Lei feng's network (search for "Lei feng's network" public attention) by writer Li Yilei, UCLA doctoral student, Qualcomm (RF), intern engineer, SI said (micro signal: silicon_talks) writer.
Bo yan
Recently, a senior technical manager at Qualcomm and sighed when I chat radio frequency integrated circuit in more than 10 years of ups and downs. He said that in the United States all during Dr (the late 90) even without a complete RFIC (RFIC) textbooks, RF Circuit then how do rely on my own. He graduated to Qualcomm, witnessed the RFIC industry in Saskatchewan a decade ago, and then suddenly RFIC became obsolete, less and less willing to do this new, old concerns will be laid off. "Too fast, which is more than 10 years," he said. Marcelo Burlon iPhone Case
Mobile phones using discrete devices radio was a symbol of fashion
RFIC industry change is indeed gloomy in this more than 10 years.
Back in the 90 's of the last century, when mobile phones in China to more than 10,000 dollars a, when consumers do not think of today, we can use the money on the phone.
When RF Circuit of mobile phone was expensive discrete components set, few people expected that before long we can use inexpensive CMOS technology RFIC which integrates all devices are on the same chip. For 2G in GSM standard at that time did not even comprehended by CMOS RFIC this thing, that phase noise CMOS circuits in the GSM standard is particularly demanding. (Today, 20 years later, while CMOS IC feature sizes have narrowed in accordance with Moore's law from a 500nm to the 16nm today, but the GSM Transmitter phase noise indicators still need to be carefully optimized to meet! ）
In fact, the CMOS RFIC did face a lot of difficulties at the time, such as: how to achieve high performance on-chip inductor? How to use noise CMOS device with poor performance for a low noise amplifier? How poor mobility of CMOS devices for high frequency oscillator? However, is the so-called hero, a pioneer of CMOS RFIC made a lot of good work, solved the problem of most CMOS RFIC.
For example, Stanford's Patrick Yue (who later co-founded Atheros, is currently in the University as a professor in Hong Kong) and Ali Niknejad of UC Berkeley (currently as a professor in UC Berkeley) realize the high quality of the silicon chip on chip inductors, UCLA Ahmad Mirzaei (now Broadcom, is also the author of the Mentor) and Asad Abidi achieved wide-band CMOS on-chip oscillator, Stanford Derek Schaeffer (now Apple) and Thomas Lee tells you how to use the CMOS low noise amplifier, and announced as CMOS feature sizes shrink Amplifier noise figure will be better! CMOS RFIC revolutionized, cheap cell phone into millions of households, and promote the prosperity of semiconductor and communications industries.
Thomas Lee classic textbooks published in 2003 also witnessed the golden age of RFIC Marcelo Burlon for iPhone 5
Today, the RFIC seniors will still be talked about in the industry of Saskatchewan at that time. In the United States, early 21st century when the Internet bubble burst, while learning circuits at that time no digital work to switch Analog/RF circuits ("older popular Analog/RF circuits, coding is for youth dinner" is exactly that time appears), and numerous IC engineer while brushing aside work so like LeetCode wants a career in computer! At the time of the communications industry's most popular, a PhD in Machine Learning graduate looking for work or have to learn for Viterbi algorithm and other communication algorithm in order to be able to enter high-throughput.
Wild growth time circuit design company, extremely short, so that some agents of wives without circuit training over a period of time will the design company layout design. In China, the circuit design of a foreign company had just landed in first-tier cities, is a large-scale with high salaries to recruit talent with a design background. In the golden age of radio frequency circuits, a master degree from LNA circuit simulation in graduate school who can easily enter the monthly salary of foreign company with 10,000 + per month.
At the time, we expected to be optimistic for the future. Communication standard regularly updated (2G->3G->4G ... ... ) Will be promoted to meet the new standards demand RFIC, as CMOS feature sizes shrink, RFIC's performance will be better and better, so as to open up new markets. Therefore, the RFIC can and the CPU according to Moore's development of the industry, the market will be growing, RF engineers can increase with age experience more and more and more and more popular.
However, the reality was not as good as expected. RFIC's performance although there has been progress, but profit margins are lower. Qualcomm, Broadcom, contentment and other industry giants have combined layoffs to save costs. RFIC engineer's life is more and more difficult, many people even while brushing aside work subject to a career programmer.
Industry giants mediatek gross margin trends, gross profit margin was not go back 50%+ time
RFIC embarrassing realities
So why RFIC ideal and the reality of so much worse? In order to answer this question, we first need to analyze consumer needs.
Asked directly if consumer demand for RFIC is probably too abstract, so let us consider the consumer demand for wireless communications. Take the WiFi, for example, more than 10 years, WiFi connection speed increased from 802.11b 2Mb/s to 802.11ac 100Mb/s. However, readers of men may wish to ask yourself, if today from 100Mb/s to 2Mb/s your home WiFi, really feel a huge difference?
May wish to take a look at your home router, it is has been used for many years without change? For basic work and play (chat, watch the video), 2Mb/s wireless connection speed is sufficient. Online games for Internet speed is required, but the claim is low latency (that is, from the remote server to the local routes unobstructed) rather than a high-bandwidth, 2Mb/s under the network condition smooth enough bandwidth to run most of the online game. Playing games which is why network card when the player will tend to call Telecom called Unicom complain about spam, but no one ever scolded produced the Broadcom wireless card sucks.
For bandwidth is only required to download, but download speed is very, very small part of the overall user experience, and, after all, do not download uncomfortable are few in number. After all, consumer demand for wireless communications speeds there is no exponential growth in ten years, in this case how can expect wireless communication hardware (including RFIC) market there is exponential growth?
802.11ac standard WiFi has supported 100Mb/s, but 8Mb/s of China Telecom broadband in 2016 will still need 1580 Yuan/year, in the upper bandwidth is only used 8Mb/s 100Mb/s what does WiFi mean?
Upstream and downstream firms unable to form a joint force mining needs
So why is consumer demand for wireless communications is growing so weak? The fundamental reason is the industry's upstream and downstream enterprises did not join forces to tap consumer demand.
Consumer demand is often hidden in the underground need to dig.
A is good of example is touch screen phone, in Apple launched iPhone zhiqian no people will thought consumers on touch screen has so big of needs, and in Apple launched iPhone yihou, whole industry are occurred has one revolution, from upstream of software industry to downstream of touch screen hardware (including screen and drive IC) are fast to embrace has this new of people machine interactive mode and profit quite feng, can be described as mining consumers needs rich everyone earned of model.
Another example of upstream and downstream enterprises work together to drive consumer demand is a PC, consumers PC replacement power comes not only from the hardware enhancements to improve the user experience (more processor chip has more memory), and comes from the software requirements (operating system updates, software version updates as well as new games need more hardware).
Ten years ago I bought a PC, the main objective was to run WinXP open Office QQ online chat and watch the news. When I smooth the computer thinking "my computer does not play the game does not run two large software, should be used for 10 years no problem." It turned out I was too Naive, I can ignore Windows Update, new colorful games I can not play, but Internet-related application force me to change the computer but I can't resist. QQ every time you upgrade the computer cards, but QQ will not upgrade for a long time because the version is too old to be landed. Internet access is the same, upgrade Web effects with Flash and Html5 technology is more and more beautiful, but the more memory and CPU are also more (now a time when Gmail mailboxes to account for hundreds of megabytes of memory, which in more than 10 years ago, the mainstream memory configuration 512MB can't imagine). Finally, I had to buy a new computer. That's upstream and downstream businesses examples of how to work together to drive consumer spending.
However, in areas related to wireless communications, upstream and downstream firms do not form a cohesive force, upstream and downstream firms keep careful thinking and sometimes even one subject.
For example, 4G has just launched, when news reports said 4G 50Mb/s high speed connection can be achieved, but the news most popular comments below are not marveling at the development of science and technology, is not looking forward to the early access to 4G, but will worry once you have chosen the 4G out of flow instantly for a month!
In other words, consumer demand for high-speed wireless communications operators a large charge to kill. RFIC engineer was in the upstream industry chain downstream of big brother--telecom operators to hang out. Another example is WiFi, as long as you are using China Telecom 2Mb/s small bandwidth pipes 100Mb/s 802.11ac network card and router not much sense to me. Naturally, app developers will design app to avoid content that requires high network bandwidth, user because the app does not require a great deal of network traffic so it doesn't realize what a handset supports high-speed wireless Internet ...
Such a result is, RFIC wireless communication module inside the phone is not so important as long as you can call, can 4G-LTE on the line, you can support the LTE category 1 (10Mb/s) or category 3 (100Mb/s) don't care, sometimes too high, you connection speed supported by consumers worried about flow too fast!
Mainstream mobile phones currently on the market running not running wireless connection speeds, very few mobile phone manufacturers will be faster Internet as a selling point, RFIC was marginalized.
By contrast, the same chip, Application Processor (AP) will be lucky. Mobile phone OS, app, etc around the AP circle, consumers like most about phone index "fast" is also determined by the AP, the AP key modules in a running race, was part of the biggest selling points. Went in various mobile phone manufacturers and engage race, hyped "nuclear war", Sun feelings when RFIC relentlessly forgotten in a corner, corners are rounded in even vendors and consumers appear to be more important than RFIC!
4G price high, with 100Mb/s speed of downloading at full speed just 1 minute less than 138 Yuan package traffic can be used. This time, RFIC good network connection speed, instead of getting to consumers but will certainly "will run out of traffic" concerns.
Engineer of RFIC is red with the blood and tears of the Red Sea
Consumers for various reasons don't care performance of wireless communications case, vendor performance requirements for RFIC is good enough, "long live the 60 minutes more waste." You finally improves receiver sensitivity of 3dB, and family consumer doesn't care, the mobile phone manufacturers will not use your chip to "mobile phone signals" to attract consumers.
Then chip designers can only do is cost down. This year a variety of belief systems by people without technical backgrounds (antivirus software, video sites, teaching English) can do mobile phones, which befuddles bosses don't know what chip indices, but is still recognizable.
However, the cost down the road is extremely cruel and it's hard to look back. In order to account for market shares at prices cut profit margins of 1000 cut 800 practices ultimately hurt the enemy, want to bleed to death before the competitors out of the target market. Even more frightening is that in this war, mobile phone manufacturers and raw materials have been used to lower-priced chips, to expect price increases after opponents away is almost impossible to improve profit margins (panel maker WinTek is unable to extricate themselves after a Red Sea of excellent examples).
At present, mediatek, Spreadtrum Communications manufacturers have in playing up to my ears in this cut-throat war now Qualcomm's intention to join the competition in the low-end market, mobile RFIC Red Sea is only going to get brutal.
