Nikon Z6 III Partially Stacked Sensor

I've moved the contentious posts into the Dumpster and deleted a couple of mine.

if you wish to see threads in the Dumpster, you have to ask me and I will enable it for you.
.

Hi,
I just wrote an article about it, which you'll find in the usual outlet.
Think of this as the same technology as AMD uses to make processors from a number of chiplets. It allows different silicon tech to be used for different parts of the circuit. The tech that is optimal for pixel arrays is along way from that which is optimal for ADCs. By stacking specialist ADC chips they can have faster ADCs. The same tech allows them to put in buffer memory, which means that ADC readings can be cleared more quickly thus making read-out faster. The stacked chips are smaller so much less costly than a full stack - which also causes yield problems, because they are made by stacking whole wafers, so a defect in any of the chips in a stack voids the whole sensor.

I doubt that the idea of a 'partially stacked sensor' itself is patentable, because 3DICs are not new, and used in a load of applications, such as for instance processors from Apple and AMD (so-called 'chiplet' processors). Just using a technique already used in the semiconductor industry to a particular type of semiconductor would be classed as 'obvious' and therefore not patentable.
The point about this is that this type of 3DIC has not been used in commodity image sensors before. Previously stacked sensors have been stacked at the wafer level, before being split into individual dice. The advantage of this is that fab lines already have wafer positioning equipment that can achieve the precision of alignment required. This technique requires that all the chips in the stack are the size of the whole sensor, so it's expensive. To stack die-to-die or wafer-to-die requires equipment that can position individual dies to the required precision, which most fab lines don't have.
Speculating on what has happened here, Sony Semiconductor Solutions (which almost certainly makes this sensor) has now included this equipment in its fab plant. Hazarding another guess, the equipment itself is based around or uses one or more Nikon Precision's alignment stations (www.nikonprecision.com/products-and-technology/alignment-stations/) and that Nikon is therefore the lead customer for this new product. As you say, Sony Imaging is another customer that might well make use of it, though I'd note that the A7 IV has a 33MP sensor, so I wouldn't expect it to use precisely the same sensor. A more likely second customer for this sensor would be Panasonic.

The term 'Partially-stacked sensor' is undoubtedly marketing jargon. However, the sensor itself does seem to have gained some performance advantages by the incorporation of specialist ADC dies.

He didn't say that they were 'just lies'. He said that the term 'partially stacked sensor' is marketing jargon, which it is. The normal technical terminology would have it that this is a 3-dimensional Integrated circuit (3DIC) functioning as an image sensor.

I don't think that the first speculation is close to the mark. In a non-stacked sensor the ADCs are right there on the sensor chip, so any heat they produce has a direct path to the pixel array. The ADC's on the helper chips will produce less heat anyway, since they will use a finer node, lower power process, and the heat has to get through a silicon-silicon bond which might approach the transmissivity of solid silicon, but won't be worse. In any case, reports I've seen so far suggest that it is a good low light performer. On the focus points, there is nothing in the technology that limits the number of focus points - Nikon tends to the lower end just by custom and the Z6 III has more focus points (299) than the Z6 II did (273). Higher MP cameras tend to have more. There's always a trade-off. The more focus points the more focus pixels need to be interpolated over and different companies take a different view on what is the best compromise between number of focus points and image quality, except for Canon, which with dual-pixel sensors can use any imaging pixel as a focus pixel. In any case, the number of focus points doesn't seem to have a great effect on focus accuracy. The focus being 'jumpy' also seems to be a Nikon feature. It's making different decisions about what is the subject. I have problems in subject recognition modes trying to photograph bowlers in cricket releasing the ball. I follow the bowler's run-up and it tracks perfectly. Then it decides that the umpire or batter is the target and switches. 3-D tracking mode works better, and allows me to track the ball rather then the bowler. Still, the technology's performance will get much better for you when Sony produces a camera using it.😉

You can't tell, can you? It depends on the speed of the ball. You can't tell whether or not the ball here is travelling at the same speed as the A7r II photos that you're comparing it with. It seems unlikely that a netball would move at anything like the speed of a football. In the end it depends on how much the ball moves in the frame, so you also have to take into account framing. In any case, if you're comparing the Z6 III electronic shutter with the A7r II's mechanical shutter, then it's unsurprising, since the Z6 III's electronic shutter has about twice the transition time as a low-end (1/125 sync) mechanical shutter. That's why Nikon put in a mechanical shutter. If you're comparing like with like, both in electronic shutter mode there must be something like the ball speed effect in play because the A7r II has a sync speed of 1/12s. Or else it's just a faulty memory and a systematic comparison would reveal something different.