Where will the COB model go in the field of consumer electronics?

COB mode has become a hot word in the field of optical communication in the past two years. After all the efforts, we need to calmly look at which COB mode is more suitable for the needs of optical communication, otherwise we will fall into the misunderstanding of industry investment.

COB, in short, means CHIP ON BOARD, which is the board mounting technology. The core of COB mode is DICE BOUND and WIRE BOUND, the former is the patch and the latter is the wire binding. We trace the origin of this and find that COB mode is not a new technology.

Before the introduction of optical communication, COB mode production was based on consumer products with low packaging performance requirements. Earlier, INTEL launched LIGHTPEAK packaging technology, which was to use an interface to unify the world of notebooks. LIGHTPEAK technology is a typical technical application of assembling VCSEL chips on PCB boards. However, INTEL's attempt was not successful and has now evolved into a Thunderbolt interface. However, USB3.0 cables and optical cables came out, and INTEL's idea could not be effectively realized. No matter how you look at it, COB was previously used in consumer electronics fields with massive demand, such as toys, computers and watches, and now it has begun to be used in industries such as mobile phones.

The optical communication industry has to pay attention to COB because it encounters a bottleneck in 40G/100G multi-channel parallel packaging. The Japanese invented the TO package that is very suitable for production below 10G many years ago, and many years ago, it was impossible to foresee the subsequent changes. High-speed optical modules are increasingly moving towards small size and high density. Now, COB seems to be imperative, especially for the packaging of 40G/100G high-speed multimode optical modules, but we must also list the following issues to arouse attention:

First, whether the current COB mode does not technically meet the basic protocol of optical modules below 10G

As a leader in domestic multimode optical module technology and a pioneer in the commercialization of active optical cable products, Gigalight has encountered a dilemma in studying the COB mode. It is very complicated for optical modules in the COB mode to implement real-time monitoring of optical power. Analysis shows that both 10G COB AOC provided by FINISAR and AVAGO do not have the DDM function of optical power, which is a complete defect for future network maintenance. Optical power monitoring can be achieved by using different COB modes. For example, the next-generation AOC that Yifeiyang is developing. However, the COB AOC used in batches on the market has already buried hidden dangers for the network. We believe that the introduction of COB should not abandon the monitoring function of optical power, because optical power DDM monitoring is equivalent to human eyes. The design concept for products participating in networking and for single products used (such as consumer COB USB3.0) should follow completely different standards.

Second, is the current use of COB mode for the production of optical modules with a rate below 10G not the original intention of the COB mode?

The COB mode is suitable for mass production, but the reliability and accuracy of optical communication products have yield problems due to reasons such as sealing glue curing. Around the COB mode, in recent years, automated equipment has been able to meet the requirements of VCSEL laser chip mounting at the micron level. However, if we introduce the backlight (MPD) monitoring function into COB, the problem becomes more complicated. COB production may have more steps than TO production mode. It is more difficult to solve. The yield of traditional TO production is almost 100%, while the modules produced by COB have a proportional failure rate in field applications even without backlight monitoring. Although people are constantly exploring how COB can be mass-produced on modules below 10G, the batch production of optical modules is much lower than that of consumer electronics. Whether it is worth spending time on COB is a question, because customers ultimately buy only the cost-effectiveness of the product rather than the production mode.

Our past concept is that COB can effectively save costs. However, when the traditional mode production reaches a high quantity, the material and process costs of the two production modes may not be advantageous. Perhaps we should analyze the subtle differences in costs more carefully, but due to the limited quantity demand of optical communications and the market being shared by multiple companies, no company can capture massive orders. When everything moves towards marginal effects, the product cost will not be essentially different, because the key to the COB mode is only: scale efficiency. There is another perhaps more important cognition: whether customers are interested in subtle cost differences. Perhaps customers are constantly weighing: brand, reliability, function and long-term.

Third, does 40G/100G optical module always refer to COB?

Gigalight started to do surface chip mounting in the research and development of 40G/100G multi-channel parallel optical module products in 2011, but we have never used the COB title. For us, COB is indeed a mass production title in different fields. However, since everyone says that 40G/100G production is COB, let's form a consensus. From the previous figure, we can see that in addition to the COB mode, FLIP CLIP is the final path. The reality is the same. Analysis shows that companies such as FINISAR mainly rely on FLIP CLIP (COG) technology in product design. Or a hybrid mode. We may say that the hybrid mode is desirable for 100G optical modules, but we can't find a word to summarize this understanding. For 40G/100G optical modules, backlight (MPD) monitoring also has problems. Currently, all products on the market do not have the function of real-time monitoring of optical power DDM, which is a product defect. The new product currently developed by Gigalight and to be launched in 2015 will completely solve this problem. However, this is really not related to COB. COB increases the complexity of backlight monitoring. In fact, COB uses one problem to bring another more difficult problem.

Fourth, has COB become a synonym for the wrong direction?

Today, we think that we should calmly but objectively welcome the COB model. COB equipment has improved accuracy, but it is not necessarily suitable for every company because this production model requires a high number of orders. 90% of companies cannot obtain high-volume orders under the COB model. If there is no volume, this automated production model will definitely not be able to play its advantages. The problem to be solved in the production of optical communications is not just automation. The particularity lies in the production and parameters of more optical module products, which the current COB does not have the ability to do at all. We should return to the technology of the optical module itself, rather than competing with a production model that can be purchased as long as you spend money. For those engaged in automated production, the COB model has existed for many years and the process is very mature. Now, should we be aware that we are traveling back to an old era to discover new weapons?