Abstract:
A configurable multiprocessor communications architecture which performs digital communications functions and which is configurable for different digital communications standards, such as various digital cellular standards. In the preferred embodiment, the multiprocessor architecture includes two or more digital signal processing cores (212), a microcontroller or microscheduler (222), a voice coder/decoder (codec), and a relatively low performance central processing unit (CPU) (250). Each of the above devices are coupled to a system memory (202). The general purpose CPU preferably performs user interface functions and overall communications management functions. A CPU local memory (262) and various peripheral devices (264) are coupled through a CPU local bus (260) to the CPU, and these devices are accessible to the CPU without the CPU having to access the main system bus. A dual port bus arbiter (252) is preferably coupled between the CPU and the system bus (214) and controls access to the system bus and the CPU local bus. The microscheduler operates to schedule operations and/or functions, as well as dynamically control the clock rates, of each of the DSPs and the hardware acceleration logic (240) to achieve the desired throughput while minimizing power consumption. The present invention thus provides a single architecture which has simplified configurability for different digital standards. The configurable digital communications architecture simplifies design and manufacturing costs and provides improved performance over prior designs.
Abstract:
A microprocessor includes an instruction cache and a branch target buffer to implement a branch prediction scheme. The instruction cache, which stores branch instructions, is organized into cache lines and sets to implement set associative caching with memory that stores instructions. The branch target buffer includes storage locations organized into lines such that instructions stored in a cache line of the instruction cache correspond to a line in the branch target buffer. The storage locations permit storage of a branch target address that corresponds to any one of the sets in the cache line of the instruction cache to permit storage of branch information for multiple branch instructions when a cache line of a set stores more than one branch instruction. Thus, the resources of the branch target buffer are shared among the sets of the instruction cache.
Abstract:
A field effect transistor comprising a semiconductor substrate (100) having a transistor trench extending downward from an upper surface of the semiconductor substrate. The trench extends to a trench depth below an upper surface of the semiconductor substrate. The transistor further includes a gate dielectric layer that is formed on floor of the transistor trench over a channel region (108) of the semiconductor substrate. A conductive gate structure (132) is formed above and in contact with the gate dielectric layer. A source/drain impurity distribution is formed within a source/drain region (140a, 140b) of the semiconductor substrate. The source/drain region is laterally disposed on either side of the channel region of the semiconductor substrate. In a preferred embodiment, the trench depth is between 100-500 nm (1000-5000 angstroms) and a thickness of the conductive gate structure is less than 500 nm (5000 angstroms) such that an upper surface of the conductive gate structure is level with or below an upper surface of the semiconductor substrate. The gate dielectric layer suitably comprises a thermal oxide having a thickness of approximately 2-20 nm (20-200 angstroms). In a lightly doped drain (LDD) embodiment, the source/drain impurity distribution includes a lightly doped region (146a, 146b) and a heavily doped region (138a, 138b). The lightly doped region extends laterally from the channel region of the transistor to the heavily doped region. In the preferred embodiment, a lateral dimension of the channel region of the transistor is approximately 100-300 nm.
Abstract:
An asymmetrical IGFET including a lightly and heavily doped drain regions and an ultra-heavily doped source region is disclosed. Preferably, the lightly doped drain region and ultra-heavily doped source region provide channel junctions. A method of making the IGFET includes providing a semiconductor substrate, forming a gate with first and second opposing sidewalls over the substrate, applying a first ion implantation to implant lightly doped source and drain regions into the substrate, applying a second ion implantation to convert substantially all of the lightly doped source region into a heavily doped source region without doping the lightly doped drain region, forming a drain-side spacer adjacent to the second sidewall, and applying a third ion implantation to convert the heavily doped source region into an ultra-heavily doped source region and to convert a portion of the lightly doped drain region outside the drain-side spacer into a heavily doped drain region without doping a portion of the lightly doped drain region beneath the drain-side spacer. Advantageously, the IGFET has low source-drain series resistance and reduces hot carrier effects.
Abstract:
Delay times are modified in full-duplex Ethernet network devices by calculating in each network station a delay interval based on a size of a transmitted data packet and a desired transmission rate. The network station waits the calculated delay time before transmitting another data packet, ensuring that the overall output transmission rate of the network station corresponds to the assigned desired transmission rate. The desired transmission rate is received as a media access control (MAC) control frame from a network management entity, such as a switched hub. Hence, each station operates at the desired transmission rate, minimizing the occurrence of congestion and eliminating the necessity of PAUSE frames.
Abstract:
A semiconductor process in which spacer structures are used to reduce the minimum dimension of a shallow trench isolation structure. First and second spacer support structures are formed over first and second active regions within the semiconductor substrate, respectively. Thereafter, spacer structures are formed on side walls of the spacer support structures such that a trench region of the semiconductor substrate remains exposed. If the spacer support structures are formed a lateral distance from each other approximately equal to a minimum dimension of the photolithography apparatus, then the lateral dimension of the exposed region between the spacer structures is less than the minimum resolvable feature size of the photolithography apparatus. The isolation trench can then be etched into a trench region of the semiconductor substrate between the spacer structures. Thereafter, the trench is filled with an isolation dielectric and first and second transistors are formed within the first and second active regions respectively of the semiconductor substrate.
Abstract:
A local oxidation of silicon (LOCOS) process directed to forming differential field oxide thickness on a single wafer in one patterning step and one growth step. When patterning the masking layer, at least two window widths are formed in the masking layer, exposing the underlying substrate and pad oxide. When one of the window widths is sufficiently small, oxidation of the substrate will be inhibited causing reduced growth and thus a reduced field oxide thickness in that window as compared to other larger windows formed in the same masking layer, creating differential field oxide thicknesses in one growth step.
Abstract:
A microcontroller (10) includes an integrated memory (30) and an integrated DMA controller (20). During testing, the DMA controller (20) is used to perform read/write tests of background patterns and complements thereof to the integrated memory (30). The integrated memory (30) need not contain the BIST logic typically required of integrated memories. By using the DMA controller (20) to perform BIST-type testing, the tests are not hardware-bound. Typically, the BIST logic included in integrated memories is designed and tested during the design of the microcontroller. This design and test time is replaced by a (possibly shorter) task of generating test vectors for the requisite DMA transfers. The risk of having an error in the BIST test is reduced using the present testing method. Formerly, if BIST logic was found to be in error, new manufacturing masks for the microcontroller were required to repair the problem. A large time and monetary investment was required. The present method allows for correction and variation in the testing by resimulating the tests and creating new test vectors. The new test vectors may then be applied to previously manufactured microcontrollers.
Abstract:
A microcontroller integrates a memory accessible by the cores included thereon. Additionally the microcontroller provides an indication upon an external bus that accesses to the integrated memory are occurring. The indication provides a ready identification of internal access cycles. In one embodiment, the indication is multiplexed with a control signal upon the external bus. The microcontroller further employs a show read bus transfer, which may be optionally enabled by the user. The show read bus transfer transmits upon the external bus the read data being provided from the internal memory to a core. The cycle is presented with identical functional timing to normal read cycles. Additionally, the A/C timings of the show read bus transfer are consistent with external read transfers. Therefore, external circuitry (such as an in-circuit emulator) may capture the data from the show read bus transfer using the same circuitry used to capture external read data. The show read bus transfer may be enabled by setting a configuration register bit within the microcontroller, or by asserting a signal upon a predefined pin at the conclusion of reset of the microcontroller. By providing the predefined pin for activating the show read mode of the microcontroller, the microcontroller may be placed into show read mode for debug purposes without changing the instruction code being executed thereon.
Abstract:
Collision delay intervals are modified in Ethernet network devices transmitting priority data requiring a guaranteed latency by multiplying an integer multiple number of slot times with a fractional coefficient (92). A network device having priority data for transmission uses the conventional Truncated Binary Exponential Backoff (TREB) algorithm during the first access attempt to calculate a collision delay interval from a randomly selected integer multiple of slot times (88). If the network device encounters another collision (102), the next randomly selected integer multiple of slot times is multiplied by the fractional coefficient (92). Use of the fractional coefficient during collision mediation on a half-duplex Ethernet provides a bounded access latency for real-time and multimedia applications by granting the network device a higher probability of successfully accessing the network media.