Imaging Resource published an interview with Masahiro Suzuki, General Manager R&D, Nikon Imaging devoted to the announcement of the new interchangeable lens Nikon 1 camera system. A significant part of this interview is devoted to its 1-inch 10.1MP image sensor based on 3.4um pixels. The sensor has 24 digital outputs, while it's not clear what is the full resolution frame rate.
Few interesting quotes:
On the 10MP resolution in relatively high-end product: "Our message will be that image resolution is not everything. We offer you additional value. The image quality from 10 megapixel sensor is excellent, good enough for the kind of use that consumers make of their pictures, even for quite big enlargements."
"We [Nikon] developed it; we engineered and developed this sensor inside Nikon. But for the production side, that is done by our partner."
"The major difference from the others is that this camera's image sensor has embedded phase detection AF, so that achieves very fast focusing."
Imaging Resource clarifies: "Nikon V1 and J1 both have a hybrid autofocus system that combines both phase-detection and contrast-detection modes. The operating mode is chosen automatically as appropriate to the shooting conditions, and a generous array of 73 phase detection AF points are available. Since there's no way to hook a separate autofocus sensor into the optical path in a mirrorless camera, Nikon has adopted a similar strategy to that used by Fujifilm in certain of its compact camera models last year. The phase detection autofocus points are placed on the image sensor itself, although it isn't currently clear how the focus points are spaced with regards to the surrounding photodiodes."
but phase detection needs contrast too ???
ReplyDeleteYes. Phase detection relies on the spacial separation of the image seen from two different optical paths, and contrast detection relies on the contrast (sharpness) difference of two images taken at different lens focus positions at different times. You have to have decent contrast to measure the separation in the first case (your point), and in the second case you need a decent contrast difference between two images to know which way to move the lens. Phase detection is nice because you know which way to move the lens after one measurement, not two, and with phase detection it is possible to have an estimate of approximately how far you need to go. The disadvantage of phase detection is that it is usually done on a separate sensor, so you have to take away light from the main sensor and you have to have a very high-tolerance/well calibrated relationship between the focus sensor and the image sensor. By putting the phase-detection on-sensor, you get the best of both worlds. The phase-detection sensor is typically read out much faster than an image sensor, so they may need special provisions to allow the phase-detection pixels to be read out quickly.
ReplyDeleteThe advantage PDAF on a dedicated sensor is the amount of light available with which to focus. The sensels on dedicated PDAF sensors are tens of microns wider than those on an image sensor, and thus much more sensitive. They are monochromatic and don't filter out 2/3 of the light falling on each pixel site because they don't use a Bayer filter. The light falling on them is collected from a wider area through the use of micro-lenses between the focus splitters and the PDAF sensor array. The on-sensor pixels must also be masked to collect light from only one side of the lens or the other, which results in a further reduction of light reaching the pixel well.
DeleteI guess that the contrast based method is needed when the DSC is in view finder mode. In this case, the phase detection will not received any light via the mirror. But when the mirror is in place, the sensor can not see the image, so the phase detection method is used. The modern optical instrument is precise enough to aligne the separated sensors.
ReplyDelete-yang ni
Nikon J1/V1 is a mirrorless camera. The reason I published this Nikon sensor post is that it has phase detection pixels embedded inside pixel array. This is quite unique feature appearing only in Nikon 1 and some Fujifim sensors.
ReplyDeleteBut the mechanical shutter is still needed ?
ReplyDelete- yang ni
Depending on the price. More expensive V1 model has mechanical shutter, while cheaper J1 has only electronic one.
ReplyDeleteI see. The Ti based mechanical shutter gives a 250fps equivalent rolling effect. If they can read out the image at 250fps, no mechanical shutter is needed at all. I guess that the low end one should give 1/125 flash sync speed.
ReplyDelete-yang ni
J1's sync speed is 1/60s. This gives about 300-350Mbps per each of the sensor's 24 digital outputs, assuming 12b per pixel. Not very fast.
ReplyDeletewhat does this mean for the sensor suppliers to Nikon like aptina or sony?
ReplyDeleteAnyone had a chance to look at Nikon's US patent 7,873,267 ("Focus Detection Device...") and comment?
ReplyDeleteRENESAS
ReplyDeleteIt has got to be OV
ReplyDeleteNikon uses the phase detect sensors for the final image. Examining raw files at first I was not able to determine the positioning of the PD pixels.
ReplyDeleteAfter that I stumbled across this file - http://images.quesabesde.com/raw.php?raw=nikon_1v1_dsc_0071.NEF
In the overexposed white areas the PD pixels are clearly visible. They are darker than the surrounding pixels. It seems Nikon are correcting the raw output to mask PD sensors, and they remain visible only on otherwise oversaturated area.
I have used dcraw to observe raw files.
Is it possible Toshiba/Fuji to produce the sensor for Nikon?
My vote is for Toshiba too. It has image sensor process in place and needs more customers to recover process development cost.
ReplyDeleteWell, this sensor may be from aptina? N company has deep relationship with aptina japan.
ReplyDelete