« on: August 07, 2014, 05:17:39 PM »
So, on the topic of DXO and their trustworthiness, since ZigZag decided to recreate his account and ruffle feathers again. I often use SensorGen.info, which uses DXO's raw data to provide unbiased statistics of cameras. I found this to be interesting:
D800 @ ISO 100:
D810 @ ISO 64:
Quantum efficiency got 10% worse with the new sensor? Or does sensitivity affect QE (hence losing 6 percentage points with the lower native sensitivity)?
Quantum efficiency describes the ability of the photodiodes to convert incident photons to free charge. It IS sensitivity. At 50%, the D810 converts 50% of the incident photons to free electron charge in the potential well. That means it is, quite literally, 6% LESS sensitive than the D800.
I don't know why, doesn't seem logical to me, hence the reason it's interesting. That, and the fact that somehow, Nikon and/or Sony figured out how to nearly double the saturation point for those pixels, despite the fact that they are the exact same size as the D800? I dunno....that smells really fishy to me...capacity is primarily related to pixel area. The pixel area is identical for those two sensors (at least, as far as I now...I guess it's possible Sony moved to an even smaller process, but that isn't going to double the photodiode area by any means...)
The pixels have roughly the same capacity at ISO 100, so it makes me think that Nikon is doing something...quirky...with ISO 64.
One thing I will say, though...I find the DR numbers of the D810 more believable than the DR numbers for the D800. There is always some overhead, even if it's a fraction of a bit. Either Sensorgen.info is really referring to information in analog space (in which case were talking about electrons in a circuit, before the ADC has kicked in), or there is some amount of error in the numbers they are displaying. If they have actual electronic measurements taken directly from the sensors pixels...well that would be pretty amazing.
I wouldn't put too much stock in the exact precision of the Sensorgen-derived QE numbers; they're based on various interconnected metrics of DxO's results rather than a more direct QE measurement method. The quoted difference is 10% not 6% (on a relative basis). As for the increase of full-well capacity for the D810, there are several established techniques for achieving that, including Aptina's DR-Pix technology, which is rumored to be what Nikon uses on the D4 sensor https://www.aptina.com/products/technology/DR-Pix_WhitePaper.pdf.