ruthenium
Offline
• • • •
Upload & Sell: On
|
 p.16 #18 · After 13 years of all Sony, I'm trying Nikon | |
Steve Spencer wrote:
Exactly. We are in agreement, but with different base ISOs which same amount of light should we compare? The max that fills the wells on the camera with higher base ISO or the max that fills the wells on the camera with lower base ISO? We could use either amount of light and compare the sensors and get different answers. When comparing the Sony A1 II and the Nikon Z9 we see that at base ISO for both cameras the A1 II has .4 stops better DR, which IMO is not very large and from ISO 640 and higher about .5 stops better DR, which still is pretty small in my book. At ISO 500, the Z9 actually does slightly better than the Sony A1 II, and it is only from ISO 100 to 400 that the difference in DR is larger. The A1 II does have a bit higher DR, but the difference isn't that large and I would expect it not to be a substantial or important difference in real world shooting in almost all situations....Show more →
Steve, this isn't exactly my area of deep expertize. The dynamic range is calculated from the ratio of the numerical value for the maximum signal to the numerical value of the minimum acceptable signal.
To your question: "but with different base ISOs which same amount of light should we compare? The max that fills the wells on the camera with higher base ISO or the max that fills the wells on the camera with lower base ISO?"
My understanding is that these two maxes should be practically the same, at or near 16383, assuming that the Analog-to-Digital-Converter is 14 bit and the light has been correctly digitized to match the strongest highlights to the largest number the ADC can handle: 16383.
If sensor A received more light than sensor B, it is tempting to claim that the signal-to-noise ratio (SNR) for A is greater than the SNR(B), and hence the DR. Let's look into this. We know that at base (and generally at low) ISO, the noise is dominated by the photon shot noise, and not by the electronic read noise. According to some calculations, the ratio of the two is about 99 to 1, and the read noise can be ignored. The shot noise increases as the square root of the signal.
For example, if the full-well capacity is 80,000 electrons per pixel, then the shot noise is at about 283.
If the full-well capacity is 60,000 then the shot noise is 245.
When the full-well capacities are digitized to match the largest number the ADC can handle, the noise is scaled proportionally.
In this example, e.g., scaling 60,000 to 80,000 means the 245 noise is scaled to 328.
Thus, I conclude that the sensor that received more light should have a better SNR than the sensor that received less light.
The DR of the former should be better than that of the latter, assuming the same sensor technology.
I hope my numbers are correct, and if there are experts who know better, please, correct me.
|
but the truth is I have not been able to see any problems in actual real-world shooting of cameras from many different brands over the years that have different DR numbers. The example here is from the Z9 and I intentionally and massively underexposed to protect the highlights, then pulled up exposure/shadows in Lightroom by 3 stops, and to me it looks good.