Monday, September 27, 2010

NoblePeak is No More

PR-Newswire: Joseph F. Finn, Jr., C.P.A., announced that assets of NoblePeak Vision Corp. of Wakefield, Massachusetts have been assigned to him for the benefit of NoblePeak's creditors.

NoblePeak holds numerous patents in germanium (Ge) on silicon processing. Its TriWave technology is the world's first Germanium-enhanced, CMOS image sensor technology, which delivers sensitivity and resolution across the visible, near infrared (NIR) and short wave infrared (SWIR) spectrum. The intellectual property, fixed assets and inventory will be sold by a sealed bid sale at 12:00 noon on November 5, 2010.

Just over a year and a half ago NoblePeak won SPIE Prism award for Photonics Innovation (Youtube):


  1. This is a shock. Seemed like a swell technology with lots of applications. Clearly there was something not quite right behind the curtain.

    See IISW paper:

  2. I believe US Export regulations has been the killer.

  3. They have (had?) a very nice paper coming up in IEDM.

  4. This shows again that the combination of "standard" CMOS with a foreign material, being Ge in this case, is extremely difficult. In the '90s also Silicon Vision tried to combine standard CMOS with amorphous silicon, in the late '80s Tohsiba tried the same on CCDs. Both concepts were very promising, but both failed. Hopefully others in the imaging field can learn from these three examples ....

  5. AT's statement is true and also false. The combination is difficult of course but finally the operation condition is also problem. NobelPeak uses Ge as SWIR sensing material, but intrinsic dark current is huge, so the device has be cooled under -100°C, that is very difficult.

    We can have a lot of very successful example: Bolometer thermal sensor, in general all the hydrized sensors based on CMOS ROIC.

    From my feeling, the silicon is an excellent material for VIS-NIR sensing. Other materials for this band have limited interest such as a-Si. It brings no fondamental advanatges but does bring many shortcomings.

    When you put a special material on Si, you have to have a very particular functionality that silicon can not provide, otherwise, it's hard.

    I guess that NobelPeak's concept will be picked up again. Because SWIR band has a lot of applications and silicon can do this. The monolithic approach is fundamental advantage against hydrized InGaAs-like solution where small pixel size and high resolution are extremely high cost and virtually impossible for volume production.

    Maybe they started too early !

  6. I think they're a victim of bad market timing - the military sensor market has been tough for new product introductions like the 1024x768 uncooled LWIR. Noblepeak probably didn't have the cash to make it through the tunnel.

    Aside from the impossibly tough fixed-focus/zoom lens design, this technology looked great.

  7. Maybe I had to be more specific. If you use a classical CMOS or CCD pixel, and you try to convert the classical photodiode into a storage diode connected to a "foreign" photo-converting material, you run into big trouble. Foreign materials and especially their electrical contact to an integrating storage diode is a nightmare as far as dark current is concerned. That is even the case for the electric contact on the floating diffusion in a classical 4T CMOS image sensor ...

  8. I understand better AT's opinion. Yes, of course, the contact on a photodiode or floating diffusion is always a problem. Maybe the main advantage of a 4T pixel is the absence of contact on the photodiode cathode. Do you agree with this, please ?

    thanks !

  9. Dark current on the floating diffusion is rarely a problem if the diode reaches full depletion quickly.

  10. One of the reasons the dark current is so low on a pinned photodiode device is the absence of a metal contact. But the main advantage of intra pixel charge transfer is the reduction of kTC noise. The 4T architecture contains both concepts and the reduction of kTC noise is the more important of the two.

    It is very difficult to achieve low readout noise when you don't have complete charge transfer, due to kTC noise processes. This is one of the reasons that that it is hard to make a competitive low lux YSNR10 device using materials other than silicon. Dark current is another big issue as Albert said above, although some material like a-Si has been shown to have low dark current.

    People have tried materials like:
    wide gap III-V (see IEDM)
    Black silicon (SiOnyx)
    organic semiconductors (e.g. Fuji and NHK)
    quantum dot film (Invisage)

    So far none have been big commercial successes.

    Still, with the right material, right manufacturing process and right readout, maybe one of these can make it to the big leagues.

  11. Thanks Eric for this clear explanation !
    Just I remember that NHK has made a super sensitive reporter camera with HARP structure and the obtained sensitivity is very high. Any progress on this subject ?

    thanks !

  12. Dark current is the more important problem. YSNR10 is already shot noise limited. Far and away, the advantage of the pinned diode is the dark current. Dark current non-uniformity is ugly, KTC noise is not. Cannot treat the numbers equally unless dark current non-uniformity has the same distribution, which it doesn't.

  13. As session chair, I have been in contact with the authors for the IEDM paper and am told that they plan to present the paper. Good news.

    Dan McGrath

  14. "YSNR10 is already shot noise limited"

    Thanks, I needed another good laugh. Yes, I guess once the read noise problem was solved via 4T then it is not a problem.

    So, the question was about the main purpose of 4T. Did you know you can make a pretty low dark current 3T device using a buried photodiode? Not as low as 4T but almost as low.

    For NoblePeak, dark current is quenched exponetially with cooling. Noise, on the other hand, perhaps goes like the sqrt of temperature.

  15. Fossum thinks that the problem with Noble Peak is KTC noise. That is not the problem with Noble Peak. There is little that Fossum does not know, just ask him.

  16. Eric, your comments on 4T pixel is a little contradictory.

    you stated that the metal contact on photodiode diffusion is the bigest dark current source.

    then you stated that 3T with a pinned photodiode, the dark current can be as low as that in a 4T. But in 3T pixel, there is a metal conttact on the photodiode and this contact can not be pinned.

    Could you please provide a comment on this ?

    Thanks a lot in advance !

  17. There are two threads running thru here. The main one being NoblePeak's technology and demise. The other is regarding the main advantage of 4T which has nothing to do with NoblePeak. I have no idea what happened to NoblePeak or what technical issues they had.

    But I do know you can make a 3T device with low dark current because we did that at Micron. But, without a metal contact you can gain another order of magnitude or more in dark current. So, there is dark current low enough for most applications, and then there are the incredibly low dark currents one can get with pinned surfaces without metal contacts.

    Here are example dark signal levels (I don't recall the actual results):

    3T no pinning: 1000 mV/sec
    3T pinned: 50 mV/sec
    4T pinned: 5 mV/sec or less

    Back to the question: Is the main advantage of 4T dark current reduction. I don't think so. I think it is noise reduction.

    kTC read noise 50 uV/e- 2 reads = 32 e- rms
    typical 4T read noise < 5 e- rms

    Photoelectrons for SNR10 with 32 e- read noise: 375
    Photoelectrons for SNR10 with 5 e- read noise: 120

  18. "3T no pinning = 1000mV/s", it seems a little bit high ? what value can we expect with Nwell photodiode ?

  19. To Eric : please quote your dark current numbers in nA/cm2 or in electrons/s and not in mV/s. It is a current that you are referring to !

  20. Anons: These numbers are off the top of my head from many years ago for illustrative purposes only, although I think they are still in the ball park of present day numbers.

    As stated, I quoted dark signal numbers, not dark current. Also nA/cm^2 is not a good measure because it suggests dark current scales well with area. It does not. To get electrons from dark signal, why don't you divide by conversion gain? (And I don't recall the CG from those days but any reasonable guess will do for illustrative purposes).

  21. NoblePeak went under substantially because of US ITAR rules. These prohibited them from making useful deals in the security markets since the cameras are made outside the US. They attempted to get defense business but that is too slow and too small. Ultimately, they ran out of money while waiting.

    Had the device been designed and built outside the US by a non-US company, it could be selling world-wide now. Check the structure of the InGaAs and microbolometer array supply. US manufacturers of these are also hampered by the same outdated ITAR rules.

  22. I wonder then how they will publish anything substantive at IEDM since Ge detectors falls under Cat XII restrictions? ITAR refers not only to devices but to public dissemination of technical data. In this case the authors and IEEE are subject to prosecution for violating ITAR unless they obtain an export license.

    Personally I support ITAR and am glad our low intensity warfare enemies don't have easy access to advanced night vision systems.

    It is hard to believe the domestic market is not big enough to support this kind of technology without having to rely on exports for a successful business plan.

    Strangely, I was unable to get information from the liquidators in order to put together a business plan for some already interested investors in an adjacent market space, even though the liquidators contacted me first. (*smacking hand to forehead and shaking my head*)

  23. The NoblePeak liquidators published another PR on the patents sale, quite similar to the previous one:

  24. A preview of their IEDM paper is here:

  25. Is this a censored blog? I posted a comment about US export regulations and it does not seem to appear?

  26. Rephrased comment:
    The technology indeed is not trivial, but the true killer was the export limitation. I am a medical researcher from a NATO-allied country that ordered several cameras and could not get them. Mind that we offered to warrant against abuse by strict procedures (similar to neutron detectors used in asphalt thickness gauges).
    I just can say: Shame on this US-policy witholding technology from medical applications!

  27. @ "Is this a censored blog? I posted a comment about US export regulations and it does not seem to appear?"

    No censoring. However, I use spam filter, mainly to filter out hidden ads in comments. I've just checked the spam folder and there is no your comment there. Are you sure you've submitted it properly?

  28. I came across this page looking for NoblePeak's site, wondering why they weren't exhibiting at FPED VIII, the premier DoD equipment demonstration trade show I'm currently exhibiting at. I've been developing an advanced pan-tilt gimbal for military and critical infrastructure since 2009,, having come across NoblePeak around October '09. In reaching military and critical infrastructure markets with the budgets for high-end sensors, the sensor/core manufacturers have to convince people like me to integrate them into rugged gimbals, which are then typically integrated by system integrators and large prime contractors into complete surveillance systems or soldier-worn kits. Although not new, SWIR and VIS-SWIR were just coming on the market with deliberate effort and marketing programs from large vendors, such as from Goodrich SUI, so NoblePeak was definitely riding the first wave trying to get adoption of vis-swir and night glow technology. The most common sensor payload in high-end surveillance is a color zoom day camera paired with a passive LWIR core of 320 to 640 H lines, so I was curious to see what unique capabilities NoblePeak permitted over thermal. I cannot discuss those applications publicly, but I can tell you that SWIR and VIS-SWIR has been snubbed by the military, so system designers like me have no deep-pocketed market to sell these cameras too. It's too expensive and does not meet their "detection" needs, which are better served by traditional MWIR-LWIR tuned to warm engines and the human body. On top of this, Noble's camera core was not compact, no convincing shock/vibration rating, and had an operating temperature range insufficient for commercial outdoor deployment, much less on an armored MRAP in Iraqi desert heat. The lens option was also disappointing; I can't see Stingray Optics doing very well after getting behind the SWIR trend that has now fizzled. Anyone interested in NoblePeak should also check out Goodrich SUI and Intevac Photonics' EBAPS cameras, also not economical or selling well to my knowledge. Please notify us of other manufacturers with similar promising technology at or near production. Also, I took two electrical physics and EM physics courses from a Prof. Eric Fossum at USC Engineering in Los Angeles 2005-2006, he was starting a company to develop small cell phone focus lenses and had to fly to the UK to get an award for being a genius and/or great inventor.

  29. Also, I agree on the need for ITAR, despite it's headache. It costs $2,275 a year, minimum, for a company to register for ITAR, along with export attorney fees. You need to apply for a license just to send a detailed datasheet or consult with foreign sales leads, so the NoblePeak white-paper should not be released to the public. ITAR and Commerce controls are currently being reformed into a common munitions list, so hopefully it gets clearer on what is regulated and how to comply. I attend export conferences and know of federal agents performing counter-espionage inside the US; the Chinese and N. Koreans are stealing everything they can, often contacting hi-tech manufacturers directly to try and buy components, such as Aegis radar components, often getting their hands on them due to small hi-tech firms being unaware they are restricted by export laws.


All comments are moderated to avoid spam and personal attacks.