All using Voigtlander 40mm Ultron II Nikon mount @ f/8, same physical exposure, tungsten lighting, shot raw and processed in LR5 with all defaults, downsampled to 8MP. For A7/A7r ISO 51,200 and 102,400, shot at max ISO of 25,600 and pushed 1EV and 2EV in LR. For Df ISO 409,600, shot at max ISO of 204,800 and pushed 1EV in LR. For A7/A7s, EFCS used for all images.The Df's output brightness in LR was a little below the A7 models and so was adjusted +.15EV in LR.
Edit 7/17/14: Here is composite #1 but with LR's color NR changed from its default of 25 to whatever value eliminates all/most color noise in uniform areas. At very high ISO's LR's default color NR of 25 is too heavy handed and destroys color fidelity unnecessarily - this new composite comparison has the color NR values turned down to the minimum values that still removes most color noise. Note that I excluded color tints associated with dynamic range loss from my threshold for NR (ie, I didn't try to remove the magenta tint otherwise it would further destroy color fidelity of the color checker chart). The new color NR values range from about 5 to 12. 100% Crop Composite #1, color NR optimized
Nice matrix, makes it easy to compare. From best to worst A7s, DF, tie between A7 and A7r... That's my take on the comparison, but your mileage may vary.
mdbrown9999 wrote:
Nice matrix, makes it easy to compare. From best to worst A7s, DF, tie between A7 and A7r... That's my take on the comparison, but your mileage may vary.
That's basically what it should be- the A7S is current-gen tuned for high-ISO, the Df is last-gen tuned for high-ISO, and the A7/A7R (and D6x0/D8x0*E) are current-gen tuned for low-ISO. I'd expect current Canon full-frames, also tuned for high-ISO, to clock in around where the Df sits.
mdbrown9999 wrote:
Nice matrix, makes it easy to compare. From best to worst A7s, DF, tie between A7 and A7r... That's my take on the comparison, but your mileage may vary.
The A7 and A7r are very close but on careful examination the A7r edges out the A7 - just a smidgen less noise in both the midtones and shadows.
A7s had some sort of noise reduction done in camera above iso100000 if you look at the camera lens in the photos you can see the smoothing compared to the df. Leads me to df>a7s >a7 >a7r
What a difference compared to when I looked at the composites the first time. As it happened I checked the first version minutes after you starting the thread. Strange results but as it was the first time for me seeing direct AB tests I had to believe the result.
Then I couldn't believe the strange comments....
kezeka, I don't understand. Are you comparing the noise or are you telling us what image you like the most? When looking at the color chart and other details the A7s wins comfortably. The "detail" in the center of the camera lens in the DF 100k ISO image is odd; you don't have it in the lower ISO images... The color and tonality differences are even more clear so in the 200+k ISO image. That is to my eyes... but those high ISO values aren't suitable for color. I also prefer the reds from Nikon compared to the A7s. That reminds me about the old Sony A900 which had more noise and better colours. Here we have the opposite?!
snapsy, are the crops 50 or 100 percent crops? Thank you for the work!
kezeka wrote:
A7s had some sort of noise reduction done in camera above iso100000 if you look at the camera lens in the photos you can see the smoothing compared to the df. Leads me to df>a7s >a7 >a7r
DxoMark tests for raw-based NR and will report it as an open circle on the SNR and DR graphs - they didn't detect any NR for the A7s. I also analyzed FFTs of the A7s raws and didn't see any of the telltale signs of NR (here is an FF comparison of the A7s and 1DX for ISO 102,400)
What a difference compared to when I looked at the composites the first time. As it happened I checked the first version minutes after you starting the thread. Strange results but as it was the first time for me seeing direct AB tests I had to believe the result.
Then I couldn't believe the strange comments....
kezeka, I don't understand. Are you comparing the noise or are you telling us what image you like the most? When looking at the color chart and other details the A7s wins comfortably. The "detail" in the center of the camera lens in the DF 100k ISO image is odd; you don't have it in the lower ISO images... The color and tonality differences are even more clear so in the 200+k ISO image. That is to my eyes... but those high ISO values aren't suitable for color. I also prefer the reds from Nikon compared to the A7s. That reminds me about the old Sony A900 which had more noise and better colours. Here we have the opposite?!
snapsy, are the crops 50 or 100 percent crops? Thank you for the work!
Yep, I screwed up the original post. First, the labels between the A7s and Df were transposed. Second, I had accidentally scaled the 100% crop to 8MP, which is what produced the huge original composite images I posted. The corrected composites contain 100% crops.
The fact that we are talking about ISO 12,800, let alone 100,000+, is simply astounding. In terms of real world usage, I'm sure there isn't much difference in the A7s and Df.
Sean Carpenter wrote:
The fact that we are talking about ISO 12,800, let alone 100,000+, is simply astounding. In terms of real world usage, I'm sure there isn't much difference in the A7s and Df.
One has nice 100% accurate EVF and one has pretty miserable OVF. Also given it should have been "manual focus" camera its pretty bad for that.. expected way more than Nikon delivered. Liked idea of that camera, thats till it was clear it isnt "that camera".
A7s over it any day.. (I tried few RAWs out of it recently, suprisingly I kinda like that output). I think Sony again tweaked CFA in different direction and this time it was probably right for me.
