Originally posted by bobmaxja Still it's was not native, it was firmware base. The sensor was not rated at ISO 80.
I've not seen the actual base sensitivity stated for the Sony IMX071 sensor (the one used in the D7000, and presumably the K-5). Where was it published? Just because Pentax, Nikon & Sony chose to use ISO100 as their Base ISO for this sensor, does not mean it's the lowest the sensor can go.
The extension in Base ISO in the K-5 is real, not fudged.
The formula for the Base ISO is:
Base ISO = 78.57 * R
g / N
sat (
aka FWC)
where R
g is the Reponsivity in the green channel, the number of photo-electrons outputted as the exposure, (the result of the scene luminance, f-stop & shutter speed - measured in lux-seconds,
lx-s), varies.
Excepting the Aptiva sensor design, (used in the Nikon V1), the responsivity (related to the conversion gain) is fixed in a sensor. The conversion gain/responsivity of the sensor does not change as you increase the ISO sensitivity - what changes is either the analogue gain between the sensor & the ADC, or the scaling of the digital number (DN) after the ADC (digital gain change).
For example, with the K-5 at ISO80 (ISO70 measured), using the
Sensorgen.info data for Pentax K-5 data (derived from the DxOMark Full SNR & ISO Sensitivity curves):
Responsivity = Base Sensitivity
(measured) * FWC / 78.57
= ISO70 * 47159 e- (electrons) / 78.57 = 42 ke-/lx-s
Comparing this with the K-5's ISO100 (ISO91 measured):
ISO91 * 36512 / 78.57 = 42 ke-/lx-s
So, as I said, sensor responsivity does not change with the ISO80 base extension.
That really only leaves N
sat/FWC. FWC does not change with increasing ISO. FWC is only achieved at the true Base ISO. At higher ISOs, (than the Base ISO), what happens is that the system "saturates", instead of the sensor. What is happening is that the digital number (DN) reaches its max. value and can go no higher. So Sensorgen's FWC column should really be titled either "Apparent FWC" or "N
sat."
Now compare the D7000 (
Sensorgen.info data for Nikon D7000) at ISO100 (ISO83 measured):
ISO83 * 49058 / 78.57 = 52 ke-/lx-s
and the A580 (
Sensorgen.info data for Sony A580) which is believed to have the same sensor as a D7000:
ISO75 * 44983 / 78.57 = 43 ke-/lx-s
Comments:
1. A580 has a similar responsivity to the K5, while the D7000 is different. This may be a measurement funny from DxOMark or a curve-fit anomaly (Sensorgen does a curve-fit on the published DxOMark Full SNR curve data, which apparently DxOMark smoothes, to solve equations to derive FWC and Read Noise values).
2. The 3 sensors have similar FWCs at their lowest ISO (measured), but the D7000 is furthest from the K-5 in FWC. Perhaps a measurement anomaly?
A570 @ ISO75: 44983 e-
K-5 @ ISO70: 47159 e-
D7000 @ ISO83: 49058 e-
Since FWC is set in the sensor design (capacitance in the sensel), and since it's unlikely that there would be differnt versions of this chip in these cameras (all released around the same time), I think the FWC would be the same for all 3 sensors.
3. However, a similar FWC in the K-5, when compared to the other cameras, occurs at ISO80
(manufacturer). At ISO100
(manufacturer) the FWC is only 36512 e-. So why the low FWC at ISO100? One possibility that occurs to me, is that the sensor performance was restricted by 12-bit raw. The K-5 is the only camera of the 3 mentioned here to use 14-bit raw. So Pentax engineers calibrated its metering system for optimum performance at ISO80
(manufacturer), where it's real FWC is revealed, but chose ISO100
(manufacturer) as the default, because it was more standardised and expected. So at ISO100, the DN is maxing out before the real FWC is reached.
4. The other possibility is that standard P&S & DSLRs have linearly-responding sensors. Perhaps the linearity degrades as FWC is approached. So instead of experiencing the non-linearity, the sensor manufacturers specify an FWC value that is a little lower. But the Pentax engineers thought that they could get a bit more out of the sensor by continuing to operate further up into this region, and then correcting the non-linearity afterwards (before storing the data to an image file).
Dan.