The fourth-generation wireless service (or 4G) has different meanings for different people.
Designers can use the current chip architecture to implement some functions.
Cost constraints and power constraints will eventually be forced to spawn some architectural innovations.
The kind of 4G that application developers dream of may require a completely new architecture.
The almost mythical 4G wireless business 4G may be the origin of the new SoC (system-on-a-chip) architecture. Or, it will only promote a simple evolution of today's baseband wireless IC. It may lead to a brand new type of mobile service for consumer customers, or it may simply handle your email better. The huge engineering challenge of 4G may become a reality before 2015, or it may happen in the next few years.
To understand the possible impact of 4G on SoC design, it is necessary to delve into the meaning of people using this term, understand some of the computing challenges related to this business support, and listen to how some system architects respond to these challenges.
Many different perspectives on the impact of 4G are a source: the lack of a clear definition. Bill Krenik, Wireless Chief Technology Officer of Texas Instruments, warned: "We must start with a definition. Because there is a lot of controversy and confusion around the noun, it has almost made it meaningless."
Krenik said that many people think that 4G means a kind of storage everywhere
In the new world of wireless connection, it is truly at any time and any place, and means that this connection can support interactive, location-based and media-rich services. Imagine that you are walking on a street unfamiliar with the city, holding a mobile phone in your hand, and watching it continuously display the moving image of the street in front of you in real time and map data, building marks and points of interest, the path to the target, and The location of people in your address book. Or imagine that the same phone can also turn the city ’s streets into a multiplayer video game. You compete with other game players ’avatars, 3D monsters, and weapons, and can show realistic damage images of virtual fighting.
There are also people who usually see 4G as a more specific vocabulary, and these people must implement the system under that term. Krenik explained: "In TI, we will not try to define a dogma for 4G, but prefer to use various real technical terms: HSPA + (high-speed packet access plus), WiMax, LTE (long-term evolution). To 3G When the United States proposed a standard, and everything else was just opinions. "He went on to say that it would be better to have an organization responsible for promoting the deployment of GSM (Global System for Mobile Communications) communications and the evolution to 3G.
Other engineers hold a more quantitative view. Alan Brown, senior RF product manager at Nokia Siemens Neetworks, said that while these engineers were defining LTE, they also promoted the 3GPP (3rd Generation Partnership Project) program. They conceived 4G as "100Mbps peak traffic for mobile devices and devices such as laptop computers. 1Gbps peak. "Each point of view puts forward different expectations for the baseband SoC for 4G mobile phones.
Development Baseband SoC
Starting with the simplest expectations (that is, LTE's expectations), mobile devices will somehow achieve a peak download link data rate of at least 100 Mbps. Ken Hansen, vice president and senior researcher at Freescale Semiconductor, said: "The baseband it requires is basically functionally different from what we use today for UMTS (Global Mobile Telecommunications System)." The function block includes the function for sample rate A hardware accelerator, a CPU core that performs MAC (Media Access Control), a security engine, and a main control interface.
The sampling rate data from the radio frequency enters analog-to-digital conversion, undergoes some front-end digital processing, and enters an FFT (fast Fourier transform) engine, which separates the OFDM (orthogonal frequency division multiplexing) signal into multiple component frequency bands. The frequency domain signal is then further digitally adjusted and enters a detector, which decodes the 64 QAM (quadrature amplitude modulation) signal on each carrier and generates a symbol from each effective carrier. These symbols are compressed with enhanced decoding.
In this architecture, the difference between 3G and 4G is the difference in quantity, not the type. Peter Carson, senior director of product management for CDMA technology at Qualcomm, pointed out: "At 3G, we extract about 1 bps per hertz of bandwidth. To achieve 100Mbps of traffic, 4G baseband must do much better than it: at least on a wider frequency band. Hertz extracts 3 bps or 4 bps. "
In practical applications, this situation means that more carrier frequencies are distributed on a 20 MHz channel, as compared to the 5 MHz channel used by UMTS 900. This may also mean using multiple antennas in a MIMO (multiple input / multiple output) structure. Today, MIMO architecture is commonly used for channel equalization: find a way to combine two antenna signals to get the best reception. But there are other things in 4G: using beamforming algorithms to make each pair of base station antenna and receiver antenna a separate channel, which can double the effective bandwidth. Hansen said: "Research shows that in the case of multiple receivers, two antennas can be used to obtain a data rate of 1.75 times."
All these functions require silicon chips. A higher sampling rate and wider channels mean a larger, more power-hungry ADC, and a faster, wider FFT engine. But the biggest problem comes from the requirement to provide a 100Mbps peak traffic, which means a faster symbol rate processor, a lot of memory, and a faster processor for MAC. Hansen said: "We see 10 times the data rate entering the MAC, and the allowable delay of some transactions is only 1/10. But considering the power consumption factor, the MAC must run at a frequency much lower than the code rate. This problem is very interesting."
Qualcomm's Carson agreed. "The peak data rate is directly converted to the chip size. One thing the architect must ask himself: Does the set peak data rate and the required chip size match the average data rate actually provided by the network."
If there is sufficient inertia to the cost of the chip, the baseband architecture at this rate can be improved. Carson said that Qualcomm's current Snapdragon architecture can still perfectly handle the peak data rate of 30Mbps to 40Mbps. This speed does not meet the LTE specification, but LTE will appear later, and some people call it a late evolution, which may give the 32 nm CMOS process time to jump out of this architecture.
Non-evolutionary design
The first challenge of the evolved architecture comes from MIMO. Dr. Thuyen Le of the functional telephone business unit of Infineon Technologies AG ’s Communications Business Group explained: “MIMO is used to improve the quality of wireless links. One idea is to use it as a diversity for transmitters and receivers to prevent attenuation. Another The idea is to use attenuation for spatial multiplexing, which allows simultaneous transmission of independent data streams through multiple transmit antennas, thus increasing the user's data rate. However, this method depends on
Is the condition of the channel matrix good? Therefore, based on two ideas, I think MIMO must be used to achieve high data rates. "
In the past, air interface designers had to use a pair of receiving antennas to improve channel balance. Now they are turning to spatial diversity multiplexing to establish multiple actual channels. In this way, the number of duplicated RF parts hardware is greatly increased. Each antenna requires its own analog front end and digital front end, and RF also needs to replicate or increase traffic for more digital basebands (Figure 1). This requirement itself does not impose architectural innovation (just very similar), but there are still power consumption issues.
All 4G architectures have a limiting factor, that is, the RF part must handle 10 times the peak data rate at the current power consumption level, and the freed up power is used for application-level processing with greatly increased energy consumption. Will Strauss, president of research company Forward Concepts, estimates that the computing power of a 4G mobile phone will eventually be 100 times that of the current 3G mobile phone. Strauss believes: "Everyone's greatest hope is in the 32 nm process, but the reality is that the energy consumption of the new process has not dropped so much. You gain in dynamic power consumption, you have to pay the price of leakage power consumption. It involves looking for novel architectures and power management methods, otherwise it is necessary to have a mobile phone battery with you. "
There are other factors that drive people to consider novel architectures. This is what I said before, one is to simply express a peak data rate (such as the LTE specification), and the other is to imagine a new mobile phone usage method (such as many daydreamers who are proclaiming 4G to investors) The time is different.
Imagine the future
Liesbet Van der Perre, Scientific Director of IMEC (Inter-University Microelectronics Center) said: "There is no doubt that there is no clear definition of 4G. But I believe we should be talking about a heterogeneous network that can support higher than the existing network Mobility and data rate. Today, if it is a real mobile application, the speed will not exceed 2 Mbps, but 4G should mean 10Mbps ~ 20Mbps of actual traffic. For good video, at least 10Mbps should be stable, not peak, for example One of the disadvantages of 3G is that it cannot provide a stable data rate for good video. "
Van der Perre and other researchers describe an environment that is more dynamic than anything wireless networks can achieve today. She commented: "Today, a mobile phone silicon supplier is faced with 30 air interfaces, multiple non-contiguous channels, and many services that are running at the same time and are very different." But one supplier's mobile phone only supports it A small subset, which greatly simplifies the complexity.
In the future, in order to ensure sufficient continuous bandwidth at the same time (imagine real-time video adjusted according to the position and direction of the mobile phone) and energy efficiency (always choose the bandwidth that is just enough and the encoding length used for the current mixed task), a Mobile devices may have to make continuous agreements with multiple vendors, all of which simultaneously use many air interfaces from multiple base stations at once (Figure 2). Burst data, video streams, control information, and the return channel of the keyboard and camera may be transmitted and switched on different services in real time. For example, to keep the camera still, H.264 can be used for motion compensation, which greatly reduces the bit rate required to connect it to the game server. Therefore, this action enables the RF controller to select an air interface with a lower bit rate.
Van der Perre said that with this world view, using today's hardware processing pipeline combined with dedicated blocks is an intermediate choice. She discovered the modularity, heterogeneous clusters of similar processors that can be set, and a configurable interconnection network that enables real-time dynamic processor configuration and task mapping. In this architecture, active energy management technologies are also possible, including fast voltage / frequency regulation, moderately fine-grained power gating of idle components, and agile movement of algorithms between software and hardware. Indeed, this solution may be the only way to meet the energy efficiency needs of true 4G terminals, despite the 32nm CMOS process.
All these projects are gradually taking shape in various research projects of IMEC, perhaps this explains Van der Perre's worldview. But this is far from an isolated view, at least in private. Various companies have publicly stated that they are focusing on pipeline-based hardware architecture, but only one well-known industry source said that some major silicon suppliers have deep involvement and large-scale investment research groups and are exploring large-scale multi-core architectures for 4G platforms. .
Most large-scale multi-core architectures have an important challenge, but large-scale multi-core architectures are not a problem here: most of the load in high-bitrate baseband processing is parallel processing that the industry finds difficult. It is difficult to assign tasks by the method of dividing data alone. But the key to successful design is system control, dynamic load balancing, and (perhaps the most important) energy management work, which are novel and complex. From this point of view, 4G is actually not revolutionary, but it is steadily advancing. In this process, designers have to forge new real-time embedded processing into shape.
I-Pulse Nozzle , in stock, original and new, fast delivery.The price is acceptable, service is guaranteed.
M1/M2:M001/M002/M003/M004/M005/M006/M013/M018
M018/N018/P018/K018
Characteristic:
1. Material: steel for body, durable for use;
2. Holder is made of china or tungsten steel;
M1/M4 series:
M001 LGO-M07701-00X
M002 LGO-M07703-00X
M003 LGO-M07705-00X
M004 LGO-M07707-00X
M005 LGO-M07709-00X
M006 LGO-M0770B-00X
M012 LGO-M0770E-00X
M013 LGO-M0770G-00X
M017 LGO-M0770H-00X
M018 LGO-M0770K-00X
M019 LGO-M0770M-00X
M020 LGO-M0770P-00X
M032 LGO-M0771K-00X
M2(N series):
N001 LC1-M7701-00
N002 LC1-M7703-00
N003 LC1-M7705-00
N004 LC1-M7707-00
N005 LC1-M7709-00
N006 LC1-M770B-00
N012 LC1-M770D-00
N013 LC1-M770F-00
N01 LC1-M770H-00
N018 LC1-M770K-00
N019 LC1-M770M-00
N020 LC1-M770P-00
N021 LC1-M770S-00
N022 LC1-M770PU-00
N032 LC1-M7714-00
Nozzles For I-pulse machine
Smt I-Pulse Nozzle
Fuji Nozzle
Smt Fuji Nozzle
SMT Nozzle
Smt Parts Nozzle
I-Pulse Nozzle
High Pressure I-Pulse Nozzle,I-Pulse Nozzle,Nozzles For I-Pulsemachine,Smt I-Pulse Nozzle
Shenzhen Srisung Technology Co.,Limited , https://www.sr-smt.com