settings to see how they actually visually compare.
    Ross
  4. Bouffie's Avatar
    Thanks, a readable explanation of what is going on inside the magic box. Begs the question why the manufacturers have not given the option to abandon the ISO standard for a variable gradation say that would allow he photographer to fine tune further - if not variable then with more steps a bit like the way White Balance can be set or varied.
  5. TonyB's Avatar
    Interesting post Reggie, whilst this may hold true for the E-3 and GH2 there are other 4/3's sensors that are not quite so good and APS-C sensors which do better. It looks to me like you may have "cherry picked" your data. Compare the Pentax K-X to the E-30 on DXO-mark and you will see the other side of the coin. In general APS-C sensors will always fare better, because they are that much larger, you can't really escape that fact. I'm not brand biased by the way, I use both and recognise they both have their strengths and weaknesses.