Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
EEWeb
EEWeb.com
EEWeb
Electrical Engineering Community
Contact Us For Advertising Opportunities
advertising@eeweb.com www.eeweb.com/advertising
1.800.574.2791
TA B L E O F C O N T E N T S
Barry Katz
SISOFT
Interview with Barry Katz - President and CTO of SiSoft
4 9
TABLE OF CONTENTS
Featured Products Optimizing Your Next Temperature Measurement Configuration Part 2: Thermocouples
BY ROBERT GREEN WITH KEITHLEY
An overview of several sources of temperature measurement error that are specific to thermocouples.
24 26 29 33
Visit www.eeweb.com
Barry
INTERVIEW
FEATURED INTERVIEW
Katz
SiSoft
President and CTO
How did you get into engineering and when did you start? I started my career at Digital Equipment Corporation (DEC) as part of the semiconductor group where I was part of a team responsible for writing system level signal integrity tools supporting the design of systems utilizing Alpha processors as well as other high-speed ASICs and supporting chipsets. In this role I was ultimately involved in the signal integrity of nearly every Alpha-based system ranging from workstations to highend servers working with signal integrity engineers to deploy signal integrity tools and methodology on their projects. Will you tell us about your extensive experience in developing tools to integrate and automate a wide range of signal integrity analysis processes? Throughout my career, I have always had a focus on automation. This started with automated postroute crosstalk analysis tools and has led to tools for performing preand post-route integrated signal integrity, timing, and crosstalk analysis. My most recent efforts have been focused on serial link analysis where my team and I are focused on signaling at rates up to 28Gbps. What have been some of your influences that have helped you get to where you are today? Ive been very fortunate in this area. My father is an EDA pioneer and entrepreneur and has provided me guidance and mentoring throughout my career. The team of grad students and advisors I had the opportunity to work with at Carnegie Mellon University were also influential. The work I did at DEC both before and after forming SiSoft as well as the projects I have had the opportunity to work on at Ericsson, Cisco, Intel, and many other companies have
Visit www.eeweb.com
INTERVIEW
been influential. Lastly, my family. My wife and two boys are my rock and support me every day. Do you have any tricks up your sleeve? I will tell you the key attribute that sets myself and my team at SiSoft apart. These days, the problems we SI engineers have to solve on a daily basis have become increasingly complex. Utilizing software tools to automate the solving of these problems has become a requirement. Many engineers will use these tools and accept the answers they get down to the 8th decimal point as gospel. My advice is dont. Build an intuition of what you expect the answers to be and test that intuition. Its almost like you are trying to predict the answer before you get it so that when you get the answer you can validate whether it makes sense. The other advice I would share with you is that whatever you do in life, maintain balance with work, family and hobbies. some of our major customers on emerging products targeting the 4G LTE market. These designs exhibit Do you have any note-worthy engineering experiences? Every year you continue to grow and you have great experiences. In looking back at all of these, I do look back to a few noteworthy experiences. Our DesignVision award for QCD was one of the great accomplishments of our team at SiSoft. Of particular note this year was the work highlighted with Ericsson that we did on Virtual Prototype Analysis where we demonstrated automated post-route simulation on thousands of links in a large backplane system. Also, recently, was the work we did in driving the new IBIS-AMI modeling standard which is key to modeling the equalization and clock recovery of SerDes devices. How do you continue to be actively involved with leading edge high-speed design? We are very lucky to have close collaborative relationships with our major system customers and semiconductor partners. Its these relationships and working closely on current and next generation projects that enable us to stay ahead of the curve. We are working on the leading edge and solving tomorrows problems today. The credit goes to my team who always rise to the challenge and the rapidly changing landscape that is signal integrity. I saw that you have played a key role in leading tool development efforts for SiSofts products. How have you done this? With any new product, standard, or capability, it always starts with a vision and recognition that
FEATURED INTERVIEW
What has been your favorite project? My favorite software development project has been the development of Quantum Channel Designer (QCD), SiSofts Multi-Gigabit Serial Link Analysis Tool, which my team has been working on for five years now and will continue to develop into the future. Technology is moving so fast in this area with emerging standards and higher data rates, it keeps the work very exciting. On the hardware side, I have been fortunate to have the opportunity to collaborate with
As a business, we are customer focused and will remain focused on close collaborative relationships with systems designers and their suppliers. We will continue to work with these customers and their semiconductor vendors to ensure high quality, highperformance, accurate models are available for simulation of their systems.
complexity on an extreme scale in terms of speed, size and density.
Visit www.eeweb.com
INTERVIEW
taking that vision to product or implementation is a team effort and requires focus. The team often spans myself, my staff as well as customers and partners. We have done this multiple times at SiSoft on both large and small scales. Driving the vision to a shared vision with ownership from the team and a collaborative mindset are the keys. Can you tell us more about SiSoft and the technology they are developing? When SiSoft was formed, we largely started as a consulting services company working with companies including DEC, Compaq, HP , and Intel. Ive always had a focus on automation so, to make our consultants more efficient, I hired my Dad and his former partner and we developed tools for automating a lot of the HSPICE simulation work we were doing. This led to the development of our first product SiAuditor and then Quantum-SI and Quantum Channel Designer. Will you tell us about SiAuditor, the first commercial software product that SiSoft launched? This was an exciting time for us at SiSoft. We were asked by Compaq to perform consulting services on a project that was spanning multiple groups and platforms with the potential for high-leverage between the SI analyses performed on these projects. I told them my team would do it, but that it had to be done using SiAuditor, a new product we were in the process of developing and needed a project to implement it on. I guaranteed our success and, given our track record with those teams, they rolled the dice. We all did. The rest is history. That project shipped on-time at speed. Compaq became the first user of SiAuditor, EMC followed shortly thereafter and we continued to roll. How did this help change the dynamic of SiSoft as a company? This was the first of many steps in transitioning from a consulting services company with a few tools and scripts to a bona fide software company with a strong consulting services arm. With technology changing at such a fast pace, what are a few of the main technologies that have helped your company in its growth and product development? Enabling Scalable Simulation Farms IBIS-AMI Modeling Standard Extraction we dont talk about simulating one net, we talk about simulation of all the nets. How does SiSoft continue to provide award-winning EDA simulation software, methodology training and consulting services for systemlevel high-speed design? Close collaborative relationships with our customers and semiconductor partners. SiSofts consulting services group enables deep relationships and our ability to drive leading edge requirements back into our core technology. What direction do you see your business heading in the next few years? As a business, we are customer focused and will remain focused on close collaborative relationships
FEATURED INTERVIEW
The need for greater and greater compute resources is ever present. Companies need a scaleable compute infrastructure that enables their engineering teams to perform the required simulations necessary for successful product design.
with systems designers and their suppliers. We will continue to work with these customers and their semiconductor vendors to ensure high quality, high-performance, accurate models are available for simulation of their systems. Whatever it takes to make them successful is where SiSoft will be. What are some new technologies we can expect to see from SiSoft in the near future? As speeds get faster, you can expect expanded capability from SiSoft in the area of Repeaters/ Retimers as well as support for emerging signaling standards. We are also looking to close the
Visit www.eeweb.com
INTERVIEW
communication gap between the SI engineers that are doing simulations and recommending system settings and the software engineers that are programming these settings in to registers in actual hardware. As simple as this may seem, many needless hours are spent debugging problems with this interface. What challenges do you foresee in our industry? Compute infrastructure, mindset, and neural processing. The fact is that things are only going to get faster. Higher speeds mean more modeling and more simulation. The need for greater and greater compute resources is ever present. Companies need a scaleable compute infrastructure that enables their engineering teams to perform the required simulations necessary for successful product design. There also needs to be recognition, by management, as to the importance of simulation and both active and passive model correlation and willingness to work with their EDA, interconnect and semiconductor vendors in collaborative ways to drive quality and capabilities. Simulation is not cheap, but it is worth it. Lastly, along the theme of neural processing, engineers need to build intuition through experience on trends or results they would intuitively expect in simulation and the lab and constantly be vigilant about validating their expectations. When there is a disconnect, drive it to closure. Video game engineering is not an option. Is there anything that you have not accomplished yet, that you have your sights on accomplishing in the near future? Ive been fortunate in my career to have the opportunity to do a lot of great things. As I reflect, I find that I enjoy both the engineering and business side of the equation. I look forward to the expansion of SiSoft and where that may lead us and I think longer term, I very much enjoy sharing my knowledge with other engineers so I see some sort of teaching in my future. On a personal level, like anyone else, I want to see my kids have their own successes in life and support them however I can.
FEATURED INTERVIEW
Visit www.eeweb.com
Avago Technologies AEDR-850x three channel reflective encoders integrate an LED light source, photo detector and interpolator circuitry. It is best suited to applications where small size and space matters! Applications include medical hand held devices, camera phones, wheel chairs, actuator, vending machine applications, just to name a few.
To request a free sample go to:
Built in Interpolator of 1x, 2x, and 4x High operating frequencies: 55 kHz at 1x interpolation Index gating -20C to 85C
1x, 2x and 4x via external pinouts Operating frequencies can be increased by external interpolator pinouts by maximum of 4x Options available for both gated and ungated versions Industrial application capable
www.avagotech.com/motioncontrol
TECHNICAL ARTICLE
Abstract
While not long ago systems had only several serial links, its now becoming common for systems to include hundreds and even thousands of such links. This paper describes the development and analysis of a large system with thousands of serial links. Due to the systems size and complexity, the design team invested in a multiyear effort to build and qualify a virtual environment capable of both verifying connectivity and simulating any and all of the channels. Problematic channels with incomplete system-level connections, poor eye openings, or high BER are quickly identified. Performance limiters such as inherent discontinuities (and their associated resonances), and Tx/Rx equalization imbalance are found and examined in detail. The virtual system is also used to guide design choices such as layer stacking, via construction, back-drilling, and trace/connector impedances. Processes to optimize and select equalization choices are also described.
as their sensitivity to system variables. While exhaustive analysis of link performance against manufacturing variations is performed in [1] in a pre-layout context, this paper illustrates the advantages of an efficient post-layout (albeit, pre-hardware) environment capable of analyzing thousands of serial links in a reasonable amount of time. Specific link-level advantages include the ability to: (a) confirm system-level connectivity, (b) balance and optimize equalization, (c) quantify design margin against performance targets, and (d) identify discontinuityinduced resonances not found by pre-layout analysis. The fact that there is typically no other way to perform these tasks further underscores the value of analyzing a virtual system as detailed herein; a capability commonly referred to as Virtual Prototype Analysis or VPA. The vision of VPA is to provide the ability to quantify electrical-layer performance of any or all of thousands of system-level links in a reasonable amount of time, prehardware. The authors are part of a larger team that has invested in a multi-year effort to enable the capabilities described. While post-layout analysis is not a new concept in the field of Signal Integrity (SI), high-speed serial links have brought significant changes in the way SI
1. Introduction
Pre-layout simulation has been demonstrated in [1] to explore the feasibility of proposed serial links as well
Visit www.eeweb.com
TECHNICAL ARTICLE
analysis is performed. In addition to the ability to simulate millions of bits, identifying and removing resonances and balancing equalization at the system level are just some of the significant process steps introduced by serial links and enabled by VPA. Once assembled, the virtual system further enables the examination of performance gains against design options. Since parameters such as backdrilling, connector choice/impedance, via construction, and layer ordering can be easily adapted, the systemlevel improvement of these cost-sensitive choices can be quantified without time-consuming hardware fabrication cycles. Innovations achieved in the past decade have enabled VPA to be applied to thousands of serial links. Necessary model formats and simulation techniques (both convolution-based and statistical) capable of resolving 10-12 Bit Error Rates (BERs) and smaller have been well-documented elsewhere [2-5]. Instead, this paper addresses a significant barrier to deploying serial link analysis in a post-layout context: the ability to rapidly solve and simulate Printed Circuit Board (PCB) vias with sufficient accuracy. As such, attention is given to detailing recent advances made in via modeling. Furthermore, as mainstream data rates push above 5 Gbps there is little eye opening left at both the Rx pin and die [6]. At these data rates it is imperative to model and measure signal performance after Rx equalization is applied, as is now required by the majority of 3rd generation serial standards [7]. This paper illustrates VPA using a system of approximately 30 large-scale PCBs interconnected through a backplane fabricated with over 50 metal layers containing thousands of serial links ranging in length from about 12 to 36. While the scale of this system is large, the concepts presented can be deployed in any system utilizing advanced high-speed serial link technologies. Though data rates will continue to increase throughout this decade, the dominant trend will be the parallelization of serial interfaces. Indeed, the presence of thousands of serial links in a single system suggests the need to significantly scale our serial link analysis in width rather than frequency. The concepts presented herein are consistent with this theme.
2. Enabling Technologies
Beyond basic serial link analysis capability, Virtual Prototype Analysis (VPA) of thousands of serial links at data rates beyond 5 Gbps has necessitated innovations detailed in this section. Specifically, these include: 1. The ability to quantify signal performance after the application of Rx equalization 2. Rapid generation and simulation of hundreds of via models with sufficient accuracy 3. Expanded capacity, throughput, and data mining capabilities
TECHNICAL ARTICLE
Tx
CHANNEL
Rx
CTLE
DFE
Clock Recovery
Rx Data Latch
Gigabit Receiver
This methodology shift has brought about changes in both Test & Measurement (T&M) equipment as well as simulators and models. Essentially, it is now necessary for T&M equipment to include simulators capable of post-processing a waveform captured at a measurement location that can be probed (such as the Rx pin) [8]. Similarly, simulators now handle increasingly complex Rx models capable of modeling the performance of the signal after it is recovered by Rx equalization. Whether these are models of component behavior provided by a
Visit www.eeweb.com
10
TECHNICAL ARTICLE
silicon vendor or modeled reference behavior provided by a tool or standard, they are increasingly available as executables in the AMI (Algorithmic Modeling Interface) format defined in the IBIS 5.0 specification [13]. To test the via length hypothesis advanced in [9], we extracted the insertion phase for the measured data for [10]. We also calculated the insertion phase predicted by our closed form model equations with the assumption that the dielectric constant is isotropic. Insertion phase is related to dielectric constant in that: 1. The dielectric constant determines the propagation velocity in the structure. 2. The group delay of the structure is the physical length of the structure divided by the propagation velocity. 3. The insertion phase is the integral of the group delay with respect to the angular frequency. As shown in Figure 2 and the modeled and measured phase data from the Simple Via Experiment [10], the data supports the hypothesis of an isotropic dielectric constant rather than an anisotropic one because the dielectric constant determined from propagation in the X and Y directions is consistent with the delay in the Z direction.
0.4 0.2
TECHNICAL ARTICLE
0 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 2E+09 4E+09 6E+09 8E+09 1E+10 1.2E+10 1.4E+10 1.6E+10 1.8E+10 2E+10
Frequency
It therefore appears that a different hypothesis is required to explain the unexpected measured electrical length of vias. To create the via model used herein, we measured pad to pad S parameters for numerous differential traces on an unpopulated backplane. Each trace measured consisted of a differential via, a differential backplane trace, and another differential via. The analysis approach was to convert the measured S parameters to Time Domain Reflectometry (TDR) plots and measure the electrical length of the via at the beginning of the trace in both single ended and
Visit www.eeweb.com
11
TECHNICAL ARTICLE
differential modes. Since the measurement was made from pad to pad, the launching via was the very first element in the electrical path, so relatively precise measurements could be made. The measurement bandwidth was 20GHz, producing an effective TDR rise time of 17pS. The measured electrical length vs. physical length is shown in Figure 3. In this Figure, the single ended measurements are shown using yellow symbols and the differential measurements are shown using red symbols. A line is extended through the measured differential values.
70 60 50
TECHNICAL ARTICLE
40 30
?
-0.0500
20 10 0 0.0000
-0.1000
0.0500
0.1000
0.1500
0.2000
0.2500
0.3000
The line in Figure 3 was drawn through the differential mode data points because they show less spread than the single ended data points, and also because differential mode is not affected by the ground return impedance. The slope of the line in Figure 3 corresponds to a dielectric constant of 3.61, compared to a dielectric constant of 3.41 for the traces on this particular PC board. This difference is within experimental error. Figure 3 indicates that the increased length appears to be a constant length offset of about 0.05 rather than a decreased propagation velocity. This offset can be explained by considering the path the current must follow. The via and trace depicted in Figure 4 consists of a top pad (typically at the top surface of the board), via barrel, bottom pad (typically at an internal routing layer), and exit trace (routing from the edge of the pad to the edge of the antipad).
The current must follow the surface of the via structure, as depicted in Figure 4. For the structures we measured, the physical distances across the upper and lower surfaces of the top pad, across the upper surface of the bottom pad, and along the exit trace add up to 0.05, the physical length shown as an offset in Figure 3. This is clearly a simplification of a complex set of electromagnetic field behaviors, but the data and physical reasoning support the hypothesis that the current paths across the top and bottom pads of a via measurably increase its electrical length. The resulting via model correlates well with measured data, making this approximation a useful one. Equally important, this via model can be calculated quickly and simulated efficiently, which are fundamental requirements of providing effective VPA. Figure 5 and Figure 6 are representative examples of the model correlation obtained using post-route extraction with closed form equations and a via model that explicitly models the electrical length of the pad. The measured results from both ends of the trace (same vias on each end) are shown in blue and the model results are shown in red. In Figure 5, the time differences between the model and the measurements are much less than the 17pS rise time of the data, and the impedance differences are less than the impedance difference between the two ends of the trace (e.g., as measured at 400pS). Thus, the TDR data is well within experimental error. Similarly in Figure 6, the model matches the frequency domain ripples of the loss characteristic precisely and even at 20GHz the difference between model and measurement is less than 3dB. The results shown are for a bare backplane (i.e., the fabricated PCB without connectors); however, the
Visit www.eeweb.com
12
TECHNICAL ARTICLE
correlation is almost as good when connectors and cards are added. As demonstrated by Figures 5 and 6, this via model is more than adequate for analyses up to 10 Gbps, and provides useful results up to 20 Gbps. In the future, it could be improved by providing direct computation of the few remaining heuristic element values, and by explicitly calculating the coupling of vias to each other and to the ground plane cavity. enough to provide useful feedback to the PCB layout team. This means taking periodic PCB database revisions, performing simulations, interpreting results and providing updated physical design rules within 48 hours. Allowing a day for interpretation of results and updating design rules, that leaves a day for setting up and performing the post-route simulation run. The simulation tool must be able to extract post-route topologies automatically, generate via models on-the-fly, and construct connector models from slice data provided by the connector manufacturer. Given the size of the simulation task, turnaround time is reduced through the combination of Statistical modeling techniques (a Statistical simulation will run in seconds, while its Time-Domain equivalent may take 50-500x longer) and acceleration though parallel processing (server farms). The simulation environment must automate the process, since complexity is high and repeatability is key to ensuring quality, as VPA will be repeated many times throughout the course of the design. Once the simulations are complete, it can be difficult to see the forest for the trees. Its not enough to produce massive sets of raw simulation data, since the data needs to be analyzed and updated design rules created
TECHNICAL ARTICLE
90.0
Ohms
80.0
70.0
60.0
0.0
100.0
200 0.0
300.0
400.0
500.0
600.0
Time (ps)
0.0
-5.0
-10.0
-15.0
DB
-20.0 -25.0 -30.0 -35.0 0.0 5.0 10.0 15.0 20.0
Hertz (GHz)
Visit www.eeweb.com
13
TECHNICAL ARTICLE
quickly. Productivity is increased when simulation output includes key performance metrics like eye height and width for each simulation case, associated with the physical characteristics of each channel. Organizing the results allows the designer to scan the data for best and worst case conditions, as well as overall trends and outliers that will require a more thorough investigation. There are two important aspects to post-route simulation data mining: automated reporting and interactive drilldown of results.
TECHNICAL ARTICLE
3.1 Connectivity
The value of the virtual systems ability to confirm systemlevel connectivity or, interconnect integrity can not be overstated. While schematic/PCB netlist forward and backward annotation capabilities ensure connectivity Automated reporting provides the first level of analysis at a single PCB level, the virtual system confirms connectivity at the system level. Furthermore, as new and includes reports like: PCBs are developed or existing PCBs are revised they Electrical integrity checks: connectivity, polarity are verified in the virtual system to confirm interconnect swizzles, Tx/Rx compatibility integrity and link performance prior to fabrication. Physical characteristics: Net lengths, lengths by layer, In other words the virtual system can serve as a testbench that verifies that new PCBs, perhaps developed potential resonant conditions by remotely located teams in different geographies and Standards compliance: insertion loss, return loss, time zones, will perform as expected when plugged into the real backplane. crosstalk ratio, eye mask pass/fail Performance metrics: Eye height, eye width, BER, Interconnect integrity can be verified physically, electrically, and operationally, and examples of each optimized EQ settings method are provided below on a typical virtual system. Automatic identification of best/worst case cases and Physical connectivity for hundreds of differential pairs associated design variables is confirmed in the Figure 7 which plots the combined Once areas are identified for further investigation, driver to receiver length of signals spanning three PCBs problems are isolated by drilling down into individual (card, backplane, card). Thousands of signals are simulation cases and performing what-if simulations shown on the X axis versus each signals length on the Y to explore potential solutions. Important capabilities axis in inches. The distribution of backplane net lengths include: is shown in red, and the corresponding total length of Correlate design metrics as a function of design each pair is plotted in green. The fact that total length always exceeds backplane length provides the first level variables or other performance metrics of confirmation that all signals connect from the Tx on Correlate any waveform at any node in the network one card across the backplane to the Rx on another card. Markers further quantify the distribution of PCB lengths, (including physically inaccessible nodes) in this case revealing there are 518 nets with backplane Overlay compliance masks on simulation plots lengths between 8 and 16, 318 nets with lengths Edit post-route topologies to support interactive what-if between 16 and 24, and so on. analysis Providing these capabilities in a way that provides reasonable turnaround with currently available computers is a tall order, but this is the problem that needs to be solved to enable full-system analysis.
Visit www.eeweb.com
14
TECHNICAL ARTICLE
40.0 38.120 35.0
30.0
Length
25.0
24.0
Plotting mask margins against channel loss in Figure 9 reveals that higher-loss channels slightly violate the mask (at very low frequencies, per the plot above right). In the plot below, red is insertion loss margin to mask and blue is the margin of the fitted attenuation to the mask. With a good sense of electrical connectivity, we next examine operational connectivity by confirming that all nets produce an eye opening with typical equalization as shown in Figure 10 (eye height in red and width in blue). The eye openings recorded have a 10-12 probability of
2.0 1.50 1.0
TECHNICAL ARTICLE
10.0
5.0 0.20
0.326k 0.40
0.845k 1.0
1.163k 1.20
Row (k)
While the data above suggests all nets are connected across the boards, we are only viewing their combined length. The plots in Figure 8 go one step further and show the differential insertion loss (left) and their associated fitted attenuation (right, an averaging of the insertion loss) of thousands of channels against typical industry masks in black. This confirms the nets are connected electrically, and demonstrates the range of system-level insertion loss we can expect to compensate for at various frequencies of operation.
DB
0.50 0.0 -0.50 0.0
-8.0
-7.0
-6.0
occurring, and hence are related to a certain BER as defined by the Rx characteristics. As two of the channels show no eye height or width (extreme right Figure 10) its possible these channels are not operating or simulating correctly. Various methods shown below exist to build confidence that those nets are simulating and producing meaningful results. At
170.0 160.0 140.0 120.0 Volts (mV) 100.0 80.0 60.0 40.0 20.0 0.0 0.20 0.40 0.60 0.80 1.0 1.20 80.0 70.0 60.0 Time (ps) 50.0 40.0 30.0 20.0 10.0 0.0
left in Figure 11 we overlay hundreds of bathtub curves and confirm that two nets fail to yield an eye below a probability of 10-11 (highlighted in red), and hence show
Visit www.eeweb.com
15
TECHNICAL ARTICLE
Volts (mV) Volts (mV)
0.0
150.0 90.0
100.0
-100.0
Additional operational connectivity tests include plotting all pulse responses in Figure 12 (at left) and eye parameters at 10-3 probability (at right, height in red, width in blue). The pulse responses reveal the range of channel delays and their corresponding attenuation,
80.0
7.0
8.0
Time (ns)
Row (k)
Figure 12: Operational Connectivity - Pulse Responses and 10-3 Eye Parameters
With connectivity and functionality confirmed, we are in position to extract design margins and resolve poorly performing channels prior to fabricating hardware.
as well as a few nets with p/n reversal or swizzle (i.e., those nets with a negative-going pulse response). The 10-3 probability simulations demonstrate much wider eye openings than the 10-12 plot shown in Figure 10, which further confirms VPA is operationally functioning as expected.
Recognizing that the worst eye heights also have the worst widths, the plots in Figure 14 allow us to study performance by plotting eye height on the X axis against system variables on the Y axis. The plot at left shows eye height versus backplane layer revealing that, in general, signals on deeper layers perform worse than those on upper layers. The plot at right shows eye height versus connector row, revealing that the worst performing signals utilize row 6 in the connector.
Visit www.eeweb.com
16
Time (ps)
a closed eye at 10-12 probability in Figure 10. At right the statistical eye for these two nets confirms they do produce an eye at lower probability, yet - for reasons to be explained later - are less stable than the others.
200.0
250.0 110.0
TECHNICAL ARTICLE
100.0
200.0 100.0
TECHNICAL ARTICLE
150.0
Volts (mV) Volts (mV)
150.0
100.0 50.0
100.0 50.0
corners as annotated above the graph. Eye heights are plotted at three different probabilities (implying three different BERs) delineated in red, green, and blue. This
160.0 140.0 150.0 180.0 160.0 140.0 120.0 100.0 100.0 80.0 50.0 60.0 40.0 20.0 0.0
TECHNICAL ARTICLE
Volts (mv)
Figure 13: Eye Height vs. Eye Width, with Signal Direction
50.0 40.0 6.0 5.0 4.0 3.0 2.0 1.0
30.0
40.0
50.0
60.0
70.0
80.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
40.0
50.0
60.0
70.0
80.0
Figure 15: Channels with Least Margin Consistent Across Signal Lengths
Unitless (y)
Unitless
plot indicates that there is a significant change in design margin as we vary silicon and interconnect corners, as is evident by viewing any vertical slice in the Figure.
20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0
20.0
40.0
60.0
80.0
100.0 120.0
140.0 160.0
Figure 14: Eye Height vs. Backplane Layer and Connector Row
The data above allows us to hypothesize that the worst case nets in this length bundle are those with the following attributes: 1. signals driven in the blue direction in Figure 13 2. signals routed on the lowest (deepest) backplane layers
Volts (mv)
3. signals connected through connector row 6 To test this hypothesis against signals with short, medium, and long backplane lengths in Figure 15 we plot all signal bundle eye height/widths below in blue and superimpose the subset of nets with these attributes in red. These plots confirm that nets with these characteristics describe the nets with the poorest performance. Thus far we have demonstrated the value of extracting thousands of performance estimates for all signals in the system, and how these datapoints can be filtered to identify which signals perform poorly. The process shown illustrates the value of quantifying specific netlevel performance across the system during a stage of development when design adjustments can be more easily made. The specific phenomenon identified above will be explained in more detail in the next section on discontinuity-induced resonance. Link performance should also be plotted against both corner-case behaviors and decreasing BER, as shown in Figure 16. In this Figure, nine corners are created by combining three silicon corners with three interconnect
300.0
200.0
100.0
Row
optimized) channel measured from the Tx to Rx. While much of the channels response is fairly flat around 100 Ohms impedance, the backplane connector locations are quite visible at ~1 nS and ~4 nS. Even though the impedance of the connector itself is roughly 100 Ohms, the connector vias create the two lower impedance dips on either side of the connector transition.
The presence of lower-impedance vias on both sides of the connectors cause a signal disturbance whose length is defined by the delay through the connector. The discontinuity causes some portion of the waves traveling
Visit www.eeweb.com
17
TECHNICAL ARTICLE
through the connector to reflect off the 2nd via and return to their source. As the energy is reflected back, some
110.0
included in the analysis, since they impose additional time delay as well.
TECHNICAL ARTICLE
100.0
Ohms
90.0
60.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
Time (ns)
Figure 17: Measured (red) and Simulated (green) Channel TDR Showing Discontinuities
portion reflects off the 1st via back into the connector, where again a portion will reflect back off the 2nd via and so on. This disturbance of defined length and the energy trapped in the structure can be referred to as a discontinuity-induced resonance or standing wave. To better understand the discontinuities related to the connector vias, we must first quantify the delays through the various rows of the connectors. Placing a model of each row in the center of an ideal transmission line the Time Domain Transmission (TDT) plot in Figure 18 reveals the delays for each row, which are tabularized in Table 1. As Table 1 shows, the standing wave (or roundtrip) frequencies suggest all rows longer than row 3 are within the range of 6 Gbps frequencies. However this explanation is focused on the connector only, while the problematic discontinuities also include via structures on both sides of the connector. These vias must be
1.20 1.0 0.80
A study of the various connector vias reveals that longest via delays are about 20pS on the cards and 60pS on the backplane. Combining these and counting half the via transit time to be within our problematic structure, in Table 2 we find that the row 6 delay is ~2 UI and hence energy from double-bit (and a portion of singlebit) patterns will resonate and degrade performance, as shown previously in the section on design margins. Further analysis shows that if a connector structure of this length is used it is helpful if at least one of the discontinuities on either side is removed. This lengthens the structure giving it more loss and therefore less pronounced standing waves. Solutions such as lower impedance traces and connectors and/or the use of specialized via structures can be used to smooth the discontinuities surrounding a connector. While pre-route analysis endeavors to exhaustively explore many connector/via
combinations and other channel discontinuities, VPA has the potential to isolate combinations that may not have been comprehended previously prior to assembling the virtual prototype of the larger system.
Volts (V)
0.60 0.40 0.20 0.0 2107.0ps 2100.0 2150.0 2200.0 2250.0 2273.0ps 2300.0
Time (ps)
Knowledge of the data stream enables a Tx to implement pre-cursor equalization, while an Rx can implement only post-cursor equalization. As such, we can hypothesize
Visit www.eeweb.com
18
TECHNICAL ARTICLE
it would be best to let the Rx handle the post-cursor to preserve as much Tx amplitude swing as possible. Conversely, since the Tx is the only device capable of performing pre-cursor equalization, it should be tasked with handling pulse spreading during those bit times. The Tx may also be tasked with post-cursor equalization if the system requires more than the Rx can provide. Using simulation to test this hypothesis, Figure 19 examines the performance of a typical net when sharing post-cursor equalization between the Tx and Rx. All eye openings are plotted at the Rx Latch with a small amount of Tx pre-cursor (-10%) and Rx CTLE applied. In the eye at left, the Tx has fully equalized the signal leaving little room for contribution from the Rx DFE. At center we evenly divide post-cursor equalization between the Tx (17%) and the Rx (18%). In this scenario the Tx eye is increased ~100mV suggesting that the extra energy delivered by the Tx is helpful. At right we see the eye shape when the Tx has only pre-cursor equalization and the Rx DFEs post-cursor is at -32%. This scenario increases the eye height another 170mV at the Rx Latch, and the eye width has improved as well. From scenario 1 to 3, the eye opening has improved by 100% simply by deploying a system-level view of channel equalization. Though the eye openings above have increased substantially, this is at the expense of additional voltage swing (and hence power, noise, and crosstalk) in the system. The plots in Figure 20 reveal that, for the same three scenarios above, the pin-level performance has changed from over-equalized to under-equalized as expected. While the third scenario yields the best internal eye, it fails to deliver an acceptable eye at the pin. However, as newer serial link standards suggest, the relevance of pin-level eye performance continues to decrease.
Post:
1.0
function in grey. While letting the Tx handle equalization (blue) successfully flattens the channel response, this is at the expense of reducing the low frequency amplitude. As such, less energy is delivered to the Rx for it to process and equalize. In green and red we see that allowing the Rx to assist in the equalization provides a more ideal response (~flat line at 0dB) up to the dashed vertical line. Green shows how sharing the post-cursor splits the difference in the low frequency response, yet the frequencies just below Fc are almost the same as when the Rx handles the post-cursor equalization (red).
0.80 0.60 0.40 5.5E-3 2E-2 0.80 0.60 0.40 3E-3 1.3E-2 0.80 0.60 0.40 2.7E-3 1.2E-2
TECHNICAL ARTICLE
Probability
Probability
1.8E-7
2.8E-8
2.2E-8
Time (ps)
Time (ps)
Time (ps)
Eye:
141mV/60Ps
150mV/60pS
56mV/21pS
Figure 20: Eye Shape Before Rx Equalization, Varying Post-Cursor Between Tx and Rx
The pulse response of the four options is shown at right, using the same color scheme. The plot in Figure 22 examines if the behaviors shown above are consistent across a larger number of nets. Eye heights for 50 nets using the three equalization options above are plotted using the same blue, green, and red colors. In this plot, grey represents the eye height at the die prior to equalization. The boxes at right quantify the band of eye heights for each of the equalization scenarios. While there is some overlap in the bands, the trend shown above is still true for this larger sampling of nets.
0.0 0.40 -10.0 0.20
Volts (V)
-20.0
DB
Tx=-35%, Rx=-2%
2.1E-2 1.0
Tx=-17%, Rx=-18%
2.1E-2 1.0
Tx=0%, Rx=-32%
1.3E-2
Fc 2875GHz
0.0 2.0 4.0 6.0 8.0 10.0
1.50
5.9E-3
1.50
5.7E-3
1.50
1.5E-2
Hertz (GHz)
Probability
Time (ps)
0.0
4.6E-4
0.0
4.3E-4
0.0
1.8E-3
Figure 21: Transfer Functions and Pulse Responses Comparing Equalization Schemes
Probability
Probability
Volts (V)
Volts (V)
Volts (V)
Time (ps)
Time (ps)
Time (ps)
Eye:
248mV/72pS
354mV/73pS
513mV/94pS
Figure 19: Eye Shape After Rx Equalization, Varying Post-Cursor Between Tx and Rx
The transfer functions for the three scenarios above are shown in Figure 21 against the unequalized transfer
Next we use VPA to examine the same three scenarios across thousands of nets in the system. As a result, over 20,000 eye shapes were derived and processed to 10-12 probability requiring ~20 hours on two CPUs. As such each simulation required about 3 seconds, or 6 seconds per CPU. An 8-CPU system could complete the task in less than 4 hours.
Visit www.eeweb.com
19
Probability
Volts (V)
Volts (V)
Volts (V)
TECHNICAL ARTICLE
Figure 23 shows the results of these simulations. At left are the distributions of eye heights (red) and widths (blue) for each simulation scenario, and at right are the corresponding eye height/width scatter graphs. As all axes scales are the same (except the blue width values
600.0 500.0 y1:(572.50mV) y2:(354.70mV) dy:(217.80mV) y1:(426.70mV) y2:(231.80mV) dy:(194.90mV) y1:(721.10mV) y2:(146.60mV) dy:(124.50mV)
Volts (mV)
700.0 600.0 500.0 150.0 400.0 300.0 200.0 100.0 0.0 0.0 0.50 1.0 15.0 2.0 2.50 3.0 3.5.0 0.0 100.0 200.0
TECHNICAL ARTICLE
Volts (mV)
150.0 400.0 300.0 200.0 100.0 0.0 0.0 0.50 1.0 15.0 2.0 2.50 3.0 3.5.0 0.0 100.0
Volts (mV)
500.0
500.0 400.0 300.0 200.0 100.0 0.0 0.0 20.0 40.0 60.0 80.0 100.0
50.0
Row (k)
Volts (mV)
400.0
300.0
Volts (mV)
200.0
100.0
500.0 400.0 300.0 200.0 100.0 0.0 0.0 20.0 40.0 60.0 80.0 100.0
10.0
20.0
30.0
40.0
50.0
60.0
50.0
ID (Compressed)
at top left), scenario 3 (Rx post-cursor) can be easily seen to perform the best.
700.0
Row (k)
While system-level equalization tuning at a conceptual level is not new, the ability to deploy the tuning with sufficient accuracy on a net-by-net basis is one of the benefits of VPA. In addition to using VPA to explore prehardware design trade-offs (see next section below), equalization exploration and tuning can also assist with firmware updates in a post-hardware context.
600.0
Volts (mV)
150.0 400.0 300.0 200.0 100.0 0.0 0.0 0.50 1.0 15.0 2.0 2.50 3.0 3.5.0 0.0 100.0
Volts (mV)
500.0
500.0 400.0 300.0 200.0 100.0 0.0 0.0 20.0 40.0 60.0 80.0 100.0
50.0
Row (k)
4. Design Trade-offs
A virtual system enables comprehensive and rapid evaluation of design options that would be too expensive and time consuming to perform on physical hardware. Examples of items that can be easily modified and investigated using VPA include: PCB layer swaps, via back-drilling, back-drill stub length / number of depths, connector/route impedance variations, length variations and constraints, data rate, alternate PCBs, and complex via structures. A few of these trade-offs are illustrated in this section.
Card_A on the Y axis shown to be performing better than Card_B on the X axis. As links are typically routed in a non-interleaved fashion for crosstalk mitigation, Tx and Rx signals tend to be on different routing layers. As such, we can also examine how performance changes when we swap signal layers (and hence transmission direction depth) within PCBs. In Figure 25 we examine eye heights for layer stacking options on a backplane at left and a plug-in card at right. In both plots signaling directions are represented by red and blue datapoints, and dots on the center diagonal imply performance is unchanged. Plots like this typically reveal that performance improves in one direction at the expense of a performance decrease in the other direction. In the Figure the blue/red performance change in the backplane (at left) are similar suggesting the layer swap is not helping overall, while on the card (at right) blue has gained more than red has lost suggesting that this swap is worth considering to improve overall systemlevel design margin.
Visit www.eeweb.com
20
TECHNICAL ARTICLE
4.2 Via Back-drilling
While back-drilling is common on thicker backplanes, another design option easily quantified by VPA is the performance benefit that comes from also back-drilling
90.0 70.0 90.0
160.0
measured at the reduced data rate on the Y axis. As such points above the line indicate an improvement in performance/margin, which is shown by the markers to be typically 37mV in height and 10 pS in width which translates to ~30% increase in design margin for an 8% reduction in data rate.
90.0 140.0 80.0
TECHNICAL ARTICLE
70.0
Time (ps)
60.0 50.0 40.0 30.0 30.0 40.0 50.0 60.0 70.0 80.0
Time (ps)
60.0
Volts (mV)
Time (ps)
70.0
60.0
50.0
Figure 24: Eye Width vs. PCB Swapping, Backplane (left) and Plug-in Card (Right)
50.0
60.0
70.0
80.0
90.0
vias on thinner plug-in cards. Figure 26 plots eye height (left) and width (right) changes with (X axis) and without (Y axis) back-drilling on the plug-in card. As such, all points below the diagonal indicate an improvement due to back-drilling. The plots below reveal that all signals either stay the same or improve some times as much as 50%. Note that while signals received by the card (in red) improve with back-drilling, improvement is even more dramatic for signals transmitted from the card (blue).
140.0 120.0 100.0 80.0 60.0 40.0 40.0 20.0 30.0 40.0 60.0 80.0 100.0 120.0 140.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 160.0 140.0 120.0 Volts (mV) 100.0 80.0 60.0
Volts (mV)
Volts (mV)
Time (ps)
140.0
Figure 25: Eye Height vs. PCB Layer Swapping, Backplane (left) and Plug-in Card (right)
The plots reveal typically 20% improvement in margin on blue nets, yet more importantly improvement generally increases on signals with the least amount of margin.
40.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 50.0 60.0 70.0 80.0 90.0
Figure 27: Eye Height and Width Improvements, 8% Data Rate Reduction
TDR plots of the original channel (lighter shades) and a channel with 300% eye improvement (in darker shades) are shown in Figure 29. Red is the direction of signal flow and blue shows the channel looking from Rx to Tx. Note that the impedance changes have not necessarily removed all the discontinuities, yet have certainly lengthened the distance between them.
Visit www.eeweb.com
21
TECHNICAL ARTICLE
5. Summary
In the coming decade, serial link analysis will be challenged not only by issues of frequency but also by issues of scale; trillions of bits will be processed
90.0 100.0 80.0 80.0
for their continued help in advancing channel design and improving the connector footprint. Additional thanks to Orlando Bell at GigaTest Labs for consistently delivering high quality measured data. Without the efforts of these and other people this work would not have been possible.
TECHNICAL ARTICLE
References
[1] Simulation Techniques for 6+ Gbps Serial Links Telian, Camerlo, Kirk, DesignCon 2010 http://www.siguys.com/ resources/2010_DesignCon_6GbpsSimTechniques_ Paper.pdf [2] New Techniques for Designing and Analyzing Multi-GigaHertz Serial Links, Telian, Wang, Maramis, Chung DesignCon 2005, http://www.siguys.com/ resources/2005_DesignCon_New_MGH_Techniques_ ISP_CA_PCIe_SATA.pdf [3] Introducing Channel Analysis for PCB Systems, 2004 Webinar, http://www.siguys.com/resources/2004_ Webinar_Introducing_Channel_Analysis.pdf [4] New Serial Link Simulation Process, 6 Gbps SAS Case Study, Telian, Larson, Ajmani, Dramstad, Hawes, DesignCon 2009 Paper Award, http://www.siguys.com/ resources/2009_DesignCon_6Gbps_Simulation_Paper. pdf [5] Adapting Signal Integrity Tools and Techniques for 6 Gbps and Beyond, Telian http://www.siguys. com/resources/2007_CDNLive_Adapting_SI_Tools_ for_6Gbps+.pdf
TIme (ps)
Volts (mV)
70.0
60.0
60.0
40.0
50.0
20.0
Row (k)
Figure 28: Eye Height and Width Variation vs. Connector/Trace Impedance Permutations
across thousands of channels. While traditional SI methodologies had succeeded in moving post-route analysis to more of a verification step [12], this paper has demonstrated how VPA greatly assists serial link development in areas of system-level equalization tuning and design trade-offs. Furthermore, due to the thousands of channel permutations analyzed, VPA has potential to identify performance limitations caused by unexpected and/or unavoidable discontinuities that may have been
120.0
110.0
Ohms
100.0 90.0
80.0 70.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Time (ns)
overlooked by pre-route analysis. While challenges related to capacity, throughput, and accuracy must be solved to assemble an efficient virtual prototype system, this paper has demonstrated one approach taken and some of the benefits associated with overcoming these barriers.
[6] Signals on Serial Links: Now you see em, now you dont. DesignCon07 Article, http://www.siguys.com/ resources/2007_Article_SignalsOnSerialLinks.pdf [7] SAS Specification, SAS-2, Project T10/1760-D, Rev 16, 18 April 2009, see also USB 3.0 etc. [8] For example, see LeCroy 10/12/11 Webinar USB 3.0 Electrical Compliance Testing http://www.lecroy. com/support/techlib/webcasts.aspx?capid=106&mid= 528&smid=663 [9] Practical Analysis of Backplane Vias for 5 Gbps and Above, E. Bogatin, L. Simonovich, C. Warwick and S. Gupta, paper 7-TA2, DesignCon2009, February 3, 2009. [10] A Simple Via Experiment, Chong Ding, Divya
Acknowledgements
The authors wish to thank Shashi Aluru, Minh Nguyen and Radu Talkad at Ericsson and Todd Westerhoff at SiSoft for their support. Many thanks to Jim Mangin and Steve Barbas for their support of this effort, and to John Lehman, Jose Paniagua and Brian Kirk at Amphenol-TCS
Visit www.eeweb.com
22
TECHNICAL ARTICLE
Gopinath, Steve Scearce, Mike Steinberger, Doug White, paper 5-TP2, DesignCon2009, February 3, 2009. [11] Practical Design of Differential Vias, Bogatin, Simonovich, Cao, PCD&F, July 7, 2010 http://pcdandf. com/cms/component/content/article/171-cur rentissue/7302-eric-bogatin-bert-simonovich-and-yazi-cao [12] An Optimized Design Methodology for HighSpeed Design DesignCon 2008 http://www.siguys. com/resources/1998_DesignCon_Optimized_SI_ Methodology.pdf [13] Demonstration of SerDes Modeling using the Algorithmic Modeling Interface (AMI) Standard, Steinberger, Westerhoff, White, paper 7-TA3, DesignCon 2008
Sisoft Video Presentation and Corresponding Powerpoint: http://www.sisoft.com/elearning/
commercially successful auto-router. Dr. Katz founded Layout Concepts and sold routers through Cadence, Zuken, Daisix, Intergraph and Accel. More than 20,000 copies of his tools have been used worldwide. Dr. Katz developed the first signal integrity tools for a 17 MHz 32-bit minicomputer in the seventies. In 1991, IBM used his software to design a 1 GHz computer. Dr. Katz holds a PhD from the University of Rochester, a BS from Polytechnic Institute of Brooklyn and has been awarded 5 U.S. Patents. Sergio Camerlo is an Engineering Director with Ericsson Silicon Valley (ESV), which he joined through the Redback Networks acquisition. His responsibilities include the Chassis/Backplane infrastructure design, PCB Layout Design, System and Board Power Design, Signal and Power Integrity. He also serves on the company Patent Committee and is a member of the ESV Systems and Technologies HW Technical Council. In his previous assignment, Sergio was VP Systems , Engineering at MetaRAM, a local startup, where he dealt with die stacking and 3D integration of memory. Before, Sergio spent close to a decade at Cisco Systems, where he served in different management capacities. Sergio has been awarded fourteen U.S. Patents on signal and power distribution, interconnects and packaging. Donald Telian is an independent Signal Integrity Consultant. Building on over 25 years of SI experience at Intel, Cadence, HP and others, his recent focus has been , on helping customers correctly implement todays MultiGigabit serial links. He has published numerous works on this and other topics that are available at his website siguys.com. Donald is widely known as the SI designer of the PCI bus and the originator of IBIS modeling and has taught SI techniques to thousands of engineers in more than 15 countries.
TECHNICAL ARTICLE
webinars/simulating-large-systems-with-thousands-ofserial-links.html
Authors Biographies
Barry Katz, President and CTO for SiSoft, founded SiSoft in 1995. As CTO, Barry is responsible for leading the definition and development of SiSofts products. He has devoted much of his efforts at SiSoft to delivering a comprehensive design methodology, software tools, and expert consulting to solve the problems faced by designers of leading edge high-speed systems. He was the founding chairman of the IBIS Quality committee. Barry received an MSEE degree from Carnegie Mellon and a BSEE degree from the University of Florida. Michael Steinberger, Ph.D., Lead Architect for SiSoft, has over 30 years experience designing very high speed electronic circuits. Dr. Steinberger holds a Ph.D. from the University of Southern California and has been awarded 14 patents. He is currently responsible for the architecture of SiSofts Quantum Channel Designer tool for high speed serial channel analysis. Before joining SiSoft, Dr. Steinberger led a group at Cray, Inc. performing SerDes design, high speed channel analysis, PCB design and custom RAM design. Dr. Walter Katz, Chief Scientist for SiSoft, is a pioneer in the development of constraint driven printed circuit board routers. He developed SciCards, the first
Visit www.eeweb.com
23
F E AT U R E D P R O D U C T S
High Current PMIC Without Charger
The AS3713 is a compact System PMU with back light driver. The device offers advanced power management functions. All necessary ICs and peripherals in a battery powered mobile device are supplied by the AS3713. It features 3 DCDC buck converters as well as 8 low noise LDOs. The different regulated supply voltages are programmable via the serial control interface. 4MHz operation with 1uH coils are reducing cost and PCB space. For more information, please click here.
LT3975 BLOCK DIAGRAM
VIN C1 INTERNAL 1.197V REF 1.02V VIN 0.5V OUT
FEATURED PRODUCTS
EN RT RT
SHDN
SWITCH LATCH R S Q
BOOST
SW
L1 D1 C2
VOUT
1.097V
VC
GND R2
FB R1
3975 BD
C5
Linear Technology Corporation announced the LT3975, a 42V step-down switching regulator that can deliver 2.5A of continuous output current and requires only 2.7A of quiescent current. Similarly, the LT3976 can operate from a 40V input, delivers up to 5A of output current and requires only 3.3A of quiescent current. Both devices offer a 4.2V to a nominal 40V input voltage range, making them ideal for automotive and industrial applications. The combination of their 16-lead thermally enhanced MSOP package and high switching frequency keeps external inductors and capacitors small, providing a compact, thermally efficient footprint. For more information, please click here.
+
3975p
Visit www.eeweb.com
24
Dave Baarman
Director Of Advanced Technologies
"Sed ut perspiciatis unde omnis iste natus error sit voluptatem accusantium laudantium, totam doloremque rem aperiam,
eaque ipsa quae ab illo inventore veritatis et quasi architecto beatae vitae dicta sunt explicabo. Nemo enim ipsam voluptatem quia voluptas sit aspernatur aut odit aut fugit, sed quia consequuntur magni dolores eos qui ratione voluptatem sequi nesciunt. Neque porro quisquam est, qui dolorem ipsum quia dolor sit amet, consectetur, adipisci velit, sed quia non numquam eius modi tempora incidunt ut labore et dolore magnam aliquam quaerat voluptatem. Ut enim ad minima veniam, quis nostrum exercitationem ullam corporis suscipit laboriosam, nisi ut aliquid ex ea commodi consequatur? Quis autem vel eum iure reprehenderit qui in ea voluptate velit esse quam nihil
ARTICLES
JOBS
COMMUNITY
DEVELOPMENT TOOLS
Ele ct ric al En w gi w ne w er .e in ew g
EE
C eb om .c m om un ity
W eb
JOIN
TODAY
PART II
Robert Green
Thermocouples
Thermocouples are far and away the most widely used type of temperature sensor because they are both rugged and relatively inexpensive. They operate based on the thermoelectric or Seebeck effect (named after physicist Thomas Seebeck). When two wires made up of dissimilar metals are joined together, a voltage is generated. The level of generated voltage is a function of temperature. As temperature changes, the voltage changes, so the thermocouples voltage output equates to a temperature reading. The linearity of a thermocouples output varies depending on thermocouple type and temperature range. Although thermocouples can cover a very wide range of temperatures (some as high as 2300C), they produce
very small output voltages, so the instrumentation used with them must provide sufficient resolution to discern small voltage changes. Several sources of temperature measurement error are specific to thermocouples: The CJC (Cold Junction Compensation) is not configured or compensatedThermocouples measure the difference in temperature between the hot junction (used for the actual measurement) and the cold or reference junction (at the instrument). If the temperature of the cold junction is not known or not compensated for (by means of cold junction compensation), the temperature readings will be inaccurate. Use of copper wire from the
thermocouple connection to the measurement instrument Connections between the thermocouple and the instrument should be made with the same kind of wire used for the thermocouple. In theory, one could use copper wire, but that would require controlling the temperature of all the wire, which is generally impractical.
The voltage measurement instrumentation is not sensitive or accurate enough for thermocouple measurementsThe output of a thermocouple is just microvolts, so be sure to choose an instrument with sufficient resolution to measure microvolts accurately. Fortunately, many modern digital multimeters (DMMs) offer built-in capabilities that make them particularly appropriate for thermocouple temperature measurements.
Visit www.eeweb.com
26
TECHNICAL ARTICLE
TECHNICAL ARTICLE
1 9 Pin 1 Pin 2 or Pin 9 8 15 Ice bath Copper wire to thermocouple wire connection (one of two)
INPUT HI INPUT LOW
For example, Keithleys Model 3706A System Switch/Multimeter supports J, K, N, T, E, R, S, and B thermocouple types. Depending on which type of thermocouple you choose, this instrument allows you to measure temperatures from -200C to 1820C at resolution levels from 0.001C (with type J) to 0.1C (with type B). When a thermocouple is connected directly to the input of a DMM, at least one of those connections will be a junction made up of two dissimilar metals, introducing a thermoelectrical voltage that will be algebraically added to the thermocouple voltage, producing an erroneous temperature measurement. To cancel the effects of this unwanted thermoelectrical voltage where dissimilar wire connections must be made, the thermocouple circuit requires a reference junction that is held at a stable, known temperature. For example, as long as the temperature of this reference (cold) junction is known, the DMM can factor in the reference temperature
to calculate the actual temperature reading at the thermocouple. Although the standard reference temperature used as the fundamental reference for NISTs voltage-to-temperature conversion tables is the ice point (0C), other known temperatures can be used. For example, some DMMs can acquire the cold junction temperature by measuring the cold junction using a thermistor or fourwire RTD, or the user can enter the known temperature value. The most accurate thermocouple measurements are achieved by using a simulated reference junction using an ice point reference (Figure 1). The copper wire to thermocouple wire connections are immersed (but electrically isolated) in the ice bath, and the user enters the 0C simulated reference temperature into the instrument. Long lengths of thermocouple wire can have a significant amount of capacitance, which is seen at the input of the DMM. If an intermittent open occurs in the thermocouple circuit, this capacitance can
produce an erroneous on-scale reading. To prevent these errors, look for an instrument that provides an open thermocouple detection circuit.
Visit www.eeweb.com
27
Ultra-Low Lux, Low Power, Integrated Digital Ambient Light Sensor with Interrupt Function
ISL29033
The ISL29033 is an integrated ambient and infrared light to digital converter with I2C (SMBus Compatible) interface. Its advanced, self-calibrated photodiode array emulates human eye response with excellent IR rejection. The on-chip 16-bit ADC is capable of rejecting 50Hz and 60Hz flicker caused by artificial light sources. The lux range select feature allows users to program the lux range for optimized counts/lux. Power consumption can be reduced to less than 0.3A when powered down. The ISL29033 supports a software and hardware interrupt that remains asserted until the host clears it through the I2C interface. The function of ADC conversion continues without stopping after interrupt is asserted. Designed to operate on supplies from 2.25V to 3.63V with an I2C supply from 1.7V to 3.63V, the ISL29033 is specified for operation over the -40C to +85C ambient temperature range.
Features
Ambient Light Sensing Simple Output Code Directly Proportional to Lux Variable Conversion Resolution Up to 16 Bits Adjustable Sensitivity Up to 520 Counts Per Lux Measurement Range: 0.0019 to 8,000Lux with Four Selectable Ranges Program Interrupt Feature Light Sensor Close to Human Eye Response - Excellent Light Sensor IR and UV Rejection 75A Max Operating Current - 0.3A Max Shutdown Current 6 Ld 2.0mmx2.1mmx0.7mm ODFN Package
Applications
Display and Keypad Dimming Adjustment for: - Mobile Devices: Smart Phone, PDA, GPS - Computing Devices: Notebook PC, Webpad - Consumer Devices: LCD-TV, Digital Picture Frame, Digital Camera Industrial and Medical Light Sensing
Related Literature
See AN1422 Light Sensor Applications
VDD 1
PHOTODIODE ARRAY
COMMAND REGISTER INTEGRATION ADC DATA REGISTER I2C/SMBus IREF fOSC INTERRUPT REGISTER 5 SCL 6 SDA
3 REXT
2 GND
4 INT
Intersil (and design) is a registered trademark of Intersil Americas Inc. Copyright Intersil Americas Inc.2011 All Rights Reserved. All other trademarks mentioned are the property of their respective owners.
TECHNICAL ARTICLE
Jonpaul Jandu
Intersil Business Development Manager
TECHNICAL ARTICLE
degrees, can offer fascinating and exotic qualities. They offer a density of visual information that is otherwise inaccessible to the naked eye or by the use of regular pinhole type cameras, even when fitted with a wideangle lens. The extreme barrel-shaped warping of the image, as if looking at a globe, is often quite acceptable even valued as a visually interesting picture. However, a different situation emerges if one considers the possible commercial or industrial applications for a fisheye image. Such applications usually depend on a regular, equidistant-matrix, rectilinear picture format. Extreme fisheye warping results in radial dislocations of the picture elements in a non-linear pattern. The pixels appear spatially shifted to the edges, away from the images centre of distortion which usually, but not necessarily, is also its focal point and visual center. One must add to that the effects of uneven luminance and colour saturation caused by the enormous directional light and shadow variations across the field of view, plus chromatic aberrations and tangential pixel dislocations introduced by the limited optical quality of the lens. Unfortunately, even though perhaps appealing at first glance, the raw data coming from a fisheye lens is not suited to computer vision applications, or the detection and recognition of objects and their movements across the picture plane. Even for visual observation
EEWeb | Electrical Engineering Community
29
TECHNICAL ARTICLE
manufacturers with a versatile fisheye correction device that they could integrate easily into all kinds of real-time surveillance cameras and systems without the needs for tedious adaptation and post-processing procedures. In May 2011, Intersil announced the TW2871, and the company is now ready to deliver the 376-pin PBGA chip in Q2 2012. conversion from RGB to YCbCr colour formats. The TW2871 delivers up to four independent split images derived from the hemispheric fisheye view. Each of them comes in the standard definition SDTV format (CVBS), NTSC or PAL, along with 720p or 1080i HD according to SMPTE. Alternatively there are two channels of 8-bit BT.656 streams, also in NTSC or PAL. The four corrected split images derived from the distorted fisheye view appear in a de-warped rectilinear format as if coming from four separate cameras of longer focal lengths. The positions of these split images can be independently controlled, moved and zoomed by the user in real time. This is done by means of an ordinary joystick console as used in typical surveillance setups. The images are shown on a split-screen monitor, along with the original super-wide-angle fisheye view as in Fig. 2. In other words, the optically corrected split images enable a targeted, zooming-in on the most interesting areas of the 180-degree wide scene. This allows for closer observation of objects of interest. Positions, directions and zoom factors of the split images are freely selectable. Thus, a surveillance system fitted with the TW2871 resembles digital panning, tilting and zooming (PTZ) a capability usually not found in todays in-door surveillance systems. They are typically mounted to the ceilings of lobbies, hallways, and entry areas of public spaces. Usually they employ a single-view camera, which must be manually zoomed and pointed at specific areas chosen for closer examination. But with the TW2871 solution, the (fisheye) camera direction is not revealed, it is hidden under a dome, appearing mirrored or black. The TW2871 generates four independent split images that can be watched as well as moved across the field of view, is the key factor in its design. It replaces the mechanical actuators used for panning, tilting and zooming that are error-prone and require maintenance. Another important feature is that by pointing the four split camera views generated by the TW2871 in the four orthogonal directions, it eliminates the nasty blind spots in the field of view as experienced with servomechanically controlled single-view cameras. If motion detection is required, the TW2871 provides the necessary algorithm to do that. It even offers a masking
TECHNICAL ARTICLE
The TW2871 enables real-time applications through a hard-wired ASIC implementation of all the necessary corrective measures but without dependence on the usual software solutions. The processor chip integrates a high-end geometrical image correction engine patented to Intersil. It achieves a video latency of 100msec maximum virtually real time. See Figures 1 and 2 for examples of how the technology corrects. The ISP connects directly to the image sensor that feeds the highly distorted raw image to the chip. An integrated DDR2 controller manages traffic from and to the off-chip DRAM, which stores the lookup tables for geometrical pixel mapping. Any fisheye lens, and any CMOS or CCD sensor of up to 8 or 10Mp of resolution at a pixel clock up to 96MHz is accepted. This feature points the way to HD video. Adaptation of the lens is extremely simple and is easily done by referring to its data sheet, by programming its relevant physical parameters (diameter, aperture, focal length, etc. everything that defines its fisheye distortion pattern) into the lookup table connected to the chip. The TW2871 processor then adjusts accordingly and automatically. No optical calibration or test imagery is needed. Its a one-time, straightforward procedure. An image signal processor (ISP) then performs a wide array of corrective actions with regards to image quality. Among these are bad-pixel, auto-exposure, auto whitebalance, wide dynamic range, 3D-noise reduction and lens shading corrections, plus Bayer alignment and
Visit www.eeweb.com
30
TECHNICAL ARTICLE
function to avoid false alarms. Masking a certain area of a split image by simply pointing and clicking from the console excludes visual objects that show regular unsuspicious motion patterns from the scrutiny of the motion detector. necessary algorithm to do that. It even offers a masking function to avoid false alarms. Masking a certain area of a split image by simply pointing and clicking from the console excludes visual objects that show regular unsuspicious motion patterns from the scrutiny of the motion detector. The four digital images delivered by the TW 2871 are de-warped and optically corrected up to XGA resolution, which is better than standard-definition (SD) TV quality. The final resolution is of course within the bounds of a digital zoom, which cannot go beyond the native pixel count of the original fisheye image. This limits surveillance applications mostly to indoor situations, at an object distance below 150 meters. However, future versions of the chip envisage HD resolution (1080i or 720p). In any case, optical zooming setups can now be relegated to outdoor surveillance situations that need to cover very large distances.
Figure 2: Example of corrected split image views plus original fisheye view
TECHNICAL ARTICLE
The four digital images delivered by the TW 2871 are de-warped and optically corrected up to XGA resolution, which is better than standard-definition (SD) TV quality. The final resolution is of course within the bounds of a digital zoom, which cannot go beyond the native pixel count of the original fisheye image. This limits surveillance applications mostly to indoor situations, at an object distance below 150 meters. However, future versions of the chip envisage HD resolution (1080i or 720p). In any case, optical zooming setups can now be relegated to outdoor surveillance situations that need to cover very large distances. The TW2871 generates four independent split images that can be watched as well as moved across the field of view, is the key factor in its design. It replaces the mechanical actuators used for panning, tilting and zooming that are error-prone and require maintenance. Another important feature is that by pointing the four split camera views generated by the TW2871 in the four orthogonal directions, it eliminates the nasty blind spots in the field of view as experienced with servomechanically controlled single-view cameras. If motion detection is required, the TW2871 provides the
Most important to the designer of surveillance systems is that the new chip provides a convenient way of implementing fisheye correction without time-consuming software-driven procedures such as panorama stitching. In most cases these are too complex and too costintensive to be done in real time. The TW2871 hardwired implementation enables image processing and correction to be done within the camera (but outside of the lens!). The camera can be installed in the immediate vicinity of the scene to be surveyed. But, as a stand-
Visit www.eeweb.com
31
TECHNICAL ARTICLE
alone solution, it doesnt need a big-box PC or other processing unit nearby, and no transporting of Gigabytes of raw image data over high-speed links. Besides surveillance systems, Intersil envisages future automotive driver-assistance applications. The concept calls for one fisheye camera mounted in the front, and another one in the back of a car. This would yield all the necessary views; front, left and right hand sides, and back, all corrected in the cameras. For this application, the current TW2871 layout may be an overkill; a superset of all the required driver assistance features. Car drivers need no PTZ, just reliable fixed views to all sides of the vehicle. More relevant in the automotive context is that the system, besides being very low-cost, should last up to 20 years. There definitely are many potential applications for the chip. Another one is video conferencing of larger groups, where the entire scene needs to be viewed, incorporating fast and flexible zooming in on participants as needed. There are also military applications in explosives seeking robots that need to view their entire surroundings. There are medical applications where it is desirable to provide multiple imagery for endoscopy procedures. There is also a home security system made simple; a single-lens camera supplying various angles and levels of digital output that comply with popular digital consumer video formats and IP transfer.
TECHNICAL ARTICLE
BeStar
INDUSTRIES
Automotive Durables Medical Industrial Mobile Fire / Safety Security Consumer Leisure
32
RETURN TO ZERO
RETURN TO ZERO
EEWeb
Electrical Engineering Community
Join Today
www.eeweb.com/register EEWeb | Electrical Engineering Community
Visit www.eeweb.com
33