The World's Largest Ultrahigh-Sensitivity CMOS Image Sensor

3kramd5

EOS 5D MK IV
Mar 2, 2012
3,077
397
HarryFilm said:
3kramd5 said:
kaihp said:
HarryFilm said:
Well now that you said it, I CAN MOST DEFINITELY SUGGEST TO YOU that indeed Canon is looking how they would layout multiple Medium format sensors of DCI 8k (8192 x 4380 pixels) of 56mm x 42mm size. Here is the below graphic used for suggested layout of the chips on 300mm silicon wafers. The aspect ratio difference of 1.89 imaging photosite vs the actual 4:3 sensor aspect ratio is PROBABLY due to the in-between-lines-of-photosites being using for DPAF or specialty edge detection or OTHER colour-filters photosites (i.e. extra Yellow channel?)
Excuse me, but that's not how chips are (normally) placed on a wafer. They are placed in a strict Manhattan corner-to-corner pattern, so you can run a (diamond) saw through on both the horizontal and vertical.

If there is enough economic incentive (and at this size of dies, there probably is), you could make a setup as shown in the graphic. It would 'just' require a very different handling of the wafer at the dicing (sawing) and picking stage.
I've just never seen a setup like this in the real world.
Here’s the thing:
Harry Film either has people trolling him, or, more likely he invents his sources and the reference material. His claims are typically outlandish and rife with technical errors, but it’s all in good fun.
---

I should note everyone that the chip layout is from my Euro-sources which THEY MADE -- their English is only decent-enough or simply outright terrible. I would expect that and even I didn't notice the mistakes pointed out by an earlier forum poster.

Anyways.... NO! I DO NOT make up my sources ....BUT.... they are -S-O-U-R-C-E-S- which means PLURAL! That also means the information is disparate and unconfirmed so I personally cannot vouch for their accuracy....BUT....very much like detective work, when you start seeing multiple "witnesses" tell the same story, it means you need to take a closer look at what is being discussed to find some commonalities.
(i.e. what "sources" are MY SOURCES using?) Reporting is all about fact-finding and correlation....AND...while correlation does not necessarily specify causation, you ALSO cannot ignore many correlations either!

Sooooo.........All I can say with reasonable confidence, is that Canon IS TESTING a medium format around 50 megapixel, high frame rate (20 fps to 25 fps burst rate) stills camera with DCI 4K full-sensor sampled 4:2:2 video! I expect an official announcement sometime in February to April 2019.

I can ALSO STATE with reasonable confidence that an update or replacement for the Canon XC-15 camera is PROBABLY headed for a Photokina 2018 announcement. It is probable that it will be a one-inch sensor and will have 10-bit 4:2:2 interframe 60 fps encoded video AND will have an interchangeable lens system.

And finally, a short-flange APS-C sensor mirrorless that looks like an Canon M5 may also be announced either at Photokina or soon after.

THOSE are what MANY of MY SOURCES are saying!

Again, WE SHALL SEE WHAT HAPPENS !!!
Perhaps you have real sources, and perhaps English isn’t their strong suit. However you represent them as engineers, and one trait engineers typically share is attention to detail. One trait Europeans typically share is fluency in multiple languages. A European engineer misspelling “format” is highly unlikely. Format in German is format. Format in Dutch is formaat. Format in French is format. Format in Italian is formato. Format in Croatian is format. Format in Spanish is formato. Format in Czech is formát. Format in Swedish is formatera. It’s never fromat. You’ll find that aspect is similarly never apsect.

If you really have someone sending you this stuff, take it with a mountain of salt; you’re probably being trolled.
 

kaihp

I'm not new here
Mar 19, 2012
847
4
HarryFilm said:
kaihp said:
Excuse me, but that's not how chips are (normally) placed on a wafer. They are placed in a strict Manhattan corner-to-corner pattern, so you can run a (diamond) saw through on both the horizontal and vertical.

If there is enough economic incentive (and at this size of dies, there probably is), you could make a setup as shown in the graphic. It would 'just' require a very different handling of the wafer at the dicing (sawing) and picking stage.
I've just never seen a setup like this in the real world.
Sony and Philips do non-square layouts on larger sensors and other DSP circuits.
You're missing the point. Wait, you not only missed the board, but the entire planet.

Chips aspect ratios are based on the ratio of IOs vs core area, and in case of sensors, they are defined by how to want the image to look like (e.g. 3:2 or 16:9 aspect). An extreme case are DRAM chips.

The point is that the saw goes through from side to side of the wafer. Now do that on your diagram, and count how make sensors are left working.
 

Don Haines

Beware of cats with laser eyes!
Jun 4, 2012
8,084
1,564
Canada
CanonFanBoy said:
Aglet said:
KeithBreazeal said:
I can see the 8x10 format making a comeback. This is awesome news!!!
you can see it here now if you have enough scratch for that itch.. ;D

http://largesense.com/

:D I thought this was really cool until I found the $106,000 price. Ouch! Still cool, but unobtainable for mortals. Neuro probably keeps one on a shelf in the laundry room. ;)
WAIT A MINUTE!

The Canon sensor is the worlds largest at 20cm square, but this 8X10 sensor would be 20cm by 25cm.....

Something does not compute
 

3kramd5

EOS 5D MK IV
Mar 2, 2012
3,077
397
Don Haines said:
CanonFanBoy said:
Aglet said:
KeithBreazeal said:
I can see the 8x10 format making a comeback. This is awesome news!!!
you can see it here now if you have enough scratch for that itch.. ;D

http://largesense.com/

:D I thought this was really cool until I found the $106,000 price. Ouch! Still cool, but unobtainable for mortals. Neuro probably keeps one on a shelf in the laundry room. ;)
WAIT A MINUTE!

The Canon sensor is the worlds largest at 20cm square, but this 8X10 sensor would be 20cm by 25cm.....

Something does not compute
Actually, if you follow the link I posted on the second page, you’ll see that it’s actually 9X11 ;)
 

Valvebounce

EOS 5D SR
Apr 3, 2013
4,199
166
52
Isle of Wight
Hi Don.
I think you missed the Ultrahigh-Sensitivity bit! ;D

Cheers, Graham.

Don Haines said:
CanonFanBoy said:
Aglet said:
KeithBreazeal said:
I can see the 8x10 format making a comeback. This is awesome news!!!
you can see it here now if you have enough scratch for that itch.. ;D

http://largesense.com/

:D I thought this was really cool until I found the $106,000 price. Ouch! Still cool, but unobtainable for mortals. Neuro probably keeps one on a shelf in the laundry room. ;)
WAIT A MINUTE!

The Canon sensor is the worlds largest at 20cm square, but this 8X10 sensor would be 20cm by 25cm.....

Something does not compute
 

HarryFilm

EOS 7D MK II
Jun 6, 2016
445
43
Don Haines said:
CanonFanBoy said:
Aglet said:
KeithBreazeal said:
I can see the 8x10 format making a comeback. This is awesome news!!!
you can see it here now if you have enough scratch for that itch.. ;D

http://largesense.com/

:D I thought this was really cool until I found the $106,000 price. Ouch! Still cool, but unobtainable for mortals. Neuro probably keeps one on a shelf in the laundry room. ;)
WAIT A MINUTE!

The Canon sensor is the worlds largest at 20cm square, but this 8X10 sensor would be 20cm by 25cm.....

Something does not compute
---

The silicon wafers used by my "Fabs" are either 200mm or 300mm ...BUT... there are also 500mm sizes for specialty systems (i.e. very large vector-based CISC/RISC/DSP chips aka Array processors) which need a very large die size. I think I remember seeing that 500mm one at one of Lockheed Martin's subsidiaries who made custom chips for government customers. Curtis Wright and of course Teledyne-Dalsa also are starting to use larger than 300mm wafer sizes.

The largest sensors can be multi-chip OR single-die slices and if I remember correctly the largest single die slice is a 340mm by 340mm for a 16k by 16k resolution satellite-specific CMOS chip -- I think it was made my Thales or Dassault for a French commercial imaging satellite. Because it used outside-of-sensor-area DSP circuits, it needed to be 34cm by 34 cm in order to fit on a 500 mm wafer. Last i heard is the company was selling 16k x 96k images at 15000+ Euros per area scan and that they are booked solid for satellite time! SOMEONE HIGH UP is most certainly booking up all the time so that no-one else gets the imagery!

---

Now those of you wondering about my sources...A source BY DEFINITION is generally unconfirmable and untrustworthy! You need MULTIPLE SOURCES to confirm ANY rumour of any type! Ergo, I like to cultivate multiple anonymous "sources" which hopefully have "real data" which can at the very least, be cross-checked and correlated to come up with a MOST-LIKELY SCENARIO. I have television media training (i.e. a Diploma in Broadcast Television) which means I was taught to DIG DEEP and CORRELATE MULTIPLE DATA SOURCE so that any given "fact" can be given a specific weight. SOME SOURCES are "Better" than others so I give them more weight. Other sources are newer and have no track record with me and I give hem much less weight.

I also have enough technical education that I can at least Quantify the legitimacy of SOME TYPES OF TECHNICAL DATA specifically relating to camera gear, imaging sensors, computer processing techniques and technology, computer memory and online/offline storage systems and aerospace/UAV/drone systems. THOSE are in my area of expertise.

Soooooo......using a simple spreadsheet and a basic 1-to-10 weighting system you give ALL facts a weight based upon the NUMBER of disparate sources that are giving said fact out (i.e. Canon is testing an MF camera!). If 7 sources out of 10 are giving me that fact, and 4 of those sources are in Japanese sensor engineering related to or known by people at Canon or its subsidiaries, then I can use a Mean and Average-based scoring system to tell me that BY WEIGHTED AVERAGE, a lot of high quality sources are telling me Canon is making a big sensor! Then If at least two OTHER sources from say....a flourite-related sub-systems manufacturer and a magnesium/aluminum tubing supplier tells me something of a certain diameter and length is going to certain factories in Japan, then I need to add that weight in too!

Ergo, Canon is a making VERY BIG LENSES for a VERY BIG SENSOR and it's for a near 65mm system....Sooooo...SEE how this works?...Correlation MAY HAVE a specific causation and by digging even further, I get to see if a specific causation is actually related to said correlations...i.e. Canon is making a large-sensor MF Stills or Video Camera and thus needs big tubed-lenses, lots of heavy indium-coated flourite glass, big magnesium bodies and big sensors !!!
 

3kramd5

EOS 5D MK IV
Mar 2, 2012
3,077
397
HarryFilm said:
Now those of you wondering about my sources...A source BY DEFINITION is generally unconfirmable and untrustworthy!
Please cite your multiple definitions which on weighted average return “unconfirmable and untrustworthy” as defining characteristics of “source.”