From the time a product concept is initiated till it reaches full production sometimes there is 14 months up to two years delay. By that time things can have changed in the market place. It is possible to buy a finished design that has gone through the manufacturing pilot runs somewhere else, and fill up the existing manufacturing capacity. Such products are usually very mature in the market and have little profit margins. The high profit margins of products that hit the beginning of a market window are so huge that every manufacturer wishes to have such golden eggs. The long development time for such golden products must be reduced to 3-4 months or even less. Even that time should really be used to develop the test jigs, mechanical design, etc., and not for doing R&D. This is only possible through a filled up R&D pipeline. The R&D groups should work on various products with a common sense of direction rather than a specific product. Not all products need to be manufactured. In many cases R&D designs can be sold, or licensed.
|
There is no doubt that the need for wireless connectivity has been the main driver of activity and growth in the high-tech sector in past few decades. Although the mobile telephony networks were able to provide the required basis for wireless voice transmission, but with the explosive growth of Internet, the 14.4kbps maximum data rate of such networks has become far from satisfactory. The envision of the so-called 3G networks was aimed at providing the required bandwidth for the mobile users. However, the 2Mbps rate of such systems was very soon overshadowed by high data-rate of the wired xDSL systems. It was in this scene that the new wireless LAN standards emerged as a major alternative for broadband connectivity. As a result, the early 802-11b standard has been able to capture a 1 billion dollar market after its deployment in the year 2000. But the full potential of WLAN technology is yet to appear in the new 802-11a products, soon to appear in the market. The new 5GHz standard with a maximum data rate of 54Mbps will take advantage of more frequency spectrum , higher data rates and more advanced modulation techniques to achieve even greater performance. The 802-11a technology can provide a broadband integrated data, voice and video platform in a single wireless access network, expanding the possibilities for content access and user applications. The extensive background of our IC design group, has enabled us to get engaged in design of the advanced circuitry for the 802-11a standard since late 2000. This project that involved the complete design of the IF section of such transceiver has been one of the major IC design projects of the company in past few years. This project has provided us with the unique opportunity of acquiring expertise in state of the art technology in such advanced and complicated technology. An expertise that will enable us to do even more challenging project in future. xDSL: The technology of choice for broadband access (March 2001) In the 1980’s leading experts assumed modulation research to be mature. A well-known expert declared the 9600-baud modem as one close to the technology limit. He said after 9600 baud the error correction overhead makes the real improvement in speed marginal. In the 1980’s the universal thinking was that fiber optic technology would replace all copper cables within 2 decades rendering the copper network antique and obsolete.
In the past 8 years with the development of low cost and powerful DSP IC’S, the use of discrete multi-tone (DMT) concept, and the high cost of fiber optic networks the way has been opened towards a new look at copper. In principle, it is possible to receive data at 8 Mbits/Sec in ADSL and 52 Mbits/Sec in VDSL technology, which leaves the 128 kbits/Sec ISDN (considered a technological break-through in the 1980’s) in the dust.
XDSL systems use the most advanced communication theory techniques, such as forward error correction, interleaving, scrambling for reduced peak to average ratios, echo cancellation and equalization using training sequences, etc. The key idea however lies in splitting the transmission band into 4 kHz sub-bands. The serial data is split into these multiple bands. In the transmission process the bit sequences act as the Fourier coefficients of the Fourier series composed of multiple sub-bands. To convert the frequency domain information to a single time domain waveform all we have to do is to take the inverse FFT, hence a DSP chip. Obviously taking the FFT on the receive side will result in the original bit sequences from each sub-band.
At high frequencies the copper cable’s attenuation becomes excessive and the interference from adjacent cables also increases. To solve this problem the number of bits assigned to each sub-band is set depending on the signal to noise ratio available in each band.
In 1999 we finished a full simulator for DMT-based ADSL and G-Lite. The software allows all options allowed in the standard to be set by the user. Various cables and hybrids can be modeled and simulated before circuit implementation.
Since March 2000, our Data Communications group has been involved in other major xDSL design projects. Development of a complete Dual-band CAP/QAM VDSL receiver technology has been the result of one of these projects. Due to the necessity of developing a fully blind equalization architecture for such receiver, the design has relied on advanced signal processing algorithms to fulfill such stringent requirements at the rate of over 50 Mbps.
The issue of loop qualification of xDSL developments has been addressed by our Data Communications group in another 2 year-long project. As the xDSL system starts to be widely deployed in Northern America and around the world, the " truck-roll" cost becomes of ever-increasing importance for the Telecom companies. Our single-ended line probing technology, fully developed and subsequently implemented in custom DSP has been successfully demonstrated in Supercomm 2001. GPS a cold war by-product has been serving the commercial airlines, shipping fleets and even weekend fishermen in the past 15 years. Conceptually GPS is very straightforward. Assume that three people at three known locations call your name at precisely 9:00 am. If you hear them 1, 2 and 3 seconds after nine, assuming 300m/Sec sound propagation speed, your distance from them will be 300, 600 and 900 meters. Since their positions are known, finding your own position including the altitude is a matter of solving 3 simple equations with 3 unknowns. The key however is to synchronize the watches of the three people with your own which may not be the same. The solution is to have a 4th person calling you from another location, thereby adding a 4th equation through which the unknown time can also be calculated.
Now assume 24 satellites instead of 4 people making rounds around the earth. At each point on earth as many as 12 satellites may be visible but all you need are 4 satellites to solve your 4 equations 4 unknowns.
Traditional GPS modules have aimed at position accuracy, velocity and deviation calculations at very high speed due to their non-civilian requirements. Their power consumption is more than one Watt. For finding your way while hiking, or knowing whether your child is at the school, or tracking your company salesmen none of those complex parameters are required.
Two methods are now pursued. One is to do correlation functions in frequency domain. By multiplying the FFT of the received signal with that of the 1 second long desired satellite PN code initial acquisition can be reduced to a few seconds. The only difficulty is the need for a DSP chip. A second method is to look for PN code synchronization still in time domain and once the signal is acquired turn off the receiver to save power and have provisions for fast re-tracking when the receiver is turned on. The key point is “human scale movements” that allow the receiver to be turned off for 10 to 20 seconds or even more.
In five years you will be tracking your children or your valuables with low cost, coin size GPS modules. Watch out! Your spouse may be tracking you as well!
In 1999, we developed complete system and circuit techniques for “human scale movement” GPS. The FPGA version of the developed technology is ready to be ported to a customized ASIC solution. While the majority of existing mobile products still depend on bipolar technology for the RF circuits, the higher speed reduced size CMOS devices have attracted much attention towards RF CMOS. As the CMOS technology speeds up, the mobile systems also move to higher frequencies such as 1.8 GHz for Dect, 2.4 GHz for Blue Tooth and 5.4 GHz for 802.11 wireless LAN. Simultaneously, higher speed bipolar devices such as Si/Ge are also opening possibilities not available before with traditional Si BJT’s. Bipolar technologies at this point continue offering more high frequency gain at less noise figure. No doubt, however that CMOS with its tremendous advantages such as integrated active filters, Delta-Sigma over sampling circuits and now higher speed is moving into the BJT territory. A single chip 1 GHz synthesizer, which in 1990 would be designed in BiCMOS with 30 MHz being the rough frequency for BJT to CMOS transition, can now be designed all in CMOS.
The second trend has been towards single chip radios. Massive integration in lower frequency circuits such as ADSL, or mobile phone baseband processors has raised the expectations for single chip systems including radios. In this case, one is confronted with basic laws of physics. In a radio, mV level signals have to be amplified all the way up to about 1 volt. With this high gain any leakage from the output to input can cause regeneration and unwanted oscillations. In today’s technology the capacitors to the substrate connect almost all nodes together at high frequency making the circuit floating on a sea of unknown substrate signals. It is easy to show that at 1 GHz 0.001 fF of leakage capacitance causes major problems in any receiver circuit. Only two solutions are possible. It may be possible to use brute force on-chip shielding techniques at the expense of chip area, or move to oxide isolated type devices in SOI technologies, in effect making “discrete” integrated devices!
A third trend, which is even more difficult than the previous two is to migrate to sub 1-Volt circuits. This would have been an easy task if MOS devices with threshold voltages of 0.5±0.1 Volt would migrate to devices with 0.15±0.03 Volt threshold. Today’s technology however only offers devices with 0.3 volts of threshold still with ±0.1-volt tolerance. While this may be acceptable for digital circuits it makes analog on-chip references, high gain amplifier chains almost impossible with traditional circuit approaches.
In 1998, we designed 1-Volt RF CMOS circuits for standard pager applications. The question is what is the DECT market window? Which technologies threaten the DECT dominance, why should we choose DECT?
To answer these questions we should split the market into In the SOHO market one emerging threat is the GSM in a box concept. In this technology the same GSM handphone which works with the system on the street is fooled by a home based low power GSM base station to switch to the home base station instead of the local city network. The home base station’s power is set to just cover the home, and not go beyond. The home base station looks for the unused channels in the cell and uses them to route the call. The normal city GSM system will work fine because it operates through other channels. The user making a call has a choice between the home pseudo GSM connection or the real GSM network. This feature is available on all existing GSM handsets. It is our opinion that this technology when developed will co-exist with the home cordless and will not threaten the home cordless market in a major way for the next 3-5 years.
The other competition comes from the spread spectrum and 900 MHz analog phones. We expect the 900 MHz analog phone to probably fade. The reason is the frequency allocation issues at 900 MHz, and the fact that the low end and the high end of the cordless market will be taken away by the 46/49 MHz, and the DECT cordless phones. We predict the 46/49 MHz to continue longer than expectations, because of lower cost and the possibility of multi-handset still at low cost. The spread spectrum cordless on the other hand will have regulatory problems in many countries. In this technology two handsets can talk to each other on the air without going through the PSTN, and without even being detected by regulatory, or security agencies. This will create havoc for these regulatory agencies causing them to oppose the spread spectrum systems as much as they can.
The second area of growth for DECT is WLL. DECT will deliver wireless phone lines to homes in rural areas, industrial towns, and will be used to expedite service delivery even in major cities before the cabling is done for permanent connection. We do not believe the huge capacity expectations will be economically feasible, or will survive the onslaught of new technologies. The competition in this area will come from XDSL systems, and from other localized systems such as PACS, and PHS, or other proprietary WLL systems. The DECT link to the PSTN is expected to be with the v5.2 standard. VSATs will also be used to give service to a whole village or township via a DECT system within days. Due to the delay in the growth of WLL’s (See IEE Communications Magazine Oct 98) emerging technologies such as B-CDMA are expected to eventually replace DECT.
We believe in this area DECT will have a dominant effect for the next few years for low-density applications such as rural areas, but for high-density environments such as metropolitan areas the BW inefficiency of DECT will cause DECT to lose its dominance. The main threat for DECT will not be the wireless systems, but XDSL technologies that almost give the same advantages to the telecom companies. The HDSL or pair gain systems allow the Telecom companies to immediately extend their service in a localized area till permanent cabling is in place.
he emerging technology that will probably replace DECT as WLL in the metropolitan areas will be a B-CDMA system followed by a distributed management system based on HDSL through standard cabling. This type of system which has attracted the interest of major telecom companies in the past 1 year will provide answers for the next 10 years. Its attraction is the optimum use of lower frequencies which makes the systems cheaper. The pico size cells will provide for optimum frequency re-use, as opposed to multi-Giga-Hertz systems which are proposed for benefiting from the available vast bandwidth at higher frequencies. The final advantage of these systems will be distributed management (unlike the GSM network that has a centralized management system). Meanwhile, the reason DECT will be replaced is its bandwidth inefficiency.
Finally, we should consider the WPABX. This technology is supposed to give service to offices in which PABX’s are already installed. A company with many internal lines has an investment in their PABX, and is not likely to change the existing system with every new turn of technology. The 900 MHz analog techniques are not suitable for this application because the number of base Tx/Rx ‘s has to be equal to the number of simultaneous calls. It is possible to multiplex a base Tx/Rx between two handsets using LO switching if a triple conversion technique is used, but the method can not be extended further. Therefore as far as cost is concerned the DECT solution has its advantages. DECT will also be able to beat CDMA systems because its standardization process is far more advanced. Same is true for the competition between PHS and DECT in the international markets. A recent study (O. Momtahen, H. Hashemi, Sharif Univ. of Tech, Dec 1998) which compares DECT, PHS, and PACS in applications where a single Tx/Rx card is used shows. | | Offered traffic for all fixed users (E/MHz/Km2) | Offered traffic for mobile users
(E/MHz/Km2) | Percent of capacity reduction | | DECT | | | | | PACS | | | | | PHS | | | | That DECT ‘s offered traffic is the highest among the three and that even if all the subscribers become mobile the available traffic is degraded by only 10%. This indicates that DECT is the best choice for the PABX applications for the next 5 years. Since digital technology found its way into music CD’s, audio circuit design has never been the same. Oversampling Delta-Sigma converters have changed things in many ways.  | Oversampling reduces in-band noise. | | The complex analog circuitry which required matching of many components has been moved to the digital domain were digital complexity is compensated for by almost sure first time functionality. | | The oversampling technique makes it possible to get almost any desirable dynamic range. | | While spice with its limited mixed-mode capability was acceptable for the Nyquist rate converters, the new designs need sampled data analysis in the Z domain. In this case Spice becomes a very weak tool. Advanced mixed mode simulation tools are needed. | Even though reduced bandwidth does not allow CD quality sound, but new phones can sound better than standard phones even though the signal is passing through a noisier channel.
In 1997 we finished the design of a 3 volts GSM Codec chip. Imagine a small to medium sized city with about 30,000 to 60,000 phones. Typically, there are about 10,000 farms and factories and other potential subscribers at the periphery that are in the range of 30 kilometers away from the city limits. Giving phone service to these areas can take about 1.5 to 2 years for proper cabling, and the cabling cost may be equivalent to the cost of cabling for more than 60,000 subscribers in the city center. With a wireless local loop system the Telecom Company can give service to the 10,000 subscribers almost within weeks. A simple calculation shows that the other-wise lost revenue from the first two years of service will cover the cost of ground cabling for these peripheral areas, plus the cost of the wireless system, and even leave some good profits in the same period.
The key point is that a WLL system is a lot simpler than a mobile system. It requires no hand-off, and because the user is fixed in position it does not even require following any of the popular mobile phone standards. There is no need for the costly base-stations and the costly MSC’s or BSC’s.
In 1997 we finished the design of a WLL subscribers radio terminal. You will definitely not be leaving home without your several smart cards in a few years. The smart card concept is indeed very simple. You won’t let any body come in to your home at night before you know who they are. Let’s assume you agree with your guest on a set of two-way passwords. If I say "sun", you say "moon". If I say "joy", you say "happy". If I say "Mars", you say "Neptune". Now imagine that there were billions of such combinations that you and your guest could exchange. This way you can safely open the door to him or her anytime he or she comes. Inside a smart card this exchange is done with a secret mathematical function. After authentication is done the smart card can exchange further information on the user’s bank account, or his or her health record. What is more is that this information exchange can be encrypted. This will allow important transactions to be done over phone lines from the comfort of one’s home. So, may be in future people will say, "Don’t stay home without your smart card".
At KavoshCom we have developed a 16 bit RISC processor with instructions optimized for the private key DES encryption algorithm. We have also developed a dedicated Math co-processor that is optimized for the public key RSA algorithm for authentication. The design is simulated down to the gate level with VHDL and is at least twice as fast as other products. Ericsson, Nokia, Philips Fizz, Philips Spark, Motorola Star Tac, Sagem, AEG, Siemens S4 or S6, ... which one is the best phone? What is the technology in a mobile phone? These and many others were answered in a major project at KavoshCom in 1996.
A mobile phone consists of three areas: | The radio section where the transmit and receive architecture, the RF power amplifier controller, the synthesizer, control of intermods, layout and shielding are the major considerations. | | The DSP section where the voice and data have to go through complex and advanced algorithms for signal compression, encoding, interleaving, encryption, burst building, modulation, equalization, etc. | | The signaling protocol section which has something of about 400 Kbytes of binary code. Indeed the GSM signaling is the most elaborate in mobile phones. In our GSM project we analyzed these issues in detail and developed several key technologies for our clients. | Real money is made in factories. Reduced inventories, fast response to customer orders, good quality control with proper tracking of all products and their constituents, proper incoming material inspection, advanced SPC (Statistical Process Control) are all techniques for improved efficiency and cost competitiveness, that have become a must for every factory. Without a proper fully integrated and automated data collection, however, system managers can not react quickly as things go wrong, and in fact truly real money will be lost in the factory.
KavoshCom ’s FMS (Factory Monitoring System) which has been developed in the last four years includes proprietary protocol and signaling schemes that are extremely resilient to strong electrical impulse noise present in all factory floors, and at the same time comes with lower cost than all competitive products. The system which is composed of custom hardware, sensors, and a dedicated software can be adapted to all industries.
In 1997 a major 2-year project in a large textile factory for weaving Polypropylene bags was successfully finished. The full factory can be monitored on a single computer screen from anywhere on the earth, 24 hours. With a click of a mouse all kinds of statistics, and production data on each machine can be observed on-line or off-line at the user’s choice. The data can be tabulated or graphed for various combinations of workers, divisions, machines, spare-parts, inventories, vs. hours, shifts, days, weeks, months, and years at the user’s choice allowing all kinds of production, or price ceilings, etc. to be enforced.
|
© 1985-2010 Entiv Data Systems, Inc ™, All Rights Reserved.
|
|
