Abstract: An integrated circuit (IC) includes multiple embedded test circuits that all include a ring oscillator coupled to a test load. The test load either is a direct short in the ring oscillator or else is a interconnect load that is representative of one of the interconnect layers in the IC. A model equation is defined for each embedded test circuit, with each model equation specifying the output delay of its associated embedded test circuit as a function of Front End OF the Line (FEOL) and Back End Of the Line (BEOL) parameters. The model equations are then solved for the various FEOL and BEOL parameters as functions of the test circuit output delays. Finally, measured output delay values are substituted in to these parameter equations to generate actual values for the various FEOL and BEOL parameters, thereby allowing any areas of concern to be quickly and accurately identified.

Abstract: On-the-fly reconfiguration of a secured CPLD. In one embodiment, a CPLD includes a novel security circuit that provides two different security control signals: an EEPROM/SRAM security signal and an EEPROM security override signal. The EEPROM/SRAM security signal prevents reading from both the EEPROM and the SRAM, and also prevents writing to the EEPROM. The EEPROM security override signal enables reading and writing for the EEPROM even when otherwise disabled by the EEPROM/SRAM security signal, but is active only when a specific set of conditions are met. These conditions can include, for example, the application of a sufficiently long erase pulse to the EEPROM array. Thus, the security on the EEPROM array is overridden only after the configuration data set stored in the EEPROM array has been erased. Reading from the SRAM is not enabled by the EEPROM security override signal. Therefore, the configuration data set is not compromised.

Abstract: A JTAG-programmable IC includes a memory array having redundant columns, a partial-width data register, and a full-width bitline register. A programming bitstream is shifted into the data register in discrete portions, with each portion being loaded into the bitline latch before the next portion is shifted into the data register. The programming bitstream portions fill the bitline latch sequentially unless a count indicator for a particular portion matches a predetermined defective column value, in which case that bitstream portion is rerouted to a region of the bitline latch associated with the redundant columns of the memory array. The count indicator is incremented with each new bitstream portion shifted into the data register. Once the programming bitstream is fully loaded into the bitline latch, the data is programmed into a selected row of the memory array in page mode.

Abstract: A method for dealing with process specific physical effects applies dimensional modifications to an IC layout to compensate for performance variations caused by the physical effects. Because the dimensional modifications harmonize the performance of the actual IC with the performance of the IC model, time-consuming re-verification operations are not required. Current drive variations caused by shallow trench isolation (STI) stress can be compensated for by adjusting the gate dimensions of the affected transistors to increase or decrease current drive as necessary. Such physical effect compensation can be applied before, after, or even concurrently with optical proximity correction (OPC). The dimensional modifications for physical effect compensation can also be incorporated into an OPC engine.

Abstract: An electrically programmable fuse includes a metal-oxide-semiconductor (MOS) programmable transistor that is gate-source coupled by a resistive element. The resistive element can comprise a gate-source coupled MOS transistor. If the MOS transistor is unprogrammed, then the resistive element ensures that the programmable transistor is turned off during read operations. However, when a programming voltage is applied across the source and drain terminals of the programmable transistor, the resistive element allows the programming voltage to be capacitively coupled to the gate of the programmable transistor from its drain. This turns the programmable transistor on, thereby reducing the snapback voltage of the programmable transistor, and hence, the required programming voltage. Once the snapback mode is entered, current flow through the programmable transistor increases until thermal breakdown occurs and the programmable transistor shorts out.

Abstract: A security circuit for a reprogrammable logic IC includes an evolved circuit that ties the performance of the security circuit to the physical properties of that particular reprogrammable logic IC. The security circuit can be a decryption and/or encryption circuit that decrypts and/or encrypts, respectively, a configuration bitstream for the IC. Because of the link between the performance of the security circuit and the physical properties of the IC, the security circuit cannot be used in other ICs. For example, an encrypted bitstream that can be decrypted by the security circuit in a first IC will typically not be decrypted by the same security circuit in a second IC, since the physical properties of the two ICs will typically be different. The evolved circuit can comprise a portion of the security circuit, such as a security key generator, or it can comprise the full security circuit.

Abstract: On-the-fly reconfiguration of a secured CPLD. In one embodiment, a CPLD includes a novel security circuit that provides two different security control signals: an EEPROM/SRAM security signal and an EEPROM security override signal. The EEPROM/SRAM security signal prevents reading from both the EEPROM and the SRAM, and also prevents writing to the EEPROM. The EEPROM security override signal enables reading and writing for the EEPROM even when otherwise disabled by the EEPROM/SRAM security signal, but is active only when a specific set of conditions are met. These conditions can include, for example, the application of a sufficiently long erase pulse to the EEPROM array. Thus, the security on the EEPROM array is overridden only after the configuration data set stored in the EEPROM array has been erased. Reading from the SRAM is not enabled by the EEPROM security override signal. Therefore, the configuration data set is not compromised.

Abstract: A high-speed optical transceiver for an integrated circuit (IC) includes a serializer-deserializer (SERDES) and a vertical cavity surface emitting laser (VCSEL) combined with a detector array. By covalently bonding the SERDES die to the IC, the two components can be processed simultaneously to produce a tightly aligned, high-speed data interface. The SERDES can be coupled to the VCSEL/detector array using a flex interconnect, or the VCSEL/detector array can also be covalently bonded to the IC or SERDES to maximize data bandwidth. The SERDES and VCSEL/detector array can also be produced in a single die using a process technology appropriate for both to maximize manufacturing efficiency.

Abstract: A serially programmable integrated circuit (IC) includes a memory array and multiple data registers daisy-chained by bypass logic. Each of the data registers is associated with a primary column grouping or redundant column grouping in the memory array. If a data register is associated with a primary column grouping that includes a defective column, the bypass logic bypasses that data register and incorporates one of the data registers associated with a redundant column grouping into the serial programming path of the IC. Therefore, when a programming bitstream is shifted into this serial programming path, defective columns in the memory array are automatically bypassed during the subsequent programming operation. To read a word from the memory array, any data stored in the redundant columns is first read out, and then the data from the primary columns is read out, bypassing the previously identified defective column groupings.

Abstract: A method for reconfiguring a complex programmable logic device (CPLD) that includes an EEPROM array and a shadow SRAM array comprises reprogramming the EEPROM array with new configuration data while the CPLD is operating in a first configuration. This relatively time-consuming operation has no effect on CPLD operation since only the SRAM array controls the configuration of the CPLD. At a desired point in time, the new configuration data from the EEPROM array can be loaded into the SRAM array to reconfigure the CPLD. Because this loading of configuration data into the SRAM array takes only microseconds to perform, normal system operation effectively proceeds without interruption. A CPLD can include multiple EEPROM arrays, each storing a different set of configuration data, thereby allowing the CPLD to rapidly switch between various configurations by loading the configuration data from different EEPROM arrays into the SRAM array.

Abstract: A structure and method to prevent barrier failure is provided. The present invention replaces a standard titanium-nitride (TiN) barrier metal layer with two separately-formed TiN layers. The two TiN layers provide smaller, mismatched grain boundaries. During subsequent tungsten deposition using WF6, the WF6 finds it difficult to penetrate through the mismatched grain boundaries, thereby minimizing any possibility of “tungsten volcano”. One embodiment includes a native or a grown oxide formed between the two TiN layers, thereby providing yet another diffusion barrier to the WF6 and acting as a glue layer between the two TiN layers. The present invention provides a thin and strong barrier metal layer with minimal barrier failures.

Abstract: A negative voltage detector including a resistor divider circuit is used to translate a negative voltage into a standard CMOS logic low or logic high value. The small area consumed by the negative voltage divider allows multiple device placement within a logic device without the consumption of much area on the logic device. Additionally, the multiple devices placed may detect different negative voltage thresholds with a simple tuning of device components.

Abstract: A one-shot system for loading a bitline shift register with a typical test pattern is described. Each bitline latch within the bitline shift register is augmented with a one-shot circuit that may pull-up or pull-down the value stored in the bitline latch. The choice of a particular memory test pattern dictates the control of the one-shot circuit.

Abstract: A core for a register-based programmable logic device includes a register configured to provide a hidden identifier in response to a secret unlock operation. The identifier is inaccessible during normal operation of the core implementation. The unlock operation is selected to be an action or set of actions that would typically not be performed during normal use of the core implementation. The logic associated with providing the hidden identifier in response to the unlock operation is configured to not interfere with normal operation of the core implementation. Therefore, the presence of this source identification capability is transparent to regular users (and unauthorized copyists) of the core implementation. The availability of the secondary identifier can be limited in duration to minimize the chances of accidental, or even intentional, discovery.

Abstract: An apparatus and method for enabling hot swapping of programmable logic devices (PLDs) and boards containing PLDs is provided. If the hot swap capability is desired, a hot swap terminal on the PLD is set to facilitate a floating state on the input/output pad of the PLD. Further, the input buffer and the output buffer of the PLD are disabled. In one embodiment, a predetermined voltage is provided on the output terminal of the input buffer. In this configuration, the hot swap circuit eliminates any leakage current, ensures no static current occurs, and provides appropriate signals to the internal circuits of the PLD.

Abstract: A method for providing a core for a programmable logic device (PLD) is provided. In this method, a vendor can designate the size and ports of a core. Using this information, a user can generate a top-level design that can accommodate the core. The user can then submit that top-level design to the vendor, or a third party designated by the vendor, to generate a complete configuration bitstream for the PLD. The user can use this configuration bitstream to program the PLD, thereby implementing the top-level design including the core. The number of bits in this configuration bitstream is typically large enough to render reverse engineering economically unfeasible. Thus, the method allows vendors to retain control over their proprietary core IP and discourages undetectable use of this IP.

Abstract: A memory cell comprises a multilayer gate heating structure formed over a channel region between source and drain regions. The multilayer gate heating structure comprises polysilicon and metal silicide layers stacked over a similarly shaped gate oxide. The gate heating structure includes a fusible portion in the metal silicide layer formed over the channel region. In an unprogrammed state, the memory cell operates as a conventional MOS transistor, with current flow between the source and drain regions being controlled by a control voltage applied to the metal silicide layer. However, when a programming voltage is applied across the metal silicide, layer, the fusible portion agglomerates, generating intense localized heating. In an embodiment of the invention, the memory cell is an NMOS device. Tie heating causes segregation of the channel dopant atoms towards the source and drain regions, lowering the threshold voltage of the device.