The sensor resolution is up to 150MP (12288 x 12288 pixel). No silicon results yet, but the tapeout was scheduled for May 2014.
Thursday, July 24, 2014
150MP HDR Sensor Thesis
Delft University publishes MSc thesis "The Design of a Stitched, High-dynamic Range CMOS Particle Sensor" by Jiaqi Zhu. The thesis presents a cooperative work with Caeleste. The DR extension is achieved by a dual-gain 6.5um pixel:
The sensor resolution is up to 150MP (12288 x 12288 pixel). No silicon results yet, but the tapeout was scheduled for May 2014.
The sensor resolution is up to 150MP (12288 x 12288 pixel). No silicon results yet, but the tapeout was scheduled for May 2014.
What is now with this work?
ReplyDeleteNice work! It is indeed challenging to implement a global shutter HDR imager. It seems to me that the readout noise of 5e might be hard to achieve, since you do not completely cancel out reset noise, and the high resolution compounds the problem. The floating diffusion node is very critical for both leakage and parasitic light sensitivity. Also I think about the MIM mismatches that may be present. I would have really liked to see how the routing of the signal lines are in this pixel!
ReplyDeleteAll these issues are common in any design. There is no performance shown on these aspects. How can you claim a nice work?
DeleteI am complimenting the student; he has done quite substantial work considering his expertise. It is his first silicon if I see his CV. I have already put my concerns of noise, mismatch, PLS regarding the product.
DeleteThe pixel design looks standard to me. The only nice part of the work is the stitching of an extreme resolution and, i guess, the capability to maintain homogeneity of the control signals and limited noise injection over the whole array.
ReplyDeleteThis is the LOFIC pixel developed by prof. Sugawa. Strange that Albert Theuwissen didn't notice there are no references to any of the many papers Sugawa wrote on LOFIC pixels. Albert gets so angry when others don't cite his works!
ReplyDeleteAlbert has invented so many things and why he should cite others? There is a small community and people cite each other. You don't exist for them, no complain just show awesome things made by you. I didn't cite this work neither :)-
Deletei think this criticism is a little bit unfair, as the student wrote this thesis, not Albert.
DeleteLOFIC....actually it looks a lot like cascaded integration, which predates LOFIC, and yet is ignored by the LOFIC authors...so, lack of references also cascades. I also noticed a lack of references in this thesis. It is too bad doing a good bibliography was not emphasized as part of the student's scholarly work.
DeleteAs a global shutter, what about shutter efficiency? With such a resolution, the readout time may be very long.
ReplyDeleteHaving a PPD with two or more TG gates is in an old Photobit patent, and sold by Micron to a patent troll. Beware! If this ever becomes successful the troll will be coming for Caeleste. And to be sure, the inventors will receive nothing.
ReplyDeleteEric, if this is true, then the troll can attack all the guys who make GS sensors, such as E2V. Since if the FD is used for image storage, then a second TG for reset is mandatory. Do you have the reference of this patent please ?
Delete-yang ni
sure, the first patent in the family is US 6,239,456. See claim 1, for example.
DeleteThe patent is now owned, I think, by Round Rock which is an NPE and who's founder does a lot of litigation for Intellectual Ventures. This patent could be of importance for TOF applications and GS applications.
BTW, the general GS application is also covered under a Caltech patent but that is a different discussion and most companies have a license already. Probably ON already has an inherited license to the dual TG patent.
Please notice that this is work done by a MSc student at Caeleste and under the full guidance of Caeleste. The student did a lot of work during the time she was with Caeleste. At this moment, silicon is being fabricated and measurement results are not yet available.
ReplyDeleteThis is not a PhD thesis, and I am pretty well aware of the fact that the thesis indeed does not have the quality of a PhD thesis, but that is not required in this case. Moreover, parts of the work are not included because of confidentiality reasons set by Caeleste.
Set a correct bibliograhic reference is not covered by the confidentiality. That is all people complain about. You have to keep this guidance as scientific advisor.
DeleteDear Albert, Arnaud, Eric and other courageous internet anonymous-and-yet-complaining users. ;)
ReplyDeleteAs Albert pointed out, this is the result of Albert's student work at Caeleste. It is a thesis project, engineering degree and NOT a PhD work. This student was embedded in our design team during the course of the project and was supervised by our team and by Albert. This is a project I followed personally and I must say I am very pleased with this student work. the student did work very hard, and the result, for a master student are quite impressive.
For obvious reasons, there is some confidentiality matter here. the pixel depicted in this thesis is not fully disclosed in this thesis. part of the pixel structure is not shown.
Of course, indeed, a more elaborate bibliography would be nice, and I am quite sure Albert will have some points to say about this later this week.
and regarding the metal routing question from the first user. well, it is indeed a tricky one ;-)
if you have any questions, do not hesitate to contact me.
cheers guys.
Benoit Dupont
Benoit, as the master thesis becomes public, everybody can read it. By not including the references, people might think this ingenious pixel has been invented by Delft U, or by Caeleste which is not the case. The student did a good job indeed but the scientific advisors should be a bit more careful while reviewing the text before final submission. After all, this is from a student of Albert! And all the works of Albert are detected by this blog administrator ;),
ReplyDeleteXA
Exactly in phase with what you said. The scientific "dynamic range" is low and the "knowledge" leakage is high for some guys in this field. Citing the works from others will not low their "inventive" work.
Delete