Pros and Cons of COB in Fiber Optical SFP Industry
COB has become a hot word in optical communication field over the last two years. While racking our brains and exerting all our wisdom to extract value from it, we also need to stay calm and figure out which kind of COB model is more adjusted to optical communication industry. Otherwise it is easy to go astray.
CHIP ON BOARD, or COB, is actually the ‘direct chip attachment on circuit board’ technology. The core processes of COB model consist of DICE BOUND and WIRE BOUND. The former refers to SMT, and the latter means cable tying. Tracing the source, we’ve been aware that COB is not a fresh technology.
The historical evolution of semiconductor assembly technology is: IC packaging → COB → Flip Chip (COG). As we can see, the packaging size is becoming smaller and smaller wherein COB is only an intermediate product for current technology.
Before entering into optical communication field, COB was merely applied to the consumable products with low encapsulation requirements. Earlier, INTEL launched the LIGHTPEAK packaging technology aiming to unify the whole notebook industry with one connector (THOUNDBOLT). In fact, LIGHTPEAK is a typical example of VICSEL Chip on PCB Board assembly technology. However, all these attempts turned out to be failed with the appearance of USB 3.O cables. Nowadays, COB technology is shifting from consumable electronics (i.e. toys, computers, watches) to the mobile phone industries.
With regard to optical communication industry, COB is of crucial importance and should be treated seriously since we have already suffered a lot concerning the 40G/100G multichannel parallel encapsulation. Many years ago, the Japanese invented TO packaging tailored to the production of less than 10G optical modules; many years later, the world changes. While high-speed optical transceivers tend to become smaller and higher-density, it seems imperative to apply COB technology, especially for 40G/ 100G high-speed multimode optical modules. Meanwhile, there are still several problems to concern with:
1. In technical sense, whether or not current COB assembly technology conforms to the protocol of less than 10G optical modules.
As the domestic leader of multimode optical transceivers and commercial pioneer of active optical cables, Infiberone has found a dilemma in the COB R&D — It is very complicated to monitor real-time output power for COB optical modules, which is definitely a defect for network maintenance in the future. According to the analysis, there is no DDM function for 10G COB AOC provided by both FINISAR and AVAGO. Moreover, the mass usage of COB AOC has buried a hidden trouble for the network. However, using a different kind of COB assembly technology can be otherwise, and Infiberone is developing such a new generation of AOC. Infiberone always believes that DDM function should not be abandoned when introducing COB by reasons that DDM monitoring for optical transceivers is like eyes for human beings. As a result, the design concept of either network products or individual products (i.e., COB USB 3.0) should follow their own standards accordingly.
2.Whether or not the current COB usage for less than 10G optical modules is the original purpose of COB assembly technology.
Needless to say, COB model is suitable for mass production whereas the reliability and accuracy remains to be imperfect due to the sealant curing. Actually, there exist yield problems for the production of optical communication products. During the recent years, in micron level precision, the automation equipments are able to meet the requirements of VCSEL laser chip assembly around the COB model. However, if we import MPD monitoring function into COB, the situation will become more complicated since COB production process includes more steps than TO production. Traditionally, yield rate of TO production is almost 100%, while COB production mode always has a certain failure rate in field practice, even without MPD monitoring. Today, although people are constantly exploring how to mass produce the less than 10G COB optical modules, the batch production of them is far lower than that of consumable electronics. Whether it is worthwhile to spend time on COB is still a problem, after all, what the customers want to buy is cost performance rather than production pattern.
Previously, it is widely believed that COB can reduce cost effectively thanks to the scale efficiency. Nonetheless, when high yield is achieved under the traditional production pattern, both the two production modes are not dominant concerning materials and process costs. On the other hand, the limited volume and market demand of optical communications has been shared by a number of players; as a result, no single company could capture mass orders and be more meticulous to the nuance between costs. When everything goes to the marginal effect, product cost won’t make essential differences. Furthermore, there is a more important cognition to note: whether customers are interested in the nuance between costs. Potentially, what the clients constantly compare with are: brand, reliability, functionality and persistence.
3. Whether or not 40G/ 100G optical transceivers will adopt ‘COB’ appellation
From 2011, Infiberone has begun the R&D in ‘direct chip attachment’ on 40G/ 100G multichannel parallel optical modules, but the COB title was never be adopted. For Infiberone, COB is indeed a batch production appellation in a different field. While since everyone regards 40G/100G production as COB, let’s achieve a consensus. The above chart shows that FLIP CLIP (COG) technology is the final path aside from COB, so is the reality. From the analysis, companies like FINISAR depend mainly on the FLIP of CLIP (COG) technology, or a mixed model, when designing their products. Perhaps we could say that the mixed pattern of 100G optical modules is desirable. The problem is we can't find a word to summarize this knowledge. For 40G/ 100G optical modules, MPD monitoring exist the same problem. At present, none of the products on the market is equipped with real-time monitoring DDM function and this is absolutely a blind side. New products developed by Infiberone will be launched in 2015, which can completely solve the problem. However, the new collections have nothing to do with COB. As we noticed, COB has increased the complexity of the MPD monitoring. Actually, COB is a question bringing another more difficult question.
4. Whether or not COB assembly technology has become synonym of a wrong direction
Until now, we think that we should welcome COB model calmly but objectively. COB improves the precision, but it’s not necessarily fit for every company, because this kind of production pattern needs high quantity of orders, which 90% of the company cannot obtain. Without volume, there is no advantage for the automatic production. Apart from automation, the current COB is not capable to handle the production and parameters of more and more optical modules. So it’s necessary to go back to the optical module technology itself rather than competing with production patterns that as long as money can buy. For people engaged in automatic production, COB model has existed for many years, and the process is highly mature. Now, should we be aware that if we are just running back to old times and discover new weapons?
Shenzhen Infiberone Technology Co., Ltd
CEO: Mr Li Zhendong