I have analized the 7D2 dynamic range using the RAWs posted by Imaging Resource. The method is based on estimating the read noise (using the masked pixels at the left of the RAW, except for the first lines) and finding the saturation level. For the first ISO levels I used real life pictures (ISO 100 seems not to fully saturate the sensor even with clipped highlights).
The results confirm that the dynamic range is identical at low ISO compared to the 7D. However, it seems that there is a noticeable improvement at high ISO.
These DR values are screen in DXO terminology. For the normalized print ones approximately DR 0.68 should be added. That is, the official DXO DR rating would be about 11.62 (compared to 11.73 in 7Dc). However, at ISO 6400 the 7D2 achieves 8.97 (compared to 7.7 in 7Dc). Note that true ISO levels may differ from that of 7Dc, though.
ISO 12800 seems to be fake, since the read noise reported by the first masked lines doubles to 12.3 compared to half of that value for ISO 100-6400 (it seems that the first masked pixels rows are stored when the sensor has not yet "powered on" the ISO amplifier... thus 12800 seems to be a digital ISO, applying a 2 multipler on the raw data).
The 7D2 sensor is a incremental upgrade as everybody suspected. The noise may have improved at high ISO settings (in the worst case, it can be definitely improved by exposing to the right thanks to the increased DR ;-).
PS: Note that the read noise really decreases with ISO... measured in electrons (we use here raw data numbers).
cgarcia wrote:
I have analized the 7D2 dynamic range using the RAWs posted by Imaging Resource. The method is based on estimating the read noise (using the masked pixels at the left of the RAW, except for the first lines) and finding the saturation level. For the first ISO levels I used real life pictures (ISO 100 seems not to fully saturate the sensor even with clipped highlights).
The results confirm that the dynamic range is identical at low ISO compared to the 7D. However, it seems that there is a noticeable improvement at high ISO.
These DR values are screen in DXO terminology. For the normalized print ones approximately DR 0.68 should be added. That is, the official DXO DR rating would be about 11.62 (compared to 11.73 in 7Dc). However, at ISO 6400 the 7D2 achieves 8.97 (compared to 7.7 in 7Dc). Note that true ISO levels may differ from that of 7Dc, though.
ISO 12800 seems to be fake, since the read noise reported by the first masked lines doubles to 12.3 compared to half of that value for ISO 100-6400 (it seems that the first masked pixels rows are stored when the sensor has not yet "powered on" the ISO amplifier... thus 12800 seems to be a digital ISO, applying a 2 multipler on the raw data).
The 7D2 sensor is a incremental upgrade as everybody suspected. The noise may have improved at high ISO settings (in the worst case, it can be definitely improved by exposing to the right thanks to the increased DR ;-).
PS: Note that the read noise really decreases with ISO... measured in electrons (we use here raw data numbers). ...Show more →
Thanks for the those findings.
Same findings as for 5D III vs 5D II. Canon's read noise at high ISO is very competitive but at low ISO it's well behind the curve.
Thanks for posting your findings.
Although, the 7D II is the successor of the 7D, it is really a 70D variant. (At least regarding the sensor)
It would be interesting to see your DR analyses comparing these 2 cameras.
super35 wrote:
If these findings line up with DXOMark, about 2/3rds of a stop better dynamic range over the 7D @ ISO 1600, and about even with the 1D4.
Oh that would make it usable. If only I could buy one.
EB-1 wrote:
Was the 7Dc the astronomical version of the 7D or the same body for a foreign market?
It's confusing. At least Nikon has a good numbering system.
Anyway as I said as ISO100, sameish to the 7D for random read noise. 5.88 read noise for masked area of 7D2 at ISO100.
However, it does seem to me that there is vastly less banding.
So the usable DR is increased a bit at low ISO compared to the 7D.
Here are my center black frame numbers from the 7D (sometimes the masked areas findings give a little bit less read noise than the main sensor areas, rarely a little more, so the masked areas stuff might be around +/- 1/4 stop or so) compared to these and accounting for the 18 vs 20 it looks like it will have a trace worse (to an entirely insignificant degree) lower ISO DR measured (but again, with basically no banding on the 7D2, effectively it will feel better) and anywhere from 1/4 to 3/4 stops better DR at high ISO measured.
Assuming the ISOs were matched to the same gain amount (that has varied a good +/- 1/3 stop at times, at least).
Seomwhere I had plots that normalized the older cams for the ISO gain differences.
You can compare his numbers above to this old stuff:
Anyway I hadn't gotten around to the high ISO DR yet, but using your numbers it seems as I said 1/4 to 3/4 stops better than 7D depending upon masking vs main sensor difference and how they turn out. So better tech than the 7D and getting close to the 5D3 style tech at high ISO for read noise but still not as good (and not even quite a match for 5D2 read noise there either, although with less ugly banding and blotching than the 5D2 at high ISO shadows, so for scenes for lots of very near black it might look better at high ISO than the 5D2, even the 7D already did, although certainly not for SNR) as that and clearly worse than the 6D tech for high ISO read noise.
Other than fixing up the banding even a little more than the 6D, the sensor seems to perform older than the 6D tech and newer than the 7D tech.
EDIT: I can also add, in addition to my numbers above, banding reports:
at low ISO:
40D fairly low banding, not bad
5D2 very bad banding in both directions
5D3 very bad banding in one direction
1Ds3 modest banding in one direction, not bad
7D both vertical gain banding (affects even bright tones) and this varies from copy to copy and in different parts of the frame and regular offset shadow banding
50D very bad banding
6D pretty low banding, not bad
1DX moderate banding, not too bad, maybe a bit worse than 1Ds3 and 6D
7D2 tentatively appears to have nearly zero banding, very good
at high ISO:
from what I vaguely recall:
20D,30D,40D,50D pretty bad banding
5D2 somewhat bad banding
7D not too bad banding, pretty good
5D3 low banding, good
6D low banding, good
1DX low banding, good
In summary, the 7D Mark II is slightly noisier (lower signal to noise) than the 1D4, but has nearly the same dynamic range while putting out more pixels per duck.
It will also have higher resolution than the 1D4 aside from increased megapixels and reach, as all Canon bodies since the 550D have been using weaker and/or better tuned AA filters and they all outperform the 1D4 in combined lens/body tests. The 70D is the best Canon APS-C yet and so expect the 7D Mark II to at least match it in that regard.
WestcoastHD wrote:
Thank you for the tables. If we put DR aside and look at noise at a given ISO, say ISO 6400, how would the new 7Dm2 compare to the 7D classic or 5Dm2.
Not known yet. That's much harder to test from random files like this. May have to wait for DxO for that.
DR is trivial to test with files like this. SNR at mid-tone, not so much.
All I can say is it seems like it at least has a pleasing 'grain' structure at high ISO.
.
