E-M5 read noise and "real ISOs"

Started Jul 10, 2012 | Discussions thread
ForumParentFirstPreviousNextNext unread
Flat view
Anders W
Forum ProPosts: 19,066Gear list
Like?
E-M5 read noise and "real ISOs"
Jul 10, 2012

If you are shooting RAW, the only major reason to crank up the camera ISO is that it may, depending on the camera/sensor design, reduce the read noise (as measured in electrons). Based on the earlier measurements reported here

http://forums.dpreview.com/forums/read.asp?forum=1041&message=41817730

we already know that the read noise on the E-M5 falls dramatically as we go from ISO 200 to ISO 400 and then falls more slowly to reach its minimum about ISO 3200, remaining practically constant from there on. Consequently, it makes a lot of sense to to increase camera ISO up to 400 if you are actually exposing for that ISO and it makes at least some sense to go a bit beyond 400 as well if you need to expose for still higher ISOs.

But one thing we don't know yet (at least I didn't until now) and which even DxO won't tell us is what happens at ISOs between the major ones, like, for example, those between 200 and 400. Depending on camera design, read noise performance may change gradually across the entire ISO scale or only at the major ISO settings.

So after finally receiving my own E-M5 a few days ago, I took a few black-frame shots (at a shutter speed of 1/4000) to determine how the camera behaves in this regard and had a look at them by means of RawDigger, resetting the black level to zero to disclose the entire read-noise distribution. Here are the results in ADUs (rather than electrons) for all ISOs between 200 and 800, computed as the weighted average across the four channels, i.e., (R + 0.5G + B + 0.5G2) / 3.

200 1.0
250 1.2
320 1.4
400 1.1
500 1.3
640 1.6
800 1.9

If read-noise performance is constant, we should expect read noise as measured in ADUs to increase in direct proportion to the ISO. If, for example, we double the ISO, the read noise would be expected to double as well. If the increase is less than that, we know that read-noise performance is improving (i.e. that there is less read noise as measured in electrons).

So what do the results tell us? First, and as expected based on earlier reports, read noise falls dramatically as we move from ISO 200 to 400. The read noise at ISO 400 is barely higher than that at ISO 200 if measured in ADUs (and nearly half if measured in electrons). This in turn implies that DR is virtually unchanged in spite of the ISO doubling.

Second, and that's a new piece of information, the intermediate ISOs of 250 and 320 perform less well than ISO 400 in this regard. However, and this is somewhat mystifying, they still perform better than would be expected if there was nothing but digital scaling going on between the two major ISOs. With digital scaling only, we would have expected a read-noise value of 1.25 at ISO 250 and 1.6 at ISO 320. Instead we have 1.2 and 1.4, respectively. While the difference between 1.25 and 1.2 is so small as to be ascribable to pure measurement error, the difference between 1.4 and 1.6 is large enough to be taken seriously.

Those who have an idea of what might be going on here are of course more than welcome to chime in. From a practical point of view, the most important news is that it's a good idea to avoid the intermediate ISOs between 200 and 400 and go directly to 400 whenever you want/need to expose for an ISO higher than base.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +21 more
ForumParentFirstPreviousNextNext unread
Flat view
Post (hide subjects)Posted by
ForumParentFirstPreviousNextNext unread
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow