News and discussions about image sensors
just read the chinese version, there is one sentence : suddenly I found that our cost is only the half of our competitors ...haha !
You hear a lot about Chinese IP pirates. This company spent next to nothing on image sensor R&D. One has to ask where they got all technical know-how to make image sensors? If you don't have to pay for R&D, then it is a lot easier to price your product at the rock bottom price.Secondly, their fab is SMIC, a known IP pirate. SMIC recently had to pay TSMC hundreds of millions of dollars for stealing TSMC fabrication IP.So, now you have an image sensor company with no R&D overhead, using a fab with no R&D overhead. Is it any wonder that they can be so cost competitive?Thirdly, shame on U.S. VC firms for funding pirates that live off the IP developed by other companies thru years of hard work. In fact, it is the ready availability of capital that makes such IP piracy possible.
good news for consumers of commodity chips.All these players today differenciate by providing lower cost CMOS image sensors: they are typically pricing 20% lower in cost compared to foreign competitor's Omnivision and Aptina Imaging.
Monkey enjoys the feeling of king when tiger is on vacation.
I would not say that the sole reason for SMIC/Galaxycore success is stealing the secrets of others. SMIC has worked on pixel and process development for a long time, may be 5 years or so. Their pixel performance gradually improved over the time until it became competitive with others. Like most others, they started from 3T and moved to 4T after mastering 3T process.Many years ago SMIC established a JV with Toppan to get access to microlens and color filter technology, so this is officially not stolen.Talking about Galaxycore part of R&D, let's say it, if one designs a low-end cheap sensor, no high speed, no HDR, there not a big effort. Normally VGA sensors include ISP - low end ISPs IP can be bought for $200K. Even the best speced ADC can be acquired for as low as $100K as silicon-proven IP. Better yet, many foundries offer ADCs as a free IP blocks for their customers, together with PLLs, I/Os, digital libraries, ESDs, etc. Tapeout in 0.15-0.18um process is quite cheap too, especially so if one goes to third party mask shops. So, all in all, $2M initial investment Galaxycore got is quite a lot, assuming their salaries are not high and CAD tools are Tanner or Dolphin rather than Cadence and Synopsys.What really matters for success is a very efficient production organization, and it seems that Galaxycore excels in that.
Image sensor world,Where can I buy a $200K ISP IPs and $100K ADCs?Please name a few.
I would rather say Galaxycore has very efficient sales channels.
@ Where can I buy a $200K ISP IPs and $100K ADCs?The ISP IP prices are negotiable. Obviously, the starting price is higher, but if you talk about bare bone entry level version, $200K is a very real number. As for the ADC, Cosmic Circuits, for instance, has a plenty of them for less than $100K, some for much less.
ISW, I think you are confused between the effort required to COPY someone else's design and process, vs. the developing your own IP. Copied IP is pirated IP. Just like when you copy a DVD.I'd like to know what Toppan thinks of the JV outcome. I always find it odd that some Chinese nationals brag about working for a US company and then going back to China to start their own company that is in the exact same field, using what they learned and low cost of money to build a knock-off business. These seems to be considered a sign of success in Chinese culture. I think it is more like industrial espionage and piracy.Defending pirates is, well, indefensible.By the way, there are some very smart upstanding Chinese researchers and entrepreneurs. I don't want anyone to think this about all Chinese high tech companies. There are just a few bad ones among many. And as mentioned above, even the pirates are supported by US VCs because they will make money. This is also immoral, in my opinion. I just hope Communist China (ha ha ha) wakes up to the importance of IP protection. This will only happen when their internal R&D efforts are pirated away by some cheaper manufacturing country.
I respectfully disagree with the production efficiency explanation, as the translated article also states the following:"Fortunately, with SMIC's platforms, GalaxyCore succeeded after testing more than 40 versions while other companies had no progress. Then SMIC decided to focus their R&D funds on GalaxyCore. While other image sensor design companies complained about the huge costs, GalaxyCore survived incredibly. “If it was our own money, the investment would have been already gone.” Zhao added."That's a pretty clear statement direct from the company head that points to low-to-no R&D costs as a principal reason for success.Also, the article states:"Before coming back to China with his patents in the field of high-end image sensors, Zhao had worked in Singapore’s Chartered Semiconductor, in ESS Technology in the United States and in UTStarcom. At the beginning, he chose to do high-end image sensors design, but because of problems in R&D, he turned to work on base band research and development. After some ups and downs, he finally started working in the field of low-end sensors."Five issued patents with Lixin Zhao as the inventor that I found on the US PTO web site were:- 6,586,789 "Pixel Image Sensor" filed by Lixin Zhao (Fremont, CA) on 10/2/2002 and issued on 7/1/2003, with no assignee.- 6,852,565 "CMOS image sensor with substrate noise barrier" filed by Lixin Zhao (Fremont, CA) on 7/10/2003 and issued 2/8/2005 with an assignee Galaxcore Inc. (Cayman Islands, KY).- 6,897,797 "Digital to analog converter with integral intersymbol interference cancellation" filed by Lixin Zhao (Fremont, CA) and Zhongxuan Zhang (Fremont, CA) on 9/29/2003 and issued 5/24/2005 with assignee UTStarcom, Inc. (Alameda, CA).- 6,982,183 "Method and system for manufacturing a pixel image sensor" citing US 6,586,789 as a related document, filed by Lixin Zhao (Sunnyvale, CA) on 5/12/2003 and issued 1/3/2006 with assignee Galaxcore Inc. (Cayman Islands, KY).- 7,551,694 "Limiter based analog demodulator" filed by Zhongxuan Zhang (Fremont, CA), Lixin Zhao (Shanghai, CN) and Steve Xuefeng Jiang (Fremont, CA) on 1/19/2006 and issued on 6/23/2009 with an assignee Marvell World Trade Ltd. (St. Michael, BB).US 6,897,797 seems the most normal to me - two inventors filing in September 2003 and assigning to UTStarcom as their employer.The other four patents don't feel right. They have filing dates well before, just before, and well after US 6,897,797, but UTStarcom doesn't seem to come into play. Instead there's assignment to offshore companies in financial secrecy havens.
ISW,I would be grateful if you name some ISP IP vendors with $200K price range also.Maybe, many of readers will be also interested.I'm quite serious and if you are not comfortable in talking about it openly, but you know some real companies, I'd like to contact you privately.
@ I would be grateful if you name some ISP IP vendors with $200K price range also.Sorry, I'm binded by NDA and can't tell you much. Generally speaking, if some vendor already sold its ISP core to few customers and recovered its development costs, it would be an easy and cheap task to configure its RTL code and synthesize it for your particular process.CDM, thank you for the interesting patent analysis. Not having personal interest in defending Lixin Zhao, I can think of different possibilities when the patents are perfectly legal. For example, he might be working as a consultant, contractor of just part time employee for Marvell and UTStarcom. Or he might be unemployed between the jobs when filing his image sensor patents.Registering the company at Cayman Islands is not that bad thing. Some very respectable companies like Aptina or GPS maker Garmin are incorporated there.Eric, I'm not sure that if one makes a cheap entry level sensor, he needs reinvent the wheel with respect to the sensor internals. All the things are quite straightforward and well explained in numerous papers. There are few generic approaches, all of them working, if implemented correctly. I'm not a patent expert, but I think at least a good part of these circuits is not defended by patents.Once Galaxycore got a good performing pixel, all the rest can be simple simple and cheap and not necessarily a theft.
ISW - It is best not to comment too much about patents but CMOS image sensors with intrapixel charge transfer, 4T for example, are covered under US Patents that I authored. Image sensors with 3T pixels may be covered by US and International patents authored by others. I believe every large player approached on this subject has acknowledged the validity of the patents and has legally licensed the IP when required by either import or sale location."quite straightforward and well explained in numerous papers" does not mean the IP is up for grabs, whether one is talking about pixels, readout circuits, or more common circuits like ADCs, and pad drivers, or more complex digital functions like ISP. Registering companies in the Caymans is a way companies only pay US taxes on the US part of their business, etc. But it is also a sort of tax dodge depending on how P&L is assigned to various locations. I have had people (lawyers, mostly) try to sell me on that concept but it was too sketchy to get comfortable with.
if people believe their patents have been infringed by GalaxyCore or whoever, they should and can go ahead to sue those "IP pirates". The "Communist China" has put the guy who distributed pirated Microsoft software into jail. People can expect the same for image sensor tech, provided they stand on solid ground."to start their own company that is in the exact same field, using what they learned and low cost of money to build a knock-off business", we should ask Gordon Moore if he felt guilty when he left Fairchild and started intel.
Mr. Fossum may have had some problems with some Chinese companies to adopt a so radical position !I'm not sure that millions of VC in Photobit have gone all to the technological development ! I saw a lot of High Tech companies speet a lot of money in unknown things, I can wonder !You can find also a lot of US companies who do the same thing, I think that the number is even higher !As for 3-T pixel structure, I think that the first suggestion can be traced back to 60's. I'm not sure that actual 4-T device is covered by the initial JPL/Photobit patents.If we follow Mr. Fossum's logic, all the sensor companies should be sued and considered as pirates regarding Photobit ! How about Hitachi's work on MOS image sensor ? It's a real pioneer work in this field. A lot to say ...
Dear Anon1: 1. Communist China put one person in jail? This is funny. They also executed a retarded man. Not so funny. 2. Regarding Intel: I have this feeling they developed a lot of their own IP. In any case, being bound by US law means they could not pirate Fairchild IP. Try to understand the difference when you give silly examples.Dear Anon2: 1. I have never had a direct problem with a communist Chinese company. I recall hearing about this campaign in China from a respected professor: "Letting a hundred flowers blossom and a hundred schools of thought contend is the policy for promoting progress in the arts and the sciences and a flourishing socialist culture in our land." Do you know it? Yeah, I have a problem with the communists and I fear for my country. I admit it. Probably the only thing that scares me more is China becoming a capitalist power.2. Photobit did not have any VC investors.3. 3T dates to Peter Noble, and contemporaries. Hitachi also did some great work. You can learn about these in the review paper I wrote about 15 years ago. By the way, Hitachi-san was an investor and partner at Photobit.4. The JPL patents are owned by Caltech. Photobit was the first company to license these patents from Caltech. In fact Caltech did sue a number of companies and the students and faculty at Caltech now benefit from that. 5. It is hard for me to respect anyone who comments anonymously about me personally. But, I have tried to respond to each point nevertheless.6. I try to live my professional life to the highest moral standards I can. I think this something we should all strive for don't you?
US government used retard men (not man) for irradiation test, does Mr. Fossum think funny too ?
Please try to refrain from purely political discussions. I'm sure there is a lot of places on the net where these discussions are much more appropriate.
Dear Prof. Fossum, I respect all your great achievement in CMOS Image sensor field, but I disagree with your comment on Galaxycore. "Stealing" technology is not only done by China, there is a long history from Taiwan and Korea, so not just "communist" thing. Also please think it in positive way, with Chinese company occupying at low-end market, the presure will push American company develop the advanced technology which is not easy to be copied, and this will help the whole imaging industry growing in a healthy way, just like in the nature world the food chain will keep the nature in balance.
These comments are coming from a patriot working for the Koreans...whatever pays the bills I guess.
Interesting comment about the food chain. Post WWII Japan started this way also, with lots of US financial support. Japan entered at the lowest end of the market and gradually worked its way to the very top by the 1980's. Korea followed the same trajectory starting in the mid 60's maybe and is now reaching the market with well valued products. In my lifetime, Korea has gone from a rice economy to very high tech. It is rather impressive. But, a major difference between Japan, Korea, and China is the the realm of IP protection, which is where this story started after ISW made some comments on GalaxyCore.
I think that Mr. Fossum's thinking reflects some mentality of western people. From my point of view, A low cost VGA sensor is no longer on the top of CIS technology, it's a commodity product.But the fact that a chinese company can get honorable market share in this market gives some strange feeling. Because these people think that chinese guys can not design it themself, if they make out someting, it's surely stealed from US companies.The comments of Mr; Fossum are quite insulting regards chinese sensor designers. Either you have very strong proves and evidences that they have stolen other company's secret, you can sue them. I agree more than 100% with you !But you can not advance such opinion on a public blog, in USA, you can be sued by justice for defamation.A general IP piracy situation can not be used in a particular case. Each particular case is different. Even 99% chinese companies pirate others IP, but you can not say that this company does this, maybe this company is just in the very 1%.This kind of discussion should be screened out from this forum. We have to be at first technical and scientific guys. The first quality of a scientist is the objectivity. You can advance a conclusion only with facts which are verified and reproducable.
Yep, the Korean pays, so they don't steal others secrets. Maybe Galaxycore should also pay a little bit ? All the world says that RMB is under-valued, so it could be a good deal ! haha !
I think Eric has a right to express his personal opinion, in this blog and elsewhere. One can agree or disagree with him. But if instead of intelligent counterargument you go to personal attacks, it does not help you to convey your point.
Once we started these IP discussions, let me ask a broader question: does the patent system make good or bad for image sensor industry as a whole, across the borders and countries? This sounds too provocative, but let me explain:Before the US patent system was established in 18th century, there was a world of trade secrets where shop owners did not tell anybody how they made their products. When the owners died, quite often their unique knowledge was gone. Obviously, this slowed down the progress. The patent system meant to be a tool to prevent this loss. The owner was offered to disclose his secrets to the general public in exchange for the government protection of his rights for 20 years or so. At the time this gave a huge boost to the US technical progress.But does this logic work now, specifically in image sensors?First, many patent lawyers learned to write patents in such a way, that it is extremely hard to understand. One can say that such a patent hardly discloses anything, and, at the same time, gives a perfect legal protection.Second, widely spread and completely legal reverse engineering leaves little place for trade secrets anyway. So, no knowledge is lost - this was one of the main justifications of the patent system.I can go on and on, but let us do a mental experiment: one morning we get up and discover there is no patent system. Would it be a better world? I mean better for image sensor industry, as I have no idea what it means for medical, software and other areas. Also, I'm talking about the global image sensor progress, not separating it to the US, Chinese or other companies.
Dear Image Sensor World,To say nominatively, a particular company has stolen the knowledge of others is not the expression of an opinion. It's the worst personal attack and legaly this is called defamation. It can be sued by US law.
From Wikipedia:-------------The history of patents and patent laws is generally considered to have started in Italy with a Venetian Statute of 1474 which was issued by the Republic of Venice. They issued a decree by which new and inventive devices, once they had been put into practice, had to be communicated to the Republic in order to obtain legal protection against potential infringers. The period of protection was 10 years.Patents, however, existed before the law. In England grants in the form of “letters patent” were issued by the sovereign to inventors who petitioned and were approved: a grant of 1331 to John Kempe and his Company is the earliest authenticated instance of a royal grant made with the avowed purpose of instructing the English in a new industry. The first Italian patent was actually awarded by the Republic of Florence in 1421, and there is evidence suggesting that something like patents was used among some ancient Greek cities. In 500 BC, in the Greek city of Sybaris (located in what is now southern Italy), "encouragement was held out to all who should discover any new refinement in luxury, the profits arising from which were secured to the inventor by patent for the space of a year." ----------So, the concept of IP protection has been around a long time in Western culture. Unauthorized use is considered a form of theft. These days patents foster creativity and innovation in small start-up companies. Otherwise large companies could immediately put small companies out of business. In a similar way, international patents should protect trade with countries with low labor costs and artifically devalued currency from theft of IP.If you want to do a thought experiment, consider a world where theft of real goods and property is not considered a crime. This eventually fosters a war-lord feudal society where only the strong and brutal are favored. The analogy to business and high tech is obvious, isn't it?
From http://www.techdirt.com/articles/20090530/1620345062.shtml------------China was a latecomer to intellectual property. Its first patent law came into effect in 1985, followed by a copyright law in 1990 (Graff 2007). However, since then, the pace of progress has been rapid; it has now joined all major international IP treaties (Maskus 2005). Its patenting activity is increasing rapidly, too, with domestic firms nearly doubling the number of patents they received in the past four years (“Chinese firms…”). China’s Patent Office now leads the world, reviewing 800,000 applications in 2008, and in 2009, domestic firms are poised to receive more patents than foreigners for the first time ever (“Battle of Ideas”). Chinese firms are also receiving more patents abroad: in 1999 they only won 90 patents in America, but by last year they had increased that number to 1,225, demonstrating a desire to use their inventions globally (“Battle of Ideas”). Chinese intellectual property, however, is still frequently critiqued. Enforcement is notoriously weak with the United States citing “rampant counterfeiting and piracy problems.” Strikingly, according to the USTR, China was the origin for 67% of seizures of counterfeit goods at the American border in 2008. In response to these and other concerns, China has recently updated its patent laws, increasing statutory damages and expanding the investigative power of the patent office (Lim 2009). ------------
No one doutes the usefulness of actual patent and copy right system. But the situation is becoming much more complicated and the science and technology are much more complexe. A single man can hardly invent all things. All the technical achievement is made by progressive and collective efforts. But the problem is that the patent is normally granted to the one person or a group of people. that is one problem.In (C)MOS imaging device, it was a long and collective technical achievement from different angles and initial applications. A lot of work have been published and a lot of concepts have been advanced by many engineers, scientifts etc.All these partial results give a lot of possible combinations, some are patented and some are not. A same device can be also thought from different points of view. For example, a pinned photodiode, it can be invented because you want to screen the surface states to reduce the dark current, it can be invented because you want to have both junctions one close to surface and the other deeper, it can be invented simply because you want to deplete the junction, etc ...As stated by the Image Sensor World, many write a lot of meaningless patent applications, simply to claim that they have a lot of patents !one webcam backend chip company in China claims that they have more than 500 (many years ago) patents !!!! So something like 100 per year, how this is possible ? And what is the value of these patents ?So today, for example, one designs a 3T APS with Bayer's colour filter can use all the public knowledges avaialble with no patent issues.For example, the first VVL sensor (PPS) has the same concept structure as that was published in IEEE by japanese engineers, LAMI (Line Amplified MOS Imager).Taking into account of the diversity of reseach activities and number of the invloved scientists/engineers, it's not that simple as a simple ambiguous patent can resolve. All of us should remain precautionnal.
@ "To say nominatively, a particular company has stolen the knowledge of others is not the expression of an opinion. It's the worst personal attack and legaly this is called defamation. It can be sued by US law.Actually, Eric put a disclaimer:"I don't want anyone to think this about all Chinese high tech companies. There are just a few bad ones among many."This can not be called defamation. Again, I view this as Eric's personal opinion. Everyone can agree or disagree. I, for one, do not agree with it.
@ "If you want to do a thought experiment, consider a world where theft of real goods and property is not considered a crime. This eventually fosters a war-lord feudal society where only the strong and brutal are favored. The analogy to business and high tech is obvious, isn't it?"IP and real goods are different things. If someone deprives you of your real goods, you do not have them anymore. IP remains in your head, even if somebody copied it.For example, why nobody in China copies hugely successful Sony 10MP BSI chip? It became a de-facto standard of low-light sensitivity, everybody else is measured against this standard. So, if some countries have weak IP law, why not copy Seony sensor? - The answer is that it's too complex to copy and copying effort is quite comparable to developing it by themselves.Talking about basic low-end sensor that Galaxycore is designing, they are simple and possible to copy.So, in a world with no IP protection, companies would run ahead a develop new things faster. Whoever wants to copy would stay far behind, in low-margin commodity world - this is not to say this is a bad business model.Talking about a small company protecting itself by patents, this is not cheap. Getting granted patent worldwide today costs as much as $400K per patent and takes many years. If this money and time is invested into R&D instead, the small company could develop better, more competitive products which are hard to copy, just like Sony sensor.
ISW: "why not copy Seony sensor? - The answer is that it's too complex to copy and copying effort is quite comparable to developing it by themselves"Well, I have to disagree about that. There are so many design splits and process splits that could be considered when you are first starting making a new structure and device. Having a successful product in your hands that you can reverse engineer makes it much simpler. Furthermore, clever ideas - patentable ideas - to make the device work or work better, are apparent in the final product. So, while copying is difficult, I would not say it is comparable at all to starting from nothing.Second, you overestimate the cost of patent protection. A single patent in the US PTO costs about $15K. If you are a US company, this protects you against the import of infringing goods, and against other US companies. Nevertheless, I agree worldwide protection is too expensive for start ups, not to mention worldwide defense of your patent(s).
"Furthermore, clever ideas - patentable ideas - to make the device work or work better, are apparent in the final product."I agree and disagree with this statement. When one can recognize a clever idea inside a complexe design, this guy may not copy other's idea.
Yes, you are right. I was thinking about layout ideas like shared RO, etc. Complex timing and operating voltage levels are more difficult to discern. Some process recipes and new material processing are also difficult to figure out - like reproducing Coca Cola.
@ "Some process recipes and new material processing are also difficult to figure out - like reproducing Coca Cola."Very true. Recall the story how TSMC tried to transfer its CIS process to Powerchip. Powerchip paid for it and TSMC sent its engineers to help. Nevertheless, the sensor performance was worse than TSMC's. It's very hard to copy process, even legally.It took SMIC and Galaxycore 5 years or so to develop competitive 0.15um/0.18um CIS process and pixel, even though they had access to all the reverse engineering (potentially). This is longer than what it took to the original developers who developed it for the first time. I'd guess it would take them at least the same time to design Sony-level BSI process.So, should Sony worry? - Not, in my opinion. Then who should worry? - Probably the companies who use a simple, easy to copy technology and not progressing fast. So, the patent system in its current form protects these slow companies and does not stimulate them to run faster. Do we need such a patent system, I mean we as an industry?
Well, all the clearly visible inventions are easily protected by patents. For me, CIS patent infringes are the easiest to demonstrate. Just thinking about the processing method implemented in a sea of gates, etc .... Like a joke by David Marr, how can you discover the FFT algorithm run on a PC by using just oscilloscope ?
(1) I have a feeling that Mr. Eric Fossum is jealous about GalaxyCore's success.(2) Mr. Eric Fossum tends to believe Chinese are not smart enough thus if a Chinese company succeeds it must be "stealing".(3) I expect Mr. Eric Fossum to distinguish "learning", "IP", and "patent". An engineer's learning goes with the engineer. I believe this is why Mr. Eric Fossum is being paid high by a Korean company.(4) IP and Patents are protected in China. I want to encourage Mr. Eric Fossum to defend his rights through legal procedures. However, please avoid spreading rumors (statements spoken as if are facts but without proof). Please do not mix "rumors" and "freedom of speech".
It's very easier to pay a US theft to steal so-called US technology, haha !
Dear Image Sensor World,I would like to verify one thing with you. I think that OV started with PowerChip and then switched to TSMC. So PSC still produces some old designs of OV on 0.25um line. Are you sure that TSMC/PSC have a CIS transfer deal ??Thanks !
Some of these cowardly anonymous personalized comments are pretty funny.(1) Nope. Don't be silly.(2) You should learn to read English more carefully, and then very carefully and slowly reread all my comments. And just in case you still missed it, I have high regard for almost all Communist Chinese-born technologists I have met or had the good fortune to work with. I have low regard for the Communist China government policies and IP-pirate companies.(3) Just because you have learned something you don't automatically have a right to make a product by copying what you learned. Is this such a complex concept to understand?(4) I don't own any patents. I am just an inventor on many patents owned by Caltech, Micron and others. But thank you for watching out for my legal rights.
@ "Are you sure that TSMC/PSC have a CIS transfer deal ??"Please see here:http://image-sensors-world.blogspot.com/2006/06/tsmc-is-having-capacity-problems.htmland here:http://image-sensors-world.blogspot.com/2007/11/omnivision-conference-call-minutes.htmlIn the second reference Omnivision says in its Q2'2008 earnings conference call:"Powerchip is a development program which was originally intended and still is planned to deliver product in the second half of 2008. It's too early to tell how this will work."
All comments are moderated to avoid spam.