Sunday, April 13, 2014

Rambus Conditional Reset Patent Application

Rambus PCT application WO/2014/055391 "Conditional-Reset Multi-Bit Read-out Image Sensor" by Thomas Vogelsang, Michael Guidash, Song Xue, and James Harris talks about a version of the old idea to expand DR by resetting pixel each time its signal value crosses a threshold during the integration time: "If the pixel signal exceeds a sampling threshold, the photosensitive element is reset. During an image capture period, digital values associated with pixel signals that exceed a sampling threshold are accumulated into image data."

51 comments:

  1. It seems that RAMBUS becomes more down to earth now.

    ReplyDelete
    Replies
    1. This idea is not new. There are several similar patents, for example: US 7,218,350 B2, so this may be invalidated by prior art if it actually comes to a court.

      Delete
  2. I don't think so. I am always amazed at Jerry Hynecek's inventiveness and the referenced patent is another good example. But, reading the claims and comparing to the Rambus application, it looks to my slightly trained eye that while the concepts are quite similar, they are substantively different in specification and claims.

    ReplyDelete
  3. This is an old idea. I remember discussing it with some colleagues at a company I haven't worked at for at least seven years. I would be surprised if patent application were granted (although anything is possible with the US patent system), and even more surprised if it stood up to a challenge.

    ReplyDelete
  4. Rambus has a history of patent trolling (or whatever you call pushing through a technique in a standards body and later patenting it), so at minimum they have enough lawyers to tie things up in the courts for a while.

    ReplyDelete
    Replies
    1. That is not true. Rambus joined JEDEC in the early 1990's - they disclosed
      the '703 patent to the standards setting body. The '703 patent's written description
      supported all of the subsequent claims that were then filed - that is perfectly legal
      under the Kingsdown precedent.

      There was NO duty to disclose pending patent claims. And even if there were, Rambus had NO pending patent claims that read on the standards being considered while Rambus was a member of JEDEC. Lastly, Rambus was not allowed to present their ideas to JEDEC - they were declined three times - that is a fact. Moreover, it was an obvious strong-arm against Rambus' clever ideas that prevented them from presenting to the standards setting body - check here:

      This is from the FTC's (Chief ALJ's) findings of fact:

      824. The chairman of the meeting, Gordon Kelley, testified that prior to the May 1992 meeting Crisp had spoken to him about the possibility of Rambus scheduling a presentation concerning DRAM design. (G. Kelley, Tr. 2553). G. Kelley also testified that he had refused to allow Rambus to present its technology for standardization at JEDEC on this and another occasion, even though he had never barred any other member company from presenting its technology. (G. Kelley, Tr. 2649-58).

      825. G. Kelley had a clear conflict of interest; he made and enforced his unilateral decision to bar Rambus from presenting its technology two weeks after he wrote in an internal company document that his company's interests were threatened by the Rambus technology and were best served if Rambus "fails to become standard." (R 279 at 7). He did not disclose this conflict to Crisp or to anyone else. (G. Kelley, Tr. 2656-57).

      And this is from the conclusions of law:

      Alled Tube does not compel a finding that Rambus's conduct before JEDEC constitutes exclusionary conduct. Here, Rambus did not at any time encourage JEDEC to promote or adopt any feature or technology for inclusion in the SDRAM standard. When asked on two occasions at JEDEC meetings if it would care to comment about its intellectual property rights, it merely declined to do so. F. 809, 855. It did not lie about its patent rights or its intention to assert them. It was not even allowed to present its technology for standardization. F. 824-25. By contrast, in Indian Head, defendant packed the annual meeting with newly registered members, by arranging and paying for people to join the industry and register as voting members, and instructed its personnel how to vote. 817 F. 2d at 947. The steel interest's recruitment of 230 members for
      purposes of casting a single vote gave it a disproportionate voice, inconsistent with the concept of consensus" standard making. Id Respondent's conduct, under the facts established in this case does not rise to the level found to constitute exclusionary conduct in Allied Tube.

      page 305 of 348 of the PDF

      http://www.ftc.gov/os/adjpro/d9302/040223initialdecision.pdf


      Delete
    2. Sorry that the other link doesn't contain the original Chief Administrative Law Judge's Initial Decision's PDF. In fact, that appears to be hard to find now. However, the
      already posted findings of fact and conclusions of law are accurate.

      This link below clearly shows that even though the the FTC's Commission didn't uphold
      their Chief ALJ's Initial Decision, the DC Circuit Court of Appeals did! A thorough
      reading of this federal appellate court's opinion (certiorari denied) provides a clear understand as to what transpired for anyone who is interested as to what really happened:

      http://cyber.law.harvard.edu/people/tfisher/2008%20Rambus.pdf

      Delete
    3. You seem to be so well aware of the JEDEC related case that I assume that you are not from Rambus’ image sensor team but probably someone from the original Rambus’ memory team and possibly someone from the highest management of Rambus. The key message of your two comments about the JEDEC case seems to be that Rambus has not only avoided a sanction but also acted in a morally correct way. At your web page you have also advertised that you are working together with inventors and giving an impression that you highly appreciate inventors (which can be also seen e.g. in Thomas Vogelsang’s business card: “Distinguished Inventor, Silver”). That was actually the reason why I came to London and wanted to meet Rambus.

      In case you truly intend to run a morally sound and legitimate business then I believe that you have not been aware of all the details related to the patent application WO/2014/055391 and that you are at the moment running an internal investigation on the matter. I believe that in a bona fide investigation it would be important at least to pose the following questions to people who are involved in the matter:
      1) Taking into account the similar operation principle of the MIG pixel structures and the pixel structures in the patent application WO/2014/055391 as well as the fact that at least three of the authors were aware of the MIG technology (two were at the same dinner and one through Kodak) then why did the authors of the patent application fail to introduce the MIG technology in the patent application WO/2014/055391 thereby jeopardizing the validity of the patent application?
      2) Why the true state of affairs related to the patent application WO/2014/055391 (invitation to meet, discussions at the dinner, awareness of the similar operation principle, etc.) was not reported to the highest management taking into account the severe juridical and image related threats that this matter exposes Rambus at?
      3) Were any Rambus employees / people Rambus is closely working with simultaneously aware of WO/2014/055391 and of any restrictions in any agreements posing limitations on patenting and/or use of the MIG technology before WO/2014/055391 was published and if so why they failed to articulate about the matter thereby posing severe juridical and image related risks to Rambus?
      4) Based on your knowledge about Rambus’ image sensor business’ royalty sharing model with your employees / people closely working with you in the field and the possibility that a large share of the royalty stream from present image sensor technologies could be directed to some of afore said people then have any written or oral agreements been made amongst at least some part of afore said people to create a win-win situation, i.e. a situation in which financial risk would be shared insuring profit irrespective of whether WO/2014/055391 provides any royalties and irrespective of the fact whether Rambus is involved in a time, money, and image consuming legal process?

      In order to show that you have acted in good faith and that you have tried thoroughly to investigate the matter I would demand written and signed answers to these questions from people who are involved in the matter (Rambus employees / people closely working with Rambus in the image sensor field).

      Delete
    4. Taking into account the circumstances and the severe threat that your business is exposed at I don’t think that there is a legal problem to ask the above questions (e.g. to ask whether there is an agreement to redistribute royalties handed out by Rambus). Actually, as I already said above in order for you to demonstrate later on that you are performing an internal investigation in good faith it would be important to demand answers to the above listed questions.

      One matter in the investigation that you should pay special attention to is that how is it possible that the four most important CMOS image sensor patents, namely the pinned photo-diode and the transistor sharing patents were sold amongst tens of other patents with only 65 M$ to OVT without any competition. The fact is that the first one greatly enhances the low light image quality and the three latter ones almost double the resolution of present CMOS image sensors. These points are of paramount importance e.g. for smart phones. You could ask your self if anybody would be interested in a smartphone if it had half the megapixels and much poorer image quality? Even by taking only the present smartphone annual market size 340 B$ (this figure was the latest estimate I’ve seen in the net) and the validity periods of afore said patents into account the worth of these patents should have been above 10 B$ unless there is another explanation for the low valuation. As I have previously explained I am convinced that the license agreements of at least the transistor sharing patents were used for blocking the MIG technology by forbidding any use, contact, publications, and patenting on the matter.

      The reason why this matter is important from patent application WO/2014/055391 point of view is that if anyone of the authors or a person working closely with Rambus were aware of any limitations on the MIG technology or even participating in the composing of the limitations then this would be legally aggravating from Rambus point of view. First of all, such limitations would have been illegal as such.

      Secondly, the knowledge of any illegal limitations would have been valuable from financial and patenting point of view – the share existence of limitations on MIG technology would be a direct evidence of how financially important and threatening the technology would have been considered by major companies in the field. On the other hand any knowledge of potential expiration of any illegal limitation would provide valuable information when the financial potential would be realized. Furthermore knowledge on any limitations on patenting would mean that the risk of similar applications would be reduced increasing the success probability of a patent application.

      Thirdly information concerning any illegal limitations would have been received under NDA (for example in an evaluation process of Kodak patents) or as a business secret. Fourthly using information concerning illegal limitations knowingly against a party suffering from the illegal limitations is hardly acceptable from legal point of view, i.e., one would have knowingly abused Pixpolar’s week position thereby increasing and lengthening Pixpolar’s suffering.

      In an internal investigation you should notice that there is a problem
      -if you were not able to get thorough and comprehensive answers to above questions an excuse being for example concerning 3) and/or 4) that one does not want to violate against an NDA or reveal a business secret, or
      -if someone has changed the story during the course of this issue or the explanations are contradictory.
      In such a case the correct way to proceed would be to withdraw the patent application or alternatively to bring the matter to the police.

      Delete
    5. START

      Please find below an analysis of the different pixel structures of WO/2014/055391 and of how they relate to prior of art.

      It is generally known that the standard three transistor (3T) pixel (figure 2 in WO/2014/055391) can be readout non-destructively. In addition it is generally known that
      -by storing the reset level of the 3T pixel in the memory,
      -by monitoring non-destructively the accumulation of the signal charge in the 3T pixel one can deduce when saturation is approaching,
      -by saving the last readout of the 3T pixel to the memory before saturation is reached, and
      -by resetting the 3T pixel immediately after the last readout
      that one can perform (without loosing any signal) non-instantaneous digital Correlated Double Sampling (CDS) by subtracting the two readout values in the memory. The benefit of the non-instantaneous digital CDS in 3T pixel is that reset noise (aka kTC noise) can be avoided.

      The reason why digital CDS is typically performed in the 3T pixel matrix is that it does not enable analog CDS unless it is provided with similar size analog memory matrix comprising in each ‘memory pixel’ two capacitors and corresponding selection transistors. Such a ‘memory matrix’ would consume an substantial amount of semiconductor area unless the ‘memory pixels’ would be placed on top of the actual pixels. The latter one would, however, necessitate the use of Back-Side Illuminated (BSI) 3T pixel matrix architecture in order not to waste semiconductor area and Fill Factor (FF). Although the ‘memory pixel matrix’ would enable also global shutter operation it is hardly ever utilized in conjunction with 3T pixels due to the higher cost.

      Irrespective of the fact whether digital or analog CDS is performed in the 3T pixel the problem is that one has to take a sample in the beginning of the integration period and in the end of the integration period meaning that there is a substantial time difference between the two correlated samples, i.e. the CDS procedure in 3T pixel is not instantaneous.

      In the 3T pixel arrangement of figure 2 it is only possible to perform row reset meaning that one would perform readout and reset when the brightest pixel of the row approaches saturation. One could also save more readout values to memory in the beginning of the integration period and just before the brightest pixel reaches saturation, i.e. one could perform Fowler sampling which is described in many e.g. infrared astronomy related publications. With the help of Fowler sampling one can remove other noise sources except 1/f noise, dark shot noise, and photon shot noise. This is important especially for pixels having only a small amount of signal charges when the readout before reset is performed.

      TO BE CONTINUED…

      Delete
    6. …CONTINUATION

      In the 3T pixels of figure 20, 22, and 23 in WO/2014/055391 an additional selection transistor is added into the pixel enabling selective (aka conditional or individual) pixel level reset. The benefit of this arrangement is that the second correlated sample to be stored to memory can be taken just below the saturation level before the pixel is reset. If at least a moderate Full Well Capacity (FWC) is assumed then 1/f noise and shot noise would dominate over other noise sources and thus only two samples would be required, i.e. there would be only little benefit of Fowler sampling – it would be nearly waste of power and memory.

      Actually whether the 1/f noise related to non-instantaneous CDS readout or the reset (aka kTC) noise related to instantaneous Uncorrelated Double Sampling (UDS) provides lower overall noise depends on FWC, on 1/f noise generation rate, and on photon induced and dark electron generation rates. Especially at low intensity levels the UDS may provide lower noise due to required long integration times. On the other hand if the shot noise at FWC level is anyway larger than 1/f or reset noise, which I believe is typically the case, then instantaneous UDS is a natural choice due to its simplicity. Based on afore said the non-instantaneous digital CDS is actually much more beneficial with regarding the 3T pixel of figure 2 in WO/2014/055391 than with regarding the 3T pixels presented in figures 20, 22, and 23 wherein instantaneous UDS would be typically the natural choice.

      The selective pixel level reset is already explained in US6,175,383 (“Method and apparatus of high dynamic range image sensor with individual pixel reset” by Orly Yadid-Pecht, Bedrabata Pain, & Eric Fossum) which comprises a similar pixel structure than in figures 20, 22, and 23 of WO/2014/055391. In WO/2014/055391 it is also described that a digital time stamp is provided when reset is performed. This is trivial since it is generally known that an image is an intensity map, i.e., an image is obtained by dividing pixel specific signals by pixel specific exposure times (see e.g. US6,115,065 “Optical imager using a method for adaptive real-time expanding of the dynamic range” by Orly Yadid-Pecht & Eric Fossum , or US6,831,689 “Optical imager using a method for adaptive real-time expanding of the dynamic range”, by Orly Yadid-Pecht).

      As already previously said, the large time difference between the two correlated samples in the non-instantaneous CDS results in a lot of 1/f noise hampering the method unless of course a deep buried channel MIG pixel were to be used (due to the lack of interface generated 1/f noise the total amount of 1/f noise would be greatly reduced). On the other hand, the reset noise related to UDS is also large. With instantaneous (analog or digital) CDS readout the amount of 1/f noise in 3T pixel could be greatly reduced but it is not enabled in a 3T pixel. The lack of instantaneous CDS readout as well as the very high level of interface generated dark noise renders the 3T pixel practically useless when compared to a modern four transistor (4T) pixel.

      TO BE CONTINUED…

      Delete
    7. …CONTINUATION

      The 4T pixel of figure 10 in WO/2014/055391 is equipped with pixel level selective reset ability in US6,175,383 fashion (extra transistor 263). This pixel is much more useful than the 3T pixels of figures 2, 20, 22, and 23 since it provides instantaneous (analog and digital) CDS readout and thus much lower 1/f noise as well as much lower level of interface generated dark noise due to the pinned photo-diode arrangement.

      In the pixel of figure 10 the surpassing of the reset threshold level could be detected by resetting the floating diffusion, then taking the first instantaneous correlated sample, next pulsing the transfer gate potential to a lower ‘threshold sensing’ potential than the transfer potential, and finally taking the second instantaneous correlated sample. In other words, one could monitor CDS wise whether there is overflow charge from the pinned photo-diode, i.e. whether the threshold is exceeded. If the ‘threshold sensing’ CDS read-out result is clearly larger than the read noise then one can pulse the transfer gate to transfer potential and after that one could take a sample replacing the afore said second instantaneous correlated sample (whether this is possible depends naturally on the Analog to Digital Converter, ADC).

      The downside of the pixel of figure 10 is that one does not enable non-destructive monitoring of signal charge amounts that are below the reset threshold level. This is especially problematic in case the dynamic range of the image is large since the pulsing rate of the transfer gate potential to the ‘threshold sensing’ potential should be set according to the brightest pixels in order to avoid saturation (aka over-exposure). When this is done it may be that the overflow charge from the most dimly lit pixels never clearly exceeds the read noise level which would lead to loss of signal and to inability to determine a pixel specific integration time which would both disable the determination of the lowest intensities.

      A way to avoid afore said problem would be after reset to continuously increase the time interval when the transfer gate is brought to the ‘threshold sensing’ potential (e.g. first time during first readout after reset, next after third readout after reset, next after seventh readout after reset, next after fifteenth readout after reset, etc.). The problem with this approach is, however, increased complexity.

      TO BE CONTINUED…

      Delete
    8. …CONTINUATION

      The benefit of the pixel of figure 15 in WO/2014/055391 is that it enables non-destructive monitoring of signal charge amounts that are below the reset threshold level as well as instantaneous (analog or digital) CDS readout. The amounts of signal charge that are below the threshold reset level can be monitored both in non-destructive non-CDS readout manner and in non-destructive instantaneous analog (or digital) CDS readout manner. Although the ability to monitor in non-destructive non-CDS readout manner an amount of signal charge that is below the threshold reset level is not directly said in WO/2014/055391 it is an inherent property of the pixel presented in figure 15 (which is self evident to any image sensor specialist) as well as of the double MIG and the double DEPFET pixels. In case it is not desired to digitize readout results below the reset threshold level it is actually not beneficial to perform non-destructive instantaneous analog CDS read-outs below the threshold reset level since in this manner higher speed and lower power consumption is resulted in.

      More precisely said, the pixel in figure 15 of the patent application WO/2014/055391 provides exactly similar inherent pixel properties as
      -single and double MIG and DEPFET (aka BCMD) pixels, namely, non-destructive non-CDS monitoring of signal charge amounts below CDS saturation level and particularly below any level that is below CDS saturation level as well as instantaneous (analog or digital) CDS readout ability when desired; and
      -double MIG and DEPFET pixels, namely, non-destructive instantaneous (analog or digital) CDS monitoring of signal charge amounts that are below CDS saturation level and particularly below any level that is below CDS saturation level (these inherent properties are self evident based on e.g. MIG patents or Stefan Wölfel’s publications on Repetitive Non-Destructive Readout, RNDR, in DEPFET sensors).
      In addition, similar floating gate readout configuration than in the pixel of figure 15 (and 19) is presented in US8,148,760 wherein this feature is described in figure 9 and in the first claim.

      One problem of the pixel of figure 15 in WO/2014/055391 is that it enables only selective row level reset (if the PG line would have been placed vertically selective pixel level reset of at least a moderate amount of signal charge would have been enabled) which is similar to standard single and double MIG or DEPFET pixels. In the pixel of figure 19 an extra transistor (412) has been added in US6,175,383 fashion in order to provide selective pixel level reset. In figure 2 of the MIG patent US8,148,760 there is, however, provided an individual (aka selective or conditional) pixel level reset mechanism featuring two transfer gates wired horizontally and vertically. The selective reset mechanism could naturally also be provided to single and double MIG or DEPFET pixels by selection transistors in the same fashion as described in US6,175,383. In the MIG technology patents it is even mentioned that the MIG pixels may comprise selection transistors which provides a direct link to the selective pixel level reset mechanism of US6,175,383.

      The main problem with the pixels of 15 and 19 is, however, the floating gate readout arrangement providing very poor coupling of the signal charge to the channel of the readout transistor and thus very poor charge conversion gain and especially so if transistor sharing between neighboring pixels were utilized. The poor charge conversion gain results in large read noise. The large read noise and the large pixel size when compared to the 4T pixel renders the pixels of figure 15 and 19 more or less useless.

      END

      Delete
    9. START

      Pleas find below further analysis on the matter.

      Based on US6,175,383 it is clear that the disclosed 3T pixels incorporating selection transistors can be monitored non-destructively in order to determine when to readout individual pixels in instantaneous destructive UDS readout fashion whereby the pixels are also reset. A fact is that non-instantaneous digital CDS based on non-destructive readouts is typically less accurate (assuming relatively large FWC, relatively low intensity, uncooled arrangement, and a surface channel source follower readout transistor) than an instantaneous destructive UDS readout and thus there is no reason why one should inevitably want to digitize the non-destructive readouts since they are practically useless.

      In US6,175,383 in the beginning of the integration time one would reset the pixel matrix in rolling reset fashion after which the pixel matrix would be readout non-destructively in rolling readout manner. Only when a pixel would approach saturation (i.e. would exceed a certain threshold) or in the end of a preset integration time it would be beneficial to readout a pixel in instantaneous destructive UDS manner.

      With the instantaneous destructive UDS readout circuitry presented in US6,175,383 one would actually obtain two subsequent non-destructive (non-UDS) samples in case reset is not performed in between the two samples, i.e. one would obtain two more or less identical non-relevant noisy samples (e.g. there would be no suppression of Fixed Pattern Noise, FPN). This fact already implicitly incorporates the idea that one may not inevitably want to digitize these useless non-destructive (non-UDS) samples since it would only increase the power consumption – e.g. one could express such a readout result by plain zeros without performing analog to digital conversion if the readout result is below threshold. The final image could be constructed in the memory simply by adding together all digitized matrix memory vectors each corresponding to one rolling readout of the entire pixel matrix, i.e. each pixel would have a certain location in the pixel matrix memory vector (in case one would store a matrix of such vectors one could afterwards decide start and end integration times individually for each pixel). Thus in US6,175,383 it would not be mandatory to provide additional digital markings for the reset time and location for a pixel that is readout in UDS fashion but it may be beneficial in some occasions. For example, if one sends to memory only the actual digitized readout values (without pauses corresponding to zero pixel values) then one would need to send also the pixel address.

      TO BE CONTINUED…

      Delete
    10. …CONTINUATION

      In Fowler sampling (of which special case is the non-instantaneous digital CDS) used in conjunction with standard 3T arrangement e.g. in astronomy it is either not beneficial to digitize the non-destructive intermediate samples in case it is desired to minimize the power consumption, memory, buffer, and/or bandwidth requirements (which probably was especially important in the early days). Instead it would be only beneficial to digitize the readout values in the beginning of the row integration period and when the brightest pixel of the row approaches saturation or in the end of a preset integration time.

      To be more precise; let’s hereby assume for simplicity
      -that only one sample is taken in the beginning and end of potentially multiple row-wise integration periods; and
      -that the same amount of time is spend for each row irrespective of the fact whether its readout results are digitized or not.
      Then one could first perform row reset, immediately after which a sample of each pixel in the row would be readout, digitized, and stored to the corresponding locations in a pixel matrix memory vector. Then one would readout the pixel matrix in non-destructive manner. In case the brightest pixel of a row would not yet approach saturation (i.e. would be below a threshold) one would not perform digitization but instead one would set all the pixel values of the row readout vector to zeros and subtract this row readout vector from the corresponding location in the pixel matrix memory vector. However, if the brightest pixel of the row would be approaching saturation then the row would be readout and reset. These non-destructive readout results of the row would be digitized and the digitized row readout vector would be subtracted from the corresponding location in the pixel matrix memory vector. One would continue to treat the pixel matrix in this rolling manner until a certain preset specification like integration time is met. The final image would be obtained by subtracting the largest positive value of the pixel matrix memory vector after which the absolute value of the pixel matrix memory vector would be taken.

      Based on afore said it is not mandatory to provide additional digital markings for the row reset time and row location when a row is digitized. In case one would spend less time for a row that is not digitized than for a row that is digitized and if one would only send digitized readout values (i.e. there would be no pauses corresponding to zero row readout vectors) to the memory then it would be important to provide both markings.

      TO BE CONTINUED…

      Delete
    11. …CONTINUATION

      In case an additional selection transistor is provided to a single MIG or single DEPFET pixel individual pixel reset is obtained. In this case one can monitor the MIG or DEPFET pixel non-destructively in non-CDS manner and when desired, to readout the pixel in instantaneous destructive CDS manner whereby the pixel is reset. This provides a direct analogy to afore said 3T pixel arrangement of US6,175,383 except that CDS readout is performed instead of UDS readout. All other previously said matters about the readout, digitization, and memory handling of US6,175,383 apply also to this case. Particularly, that one may not inevitably want to digitize the useless noisy non-destructive non-CDS readouts and that it is not mandatory to provide additional digital markings for the reset time and location for a pixel that is readout in instantaneous destructive CDS fashion.

      In case of a minimum size single MIG or single DEPFET pixel (not comprising a selection transistor) one can only perform row reset. In this case one can monitor a MIG or DEPFET pixel row non-destructively in non-CDS manner and when the brightest pixel of a row approaches saturation, one can readout the whole row in instantaneous destructive CDS manner thereby resetting the pixel row.

      To be more precise; let’s hereby assume there may be multiple row-wise integration periods as well as for simplicity that the same amount of time is spend for each row irrespective of the fact whether its readout results are digitized or not. In the beginning one could first perform row reset and then one would roll readout the pixel matrix in non-destructive non-CDS manner. In case the brightest pixel of a row would not yet approach saturation (i.e. would be below a threshold) one would not perform digitization but instead one would set all the pixel values of the row readout vector to zeros and add this row readout vector to the corresponding location in the pixel matrix memory vector. However, if the brightest pixel of the row would approach saturation then the row would be readout in instantaneous destructive CDS manner and these readout results would be digitized and the digitized row readout vector would be added to the corresponding location in the pixel matrix memory vector. One would continue to treat the pixel matrix in this rolling manner until a certain preset specification like integration time is met. In this manner the pixel matrix memory vector would comprise the final image.

      Based on afore said one may not inevitably want to digitize the useless noisy non-destructive non-CDS readouts and besides it is not mandatory to provide additional digital markings for the row reset time and row location when a row is digitized. In case one would spend less time for a row that is not digitized than for a row that is digitized and if one would only send digitized row readout values (i.e. there would be no pauses corresponding to zero row readout vectors) to the memory then it would be important to provide both markings.

      END

      Delete
    12. One more side note. If one were to digitize also the non-destructive non-CDS or non-destructive non-UDS readouts one could multiply these digitized noisy results with a zero and thus one could utilize exactly the same memory handling procedures as described in the above described four cases. In other words, one could digitize all readout results and when a pixel / row readout would be stored to memory it would still not be mandatory to provide additional digital markings for whether pixel / row reset took place, for pixel / row reset time, and/or for pixel / row location.

      Delete
    13. A qualification to afore said. Irrespective of the fact whether digitization is performed or not to a pixel readout result wherein reset is not performed (non-destructive non-CDS or non-destructive non-UDS readout) one could utilize the same integrated vertical selection circuitry that made the decision not to reset the pixel in the readout process to set the corresponding pixel specific readout vector to zeros.

      Delete
    14. Yet another point is that the integrated vertical selection circuitry could take instructions from a circuitry/chip integrating together e.g. pitch and yaw information provided by an angular velocity sensor. In case the camera roll motion would increase above a certain level afore said circuitry/chip could instruct the vertical selection circuitry to provide instantaneous destructive CDS readout, instantaneous destructive UDS readout, or reset after instantaneous non-destructive CDS readout. In case the camera roll motion would have exceeded a larger limit of resulting in too much image blur afore said circuitry/chip could instruct the vertical selection circuitry to provide zeros for the pixel readout vector irrespective of the fact whether digitization is performed or not.

      Delete
    15. START

      I think that it is very likely that the highest management of Rambus has not been aware of all details around the patent application WO/2014/055391 since I believe that what has happened is not in the best interest of Rambus. For your information please find below an assessment of potential problems related to the matter.

      Based on my previous posts there are plenty of problems related to the patent application WO/2014/055391. First of all, there is a lot of prior art that has not been introduced in the patent application. The fact that US6,175,383 is not referred to in WO/2014/055391 seems particularly strange to me taking into account
      -that Eric has been working together with Rambus image sensor team, which can be clearly deduced from Eric’s comment: “They have lived up to all their promises and we have worked well together. Their technical team members and partners are all good guys.”;
      -that Eric is one of the inventors of US6,175,383;
      -that the first two inventors (Orly Yadid-Pecht & Bedrabata Pain) of WO/2014/055391 have been working for Eric; and
      -that Eric did not mention about US6,175,383 although he even commented the content of WO/2014/055391 by writing: “But, reading the claims and comparing to the Rambus application, it looks to my slightly trained eye that while the concepts are quite similar, they are substantively different in specification and claims.”

      Based on Eric’s comment: “I am always amazed at Jerry Hynecek's inventiveness and the referenced patent is another good example.” it seems that Eric has been well aware of Jaroslav Hynecek’s work, which I believe is also the case with Michael. Since Jaroslav Hynecek’s BCMD pixel’s inherent properties are also introduced as innovative features and covered by the claims of the patent application WO/2014/055391 the BCMD patents should also have been introduced as prior art in WO/2014/055391. Besides, the Repetitive Non-Destructive Readout (RNDR) DEPFET pixel and Fowler sampling belong clearly also to prior art.

      Secondly, taking into account
      -that some of the pixel embodiments in WO/2014/055391 comprise similar inherent properties than MIG pixels;
      -that the claims of the patent application WO/2014/055391 cover inherent MIG pixel properties;
      -that at least three of the authors (two were at the same dinner and one through Kodak) of WO/2014/055391 were fully aware of the MIG technology already before filing of the priority application; and
      -that the authors failed to present MIG technology as prior art in the patent application WO/2014/055391;
      makes it without any doubt a deliberate non-disclosure to patent authorities.

      Thirdly taking into account
      -that what I told at the dinner about inherent MIG pixel properties to Thomas, Song, and Jay is presented as an innovative feature in WO/2014/055391; and
      -that Thomas and Song are listed as inventors in WO/2014/055391;
      means that the list of inventors is incomplete regardless of the fact whether WO/2014/055391 is considered inventive or not.

      Fourthly, I’ve been openly discussing about afore said inherent MIG pixel properties to various parties. In addition, I’ve also composed multiple presentations wherein afore said inherent MIG pixel properties are presented and I’ve sent these presentations by email way before the priority date of WO/2014/055391 and without any NDA e.g. to big mobile phone manufacturers, which have further presented the material to their image sensor suppliers. If you don’t believe go ahead and ask them – since the presentations are not behind NDA there should be no legal obstacles for them to handout the presentations to you.

      Fifthly, the first to invent principle was still effective in US when the priority applications of WO/2014/055391 were filed in US. Even though the PCT application WO/2014/055391 is filed at EPO this fact probably still has an influence on the matter in US.

      TO BE CONTINUED…

      Delete
    16. …CONTINUATION

      Above presented facts should have a disabling effect on the patent application WO/2014/055391. Based on this and on my previous posts my judgment is that WO/2014/055391 should not pose a significant threat on Pixpolar’s freedom to operate. However, you naturally never know for sure about the outcome of a patenting process and especially so in US wherein you can even apply for new independent claims whenever you want. Therefore you might still analyze from business perspective how likely it would be that WO/2014/055391 could significantly limit Pixpolar’s freedom to operate. Furthermore you are probably also carrying out an assessment evaluating the legal risks related to the matter in case you decide to pursue the patent application further.

      As I have already pointed out, in order for you to be able to demonstrate in court that you have acted in good faith you should perform a thorough internal investigation on the matter whereby you should demand written and signed questionnaires/reports on the matter from people that are or could be related to the matter and that are your employees or people with whom you are working closely together.

      In the reports it would be important to address whether anybody of afore said people were aware of any restrictions on Pixpolar (or on me, or on anything that would be prohibitive from MIG technology point of view) in Kodak’s and/or Caltech’s agreements including license agreements and if so whether they told about it to anyone inside Rambus or to anyone working closely together with Rambus. This should naturally apply to all authors as well as direct and indirect beneficiaries of WO/2014/055391, and besides to anyone who could have insider knowledge about the suspected restrictions and who would be your employee or person working together with Rambus. In addition it would be important to address in the reports of direct and indirect beneficiaries of WO/2014/055391 whether any redistribution of royalties handed out by Rambus is agreed on either verbally or written. Furthermore the direct and indirect beneficiaries of WO/2014/055391 should acknowledge in their reports that in case any redistribution of royalties handed out by Rambus takes place or is agreed on that this would be immediately reported to the highest management of Rambus.

      An important point is also that if there had been illegal restrictions on Pixpolar e.g. in Kodak’s license agreements then there would be multiple parties who would know about that and the truth might come out someday. We would naturally give from our side a full immunity to anyone directly involved in the matter and publishing it, be it either an image sensor manufacturer or any former or present employee of an image sensor manufacturer or Rambus. Anyhow,
      -if there had been such illegal restrictions;
      -if Rambus’ employees or people working closely to Rambus would have known about such restrictions and deliberately exploited such information; and
      -if Rambus would not have performed in good faith a thorough internal investigation on the matter even though being aware of a suspicion about illegal restriction;
      then it would be legally rather problematic from Rambus’ point of view.

      TO BE CONTINUED…

      Delete
    17. …CONTINUATION

      Yet another point is that in case this matter would have to be settled in court then it could have negative impact on Rambus’ image and business. I am not entirely sure whether it is related to this issue but when the matter about WO/2014/055391 came up around a month ago or so you were still advertising on your web pages that you are working together with inventors but now you have removed this slogan from your webpages. It was actually because of your inventor friendly attitude on your webpages why I wanted to meet Rambus in London at 2012. Just out of curiosity, what was the reason on Rambus side to invite me to London?

      When reading through the patent application WO/2014/055391 I could not avoid the feeling that it was triggered / initiated by and constructed around what I told at the dinner to Thomas Vogelsang about an inherent MIG pixel (as well as a DEPFET / BCDM pixel) property, namely, that the pixel can be read-out non-destructively to determine when the signal is big enough (i.e. exceeds a certain value / threshold) after which the pixel can be readout Correlated Double Sampling (CDS) wise. The especially problematic point of WO/2014/055391 is that the claims are written such that they cover inherent MIG (and DEPFET) pixel properties and thus it seems to me that WO/2014/055391 is directly targeted against the MIG technology. Another point is that I have the impression that it may not have been intended for me to find out about the patent application WO/2014/055391. Now the question is whether this is just the tip of the iceberg that we are seeing. In case you are truly aiming to corner or block our business I would like to remind you that as a counteract we could also try to do the same for you by licensing only to mobile and/or image sensor manufacturers not being in a business relationship with Rambus including their clients and suppliers.

      The reason why I take the matter so seriously is that I have invested basically everything (more than ten years by now) in the development of the MIG technology and I would hate to see it being amputated or aborted by information on its inherent properties that I have told about and/or that is publicly available. However, even if this were the case in western countries we could probably continue our business in other countries and invest all the profits into litigations in western countries.

      To conclude – based on this and on my previous posts related to WO/2014/055391 I would strongly encourage Rambus to withdraw the patent application WO/2014/055391 and to stop cornering our business.

      END

      Delete
    18. Artto

      Even if valid posting arguments such as these on a public blog is ripe for a defamation suit. I would suggest partnering with one of the big players even if it means giving most of the value away. If Rambus is really doing what they say you're doing you're not going to have the resources to come out on top.

      Delete
    19. Aurola has defamed, without doubt, the character of Guidash, Vogelsang, Endsley and Fossum, He has implied unethical or improper conduct of these guys. If this was in the US, he would risk losing every penny he has in a defamation suit, and for what real purpose? To complain to the world that no one took his invention seriously or thought it was worth developing, and to argue to the world why he invented everything first? I think the owner of this blog is also possibly at fault for allowing such a posts to be posted. It really has little value to most people in the business and is just pathetic.

      Delete
    20. Every thriving community has all kinds of people. Some of them view things differently from the most and differently from myself. Although I can use my moderating rights to block them from posting comments, I normally do not do that unless somebody complains. So far, I got no complaints on this specific thread.

      Delete
    21. I could complain, but what is the point? It is like the old Soviet-era style of disinformation.

      I remember Bill Mandl of Aerojet had proposed non-destructive readout of a CCD, during integration, coupled to a sigma delta ADC. The "dinosaur" paper had NDRO multiple sample CMOS APS concept in it. These ideas have been around a long time, and it is only the exact, specific implementation features that make them patentable. The MIG concept was interesting, worth some papers, patentable I guess, but apparently not compelling for the commercial sector. The last part is key. It is too bad, but you have to move on to the next thing and not dwell on lost dreams.

      Delete
    22. Albert TheuwissenMay 27, 2014 at 8:26 PM

      Anonymous wrote : "I think the owner of this blog is also possibly at fault for allowing such a posts to be posted."
      I absolutely disagree ! The owner of the blog puts an incredible amount of effort, and private time in this great blog. We all should be happy that someone is doing this free of charge, just to support our imaging community. If we do start blaming Vladomir for allowing posts of other people (sometimes anonymous), then this blog could quickly come to an end, and that is something that none of us would be happy about. Everyone is responsible for his/her own comments that he/she puts on the blog !

      Delete
    23. Thank you for your warm words, Albert! Indeed, my intention is to let everybody express his/her opinion with minimal restrictions.

      Eric, if you of anybody else feel offended, just ask me to delete these comments in private mail.

      On the technical side, the non-destructive readout was used in CID since the beginning of 70s, I think. While quite attractive in theory, in case of CID it was not able to overweight its other deficiencies:

      http://www.cis.rit.edu/research/CID/a_cid_is.htm
      http://www.thermo.com/eThermo/CMA/PDFs/Product/productPDF_8997.pdf

      So, in terms of prior art, the non-destructive readouts have been around for a long, long time, both in the labs and in many commercially available devices.

      Delete
  5. FYI
    I was contacted by RAMBUS before the image sensor 2012 conference in London took place since they wanted to meet me at the conference. During the course of the conference we had dinner together where I told to the RAMBUS people that one of the advantages of the MIG sensor is that the MIG pixel can be read-out non-destructively multiple times in order to see when the signal is big enough for CDS read-out. Thomas Vogelsang commented on this matter by saying that it is an interesting way to increase the dynamic range. Although the pixel structure proposed in the patent application is very much different from the MIG pixel, the ability to perform non-destructive read-out is nevertheless the same.

    ReplyDelete
    Replies
    1. You are implying that not only did you drive Kodak out of business with your idea, but also Rambus stole the non-destructive readout idea from you in March 2012? I guess they read your brain waves when they filed their patent applications in 2011. Better start wearing a foil cap in case someone else steals your ideas.

      Delete
    2. According to the patent application it claims priority to an application filed in late 2012.
      The discussion took place and the content was as I explained.
      If you have any remarks on the Kodak matter please comment to the posts wherein I provided a synthesis of it (starting from my posts on April 4, 2014 at 7:51 AM / Pixpolar Patent Case Study, Friday, March 28, 2014).

      Delete
    3. Actually Rambus has filed many patents so if you take the time to check, instead making implied accusations, you will find that some date back at least as early as 2011.
      e.g. (WO2013070932) CONDITIONAL-RESET, TEMPORALLY OVERSAMPLED IMAGE SENSOR has a 2011 priority date.

      Delete
    4. Conditional reset and non-destructive read-out without CDS ability has been around since at least US6,175,383 by Orly Yadid-Pecht. The patent application WO2013070932 is about non-destructive read-out and conditional reset – the term CDS (i.e. Correlated Double Sampling) has not been mentioned even once in the patent application. The patent application WO/2014/055391 is about conditional reset, non-destructive read-out, and correlated double sampling – the latter most term being present everywhere in the patent application.

      The floating gate read-out of WO/2014/055391 has in general very week coupling to the channel of the read-out transistor, i.e., the charge conversion gain is so poor that no one really uses it. The first claims of the patent application WO/2014/055391 do not, however, specify the pixel architecture.

      Delete
    5. In a single transistor MIG pixel you have an inherent, built-in ability to read-out the signal charge in non-destructive non-CDS manner. By performing such read-outs you have the intrinsic ability to decide when you want to perform CDS read-out of a row of pixels (or of whole pixel matrix) in single chip applications (during the CDS read-out of the single transistor MIG pixel the signal charge is reset). In case you add selection transistors to afore described pixel you can monitor the build-up of the signal charge in non-destructive non-CDS manner and selectively read-out in CDS manner any pixel that is approaching saturation.

      In a double MIG pixel you have the inherent, built-in ability to read-out the signal charge in non-destructive CDS manner. By performing such read-outs you have the intrinsic ability to decide when you want to reset a row of pixels (or the whole pixel matrix) in single chip applications. In case you add selection transistors you can reset the double MIG pixels individually.

      The reason why I have been freely speaking about MIG pixel advantages like e.g. about the ability to choose when to perform CDS read-out stems from the fact that these are inherent, built-in properties of the MIG pixel.

      In the pixel structure of FIGs 15 and 18A – 18G of WO/2014/055391 a pixel structure is presented that is capable of performing read-out in non-destructive CDS manner and to perform selective reset (with the aid of VPG and PG lines in case they are adjusted perpendicularly). Thus the operation of this pixel is similar to the double MIG pixel comprising selection transistors.

      In case, however, the performance is compared, the pixel of FIGs 15 and 18A – 18G is clearly inferior to the double MIG pixel comprising selection transistors. The MIG pixel has 100 % Fill Factor (FF) in BSI configuration, decent Charge to Current Conversion Gain (CCCG) can be obtained, there is no interface induced 1/f noise (deep buried channel operation), small read noise (decent CCCG and low 1/f noise), current mode read-out can be utilized (linear output) meaning that there is no cross-talk due to capacitive coupling, there is no interface generated dark noise minimizing also the amount of Fixed Pattern Noise (FPN), it is easy to manufacture (simple layered structure), fully depleted pixel configuration is easy to obtain enabling high Quantum Efficiency and low cross-talk (no diffusion cross-talk), and blooming protection is inherently present (vertical anti-blooming).

      The pixel of FIGs 15 and 18A – 18G bears none of these advantages; the FF is way below 100 % (both in BSI and FSI configuration), the coupling of the signal charge to the channel of the read-out transistor is very poor resulting in poor Charge Conversion Gain (CCG), there is lots of 1/f noise (deep buried channel cannot be utilized since otherwise the poor CCG would be even worse), the read noise is very large (poor CCG, large 1/f noise), current mode operation is not really feasible and thus cross-talk due to capacitive coupling cannot be removed without further reducing the CCG, interface generated dark noise is resulted in at least during read-out (there will be also an additional interface generated charge induced component in read noise) which increases also the FPN, the pinned photo-diode is difficult to design, fully depleted pixel configuration is very difficult to realize resulting in diffusion cross-talk, blooming protection is difficult to achieve without increasing the dark noise, and besides six gates are required per pixel.

      Actually, even when compared to a standard 4T shared CMOS pixel architecture the pixel of FIGs 15 and 18A – 18G is probably inferior, since in the latter one the read noise is much bigger (poor coupling of the signal charge to the channel of the read-out transistor), the dark noise and FPN are larger, the pixel size is larger because the amount of gates required per pixel is larger, and the FF is poorer.

      Afore described benefits and disadvantages should be self evident to image sensor professionals.

      Delete
    6. As already said the fact that a single MIG pixel can be read-out non-destructively in non-CDS manner and that based on this information one has the ability to choose when to perform CDS read-out is an inherent built-in property of the single MIG pixel in the same manner that it is an inherent built-in property of internal gate devices (aka DEPFET, BCMD, etc.). In addition, as already said, the fact that a double MIG pixel can be read-out non-destructively in CDS manner and that based on this information one has the ability to choose when to perform reset is an inherent built-in property of the double MIG pixel. These MIG pixel properties apply irrespective of the fact whether selective reset is utilized or not (i.e. whether individual pixels are reset when convenient or whether the whole matrix/row is reset when e.g. the brightest pixels approach saturation). Besides the ability to perform Non-Destructive CDS Read-out (NDCDSR) has been demonstrated already a long time ago in the floating gate configuration (e.g. Janesick / skipper amplifier) and by Max Planck Institute in a double DEPFET detector (instead of NDCDSR the term Repetitive Non Destructive Readout, RNDR, is used). In addition, the selective reset (Orly Yadid-Pecht) has been also around for a long time.

      Afore described inherent MIG pixel properties as well as the previously described MIG pixel advantages and the disadvantages of the pixel of FIGs 15 and 18A – 18G of WO/2014/055391 are self evident to image sensor professionals. In case the authors of the patent application WO/2014/055391 would not have been aware of the MIG technology there would be prior art publications that the application would have to face but I would not have a problem with it. However, I had been in a discussion about the MIG technology with Micheal Guidash (the second author of patent application WO/2014/055391) already in 2006 while he was still working for Kodak. Besides Rambus sent me emails before the image sensor 2012 conference in London inquiring whether they could meet me at the conference. During the conference I was invited to a dinner with Jay Endsley, Director of Engineering, Image Capture, with Dr. Thomas Vogelsang, Senior Principal Engineer (first author of patent application WO/2014/055391), and with Song Xue, Ph.D., CIS principal Architect, Imaging Division (third author of patent application WO/2014/055391), so they were also well aware of MIG technology.

      Taking also into consideration the serious drawbacks of the floating gate arrangement, the generic nature of the claims of the patent application WO/2014/055391, and the content of the discussion I had with Thomas Vogelsang during the dinner, the question arises whether the floating gate pixel of patent application WO/2014/055391 was ever intended for actual use or whether inherent MIG pixel properties were deliberately copied to an prior art floating gate arrangement in order to go after royalties from any party utilizing the MIG technology. If the latter case were true it could be regarded as a serious misuse of the patent system and it could be considered as a deceptive act in a criminal court, which could lead to punitive compensations and even to imprisonment, i.e., it could turn out to be a far more serious case than the previous trials on the matter whether the non-disclosure to JEDEC in standardization process was a deceptive act or not. Yet another point is that if the second case were true it could seriously deteriorate Rambus’ image amongst inventors, which could have negative effects from business point of view.

      In order to avoid the potential problems related to the patent application WO/2014/055391 I would strongly encourage Rambus to withdraw it.

      Delete
    7. Michael Guidash,
      it is roughly 8 years ago since we spoke last time. At that time you were still working at Kodak but it seems that you are now working for Rambus since you are one of the authors of the patent application WO/2014/055391. Thus I wonder whether you could highlight couple of below listed matters.
      1) Eric Fossum has argued that it was actually he who invented the 4T pinned photo-diode CMOS pixel configuration and not your team from Kodak – could you please comment on the matter?
      2) Around at the same time when we had the discussion with you and your team at Kodak the pinned photo-diode and shared pixel architecture was licensed to image sensor manufacturers supplying for mobile phones. Were you participating in these license negotiations and if so were there any restrictions in the license terms concerning Pixpolar?
      3) Where you already working for Rambus when the meeting between me and Rambus took place in London?
      4) While you were participating in the drafting of the patent application WO/2014/055391 were you aware of the meeting between me and Rambus and were you aware of the content of the discussions during the meeting?

      Delete
    8. Michael, one more question that I forgot to ask you – if there were restrictions concerning Pixpolar in the Kodak license agreements and if you were aware of that did you share this information to your colleagues at Rambus?

      Delete
    9. Artto, now you are being delusional about what I said in item 1. I never said this. What I think is that JPL members of the collaborative effort with Kodak were left off the patent, me in particular, since I feel I made inventive contributions to at least one of the claims. I think this is a weakness in the monetary value of this IP.

      Delete
    10. I would also have couple of questions to you Thomas and I would appreciate if you could give answers to them.
      1) Could you confirm that you contacted me by email before the conference asking whether you could meet me because you were interested to talk about Pixpolar’s technology?
      2) Could you confirm that during the conference you contacted me by email because you didn’t see my name on the lists of participants and asked whether I am in London?
      3) Could you confirm that me, you, Jay Endsley, and Song Xue had dinner together in an Italian restaurant located along the same street than the conference hotel? (Since Rambus paid the dinner you should find the restaurant bill from Rambus’ book keeper.)
      4) Do you agree that during the dinner we discussed about Pixpolar’s technology?
      5) Do you agree that I told you during the dinner that the MIG pixel can be read-out non-destructively multiple times in order to see when the signal is big enough for CDS read-out and that you commented on this matter by saying that it is an interesting way to increase the dynamic range?

      Delete
    11. Eric, you commented once that when you had been eating together with Kodak people you had said that the photogate could be replaced with a pinned photodiode. Do you remember this?
      I assume that the claim that you are referring to is the first claim?

      Delete
    12. Eric, you have previously said that: “the sale of the QIS fundamental patent to Rambus was a challenge. US 8,648,287 (priority date of 2005, issued 2014 - almost 9 years!)”. On the other hand, the earlier mentioned Rambus patent application WO2013070932 by Thomas Vogelsang et al has a priority date of 8th of November 2011 and it seems to be very closely related to your patent US8,648,287. To me it seems that the situation could have been problematic in these two cases:
      1) If the agreement between you and Rambus were signed after 8th of November 2011, if there were in the agreement between you and Rambus a royalty / revenue share model, and if Rambus had not disclosed the existence of the priority applications / application WO2013070932 to you before signing of the agreement.
      2) If the priority applications / application WO2013070932 are based on information you gave to Rambus, unless the agreement is signed before 8th of November 2011 and this is agreed on in the agreement.
      Could you please answer whether either one of the cases 1) or 2) took place?

      Delete
    13. Artto, first I am satisfied in all my dealings with Rambus to date. They have lived up to all their promises and we have worked well together. Their technical team members and partners are all good guys. Second, the details of my business are not your business.

      Delete
    14. Fair enough, but you probably understand that I don’t share your opinion on Rambus. Anyhow, part of your business may have directly affected me and that’s why I would still have couple of questions to you.

      When we met at the 2012 Image Sensors Conference you mentioned to me and to couple of others that all the image sensor manufacturers were already paying to Caltech for your CMOS patents (there had been a dispute with at least one image sensor manufacturer; e.g. about whether image sensors featuring pinned photodiode are actually CMOS or not). On the other hand, you have said on a video that Caltech’s image sensor patents provide Caltech with probably the highest revenue stream from all of its patented technologies. Based on this information it seems that you have an in depth knowledge of Caltech’s image sensor license royalties which could of course stem from the fact that in US research institutes and universities typically give a part of the royalties to the inventors. However, by acknowledging your in depth knowledge of image sensor technology and industry I would not be surprised at all in case you had been involved in Caltech licensing negotiations e.g. as an technical adviser.

      Taking into account that there has been a disagreement between you and Kodak about the pinned photodiode patent and that there has been a disagreement between Caltech and at least one image sensor manufacturer on whether Caltech’s CMOS patents apply to it or not, it would seem likely to me that the image sensor manufacturers which were willing to license (second half of 2006 or 2007) Kodak’s transistor sharing patents would also have simultaneously wanted to agree on the Caltech CMOS image sensor patents and that there could have been negotiations between Kodak and Caltech about share of royalties and license terms. Thus I would like to make the following questions to you:
      1) Were you participating in Caltech’s image sensor license negotiations?
      2) If so were there joint negotiations between Caltech and Kodak including potentially also image sensor manufacturers about the licensing of Kodak’s shared pixel architecture?
      3) If this is true were you aware whether there had been any restrictions concerning Pixpolar in Kodak’s and/or Caltech’s license agreements?

      You commented previously about patents (including transistor sharing) that Kodak sold to OVT: “I also cannot explain why OVTI paid $65M for the Kodak patents. I consulted for two companies on that portfolio and could not put the valuation higher than about $20M”. Taking into account the fundamental importance of the transistor sharing patents I would judge the worth of the transistor sharing patents to be way above 1 billion dollars unless they had been compromised with poor license agreements (e.g. if the license agreements on transistor sharing patents were aimed to pose restrictions on Pixpolar’s technology and not to maximize potential royalties). Thus I would like to ask you also these questions:
      4) Taking into account your poor evaluation on the transistor sharing patents, did you have access also to the license agreements of the transistor sharing patents when you were evaluating them?
      5) If so were there any restrictions on Pixpolar in these license agreements?

      Delete
    15. 1. I have had zero interaction with Caltech when it came to it licensing negotiations and its portfolio. I should probably say the probably the licensing stream is among the highest but I dont know the current status. Caltech has a lot going on. Caltech has been quiet about all these things.
      2. There was no disagreement between me and Kodak regarding the PPD aside from what I have said. Patents at Kodak are not handled by the technical guys I worked with and patents tend to develop a life of their own, including what is claimed in the final patent.
      3. Frankly, I can't recalling hearing about Pixpolar technology from a licensing point of view from anyone except what you have published yourself over the years.
      4. The pixel sharing patent is a good one. But, quit putting words in my mouth or taking comments out of context.
      5. I will not respond to any other posts from you starting now. Period.

      Delete
    16. Eric, one more question to you: if there were restrictions on Kodak and/or Caltech license agreements and you were aware of them then did you share this information to Rambus?

      Delete
    17. Eric, unfortunately, due to the acceptance delay I didn’t see your answers before I posted my latest question – I would anyhow like to thank you for your answers. I understand also well if you have felt uncomfortable with regarding my questions and I am sorry for that. Bear in mind, however, that I have very much at stake in here. It is about my lifework – 10 years of work with the MIG technology and before that 16 years of hard work starting from the high school as well as about the future of my family and children.

      So far Pixpolar hasn’t profited anything from the image sensor business and I would be totally fine with that if it were for legitimate reasons. The fact that after we approached image sensor manufacturers (including Kodak) in 2006 Kodak’s transistor sharing patents were licensed into mobile phone market while around at the same time the major image sensor manufacturers cut contact to us (except brief chats at conferences) tells me, however, otherwise. To me this is a clear indication of a violation against the anti-trust laws.

      It is a fact that if MIG technology would have been widely adopted Kodak’s royalty stream stemming from a number of CMOS image sensor patents (including the pinned photodiode and transistor sharing patents) would have been greatly reduced. Another fact is that in a single MIG pixel Correlated Double Sampling (CDS) readout can be done in a similar manner as in DEPFET (aka BCMD) pixels, i.e. intra-pixel charge transfer is not required. Thus the MIG technology could have been regarded as posing a threat to at least a part of your intra-pixel charge transfer related Caltech patents.

      I am convinced of Kodak’s role in the matter but so far I haven’t been fully convinced that Caltech was totally detached from the matter. In your answer to the third question you said that you can’t recall hearing about Pixpolar technology from a licensing point of view except what I have published myself over the years. Based on your answer I assume, and please correct me if I am wrong, that you are not aware (neither heard nor read) that there were any restrictions on licensing terms concerning me, my patents, Pixpolar, geographic origin, or anything that would be prohibitive from MIG technology point of view. Secondly, it is the first time this spring that the issue with the Kodak licensing terms became public.

      Afore said points are, however, not the only reason why I decided to pose the somewhat uncomfortable questions to you. At the 2012 Image Sensors Conference I managed to agree on a dinner (on another evening than the dinner with Rambus took place) with a number one technology director of a major image sensor manufacturer. I agreed to meet him in the lobby where he received a rather lengthy telephone call after which he said somewhat apologetically to me while you Eric were present that we could not have dinner together because he is actually supposed to have dinner with you Eric and continued that we could, however, discuss the matter briefly in the conference area. I was just able to very shortly present the pixel design after which he had to leave. During the discussion you were waiting nearby and I could not avoid the feeling of you acting as a watchdog for him. Anyhow, the dinner would have been for me the first possibility after 2006 to enter into a comprehensive discussion with a key representative of a major image sensor manufacturer and I bet you have had before, during, and after the conference several possibilities to discuss with him. Due to your deliberate action I missed at that time the opportunity to break the ‘embargo’ around me.

      Delete
    18. I would still like to explain more closely one detail of the brief discussion that took place at the conference area with the number one technology director of the major image sensor manufacturer. The reason why he wanted to end the discussion was due to the fact that you Eric came to stand next to our table and indicated that you guys should start going.

      Delete
  6. Every inventor who licenses can be considered to be a patent troll. Patent troll is a term invented by companies who does not want to pay for others inventions but they want others to pay for their own inventions. This was why the term patent troll was coined. There are already laws in place to deal wth patent trolls as they will have to pay for court costs if they lose in court. So, no more need to label anyone as patent trolls anymore. All universities are patent trolls also.

    ReplyDelete
    Replies
    1. I don’t think that it is fair to describe as a patent troll a party that has developed a technology from scratch and that has invested heavily in the development and marketing work despite this party would not have its own production capacity. It is generally known that ‘carrot-licensing’ (i.e. commercialization of a novel technology through licensing) is difficult, expensive, and time consuming and especially so in replacement sale type of arrangement when your customers are already using another technology.

      I would rather describe as a patent troll a party that wants to exploit from other parties’ work with minimum effort, i.e., a freeloader party investing only in prohibitive patenting to be used for limiting or blocking the freedom to operate of technologies in which other parties have already performed the heavy initial work.

      One characteristic feature of patent trolls is also that they typically aim to be unnoticed by their target until it has made the heavy investments to the manufacturing and started selling products to its customers. In an ideal situation from the patent trolls point of view the target has made already binding long-term supply contracts with its customers before the patent troll initializes the licensing negotiations with the target. In this kind of ‘stick-licensing’ scheme there is basically no other safe possibility for the target than to pay, i.e. ‘stick-licensing’ enables high profits with much lesser effort and risk than ‘carrot-licensing’.

      Unfortunately, the US patent laws offer especially benign ground for the patent trolls to operate since based on material presented in an existing patent (or patent application) one can apply for new independent patent claims whenever desired. An added bonus is also that in case the patenting process of these new split patents takes a long time one can even get extension to the validity period. Thus one cannot know based on the claims of existing patents or patent applications whether there is actually freedom to operate in US or not, but instead one has to know every detail of all patent applications in the field and to estimate how long a possible extension period could be. An additional problem in US is that many times complex patent disputes are settled in courts by juries not having adequate legal or technical background. Yet another problem is that if there were no freedom to operate in US the manufacturers are often not willing to utilize the technology even outside US (one product line is much less expensive than two).

      By filing very comprehensive and detailed US patent applications or by buying access to content of patents or patent applications of other independent parties the patent trolls are enabled to pose a Damocles sword type of situation to parties practicing their own manufacturing or carrot-licensing type operation. The downside is that the risk of investing to novel technologies is increased which is not in the best interest of the society.

      Delete
  7. Artto, I do not know you personally... but I find it quite strange that you try to solve a complex IP (and lost business) issue involving Kodak, Rambus and tens of other companies on a public forum, citing even private discussions you had with people and asking one-on-one questions. C'mon! Who in the world will answer your questions on a public forum to invite lawsuits? If you feel that all these companies conspired against you, go to a court with your lawyer. Please understand that this forum is not a place to win IP battles.

    ReplyDelete
    Replies
    1. I understand well enough that this forum is not a place where IP matters can be settled. However, I disagree totally about your opinion that I should keep my mouth shut – I’ve done that so far for 8 years and it hasn’t helped me at all. The point is that when one either goes to court or files a patent application about the content of a private conversation the content will unavoidably become public. So is it according to your moral OK
      -if the content of a private conversation becomes public through a patent application of the receiving party, or
      -if the content becomes public when the disclosing party takes the matter into court provided that it has the money to sue;
      but that it is morally not OK
      -if the disclosing party publishes the matter in a blog post?
      Consequently, it would not be morally OK to publicly complain about a private conversation if you would not have the money to sue when the receiving party’s patent application becomes public, i.e. the degree of freedom to speech would depend on the thickness of your wallet?

      Another point is also that in this forum plenty of material related to patent disputes has been published and widely discussed. Yet another point is the share value of the information – if you were an inventor planning to establish a company in the image sensor field, wouldn’t it be beneficial for you to be aware of the matter? Or wouldn’t this information be valuable e.g. for a company that is considering to license MIG technology?

      The fact that I’ve been asking one-to-one questions and that I haven’t received any answers (except from Eric) is at least to me a clear indication. Anyhow, if the persons that I’ve been asking questions about are willing to publicly ask me any questions related to the matter I am fully prepared to publicly answer them.

      On the other hand, the business is by far not lost yet since the MIG patents are still valid for 10 years. This Rambus patent application could, however, endanger at least part of the business particularly in US.

      Delete

All comments are moderated to avoid spam and personal attacks.