16Gb/s and Beyond with Single- Ended I/O in High-Performance Graphics Memory
|
|
- Jeffery Lynch
- 6 years ago
- Views:
Transcription
1 DesignCon Gb/s and Beyond with Single- Ended I/O in High-Performance Graphics Memory Tim Hollis, Micron Technology, Inc. [ Salman Jiva, Micron Semiconductor Products Martin Brox, Micron Semiconductor GmbH Wolfgang Spirkl, Micron Semiconductor GmbH Thomas Hein, Micron Semiconductor GmbH Dave Ovard, Micron Technology, Inc. Roy Greeff, Micron Technology, Inc. Dan Lin, Micron Technology, Inc. Michael Richter, Micron Semiconductor GmbH Peter Mayer, Micron Semiconductor GmbH Walt Moden, Micron Technology, Inc. Maksim Kuzmenka, Micron Semiconductor GmbH Mani Balakrishnan, Micron Semiconductor GmbH Milena Ivanov, Micron Semiconductor GmbH Manfred Plan, Micron Semiconductor GmbH Marcos Alvarez Gonzalez, Micron Semiconductor GmbH Bryce Gardiner, Micron Technology, Inc. Dong Soon Lim, Micron Technology, Inc. Jörg Weller, Micron Semiconductor GmbH 1
2 Abstract GDDR5 has emerged as a leading DRAM interface for applications requiring high system bandwidth like graphic cards, game consoles and high-performance compute systems. However, the requirements of newer applications drive even higher memory bandwidth. The paper discusses the development of GDDR6 as a lower-risk and more cost-effective solution as compared to other high-bandwidth memory solutions. We further introduce GDDR6 as offering a 2x increase in per-pin bandwidth over GDDR5, while maintaining compatibility with the established GDDR5 ecosystem. Circuit and channel performance scaling will be discussed and validated through measurement to demonstrate the potential for scaling GDDR6 to 16Gb/s. Authors Biography Tim Hollis received the B.S. degree in Electrical Engineering from the University of Utah, Salt Lake City, UT in 2003 and the Ph.D. in Electrical Engineering from Brigham Young University, Provo, UT in 2007, focusing on channel equalization and jitter attenuation circuits for high-speed serial interconnects. He joined Micron Technology, Inc. s Advanced Architecture Group in 2006 and currently serves as the Technical Lead for Micron Technology, Inc. s Signal Integrity R&D team, focusing on future I/O development. Dr. Hollis holds 122 issued US and International patents and is a Senior Member of the IEEE. Salman Jiva is a segment marketing and business development manager for Micron Technology, Inc. s Compute and Networking Business Unit. He leads the product and customer strategy efforts for Micron Technology, Inc. s networking segment and partners with Micron Technology, Inc. s ecosystem to enable technologies with key semiconductor and system partners. Before coming to Micron Technology, Inc., Salman spent over seven years at Altera Corporation, where he held several positions in product management, product marketing and technical marketing for their high end FPGAs, SoCs and respective technologies. Salman also spent 6 years at Cisco Systems as a hardware and systems engineer designing enterprise class switches and routers. Mr. Jiva earned his M.S degree in electrical engineering from Santa Clara University. Martin Brox received Dipl. and Dr. degrees from the University Münster in 1988 and 1992 respectively. In 1988 he joined Siemens Corporate Research and in 1992 moved to the IBM/Siemens/Toshiba DRAM development project. In 1997 he returned to Siemens/Munich which later became Infineon and Qimonda responsible for RDRAM, GDDR3 and GDDR5. In 2009 he joined Elpida (now part of Micron Technology, Inc.) focusing on GDDR5, GDDR5X and GDDR6. Dr. Brox served on the program committee of ISSCC and is serving on ESSCIRC/ESSDERC. 2
3 Wolfgang Spirkl received a diploma degree in Physics from the Ludwig-Maximilians- University (LMU, Munich, Germany) in 1986, and a Ph.D. degree in Physics as well from the LMU in From 1998 on he worked for Siemens (later Infineon, Qimonda) in the verification of embedded, network, commodity and graphic DRAM. He is currently working in the Micron Technology, Inc. s Munich Design Centre on product and test for high speed graphic DRAM. Thomas Hein received the Diploma in information technology from the Technical University of Dresden in In 1995 he joined Siemens Semiconductors, which became Infineon Technologies and later Qimonda AG, where he led the design of multiple Multi-Bank MDRAM, SGRAM, GDDR1/3/4/5 designs. From 2009 to 2014 he worked for Elpida Memory (Europe) GmbH. In 2014, he joined Micron Semiconductor (Deutschland) GmbH where he currently working on definition and design and of various high-speed GDDR5, GDDR5X and GDDR6 DRAMs. He was the design lead of the 8G GDDR5X. His interests in DRAM design includes new DRAM architectures, chip packaging and high-speed interfaces. Dave Ovard received a B.S. degree in electrical engineering from the University of Utah (Salt Lake City, Utah USA) in He has been a member of the Signal Integrity R&D Group at Micron Technology, Inc. (Boise, Idaho USA) since Prior to joining Micron Technology, Inc., he worked at Unisys, developing military data links; Omnipoint Communications, developing early CDMA cell phone technology; and Micron Communications, developing RFID technology. Mr. Ovard is a member of the IEEE and holds more than 60 patents. Roy Greeff received a B.S. degree in electrical engineering from the University of Utah (Salt Lake City, Utah USA) in Presently he works at Micron Technology, Inc. (Boise, Idaho USA) as a Senior Signal Integrity Engineer. Previously he worked as Microwave Communications Engineer in the defense industry. Mr. Greeff has more than 40 patents. Feng (Dan) Lin received the Ph.D. degree in Electrical Engineering (EE) from the University of Idaho, Boise, ID in 2000, and M.S. and B.S. degrees in EE from the University of Electrical Science and Technology of China in 1995 and 1992 respectively. He joined DRAM Design R&D at Micron Technology, Inc., Boise in 2000, focusing on the development of high-speed, leading edge DRAM for high-performance computing. He most recently serves as the lead designer in Signal Integrity R&D team for advanced memory architecture and future I/O development. Dr. Lin is a co-author of the textbook DRAM Circuit Design, Fundamental and High-Speed Topics (Wiley-IEEE Press, 2008). Dr. Lin holds over 117 U.S. and foreign patents to date. 3
4 Michael Richter received a diploma in electrical engineering from the Technical University Munich (Munich, Germany) in He joined Siemens Semiconductors in 1984 where his work included ASIC design and design support for ASIC customers. Later, he joined Infineon Technologies AG where he served as a Program Manager of a high-speed crypto IC project. With Qimonda AG he was responsible for product definition and standardization of GDDR5, which he continued with Elpida Memory and now with Micron Technology, Inc. on GDDR5X and GDDR6. Peter Mayer made a professional training as radio and TV technician and received a Diploma in electrical engineering from the University of Erlangen (Germany) in In 1997 he joined Siemens Semiconductor in Munich which became Infineon Technologies and later Qimonda. There he served in product engineering team and lead graphics memory (SGRAM, GDDR1 to GDDR5) system team. In 2009 the graphics group joined Elpida (now Micron Technology, Inc.), where he works on GDDR5, GDDR5x and GDDR6 system enabling, system design analysis and technical customer support. He is author or co-author of several US and international patents. Walter Moden received a B.S. and 2 M.S. degrees from the University of Idaho (Moscow, ID) in 1984, 1986, and He has been at Micron Technology, Inc. (Boise, Idaho USA) since 1988, holding various positions in the Assembly and Packaging Groups. Currently working as a Principle Package Design Engineer focusing on high speed memory. Mr. Moden holds over 140 U.S. and foreign patents. Maksim Kuzmenka received a diploma in electronics from the Belarusian State University of Informatics and Electronic (Minsk, Belarus) in In 1993, he joined BELOMO (Belarusian Optic-Mechanics Association) where he worked in the area of pulse mode power supplies design. In 2001, he joined the Memory Products division of Infineon Technologies which was later changed to Qimonda and Elpida (now Micron Technology, Inc.). He participated in the development of IO and multiple other analog and mixed mode circuitry. Mr. Kuzmenka is the (co-)author of several European and U.S. patents Mani Balakrishnan received a B.S. degree from the Coimbatore Institute of Technology (Coimbatore, India) in 2000 and an M.S. degree from the University of Southern California (Los Angeles, California USA) in He worked with Biomorphic from 2004 to 2006 in CMOS Image Sensor and Intel India from 2006 to 2012 in Design and Development of High-speed IO blocks for the Intel Architecture Group. In 2013, he joined Elpida (now Micron Technology, Inc.). He is working on High-speed PLL and Receiver Blocks. Milena Ivanov received Master degree in telecommunications from the Technical University in Sofia, Bulgaria, in 1996, and the Diploma in electrical engineering from the 4
5 Technical University of Munich, Germany in She joined the Memory Products Division of Infineon Technologies (which became Qimonda AG) in 2005 and was active in the development of emerging memories, in particular, conductive bridging RAM. In 2009 she joined Elpida (now Micron Technology, Inc.) and was engaged in the development of GDDR5, GDDR5X and GDDR6. Ms. Ivanov is author as well as coauthor of several US-patents. Manfred Plan received a diploma in electrical engineering from the Technical University of Munich (Munich, Germany) in In 1988, he joined Siemens AG, received admission to a trainee program (Siemens Graduate Program), and worked in different product development divisions (Berkeley, California USA and Villach, Austria). From 1990 until 1998, he was active in chip and memory design for the Consumer Development division and until 2000 in microprocessor design for the Chipcard Development division at Siemens/ Infineon AG. In 2000, he moved to the Memory Product Development division at Infineon/Qimonda AG where he worked on embedded DRAM, RLDRAM, and GDDR5 until 2009, and he then joined Elpida (now Micron Technology, Inc.). Mr. Plan is a member of FEANI, co-author of several publications, and holds some national patents as well as one U.S. patent. Marcos Alvarez Gonzalez received a B.S. and M.S. degree in computer science and engineering from the Universidad de A Corunna (Spain) in 2008 and a M.S. degree in computer science and artificial intelligence from Arizona State University (USA) in He joined Micron Technology, Inc. (Boise, ID, USA) in 2010 as a quality and reliability assurance engineer. In 2015 he started working for Micron Semiconductor Deutschland (Munich, Germany) as a senior product engineer. Bryce Gardiner received a BS degree in electrical engineering, from the University of Utah, Salt Lake City, Utah in He is currently a Senior Signal Integrity Engineer in the Systems Signal Integrity group at Micron Technology, Inc. (Boise, ID USA). His current work at Micron Technology, Inc. is in providing power and signal integrity analysis by optimizing test boards, characterization boards and systems across Micron Technology, Inc. s entire product portfolio. Before joining Micron Technology, Inc. in 2010, he worked for Northrop Grumman in their Electronic Navigation Systems Division, developing Electronic Guidance Systems. Mr. Gardiner is a member of IEEE. Dong Soon Lim received a B.S. degree in electrical engineering from the University of Dong-A (Busan, Korea) in He is working as a senior signal integrity engineer at Micron Technology, Inc. (Boise, Idaho USA). He is responsible for the electrical models, substrate design optimization, SI and PI performance of high speed memory packages. His interests include high speed package design and system level analysis for high-speed products, such as LPDDR4 and graphic DRAM. Before joining Micron Technology, Inc., 5
6 he worked as senior signal Integrity engineer in the area of wireless network system, mobile chip & package, and mobile product at Samsung Electronics (Suwon, Korea). Jörg Weller (M 07) was born in Munich, Germany in He received a diploma in electrical engineering (M.S.E.E.) from the Technical University of Munich in He joined Siemens AG in 1994, where he was responsible for the design and layout of a 4 M-bit DRAM, continuing with production wafer testing for RDRAMs in In 2001, he started design analysis for graphics DRAMs with Infineon AG. Following the spin-off of Qimonda AG in 2006, he continued design analysis work on GDDR3 and GDDR5 interfaces. In 2009, he joined Elpida Memory Europe and was involved in graphic DRAM development. With the transition to Micron Technology, Inc. in 2012 he continued to work on GDDR5, GDDR5x and GDDR6 graphic DRAM. 6
7 Introduction GDDR5 has been an enabler of high performance applications for nearly 10 years. While the first devices were introduced at a data rate of 6 Gb/s/pin [1], today, cards reaching 8 Gb/s/pin are readily available in the marketplace. Yet, even the fastest GDDR5 claimed is only running at 9 Gb/s/pin [2]; barely exceeding the speed of parts already commercially available. Thus, as successful as GDDR5 has been, this apparent slow-down in bandwidth scaling needed to be addressed. To preempt the inevitable gap between market bandwidth demands and available memory component performance, the industry has pursued two parallel paths. One approach has been to develop a completely new memory architecture, coincidently named High Bandwidth Memory (HBM) [3]. With the understanding that memory interface performance is limited primarily by the slower DRAM transistor process, as well as the chip-to-chip interconnect, HBM chose to completely redefine the problem by adopting an extremely wide (many wires) interface, thus allowing the DRAM input/output (I/O) circuitry to operate well within the DRAM process capabilities. That redefinition, however, requires a substantial enhancement in the supporting technologies; namely the reliance on Silicon Interposer and Through-Silicon-Via (TSV) technologies to couple the GPU/ASIC with the DRAM. HBM, as it has been defined, thus consists of a TSV-interconnected stack of DRAMs resting on a logic buffer die, which communicates with the GPU/ASIC across a finepitched Si interposer-based channel at relatively low per-pin data rates. While the HBM architecture offers many advantages, its complexity, in terms of testability, stability, durability and overall cost have limited its adoption to only the highest-tier applications. In parallel with the development of HBM, the more evolutionary path of the GDDR family of DRAMs has continued to scale in performance, offering a more cost-effective and flexible alternative. Without redefining the system, GDDR5X directly addressed the key bandwidth limiters through internal data path and clocking enhancements [4], thus providing for less risky adoption in more cost-sensitive applications (e.g., gaming, etc.,). GDDR5X is currently found in the marketplace reaching per-pin data rates of up to 11.4 Gb/s [5]. Even as GDDR5X continues to make incremental improvements, the next natural step on the GDDR path was to develop a GDDR6 standard capable of supporting per-pin data rates doubling GDDR5. Presently the official GDDR6 JEDEC standard covers a range from 12Gb/s 14Gb/s, but in this paper, we will demonstrate GDDR6 scalability to at least 16Gb/s/pin. The remainder of the paper is divided into three primary sections: silicon changes (e.g., circuit and architecture) and channel enhancements required to support scaling the single-ended GDDR interface all the way to 16Gb/s, along with performance measurements of Micron Technology, Inc. s first 8Gb, 16Gb/s/pin GDDR6 offering. 7
8 Figure 1: Die photo of Micron Technology, Inc. s first 8Gb, 16Gb/s-capable GDDR6 offering Silicon Changes An image of Micron Technology, Inc. s first GDDR6 die is shown in Fig. 1. While circuit and architectural changes were needed to achieve the new bandwidth target, a primary goal of the GDDR6 component definition was to stay close to the evolutionary path, and thus take advantage of the advanced GDDR5 and GDDR5X infrastructure, including established packaging, handling and testing methods. As a result, external features like the general command protocol have been held over from the previous standards. The most prominent new feature, in terms of system application, has been the added support for dual-channel (2 x16) operation. Another change, visible to the system, is support for stronger on-die termination and output driver pull-up strength of 48. While the output driver remains somewhat asymmetric (60 /40 or 48 /40 ), this support for 48 improves signal symmetry, while providing a better match to the typical characteristic impedance of the package + printed circuit board (PCB) channel for enhanced signal integrity. Aside from these changes, and some specification extensions related to Phase Locked Loop (PLL) operation and clocking flexibility, GDDR6 speeds are enabled through the accumulation of several incremental improvements over the GDDR generations. Data Bus Inversion (DBI) was introduced as early as GDDR4, and has continued to prove effective at mitigating simultaneous switching output (SSO) noise while lowering signaling power [6]. Decision Feedback Equalization (DFE) for channel impairment compensation, PLL for jitter filtering, regulated voltage supplies to minimize Power 8
9 Supply Induced Jitter (PSIJ) in the high-speed Write Clock (WCK) distribution were all included in early GDDR5 designs [1]. Output driver boosting (similar to pre-emphasis) was incorporated into GDDR5 to open the data eye as speeds continued to push, while at the same time, innovative internal modes of operation, like Frequency Controlled Switching (FCS) of charge pumps, used to set the varied voltage levels throughout the DRAM, proved helpful in reducing internal supply variation for greater stability along the data path [7]. Duty cycle correction and input clock equalization, along with inductive-capacitive (LC) resonant clock tree termination and a more prevalent use of current-mode-logic (CML) circuit design enabled yet higher speeds through reducing jitter in the DRAM clock distribution [2]. To overcome limitations in DRAM array timing, GDDR5X doubled the data prefetch, increasing the number of bits presented to the output driver with each data access. To double the datarate out of the DRAM without increasing the memory system clock frequency, GDDR5X supports both Double Data Rate (DDR) and Quad Data Rate (QDR) modes of operation. A PLL on the DRAM provides a degree of jitter filtering, and at the same time may be used as a WCK frequency multiplier in QDR mode, though the GDDR6 JEDEC specification provides flexibility in PLL usage, supporting high-speed operation without the PLL. To improve signal integrity when reading data from the memory, tunable de-emphasis was incorporated into the output driver structure; a nontrivial design due to the inherent asymmetry in the unbalanced pull-up/pull-down drive strength definition. Further, to guarantee a more robust interface, increasingly complex and accurate training and calibration of the interface timing and voltage margins has become essential [8], and, as needed, GDDR5 and beyond support per-pin de-skew of the data bus. Finally, to ensure that errors are not introduced during chip-to-chip communication, the results of Cyclic Redundancy Check (CRC) calculations are transmitted from DRAM to the GPU/ASIC over an Error Detect Code (EDC) pin at half of the data rate [4,7]. Nearly all of these incremental advances developed or adopted throughout the history of GDDR find their place within the GDDR6 architecture, and in a later section of this paper corresponding benefits will be quantified through characterization of Micron Technology, Inc. s GDDR6 silicon. Channel Enhancements While the chip-to-chip interconnect consists of several key components, many of which are the responsibility of the system architect, one key channel component which is defined within the JEDEC standard is the ball assignment for the ball grid array (BGA) of the DRAM package. Not only does the ball assignment influence signal integrity within the DRAM package, but it can either facilitate or hinder the routing of the main PCB channel. This is because the relative proximity of signals in the ball grid extends into the PCB through the vertical via transitions to the planar routing layers, which could be 100s of microns deep within the PCB substrate. The ball grid often represents a compromise between the needed signal integrity of the channel, the spatial requirements of the die physical layer (PHY), the greater die architecture and cost. The ball grid should limit 9
10 unwanted signal coupling (crosstalk) in the vertical interconnect, while promoting clean routing of signals, including sufficient and consistent signal return paths to their optimal locations at the silicon interface. Figure 2: Comparison of the JEDEC-specified BGA ball assignments for GDDR5X and GDDR6 (upper-left quadrant, single-byte only). Fig. 2 compares the upper-left quadrant of the JEDEC-specified GDDR5X and GDDR6 package ball assignments; each of the four quadrants supports a single byte of data lines. While not shown in this format, the transition from GDDR5X to GDDR6 included a slight increase in ball pitch from 0.65mm to 0.75mm. On the other hand, as shown, the newly-defined GDDR6 ballout distributes the high-speed data signal balls over a larger area within the grid by extending into the fourth column from the center of the package (column 2), leading to several advantages. For example, VSS balls are more equally distributed across the ball matrix resulting in better signal returns. Coupling between data lines and the even more critical WCK lines is reduced, as well as coupling between data and EDC pins. The proximity of uni-directional (WCK, EDC, etc.) and bi-directional (DQ, DBI, etc.) signals in the package produces distinct coupling conditions during DRAM Read and Write operations, and warrants careful analysis. A qualitative evaluation of the worst-case data ball positions in the GDDR5X and GDDR6 definitions projects that the DQ3 signal in the GDDR5X case will experience far end crosstalk (FEXT) from DQ0, DQ1, DQ2 and WCK, with even more detrimental near end crosstalk (NEXT) from the EDC signal. (NEXT is expected to be a greater concern, as much of the FEXT will be mitigated through stripline routing of the high-speed signals.) In the GDDR6 case, the DQ2 signal experiences FEXT from DQ0, DQ1 and DQ3 and NEXT from EDC. Thus, the high-speed clocks have been spatially separated from the single-ended data lines. Additionally, at least one aggressor has been removed 10
11 from proximity to the EDC ball, which further increases the robustness of the interface. Of course, this qualitative discussion does not account for the pin assignments at the far (GPU) end of the channel, but the GDDR6 DRAM package ballout at least promotes cleaner channel routing between chips. Figure 3: Comparison of crosstalk in GDDR5X and GDDR6 DRAM packages. Fig. 3 adds some data to this qualitative discussion of coupling, by presenting the sum of a crosstalk within the DRAM package onto the worst-case DQ lines (left), DQ3 for GDDR5X and DQ2 for GDDR6, along with the sum of all crosstalk onto the EDC lines (right). A DRAM Write operation is assumed for both cases, and thus far-end coupling from all signals onto the DQ of interest, with the exception of the EDC, is summed and then combined with the corresponding near-end coupling from the EDC line in the left plot. In the right plot, near-end coupling from all signals onto the EDC line is accumulated. While the package models were both extracted out to 40GHz, the GDDR5X extraction accounted for half of the physical package, while the GDDR6 extraction only accounted for a quadrant of the package, which may help to explain the apparent differences in smoothness of the response curves. It is clear, however, from the left side of Fig. 3, that the accumulated coupling onto the worst-case DQ is improved in the GDDR6 package, consistent with our intuition-based comparison of the two ball-outs. The improvement in the EDC response, on the right, may not be as clear, but poses less of a problem as the specification allows for the EDC signal to toggle at half-rate. As a developer of memory technologies, Micron Technology, Inc. does not often delve into the world of system architecture and design. Yet, as all high-speed interface designers know, it is difficult to produce an optimized composite channel when developing the distinct component packages and main substrate connectivity in isolation. Thus, to facilitate better DRAM packaging and I/O characteristics, Micron Technology, Inc. purchased recent off-the-shelf high-speed graphics cards to study the typical DRAM to GPU interconnect as a point of reference for channel optimization studies (see 11
12 Fig. 4). Such an approach has greatly increased our modeling confidence, in that our assumptions (channel length, pitch, stack-up, etc.) are justified through commercially available technologies. Figure 4: Exemplary GPU-to-DRAM channel from an off-the-shelf graphics card. Figure 5: End-to-end channel model producing all simulation results throughout this paper. That said, none of the simulation results shared in this paper correspond directly to an analyzed graphics card channel. Rather, we have chosen to share results based on simulations couched in distinct, yet reasonable assumptions. Our end-to-end channel 12
13 model is symmetric, assuming the via transitions below the DRAM package, as well as the DRAM package routing / construction, at both ends of the link, as shown in Fig. 5. As indicated in the figure, the channel is broken into segments to allow for more accurate 3D modeling of the vertical interconnect and package, while the main PCB route is represented by 2D models for flexibility in studying the impact of channel length, data line width, routing pitch, stack-up, etc. Both ends of the channel are terminated, depending on the direction of the signaling operation, through 48Ω or 60Ω to a 1.35V VDDQ supply. The parasitic capacitive loading at the die pad is assumed to be 0.5pF at each end of the channel. While countless transistor-level simulations have been completed for both the output and input paths of the DRAM, all simulation results shown in this paper are based on linear driver modeling to enable rapid, peak distortion analysis (PDA) of worst case pattern conditions [9]. We acknowledge that the unmatched pullup/pull-down characteristics of the driver cannot be perfectly captured through linear modeling, but we remain confident, based on internal modeling correlation, that this assumption does not significantly alter the results of the paper. As a goal of this paper is to demonstrate bandwidth scalability in GDDR6, Fig. 6 presents the worst-case data eye openings at 14Gb/s and 16Gb/s for the baseline model. It is important to note that the platforms, on which the exemplary channel models were based, were not designed to support 16Gb/s. Thus, the cases shown in Fig. 6 potentially push beyond the boundaries of expected performance. Nevertheless, it is observed that with the available DFE functionality incorporated into GDDR6, this channel delivers an open eye at 14Gb/s. At 16Gb/s, however, the received data eye is completely closed, even after equalization. Figure 6: Baseline channel simulation employing available single-tap DFE, but without any other channel enhancements. 13
14 What steps then may be taken to enable 16Gb/s signaling over this basic channel structure (e.g., materials, distance, routing cross-section, etc.)? One option, not incorporated into the model that produced the eyes in Fig 6, is the well-known back drilling of vias to mitigate impedance discontinuities and crosstalk in the vertical interconnects below the component packaging. In the simulation world, it is straightforward to evaluate what impact a process like back drilling would have on overall performance. Much can be learned from the channel pulse response, as presented in Fig. 7. First, a few details. For simplicity, all PDA-based calculations are done on a sample basis (e.g., x samples per unit interval (UI)) rather than on absolute time. While the pulse responses shown are labelled as volts and sample, final eye measurements are output as absolute voltage and time. All cursors highlighted in red are separated by 1 UI (62.5ps) and are shifted such that the main cursor aligns with the center of the resulting data eye. While only 13 post cursors are shown, several additional cursors are captured and included in all eye closure calculations. The number of cursors must be chosen so as not to ignore perturbations later in the pulse tail. This is particularly true for a high-speed graphics channel, which, being relatively loss-less, may support lingering reflections resulting from imperfect channel termination, as well as other discontinuities. Figure 7: Pulse response comparison of the channel without and with via back drilling. Fig. 7 compares the 16Gb/s pulse responses for a common data line with and without back drilling. Qualitatively it is observed that back drilling positively impacts the channel response in a variety of ways. First it increases the main cursor value (amplitude), while reducing the first three post cursors significantly. Further it greatly reduces the magnitude of the signal reflection dispersed over post cursors Based on the superposition of the pre-cursor and first 13 post-cursors alone, PDA projects vertical eye openings of mV and mV with and without back drilling, respectively. While one might 14
15 expect mV to provide sufficient margin, in the presence of crosstalk the closed data eye shown on the right side of Fig. 5 is not necessarily surprising, even with DFE. To validate this last statement regarding the degree of crosstalk observed across the channel, Fig. 8 compares the un-equalized back-drilled case, with and without aggressors (e.g., all remaining DQ lines in the Byte, along with the EDC and DBI signals). Based on this simulation, about 238mV of crosstalk is expected in the cleaner (e.g., back-drilled) of the two channel environments. More crosstalk would be expected in the absence of back drilling. Thus, as a goal of this paper is to demonstrate a practical path to 16Gb/s, all remaining simulations assume back drilling of vias in the PCB as a foundation for additional enabling steps, including equalization. Figure 8: Simulated data eye openings at 16Gb/s ISI only (left) and with additive crosstalk from the remaining high-speed data lines, EDC and DBI signals (right). Based on a further review of the raw pulse response from the right side of Fig. 7, a practical, power-efficient equalizer solution might only need to address the 1 st post cursor. The GDDR6 I/O incorporates both tunable single-tap de-emphasis into the output driver and tunable single-tap DFE within the input path, both designed to operate on the 1 st post cursor. Fig. 9 compares the relative effectiveness of the available de-emphasis and DFE. As shown, the de-emphasis improves the eye height by 6mV, while degrading the eye width by 1ps. The DFE, on the other hand, improves the eye height by 65mV without degrading the eye width. It is important to note that the results shown are channel-specific and are insufficient to make a universal assessment of the relative value of either equalization method, though some qualitative observations can be made when comparing the corresponding pulse responses, as demonstrated by Fig
16 Figure 9: Simulated data eye openings at 16Gb/s No equalization (left), single-tap deemphasis (center) and single-tap DFE (right). As highlighted in Fig. 10, de-emphasis-based equalization (green and blue curves) reduces the overall amplitude of the signal while reducing the 1 st post cursor. As a result, the optimal amount of de-emphasis corresponds to a balance between signal amplitude and ISI cancellation. For the channel under consideration, 3dB of de-emphasis (blue curve) nearly reduces the 1 st post cursor to zero, yet, as will be shown, a larger eye opening is possible with only 1dB of de-emphasis (green curve). This is because 3dB of de-emphasis does not leave enough of the main cursor to provide a net increase in eye opening, while 1dB of de-emphasis, on the other hand, results in a net positive of 6mV. Intuitively, because the DFE zeros out the 1 st post cursor without reducing the signal amplitude, one would expect a better overall result, which is clearly demonstrated in Fig. 9. One other nuance captured in Fig. 10 is that de-emphasis, while primarily addressing the 1 st post cursor, may impact additional post cursors for better or worse. As shown in this particular case, the 2 nd post cursor is degraded slightly by de-emphasis, while this behavior does not occur with DFE. However, this same fact that de-emphasis may affect more than just the tap in question could produce much better results under different channel conditions. 16
17 Figure 10: Overlay of channel pulse responses comparing various equalization methods. Fig. 11 presents two additional equalization conditions. As shown on the left, when combining the best amount of de-emphasis, namely 1dB, with a corresponding optimized tap of DFE (to cancel the remaining 1 st post cursor ISI), the resulting eye is smaller than that achieved by applying DFE alone (see right side of Fig. 9). This is because the de-emphasis unnecessarily reduces the signal amplitude and the DFE offers no gain to compensate for that reduction. The eye diagram on the right of Fig. 11, corresponding to 3dB of de-emphasis, is also interesting. Recalling from the pulse responses of Fig. 10 that even though 3dB of de-emphasis would almost perfectly zero out the 1 st post cursor, the resulting eye height remains identical to the un-equalized case (while the timing degrades by 3ps). Comparison of this eye with the original un-equalized eye in Fig. 9 reveals that the ISI is, indeed, reduced by the de-emphasis, but the overall signal amplitude is reduced by a similar amount (at least when all of the crosstalk and reflections are accounted for). 17
18 Figure 11: Simulated data eye openings at 16Gb/s Combined de-emphasis and DFE (left) and stronger (3dB) de-emphasis (right). Two final observations regarding equalization. It is worth noting that none of the equalization methods described herein improve eye width. Thus, every effort should be made to minimize crosstalk across these high-speed parallel interconnects. It is also important to understand that while additional equalization methods could be employed in this application, such are not explicitly called for by the JEDEC GDDR6 specification, and therefore are not evaluated here. Nevertheless, 1-tap of DFE, coupled with the backdrilling of PCB vias, appears sufficient to support 16Gb/s signaling. GDDR6 Performance Measurements As it is generally helpful to increase confidence through complimenting simulation with measured results, ATE-based characterization of Micron Technology, Inc. s first GDDR6 offering is shared, beginning with Fig. 12, which compares the measured link margin at 16Gb/s and 16.5Gb/s, based on shmooing the DRAM and tester reference voltages along with the phase of the data relative to the data clock and strobe. Green and Red points distinguish between error-free operation, and the detection of errors, respectively. As shown, GDDR6 s support for the stronger 48Ω termination is expected to improve signaling margins, especially at higher speeds. 18
19 Figure 12: Measured link margin shmoos at 16Gb/s/pin and 16.5 Gb/s/pin for 60Ω and 48Ω line termination. Fig. 13 presents the impact of DFE from two perspectives. First, the maximum achievable data rate (x-axis), as determined by an agreed-upon degree of eye opening (height and width), is plotted against an increasing amount of DFE compensation (yaxis). There are at least two key take-aways. First, the observation that, in spite of the relatively clean tester environment, there is a clear benefit to be gained in optimizing the DFE coefficient selection, above and below which the maximum achievable data rate is degraded. And second, 16Gb/s is nearly achievable without DFE, and thus the equalization adds margin and reliability to the interface. 19
20 Figure 13: Measured achievable data rate shmoo (left) and corresponding link margin shmoos for three DFE settings: no equalization (bottom-right), optimal DFE (centerright) and maximum DFE (top-right). For a deeper comparison, the right side of the figure presents three measured link margin shmoos corresponding to no equalization, optimal DFE setting, and maximum (not optimal) DFE. Interestingly, the non-optimized, maximum DFE setting does not degrade the results substantially, but the optimal setting clearly represents the best solution, in terms of symmetry and overall eye height. Fig. 14 presents the impact of enabling deemphasis. Based on these results, de-emphasis appears to provide substantial benefit over the ATE channel. Figure 14: Measured link margin shmoos at 16Gb/s without and with single-tap deemphasis enabled. 20
21 Figure 15: Measured 20Gb/s data eye based on a PRBS6 pattern While the preceding results demonstrate full DRAM functionality up to as high as 16.5Gb/s, it is possible for the overall performance of an architecture to be capped by timing limitations in the memory array itself. To determine if this GDDR6 interface could extend beyond the 16.5Gb/s range, the device was placed into a mode of operation which exercises only the I/O while bypassing the memory array. The oscilloscope measurement presented in Fig.15 confirms that when bypassing the memory array, and with a small, but helpful, boost in I/O supply voltage, it is possible to push Micron Technology, Inc. s GDDR6 I/O as high as 20Gb/s. Summary As compute systems continue to advance, their efficacy often depends on the accessibility of memory. While some high-tier applications can absorb the high cost and complexity of HBM, the performance of GDDR DRAM continues to scale, providing a more flexible, low-risk, cost-effective alternative. Through reviewing the current state of GDDR5X and ATE-based measurements of Micron Technology, Inc. s first GDDR6 offering, along with known circuit and channel enhancements (namely an improved DRAM package ball out definition with looser pitch and via back drilling within the PCB), we are confident in claiming that GDDR6 data rates will extend beyond the 14Gb/s/pin target defined by JEDEC all the way to 16Gb/s/pin. As a result, GDDR6 looks to be an attractive compliment for generations to come. 21
22 References: [1] S. J. Bae et al., "A 60nm 6Gb/s/pin GDDR5 Graphics DRAM with Multifaceted Clocking and ISI/SSN-Reduction Techniques," ISSCC-2008, pp [2] H. Y. Joo et al., "A 20nm 9Gb/s/pin 8Gb GDDR5 DRAM with an NBTI monitor, jitter reduction techniques and improved power distribution," ISSCC-2016, pp [3] D. U. Lee et al., "1.2V 8Gb 8-channel 128GB/s high-bandwidth memory (HBM) stacked DRAM with effective microbump I/O test methods using 29nm process and TSV," ISSCC-2014, pp [4] M. Brox et al., An 8Gb 12Gb/s/pin GDDR5X DRAM for Cost-Effective High- Performance Applications, ISSCC-2017, pp [5] NVIDIA TITAN Xp Design & Visualization, Accessed November 3, 2017 [6] T. M. Hollis, Data Bus Inversion in High-Speed Memory Applications, IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 56, no. 4, April [7] R. Kho et al., "A 75 nm 7 Gb/s/pin 1 Gb GDDR5 Graphics Memory Device with Bandwidth Improvement Techniques," in IEEE Journal of Solid-State Circuits, pp , Jan [8] A. Shiloy, GDDR5X Standard Finalized by JEDEC: New Graphics Memory up to 14 Gbps, January 22, 2016., Accessed November 3, [9] B. K. Casper, M. Haycock and R. Mooney, "An accurate and efficient analysis method for multi-gb/s chip-to-chip signaling schemes," 2002 Symposium on VLSI Circuits. Digest of Technical Papers (Cat. No.02CH37302), Honolulu, HI, USA, 2002, pp
Verification of HBM through Direct Probing on MicroBumps
Verification of HBM through Direct Probing on MicroBumps FormFactor Sung Wook Moon SK hynix Outline HBM market HBM test flow Device structure overview Key test challenges addressed Signal delivery and
More information100Gb/s Single-lane SERDES Discussion. Phil Sun, Credo Semiconductor IEEE New Ethernet Applications Ad Hoc May 24, 2017
100Gb/s Single-lane SERDES Discussion Phil Sun, Credo Semiconductor IEEE 802.3 New Ethernet Applications Ad Hoc May 24, 2017 Introduction This contribution tries to share thoughts on 100Gb/s single-lane
More informationSwitching Solutions for Multi-Channel High Speed Serial Port Testing
Switching Solutions for Multi-Channel High Speed Serial Port Testing Application Note by Robert Waldeck VP Business Development, ASCOR Switching The instruments used in High Speed Serial Port testing are
More informationHalf-Rate Decision-Feedback Equalization Di-Bit Response Analysis and Evaluation EDA365
DesignCon 2008 Half-Rate Decision-Feedback Equalization Di-Bit Response Analysis and Evaluation Jihong Ren, Rambus Inc. jren@rambus.com Brian Leibowitz, Rambus Inc. Dan Oh, Rambus Inc. Jared Zerbe, Rambus
More informationBrian Holden Kandou Bus, S.A. IEEE GE Study Group September 2, 2013 York, United Kingdom
Simulation results for NRZ, ENRZ & PAM-4 on 16-wire full-sized 400GE backplanes Brian Holden Kandou Bus, S.A. brian@kandou.com IEEE 802.3 400GE Study Group September 2, 2013 York, United Kingdom IP Disclosure
More informationC65SPACE-HSSL Gbps multi-rate, multi-lane, SerDes macro IP. Description. Features
6.25 Gbps multi-rate, multi-lane, SerDes macro IP Data brief Txdata1_in Tx1_clk Bist1 Rxdata1_out Rx1_clk Txdata2_in Tx2_clk Bist2 Rxdata2_out Rx2_clk Txdata3_in Tx3_clk Bist3 Rxdata3_out Rx3_clk Txdata4_in
More informationSpatial Light Modulators XY Series
Spatial Light Modulators XY Series Phase and Amplitude 512x512 and 256x256 A spatial light modulator (SLM) is an electrically programmable device that modulates light according to a fixed spatial (pixel)
More informationDebugging Memory Interfaces using Visual Trigger on Tektronix Oscilloscopes
Debugging Memory Interfaces using Visual Trigger on Tektronix Oscilloscopes Application Note What you will learn: This document focuses on how Visual Triggering, Pinpoint Triggering, and Advanced Search
More informationIEEE Santa Clara ComSoc/CAS Weekend Workshop Event-based analog sensing
IEEE Santa Clara ComSoc/CAS Weekend Workshop Event-based analog sensing Theodore Yu theodore.yu@ti.com Texas Instruments Kilby Labs, Silicon Valley Labs September 29, 2012 1 Living in an analog world The
More informationOptimizing BNC PCB Footprint Designs for Digital Video Equipment
Optimizing BNC PCB Footprint Designs for Digital Video Equipment By Tsun-kit Chin Applications Engineer, Member of Technical Staff National Semiconductor Corp. Introduction An increasing number of video
More informationCombating Closed Eyes Design & Measurement of Pre-Emphasis and Equalization for Lossy Channels
Combating Closed Eyes Design & Measurement of Pre-Emphasis and Equalization for Lossy Channels Why Test the Receiver? Serial Data communications standards have always specified both the transmitter and
More informationCombating Closed Eyes Design & Measurement of Pre-Emphasis and Equalization for Lossy Channels
Combating Closed Eyes Design & Measurement of Pre-Emphasis and Equalization for Lossy Channels Why Test the Receiver? Serial Data communications standards have always specified both the transmitter and
More informationRFI MITIGATING RECEIVER BACK-END FOR RADIOMETERS
RFI MITIGATING RECEIVER BACK-END FOR RADIOMETERS Phaneendra Bikkina 1, Qingjun Fan 2, Wenlan Wu 1, Jinghong Chen 2 and Esko Mikkola 1 1 Alphacore, Inc., 2 University of Houston 2017 CASPER Workshop Pasadena,
More informationPractical Receiver Equalization Tradeoffs Applicable to Next- Generation 28 Gb/s Links with db Loss Channels
DesignCon 2013 Practical Receiver Equalization Tradeoffs Applicable to Next- Generation 28 Gb/s Links with 20 35 db Loss Channels Edward Frlan, Semtech Corp. (EFrlan@semtech.com) Francois Tremblay, Semtech
More informationData Converters and DSPs Getting Closer to Sensors
Data Converters and DSPs Getting Closer to Sensors As the data converters used in military applications must operate faster and at greater resolution, the digital domain is moving closer to the antenna/sensor
More informationPAM4 signals for 400 Gbps: acquisition for measurement and signal processing
TITLE PAM4 signals for 400 Gbps: acquisition for measurement and signal processing Image V1.00 1 Introduction, content High speed serial data links are in the process in increasing line speeds from 25
More informationScan. This is a sample of the first 15 pages of the Scan chapter.
Scan This is a sample of the first 15 pages of the Scan chapter. Note: The book is NOT Pinted in color. Objectives: This section provides: An overview of Scan An introduction to Test Sequences and Test
More informationMMI: A General Narrow Interface for Memory Devices
MMI: A General Narrow Interface for Devices Judy Chen Eric Linstadt Rambus Inc. Session 106 August 12, 2009 August 2009 1 What is MMI? WLAN BT GPS NOR S/M Baseband Processor Apps/Media Processor NAND M
More informationSPATIAL LIGHT MODULATORS
SPATIAL LIGHT MODULATORS Reflective XY Series Phase and Amplitude 512x512 A spatial light modulator (SLM) is an electrically programmable device that modulates light according to a fixed spatial (pixel)
More informationDraft Baseline Proposal for CDAUI-8 Chipto-Module (C2M) Electrical Interface (NRZ)
Draft Baseline Proposal for CDAUI-8 Chipto-Module (C2M) Electrical Interface (NRZ) Authors: Tom Palkert: MoSys Jeff Trombley, Haoli Qian: Credo Date: Dec. 4 2014 Presented: IEEE 802.3bs electrical interface
More informationEVALUATION KIT AVAILABLE 12.5Gbps Settable Receive Equalizer +2.5V +3.3V V CC1 V CC. 30in OF FR-4 STRIPLINE OR MICROSTRIP TRANSMISSION LINE SDI+ SDI-
19-2713; Rev 1; 11/03 EVALUATION KIT AVAILABLE 12.5Gbps Settable Receive Equalizer General Description The driver with integrated analog equalizer compensates up to 20dB of loss at 5GHz. It is designed
More informationImplementation Challenges and Solutions of Low-Power, High-Performance Memory Systems
WHITE PAPER Implementation Challenges and Solutions of Low-Power, High-Performance Memory Systems 1050 Enterprise Way, Suite 700 Sunnyvale, CA 94089 Phone: + 1 408 462 8000 Fax: + 1 408 462 8001 www.rambus.com
More informationDesignCon Pavel Zivny, Tektronix, Inc. (503)
DesignCon 2009 New methods of measuring the performance of equalized serial data links and correlation of performance measures across the design flow, from simulation to measurement, and final BER tests
More informationECEN689: Special Topics in High-Speed Links Circuits and Systems Spring 2011
ECEN689: Special Topics in High-Speed Links Circuits and Systems Spring 2011 Lecture 9: TX Multiplexer Circuits Sam Palermo Analog & Mixed-Signal Center Texas A&M University Announcements & Agenda Next
More informationDesigning High Performance Interposers with 3-port and 6-port S-parameters
DesignCon 2015 Designing High Performance Interposers with 3-port and 6-port S-parameters Joseph Socha, Nexus Technology joe.socha@nexustechnology.com Jonathan Dandy, Tektronix jonathan.s.dandy@tektronix.com
More informationSystem-Level Timing Closure Using IBIS Models
System-Level Timing Closure Using IBIS Models Barry Katz President/CTO, SiSoft Asian IBIS Summit Asian IBIS Summit Tokyo, Japan - October 31, 2006 Signal Integrity Software, Inc. Agenda High Speed System
More informationA 5-Gb/s Half-rate Clock Recovery Circuit in 0.25-μm CMOS Technology
A 5-Gb/s Half-rate Clock Recovery Circuit in 0.25-μm CMOS Technology Pyung-Su Han Dept. of Electrical and Electronic Engineering Yonsei University Seoul, Korea ps@tera.yonsei.ac.kr Woo-Young Choi Dept.
More informationInnovative Fast Timing Design
Innovative Fast Timing Design Solution through Simultaneous Processing of Logic Synthesis and Placement A new design methodology is now available that offers the advantages of enhanced logical design efficiency
More informationPrecise Digital Integration of Fast Analogue Signals using a 12-bit Oscilloscope
EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH CERN BEAMS DEPARTMENT CERN-BE-2014-002 BI Precise Digital Integration of Fast Analogue Signals using a 12-bit Oscilloscope M. Gasior; M. Krupa CERN Geneva/CH
More informationCDAUI-8 Chip-to-Module (C2M) System Analysis #3. Ben Smith and Stephane Dallaire, Inphi Corporation IEEE 802.3bs, Bonita Springs, September 2015
CDAUI-8 Chip-to-Module (C2M) System Analysis #3 Ben Smith and Stephane Dallaire, Inphi Corporation IEEE 802.3bs, Bonita Springs, September 2015 Supporters Ali Ghiasi, Ghiasi Quantum LLC Marco Mazzini,
More informationVersatile IO Circuit Schemes for LPDDR4 with 1.8mW/Gbps/pin Power Efficiency. Kyoung-Hoi Koo
Versatile IO Circuit Schemes for LPDDR4 with 1.8mW/Gbps/pin Power Efficiency Kyoung-Hoi Koo Outline Introduction for LPDDR4 Channel Sensitivity Analysis Backward Compatibility Summary Rapid Technology
More informationMicrowave Interconnect Testing For 12G-SDI Applications
DesignCon 2016 Microwave Interconnect Testing For 12G-SDI Applications Jim Nadolny, Samtec jim.nadolny@samtec.com Corey Kimble, Craig Rapp Samtec OJ Danzy, Mike Resso Keysight Boris Nevelev Imagine Communications
More informationPerformance Modeling and Noise Reduction in VLSI Packaging
Performance Modeling and Noise Reduction in VLSI Packaging Ph.D. Defense Brock J. LaMeres University of Colorado October 7, 2005 October 7, 2005 Performance Modeling and Noise Reduction in VLSI Packaging
More informationBenchtop Portability with ATE Performance
Benchtop Portability with ATE Performance Features: Configurable for simultaneous test of multiple connectivity standard Air cooled, 100 W power consumption 4 RF source and receive ports supporting up
More informationPowerful Software Tools and Methods to Accelerate Test Program Development A Test Systems Strategies, Inc. (TSSI) White Paper.
Powerful Software Tools and Methods to Accelerate Test Program Development A Test Systems Strategies, Inc. (TSSI) White Paper Abstract Test costs have now risen to as much as 50 percent of the total manufacturing
More informationDual Link DVI Receiver Implementation
Dual Link DVI Receiver Implementation This application note describes some features of single link receivers that must be considered when using 2 devices for a dual link application. Specific characteristics
More informationCDAUI-8 Chip-to-Module (C2M) System Analysis. Stephane Dallaire and Ben Smith, September 2, 2015
CDAUI-8 Chip-to-Module (C2M) System Analysis Stephane Dallaire and Ben Smith, September 2, 2015 Introduction (1) Follow-up to previous ad hoc contribution on the merits of various reference receiver architectures
More informationSharif University of Technology. SoC: Introduction
SoC Design Lecture 1: Introduction Shaahin Hessabi Department of Computer Engineering System-on-Chip System: a set of related parts that act as a whole to achieve a given goal. A system is a set of interacting
More informationNew Techniques for Designing and Analyzing Multi-GigaHertz Serial Links
New Techniques for Designing and Analyzing Multi-GigaHertz Serial Links Min Wang, Intel Henri Maramis, Intel Donald Telian, Cadence Kevin Chung, Cadence 1 Agenda 1. Wide Eyes and More Bits 2. Interconnect
More informationThe EMC, Signal And Power Integrity Institute Presents
The EMC, Signal And Power Integrity Institute Presents Module 12 Pre-emphasis And Its Impact On The Eye Pattern And Bit-Error-Rate For High-Speed Signaling By Dr. David Norte Copyright 2005 by Dr. David
More informationComparison of NRZ, PR-2, and PR-4 signaling. Qasim Chaudry Adam Healey Greg Sheets
Comparison of NRZ, PR-2, and PR-4 signaling Presented by: Rob Brink Contributors: Pervez Aziz Qasim Chaudry Adam Healey Greg Sheets Scope and Purpose Operation over electrical backplanes at 10.3125Gb/s
More informationDual Link DVI Receiver Implementation
Dual Link DVI Receiver Implementation This application note describes some features of single link receivers that must be considered when using 2 devices for a dual link application. Specific characteristics
More informationA MISSILE INSTRUMENTATION ENCODER
A MISSILE INSTRUMENTATION ENCODER Item Type text; Proceedings Authors CONN, RAYMOND; BREEDLOVE, PHILLIP Publisher International Foundation for Telemetering Journal International Telemetering Conference
More informationNew Serial Link Simulation Process, 6 Gbps SAS Case Study
ew Serial Link Simulation Process, 6 Gbps SAS Case Study Donald Telian SI Consultant Session 7-TH2 Donald Telian SI Consultant About the Authors Donald Telian is an independent Signal Integrity Consultant.
More informationAnalyzing 8b/10b Encoded Signals with a Real-time Oscilloscope Real-time triggering up to 6.25 Gb/s on 8b/10b encoded data streams
Presented by TestEquity - www.testequity.com Analyzing 8b/10b Encoded Signals with a Real-time Oscilloscope Real-time triggering up to 6.25 Gb/s on 8b/10b encoded data streams Application Note Application
More informationECE 5765 Modern Communication Fall 2005, UMD Experiment 10: PRBS Messages, Eye Patterns & Noise Simulation using PRBS
ECE 5765 Modern Communication Fall 2005, UMD Experiment 10: PRBS Messages, Eye Patterns & Noise Simulation using PRBS modules basic: SEQUENCE GENERATOR, TUNEABLE LPF, ADDER, BUFFER AMPLIFIER extra basic:
More informationRX40_V1_0 Measurement Report F.Faccio
RX40_V1_0 Measurement Report F.Faccio This document follows the previous report An 80Mbit/s Optical Receiver for the CMS digital optical link, dating back to January 2000 and concerning the first prototype
More informationReport on 4-bit Counter design Report- 1, 2. Report on D- Flipflop. Course project for ECE533
Report on 4-bit Counter design Report- 1, 2. Report on D- Flipflop Course project for ECE533 I. Objective: REPORT-I The objective of this project is to design a 4-bit counter and implement it into a chip
More informationFuture of Analog Design and Upcoming Challenges in Nanometer CMOS
Future of Analog Design and Upcoming Challenges in Nanometer CMOS Greg Taylor VLSI Design 2010 Outline Introduction Logic processing trends Analog design trends Analog design challenge Approaches Conclusion
More informationAltiumLive 2017: Effective Methods for Advanced Routing
AltiumLive 2017: Effective Methods for Advanced Routing Charles Pfeil Senior Product Manager Dave Cousineau Sr. Field Applications Engineer Charles Pfeil Senior Product Manager Over 50 years of experience
More informationISSCC 2006 / SESSION 18 / CLOCK AND DATA RECOVERY / 18.6
18.6 Data Recovery and Retiming for the Fully Buffered DIMM 4.8Gb/s Serial Links Hamid Partovi 1, Wolfgang Walthes 2, Luca Ravezzi 1, Paul Lindt 2, Sivaraman Chokkalingam 1, Karthik Gopalakrishnan 1, Andreas
More informationA Briefing on IEEE Standard Test Access Port And Boundary-Scan Architecture ( AKA JTAG )
A Briefing on IEEE 1149.1 1990 Standard Test Access Port And Boundary-Scan Architecture ( AKA JTAG ) Summary With the advent of large Ball Grid Array (BGA) and fine pitch SMD semiconductor devices the
More informationEL302 DIGITAL INTEGRATED CIRCUITS LAB #3 CMOS EDGE TRIGGERED D FLIP-FLOP. Due İLKER KALYONCU, 10043
EL302 DIGITAL INTEGRATED CIRCUITS LAB #3 CMOS EDGE TRIGGERED D FLIP-FLOP Due 16.05. İLKER KALYONCU, 10043 1. INTRODUCTION: In this project we are going to design a CMOS positive edge triggered master-slave
More informationTiming EECS141 EE141. EE141-Fall 2011 Digital Integrated Circuits. Pipelining. Administrative Stuff. Last Lecture. Latch-Based Clocking.
EE141-Fall 2011 Digital Integrated Circuits Lecture 2 Clock, I/O Timing 1 4 Administrative Stuff Pipelining Project Phase 4 due on Monday, Nov. 21, 10am Homework 9 Due Thursday, December 1 Visit to Intel
More informationEfficient 500 MHz Digital Phase Locked Loop Implementation sin 180nm CMOS Technology
Efficient 500 MHz Digital Phase Locked Loop Implementation sin 180nm CMOS Technology Akash Singh Rawat 1, Kirti Gupta 2 Electronics and Communication Department, Bharati Vidyapeeth s College of Engineering,
More informationOF AN ADVANCED LUT METHODOLOGY BASED FIR FILTER DESIGN PROCESS
IMPLEMENTATION OF AN ADVANCED LUT METHODOLOGY BASED FIR FILTER DESIGN PROCESS 1 G. Sowmya Bala 2 A. Rama Krishna 1 PG student, Dept. of ECM. K.L.University, Vaddeswaram, A.P, India, 2 Assistant Professor,
More informationGuidance For Scrambling Data Signals For EMC Compliance
Guidance For Scrambling Data Signals For EMC Compliance David Norte, PhD. Abstract s can be used to help mitigate the radiated emissions from inherently periodic data signals. A previous paper [1] described
More informationElectrical Sampling Modules Datasheet 80E11 80E11X1 80E10B 80E09B 80E08B 80E07B 80E04 80E03 80E03-NV
Electrical Sampling Modules Datasheet 80E11 80E11X1 80E10B 80E09B 80E08B 80E07B 80E04 80E03 80E03-NV The DSA8300 Series Sampling Oscilloscope, when configured with one or more electrical sampling modules,
More informationStatic Timing Analysis for Nanometer Designs
J. Bhasker Rakesh Chadha Static Timing Analysis for Nanometer Designs A Practical Approach 4y Spri ringer Contents Preface xv CHAPTER 1: Introduction / 1.1 Nanometer Designs 1 1.2 What is Static Timing
More informationTechnical Article MS-2714
. MS-2714 Understanding s in the JESD204B Specification A High Speed ADC Perspective by Jonathan Harris, applications engineer, Analog Devices, Inc. INTRODUCTION As high speed ADCs move into the GSPS range,
More informationWhat is sync? Why is sync important? How can sync signals be compromised within an A/V system?... 3
Table of Contents What is sync?... 2 Why is sync important?... 2 How can sync signals be compromised within an A/V system?... 3 What is ADSP?... 3 What does ADSP technology do for sync signals?... 4 Which
More informationIC Mask Design. Christopher Saint Judy Saint
IC Mask Design Essential Layout Techniques Christopher Saint Judy Saint McGraw-Hill New York Chicago San Francisco Lisbon London Madrid Mexico City Milan New Delhi San Juan Seoul Singapore Sydney Toronto
More informationDe-embedding Techniques For Passive Components Implemented on a 0.25 µm Digital CMOS Process
PIERS ONLINE, VOL. 3, NO. 2, 27 184 De-embedding Techniques For Passive Components Implemented on a.25 µm Digital CMOS Process Marc D. Rosales, Honee Lyn Tan, Louis P. Alarcon, and Delfin Jay Sabido IX
More informationSDA 3302 Family. GHz PLL with I 2 C Bus and Four Chip Addresses
GHz PLL with I 2 C Bus and Four Chip Addresses Preliminary Data Features 1-chip system for MPU control (I 2 C bus) 4 programmable chip addresses Short pull-in time for quick channel switch-over and optimized
More informationFDTD_SPICE Analysis of EMI and SSO of LSI ICs Using a Full Chip Macro Model
FDTD_SPICE Analysis of EMI and SSO of LSI ICs Using a Full Chip Macro Model Norio Matsui Applied Simulation Technology 2025 Gateway Place #318 San Jose, CA USA 95110 matsui@apsimtech.com Neven Orhanovic
More informationLossless Compression Algorithms for Direct- Write Lithography Systems
Lossless Compression Algorithms for Direct- Write Lithography Systems Hsin-I Liu Video and Image Processing Lab Department of Electrical Engineering and Computer Science University of California at Berkeley
More informationLOW POWER DIGITAL EQUALIZATION FOR HIGH SPEED SERDES. Masum Hossain University of Alberta
LOW POWER DIGITAL EQUALIZATION FOR HIGH SPEED SERDES Masum Hossain University of Alberta 0 Outline Why ADC-Based receiver? Challenges in ADC-based receiver ADC-DSP based Receiver Reducing impact of Quantization
More informationVGA Controller. Leif Andersen, Daniel Blakemore, Jon Parker University of Utah December 19, VGA Controller Components
VGA Controller Leif Andersen, Daniel Blakemore, Jon Parker University of Utah December 19, 2012 Fig. 1. VGA Controller Components 1 VGA Controller Leif Andersen, Daniel Blakemore, Jon Parker University
More informationGood afternoon! My name is Swetha Mettala Gilla you can call me Swetha.
Good afternoon! My name is Swetha Mettala Gilla you can call me Swetha. I m a student at the Electrical and Computer Engineering Department and at the Asynchronous Research Center. This talk is about the
More informationAt-speed testing made easy
At-speed testing made easy By Bruce Swanson and Michelle Lange, EEdesign.com Jun 03, 2004 (5:00 PM EDT) URL: http://www.eedesign.com/article/showarticle.jhtml?articleid=21401421 Today's chip designs are
More informationA low jitter clock and data recovery with a single edge sensing Bang-Bang PD
LETTER IEICE Electronics Express, Vol.11, No.7, 1 6 A low jitter clock and data recovery with a single edge sensing Bang-Bang PD Taek-Joon Ahn, Sang-Soon Im, Yong-Sung Ahn, and Jin-Ku Kang a) Department
More informationNotes on Digital Circuits
PHYS 331: Junior Physics Laboratory I Notes on Digital Circuits Digital circuits are collections of devices that perform logical operations on two logical states, represented by voltage levels. Standard
More informationCombining Dual-Supply, Dual-Threshold and Transistor Sizing for Power Reduction
Combining Dual-Supply, Dual-Threshold and Transistor Sizing for Reduction Stephanie Augsburger 1, Borivoje Nikolić 2 1 Intel Corporation, Enterprise Processors Division, Santa Clara, CA, USA. 2 Department
More informationPresentation to IEEE P802.3ap Backplane Ethernet Task Force July 2004 Working Session
Presentation to IEEE P802.3ap Backplane Ethernet Task Force July 2004 Working Session Title: PAM-4 versus NRZ Signaling: "Basic Theory" Source: John Bulzacchelli Troy Beukema David R Stauffer Joe Abler
More informationMemory Interfaces Data Capture Using Direct Clocking Technique Author: Maria George
Application Note: Virtex-4 Family R XAPP701 (v1.4) October 2, 2006 Memory Interfaces Data Capture Using Direct Clocking Technique Author: Maria George Summary This application note describes the direct-clocking
More informationAt-speed Testing of SOC ICs
At-speed Testing of SOC ICs Vlado Vorisek, Thomas Koch, Hermann Fischer Multimedia Design Center, Semiconductor Products Sector Motorola Munich, Germany Abstract This paper discusses the aspects and associated
More informationAgilent Technologies Pulse Pattern and Data Generators Digital Stimulus Solutions
Agilent Technologies Pattern and Data Generators Digital Stimulus Solutions Leading pulse, pattern, data and clock generation for all test needs in digital design and manufacturing Pattern Generators Agilent
More informationECEN620: Network Theory Broadband Circuit Design Fall 2014
ECEN620: Network Theory Broadband Circuit Design Fall 2014 Lecture 12: Divider Circuits Sam Palermo Analog & Mixed-Signal Center Texas A&M University Announcements & Agenda Divider Basics Dynamic CMOS
More informationMassachusetts Institute of Technology Department of Electrical Engineering and Computer Science Introductory Digital Systems Laboratory
Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.111 - Introductory Digital Systems Laboratory How to Make Your 6.111 Project Work There are a few tricks
More informationPICOSECOND TIMING USING FAST ANALOG SAMPLING
PICOSECOND TIMING USING FAST ANALOG SAMPLING H. Frisch, J-F Genat, F. Tang, EFI Chicago, Tuesday 6 th Nov 2007 INTRODUCTION In the context of picosecond timing, analog detector pulse sampling in the 10
More informationMemory-Depth Requirements for Serial Data Analysis in a Real-Time Oscilloscope
Memory-Depth Requirements for Serial Data Analysis in a Real-Time Oscilloscope Application Note 1495 Table of Contents Introduction....................... 1 Low-frequency, or infrequently occurring jitter.....................
More informationLow-Power Decimation Filter for 2.5 GHz Operation in Standard-Cell Implementation
Low-Power Decimation Filter for 2.5 GHz Operation in Standard-Cell Implementation Manfred Ley, Oleksandr Melnychenko Abstract A low-power decimation filter for very high-speed over-sampling analog to digital
More informationTiming Error Detection: An Adaptive Scheme To Combat Variability EE241 Final Report Nathan Narevsky and Richard Ott {nnarevsky,
Timing Error Detection: An Adaptive Scheme To Combat Variability EE241 Final Report Nathan Narevsky and Richard Ott {nnarevsky, tomott}@berkeley.edu Abstract With the reduction of feature sizes, more sources
More informationNovel Low Power and Low Transistor Count Flip-Flop Design with. High Performance
Novel Low Power and Low Transistor Count Flip-Flop Design with High Performance Imran Ahmed Khan*, Dr. Mirza Tariq Beg Department of Electronics and Communication, Jamia Millia Islamia, New Delhi, India
More informationSignalTap Plus System Analyzer
SignalTap Plus System Analyzer June 2000, ver. 1 Data Sheet Features Simultaneous internal programmable logic device (PLD) and external (board-level) logic analysis 32-channel external logic analyzer 166
More informationWhat really changes with Category 6
1 What really changes with Category 6 Category 6, the standard recently completed by TIA/EIA, represents an important accomplishment for the telecommunications industry. Find out which are the actual differences
More informationML6428. S-Video Filter and 75Ω Line Drivers with Summed Composite Output. Features. General Description. Block Diagram Σ BUFFER.
www.fairchildsemi.com ML S-Video Filter and Line Drivers with Summed Composite Output Features.MHz Y and C filters, with CV out for NTSC or PAL cable line driver for Y, C, CV, and TV modulator db stopband
More informationEECS150 - Digital Design Lecture 18 - Circuit Timing (2) In General...
EECS150 - Digital Design Lecture 18 - Circuit Timing (2) March 17, 2010 John Wawrzynek Spring 2010 EECS150 - Lec18-timing(2) Page 1 In General... For correct operation: T τ clk Q + τ CL + τ setup for all
More informationA Modified Static Contention Free Single Phase Clocked Flip-flop Design for Low Power Applications
JOURNAL OF SEMICONDUCTOR TECHNOLOGY AND SCIENCE, VOL.8, NO.5, OCTOBER, 08 ISSN(Print) 598-657 https://doi.org/57/jsts.08.8.5.640 ISSN(Online) -4866 A Modified Static Contention Free Single Phase Clocked
More informationA FOUR GAIN READOUT INTEGRATED CIRCUIT : FRIC 96_1
A FOUR GAIN READOUT INTEGRATED CIRCUIT : FRIC 96_1 J. M. Bussat 1, G. Bohner 1, O. Rossetto 2, D. Dzahini 2, J. Lecoq 1, J. Pouxe 2, J. Colas 1, (1) L. A. P. P. Annecy-le-vieux, France (2) I. S. N. Grenoble,
More informationT ips in measuring and reducing monitor jitter
APPLICAT ION NOT E T ips in measuring and reducing Philips Semiconductors Abstract The image jitter and OSD jitter are mentioned in this application note. Jitter measuring instruction is also included.
More informationIs Now Part of To learn more about ON Semiconductor, please visit our website at
Is Now Part of To learn more about ON Semiconductor, please visit our website at www.onsemi.com ON Semiconductor and the ON Semiconductor logo are trademarks of Semiconductor Components Industries, LLC
More informationBelden IBDN System 10GX The next level of cabling performance
Belden IBDN System 10GX The next level of cabling performance by Paul Kish Director Belden IBDN Systems and Standards Introduction There is a new standard under development in the IEEE 2.3an task force
More informationL11/12: Reconfigurable Logic Architectures
L11/12: Reconfigurable Logic Architectures Acknowledgements: Materials in this lecture are courtesy of the following people and used with permission. - Randy H. Katz (University of California, Berkeley,
More informationA 400MHz Direct Digital Synthesizer with the AD9912
A MHz Direct Digital Synthesizer with the AD991 Daniel Da Costa danieljdacosta@gmail.com Brendan Mulholland firemulholland@gmail.com Project Sponser: Dr. Kirk W. Madison Project 11 Engineering Physics
More informationLMH0344 3Gbps HD/SD SDI Adaptive Cable Equalizer
3Gbps HD/SD SDI Adaptive Cable Equalizer General Description The 3Gbps HD/SD SDI Adaptive Cable Equalizer is designed to equalize data transmitted over cable (or any media with similar dispersive loss
More informationClock Generation and Distribution for High-Performance Processors
Clock Generation and Distribution for High-Performance Processors Stefan Rusu Senior Principal Engineer Enterprise Microprocessor Division Intel Corporation stefan.rusu@intel.com Outline Clock Distribution
More informationMeasurements and Simulation Results in Support of IEEE 802.3bj Objective
Measurements and Simulation Results in Support of IEEE 802.3bj Objective Jitendra Mohan, National Semiconductor Corporation Pravin Patel, IBM Zhiping Yang, Cisco Peerouz Amleshi, Mark Bugg, Molex Sep 2011,
More informationVLSI Chip Design Project TSEK06
VLSI Chip Design Project TSEK06 Project Description and Requirement Specification Version 1.1 Project: High Speed Serial Link Transceiver Project number: 4 Project Group: Name Project members Telephone
More informationClocking Spring /18/05
ing L06 s 1 Why s and Storage Elements? Inputs Combinational Logic Outputs Want to reuse combinational logic from cycle to cycle L06 s 2 igital Systems Timing Conventions All digital systems need a convention
More information