Multiplex communications – Data flow congestion prevention or control – Control of data admission to the network
Reexamination Certificate
1997-06-17
2001-02-13
Pham, Chi H. (Department: 2731)
Multiplex communications
Data flow congestion prevention or control
Control of data admission to the network
C370S280000, C370S442000, C370S463000, C370S465000, C375S222000, C379S093010
Reexamination Certificate
active
06188669
ABSTRACT:
BACKGROUND
1. Field of the Invention
The present invention relates to systems for connecting telephone subscriber lines to computer networks, and more particularly to a digital subscriber loop modem located at the central office side of a plurality of subscriber loops, which uses statistical multiplexing and flow control for communications between computing resources within the central site modem and the plurality of subscriber loops.
2. Related Art
With the advent of computer networking and personal computers, public switched telephone networks (PSTNs) are increasingly being used to connect computer systems to other computer systems, and to connect computer systems to computer networks, such as the Internet. Modems are presently used to transmit data from remote computer systems through telephone switching equipment, but these modems are presently constrained by limitations of the switching circuitry in telephone system central offices. Digital subscriber loop (DSL) modems have been developed to overcome this problem by placing a modem at the telephone central office side of a subscriber loop to receive signals from a remote modem. This allows the telephone central office switching equipment to be bypassed, thereby facilitating more rapid transmission of data between a remote modem and a modem located at the central office side of a subscriber loop.
In order to build a central site modem which serves multiple subscriber loops at a telephone central office or at a private branch exchange (PBX), it is necessary to either dedicate a modem to each subscriber loop, or to find some method of allowing multiple subscriber loops to share a common pool of modem resources. Permanently attaching a modem to each subscriber loop at the central site is wasteful because the modem can only be used by the single subscriber loop to which it is permanently connected. When a subscriber is not using the modem, the modem is idle. It is therefore desirable to provide a system in which a large number of subscriber loops share a common pool of modem equipment. This allows the cost of the modem equipment to be divided across a large number of subscriber loops. This also leads to greater modem utilization because when one subscriber loop goes idle, the modem equipment to which it is connected can be switched to another active subscriber loop.
A number of possible schemes for interconnecting a plurality of subscriber loops to a common pool of modem equipment exist, but these schemes have significant drawbacks. A simple scheme is to use a collection of relays to connect the modem equipment to the subscriber loops. In general, the cost of these relays increases as the product of the number of subscriber lines multiplied by the number of modems to which they are connected. If the number of subscribers and the modems to be connected together increases significantly, the cost of such a switching system becomes greater than the cost of simply connecting a modem to each subscriber loop.
Another possibility is to use frequency division multiplexing to connect subscriber loops to modems. In this type of a system, each subscriber loop is assigned a unique frequency slot, and each subscriber loop's baseband signal is modulated up to the frequency slot, for example, 500 MHz. When a subscriber loop needs to be connected to one of the shared modems, a frequency/agile down converter is locked onto that user's frequency slot to demodulate the signal back down to baseband. Note that baseband for DSL signals is about 1 MHz. This type of system at first appears to be low cost because inexpensive cable modem parts can be used implement it. However, the problem with such a system is that both frequencies and phase relationships of up and down converters must be precisely matched. Any disturbance of these phase relationships will destroy the DSL signals. Because of these complications it does not appear such a system would be either easy to build or inexpensive. Even if such a frequency system could be built, there are other issues of providing plain old telephone service (POTS), which would add to the cost.
Another possible method of line concentration is to digitize the analog signals from subscriber loops on the line card, and then to send these signals onto a time division multiplexed (TDM) bus. The cost per subscriber is then limited to the line card functions at about $10 per line. These line card functions include battery, over-voltage, ring, supervision, codec, hybrid and test. For normal voice telephone service, such a scheme is completely practical. When the subscriber is on hook, only the minimum line card components are idle.
This TDM line concentration method will not work well for DSL modem transmissions because of the high bandwidth required to transmit a DSL signal. In order to capture a DSL signal properly, a DSL signal must be sampled at several times the highest frequency. DSL signals can vary in bandwidth depending upon which version of DSL is considered. The T1.413 standard has frequency components up to 1 MHz. Fractional capabilities may only extend up to a maximum frequency of 100 KHz. Even at drastically reduced DSL bit rates, a sampling rate of 256 K samples per second must be used. Assuming a 16-bit sample size, this results in a data rate of 4 M bits per second. This tremendous data rate must be accommodated by a digital signal processor which processes the DSL signal. For example, the Texas Instruments C
6
digital signal processor, manufactured by the Texas Instruments Company, can accept at most 53 M bits per second. This means a single C
6
digital signal (DSP) can accommodate about 12 sessions. This rate corresponds to the DSP capacity of approximately 12 V.34 modem simultaneously active on a single processor. A session is defined as a subscriber being simply off-hook and capable of transferring data. It does not mean that any data is actually being moved. The problem with a simple TDM design is that many users want immediate access to modem resources, but these users are not in fact sending data most of the time. The equipment is mostly idle.
There are two problems with the above-described TDM system. First, 12 users is not a statistically large enough size to get good averaging results. Second, the modem equipment is being wasted just as if it were permanently attached to each subscriber. Of course, it is possible to build a TDM switching network, and assign many more DSPs to get a larger averaging pool. However, this expands the cost in a way that may exceed the advantages gained by sharing modems on a call-to-call basis. To illustrate this point further, a V.34 modem consumes about 30 million instructions per second (MIPS) of DSP power and about four megabits per second of bandwidth in data transmissions between the codec and the DSP, while the DSP is operating in DSL mode. Data is not actually being transferred most of the time. However, resources are still required. The only purpose of keeping all this equipment running is to maintain synchronization of the clock between the two communicating modems. V.34 modems use crystal oscillators which are accurate to about 200 parts per million. The phase relationship between the two modems can drift at about a 30 hertz rate. Therefore, most of the time the 4 megabits of bandwidth is being used to carry a 30 bit signal, and the 30 MIPS of DSP computing power is being used to replace a simple phase lock loop. This is very wasteful.
What is needed is a modem resource sharing scheme wherein a subscriber loop is automatically coupled to modem resources only when data is actually being transferred across the subscriber loop. For example, if on average a subscriber is actually transferring data only {fraction (1/0)}th of the time, then a single DSP should be able to support about 10 times as many subscribers if the DSP resources are only used when actually needed.
SUMMARY
The present invention provides a method and an apparatus for connecting a plurality of subscriber loops to a pool of modem resources in a way that limits the amount of modem
3Com Corporation
Duong Frank
Pham Chi H.
Wilson Sonsini Goodrich & Rosati
LandOfFree
Apparatus for statistical multiplexing and flow control of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus for statistical multiplexing and flow control of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus for statistical multiplexing and flow control of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2607906