One of the few areas where Sony is doing really well is imaging sensor production. And today Sony announced the acquisition of the Tsuruoka fab from Nintendo! The fab used to make chips for Nintendo game console (Source: Image Sensor World). Sony reports:
This planned investment forms part of Sony’s mid- to long-term plan to increase its total production capacity for image sensors from the current capacity of approximately 60,000 wafers per month to approximately 75,000 wafers per month.Following the acquisition, Sony intends to convert part of the semiconductor manufacturing equipment that Sony is scheduled to acquire from Renesas Yamagata Semiconductor to CMOS image sensor manufacturing equipment, and also to install new manufacturing equipment for CMOS image sensors in Yamagata TEC. Yamagata TEC will primarily engage in the manufacture of photodiodes and wiring processes for stacked CMOS image sensors.
Now what we need from Sony is an increased lens production capability too and some really great new camera to put Nikon and Canon in the shadow.
0 Responses
I want a Nintendo menu style on next RX1! 🙂
Princess Peach is definitely an E-Mount user.
Grrrr… A-mount big man camera.
*flexes biceps*
well first of putting canikon in shadow you must win Fuji my dear SAR LOL and with xt1 prepare to sink
Keep shooting JPG mate, or spend your time finding better RAW converter.
@magnus what’s your problem? 🙂 I think you’re posting on the wrong place..
Wow, amazing, no more nintendo? 🙁
The article does not say that, try another castle.
The press release says STACKED sensor production, does this mean they are definitely going to make a foveon type sensor?
Yay! Can play Super Mario on Sony A9
…. and we could have that mind boggling repetitive music playing while we look through the menus!
Cool. Finally… we can save the kidnapped princess right on our cameras…
yay i cant wait for full frame mirrorless nintendo ds!
It’s kind of amazing to see that the core technology production of all that silicon stuff is so similar that it’s cheaper to buy factories from a completely different branch than to build a new one.
Its good that sony will now make stacked sensors
Now lets see them in our cameras.
54mp x 3 layers for ff
36mp. X 3 layers for. c.
Even 24mp x 3 layer for C would be ok.
Stacked sensors actually all 4x the effective res of bayer because they produce actual full color pixels rsther than interpolated.
That means that the same lenses currently in use will produce 4x the effective resolution they are capable of with bayer.
The bad news is that since they just bought the factory it will be quite a while before they reach the cameras.
Actually with 54mpx 3 layer ff I wont need C.
If you really want top noise performance, a 105MP x 3 stacked APS-C sensor is absolutely necessary. Totes McGotes.
So how much do you plan on spending on external storage with this uber huge MP camera?
Storage really isn’t a problem, 4 TB drives are relatively inexpensive. A decent processor will be needed though for getting through the RAW files in a reasonable amount of time. I wouldn’t attempt to process a shoot of 50 to 1000 96MP images on my 2011 computer.
Should have read, “500 to 1000.”
$100. for a 4tb drive…get real.
I was making a little joke. I think we’re already very close to the point of diminishing returns with big MP sensors.
Still, I’m not a camera sensor engineer, so won’t claim to know what the practical result of any of this will be.
You won’t mind the loss of the additional 1.5x reach or narrower DoF?
If I crop to C size..same dof..I still have 24mp x 3 layers..the equivelent of 96mp bayer…I can live with that.
So you take the single stack resolution and multiply it by 4 than not 3? Once you crop down to aps-c, wouldn’t a 24x4mp FF sensor go down to 64MP?
They may already be setting up another factory for stacked sensors, wasn’t there a big investment & expansion to an existing semiconductor factory announced last year?
Where did you take those 4×? The difference is less than 1.5 times, because 36Mp bayer sensor shows better resolution than 16(15?)Mp foveon.
My response to you printed on the bottom of the page.
Is there such thing in production or should these lines start out as “In theory”?
Previous rumors stated that the cameras with new sensors will be released in 2015 at the earliest. It sounds like they are in the stages of prepping for production. Since it is a new process this prep likely takes a lot longer than for a new standard CMOS design.
Correct. Major breakthrough coming in 2015. Not this year.
Maybe I’m not fully clear on the tech behind the stacked sensor, but I’m skeptical that each stack of early sensors will equal or exceed the highest resolution traditional CMOS sensors (per stack). I doubt we will see 36 x 3 for 108MP Sensors at the on-set. I’m sure it will eventually scale up to that, but I would expect (not hope for or vote for) something more along the lines of 8 to 16 mp per stack for better low light performance, expense to produce and for future upgrade path.
Given that they will be changing processes I’m skeptical that the resolution envelope will be pushed exponentially over current top res sensors to start out with (see Intel’s tick-tick strategy of upgrading chips). I also don’t see the total resolution exceeding their new MF format sensor only a year after the MF sensor was announced.
My personal opinion is that I would be pretty satisfied with a full frame sensor with 16 x 3 MP if it offered a great balance of speed, buffer, resolution and solid high ISO performance up to 6400 (at least 1 to 1.5 stops better than the a99 noise levels).
sure about your interpretation of “stacked sensor”?
Sony made stacked sensors for cell phones since at least 2010, but this devices are stacked chips of the photo diode layer and a computation layer bound directly behind.
I think it has nothing to do with foveon like sensor design.
” “Stacked CMOS image sensors” are CMOS image sensors in a stacked structure that layer the pixel section containing back-illuminated structure pixels onto chips containing the circuit for signal processing, as an alternative to the supporting substrates used in conventional back-illuminated CMOS image sensors.”
look at http://www.sony.net/SonyInfo/News/Press/201401/14-016E/index.html footnote *3
Might as well put Mario and Luigi in charge of Sony, they couldn’t do any worse than the current idiots there.
+1
Actually Yamagata Tech will be making the stacked sensors. The new fsctory will make current tech sensors to fill the production shortfall.
Possibly using some of Yamsgatas old equipment.
Maybe while Mario’s stuck in my camera, he can surf the damn pipes and install some of those firmware updates I’ve been waiting on.
Square CMOS?
The big technology in sensors in the near future is Pansonic and Fuji’s ORGANIC cmos tech. Its a dramatic improvement in signal to noise ratio and dynamic range.
This will make the current tech obsolete.
I hope Sony has their own competing tech.
Bayer sensors..example 24mp…consiststt of 24 monochrome mp..12mp green and 6mp red and blue.
In a process called demosaicing 4 monochrome pixels are interpolated into a single 3 color pixel at the location central to the 4.
That yields a 6mp file.
To increase the resolution this process is repeated 4x with 4 different combinations of 4 adjacent pixels therfore yielding a 24mp file interpolated from 6mp r, 6mp b and 12mp g.
A 3 layer 24mp stacked sensor would also produce a 24mp color file, however these would be real actual pixels..not interpolated pixels.
If these pixels are interpolated at the same level they would yield a 96mp file of equal pixel quality to the 24mp bayer sensor pixels.
Oh okay, so it multiplied by 4 than not 3. Thanks.
Bayer pixels are interpolated 4x to produce the same number of color pixels as the original monochrome pixels.
We would interpolate the 3 layer pixels athe the same 4x level to produce acomparable file.
In your example 16mp x 3 layers= 48mp monochrome pixels.
When we interpolate the original 16mp colr file we get 16mp x 4 = 54mp.
The comparable file produced by a stacked sensor is actually 1/3 more pixels than the one produced by a bayer sensor with the same number of pixels.
It is quite possible that the rumored new 54mp stacked sensor would actually be 18mp x 3 layers and produce a 18mp raw color file that can then be interpolated to 74mp color file (Bayer equivelent)
So, let’s say Sony released a 6mp x 4 stacked sensor for an interpolated 24mp file. The a99 raw files are 6000 x 4000. Would the stacked sensor files be 6000 x 4000 or 8000 x 3900 ( 1/3 more) or would it show up as only 3000 x 2000 when opened in camera raw software?
A 24mp stacked sensor would be 8mp x 3 layers.
The raw file would be 8mp color pixels.
If you only wanted 8mp you would have a very high quality 8 mp file.
If you wanted higher res you would double the resolution…4x the pixels…and you would have a 36mp file from a 24mp sensor.
If you started with a 6mp x 3 layer 18 mp stacked sensor you would end up with a 24 mp file.
Barry, honest question (since you seem to know a lot about sensor technology):
Why does FOVEON sensor S*CK at high ISOs? Is it a kind of limitation of current 3-layered-sensor technology or simply Sigma doesn’t know how to do things right?
Also, when you say about “stacked sensor” you mean actually three sensors stacked together or 3-layered-sensor like Foveon?
The difficulty is that there is no hard cutoff in light as you go deeper into the sensor. Different wavelengths go to different depths, giving some degree of crosstalk between the color sensing layers. In a traditional sensor you filter out all but green, red and blue, so there is no crosstalk. To solve the crosstalk you need some serious post exposure processing to figure out what belongs where. With high ISO, the noise screws up this processing, and tends to amplify the noise.
Foveon is a stacked ORGANIC sensor.
Organic materiels are great at absorbing light but not so good at converting and transmitting electric as l signals.
Silcon…as used in current chip technology such as the cmos bayer array sensor..is very good at precessing and transmmitting electric as l signals and therfore capable of much better high iso performance than organic.
One of the new Sony patents is for a hybrid organic/silicon chip where the light absorbing portions are organic and the electrical portions are silico.
Noe..organic uses carbon instead of silicon..
This hybrid has the potential to offer the best of both technologies and resolve some of the organic problems.
My best GUESS is that this is the technology Sony will use however it is only my best guess.
A stacked sensor will have 3 sutractive color layers.
Yellow,cyan,magenta.
Probably in that order because yellow is the cleanest colr and magenta is the least clean colr.
Each subtractive color absorbs a single primary light color..blue,red ,green.
And allows all other colrs to pass through.
Each of these colors are placed on top of each other sol they are perfectly allighned. These need only be combined to create a recognizable color image.
Bayer utilizes 3 colors alongside each other and must interpolate them in order to create a recgnizable colr image
Nick
Crosstalk from side to side as you say is fairly accurate and particularly relevent in bayer wher it crosses to pixels of different colors.
It is not a phenomenon related to depth in stacked sensors..see my explaination below.
Although crosstalk is actually more significant in a stacked sensor due to the greater depth it is less noticeable because the crossover is to adjacent pixels of the same color wheras in bayer the crossover is to pixels of a differrent color.
Mario was an Alpha user all this time and we never noticed. Even if it was so obvious when he started saying: “It’s A me: Mario!”
Better Mario than a piece of cheese.
“Now what we need from Sony is an increased lens production capability too and some really great new camera to put Nikon and Canon in the shadow.”
We’ve been hearing this for the past 5+ years and still… Nothing