Abstract Getting logical designs to meet specifications is the first step in creating a manufacturable design. Getting the physical design to work is the next step. The physical effects of PCB materials, transmission lines, vias, and trace coupling can all cause a working logical design to fail. However, you can predict these effects before getting to the prototype stage. This module uses Agilent EEsof EDA s physical design tools to characterize the PCB design in order to minimize prototype turns.
Gaining the Wireless Edge 2000 Techniques for RF and High-Frequency Wireless Design Modeling Physical PCB Effects
Overview Fitting physical verification into the overall design flow Maintaining component specifications after placement Generating an ADS model of a Motorola duplexer 2D EM characterization of layout 3D EM characterization of layout Pinpointing the design problem and implementing a fix Summary Page 3
Design Flow Steps Illustrated in this Module I I N SS TT RR U M EE N TT A TT I I O N D EE SS I I G N FF EE EE D BB A CC KK Concept Concept System System Design Design Circuit Circuit Design Design Layout Layout EM EM Simulation Simulation Manufacturing Manufacturing Logical Design Synthesis, Synthesis, Utilities Utilities and and Optimization Optimization Data Data Processing Processing EM EM Visualization Visualization & & Optimization Optimization Physical Design Supporting Infrastructure System System Library Library Design Design Libraries Libraries Vendor Vendor Libraries Libraries Transmission Transmission Line Line Models Models Foundry Foundry Libraries Libraries Layout Layout Footprints Footprints Custom Custom Models Models Page 4 Hardware Deliverable Customers Customers Simulation Deliverable In this paper we will focus on the indicated steps in the design flow.
Solution -- Advanced Design System 1.3 Unprecedented Productivity Bridging the Gap Simulation Technology Linear Nonlinear Circuit Envelope Time Domain Agilent Ptolemy Electromagnetic Others CUSTOMIZATION RF IP Encoder LIBRARIES Design Libraries APPLICATION CONTENT DesignGuides SYNTHESIS RF Compiler/E-Syn/LineCalc AUTOMATION Layout to HFSS Link Applications Amplifiers Filters Mixers Oscillators Passives System Mods/Demods Packaging Others Page 5
Main Challenges in Designing PCS Phones Smaller Phone Size, Decreased Thickness (Folding Phones) Smaller Duplexer Denser Layout Poor Rx rejection at Tx Band Longer Battery Life, Longer Standby Time Improved Power Management (MSM3000 or MSM3100) Lower Current Consumption of LNA Lower IP3 Both Needs Affect The Single-Tone Desensitization Specification! Page 6 With the rapid expansion of the CDMA/PCS wireless market, cell-phone manufactureres are aggressively developing lighter and more competitive handset models. In particular, folding phones are quite popular due to their small size and transportability. However, folding phones bring several design challenges with them. The main challenge is reducing the thickness of the phone. The designer needs to reduce the number of PCBs, eliminate the keypad PCB, and use thin components. Duplexers are some of the most difficult components to design into compact PCBs, not only because they are usually the physically largest components in handsets, but also because they play a critical role in the performance of the phone. They require plenty of ground plane and wide separation between each port to maintain their original performance specs. The single-tone desensitization specification of a phone is mainly affected by the IIP3 of the LNA. However, the receiver rejection at the transmitter band of the duplexer is another key attribute of that specification. This paper will examine the design of a space-efficient PCB around the duplexer that maintains the original duplexer performance specs.
Single-Tone Desensitization Required Rx rejection at the Tx Band of the Duplexer is typically 55 db minimum Tx Band -24(28) dbm 1750~1780 MHz Jammer -30 dbm Rx Band -101 dbm 1840~1870 MHz 90 MHz 12.5 MHz Page 7 The Tx band is 1750 ~ 1780 MHz, and the Rx band is 1840 ~ 1870 MHz. To perform a single tone desensitization test, a very strong jammer signal with 12.5 MHz offset and a receiver signal of -101 dbm are injected into the antenna port. With these inputs, the transmitter power amplifier will generate its maximum power of 28 dbm according to the power control of the baseband MODEM. Because the Rx-Tx isolation of duplexer is not perfect, some amount of Tx leakage signal is fed to the LNA input. The leakage signal level is determined by the Rx rejection at the Tx band of the duplexer. Typically, 55 db minimum of Rx rejection at the Tx band is specified for the duplexer.
An Example of a Handset PCB Layout Top-Layer View: Antenna Connector Example: A real board currently in mass production Isolator Duplexer LNA Page 8 The PCB example in this presentation is based on a real case. The example board is in mass production for an Asian vendor. As shown in the figure, the PCB is very complex. All RF, logic, audio, and power circuits are on the top side of the PCB. The keypads are implemented on the backside of the same PCB. This implementation helps reduce the phone thickness significantly; however, this causes the antenna connector, LNA, PA, and duplexer to be mounted on the same plane of the PCB. All the connection lines (Ant-duplexer, LNA-duplexer, PA-duplexer) are located very close to each other. The main challenge now is to obtain the best performance given these PCB layout restrictions.
Monoblock Duplexer KFF6614, 3mm height, Motorola: Specifications from datasheet Parameter Frequency Typical Spec (MHz) (@25 o C) Ant to Tx Response Pass Band Insertion Loss 1750-1780 1.7 db Pass Band Return Loss 1750-1780 15.0 db Rejection @ Rx Band 1840-1870 45.0 db Ant to Rx Response Pass Band Insertion Loss 1840-1870 2.5 db Pass Band Return Loss 1840-1870 15.0 db Rejection @ Tx Band 1750-1780 60.0 db Tx to Rx Response Rejection @ Tx Band 1750-1780 61.0 db Rejection @ Rx Band 1840-1870 46.0 db Focus on this Specification: Rejection at Tx Band 1750-1780 MHz is 60 db Page 9 The duplexer used in this design is a monoblock type. It's only 3 mm thick, yet it is still the thickest component. The specifications in the datasheet are good, but this is de-embedded data and does not include the performance degradation when mounted on a general PCB.
Building a 3-Port S-Parameter Model KFF6614 (Monoblock Duplexer, Motorola) Use Data Access Component and Measurement Data Page 10 The first step of the PCB analysis and redesign is to build a 3-port duplexer model. As the main purpose is not to design a duplexer but to design a PCB layout around it, the best solution is to use the measured S-parameter representation of the duplexer. Since the duplexer is a 3-port device and the general Vector Network Analyzer (VNA) is a 2-port measurment device, three sets of measured 2-port S-parameter data were used to construct a 3- port duplexer model. The measured 2-port S-parameters were supplied by Motorola, and using the ADS DataAccessComponent and Equation- Based n-port S-parameter model, a 3-port duplexer model is constructed. The connector effects and transmission line of the test jig were calibrated. The measurement based 3-port duplexer model will next be used in analyzing and redesigning the PCB layout.
Insertion Loss: Antenna to Rx, Antenna to Tx Measurement-Based 3-Port Duplexer Model Tx Band Rx Band Original Rx rejection @ Tx Band >60 db Ultimate rejection is 62 db and the insertion loss is 3 db The total isolation required on the PCB is 65 db min Page 11 The graph above shows the transmission characteristics of the duplexer model. Because the mounting effects are not included, the Rx rejection at the Tx band is very good (below -62 db). To maintain this performance after mounting, the port isolation of the board should be greater than 65dB. The suppression of PCB feedthrough is the main focus. The required board isolation of 65 db is possible with a careful design of the PCB layout.
PCB Layout around the Duplexer(I) For Momentum Simulation Tx Antenna Duplexer Mount Rx Page 12 To analyze the board coupling, a planar EM simulation is done using Momentum. For this, the layout should be either created in or imported to ADS. Mentor or Cadence layouts can be exported using the intermediate file format (IFF); these layouts can then be imported to ADS. Once the required simulation setup parameters such as substrate thickness and mesh size are specified, a Momentum simulation is run and saved to a 6-port s-parameter dataset. The data obtained is pure board simulation data.
Isolation for PCB Layout (I) Ant to Rx Ant to Tx Tx to Rx Tx Ant Rx Antenna to Rx isolation is about 58 db Not enough for ultimate rejection of 62 db Required isolation on PCB is about 65 db min Page 13 The above graph shows port isolation of the board. The Tx and Rx ports are well separated, as are the Tx and Antenna ports. Hence, the simulated port isolation of Tx-to-Rx and Ant-to-Tx is good. However, the Antenna and Rx ports are in close proximity, and a performance problem could result. Indeed, the simulation results show about 58 db of Ant-to-Rx port isolation. This is not enough to obtain the 60 db of Rx rejection at the Tx band of duplexer that is desired.
Simulation Results for Layout (I) Duplexer Model & Momentum 15.6 db Page 14 To characterize the overall performance of the duplexer, including the PCB mount effects, schematic is created with the 3-port duplexer model and 6-port layout obtained from Momentum, and an S-parameter simulation is performed. As already predicted in the previous slide, the Rx rejection at the Tx band was degraded by about 15 db when mounted on board.
PCB Layout around the Duplexer (II) Via Holes for Preventing Field Coupling through the Substrate Antenna Line Rx Line Duplexer Mount Page 15 The main reason for port isolation degradation is feedthrough. That is, some amount of field at the Antenna line is coupled to the field of the Rx line at some location in the board. But at this moment, it is not known where this coupling is located. Most likely, there is coupling through the inside of the substrate. To prevent this, via holes are added between the Antenna line and Rx line. These via holes should prevent some degree of field leakage.
Isolation for PCB Layout (II) Ant to Rx Tx Ant Ant to Tx Tx to Rx Rx Antenna to Rx isolation is still 58 db Not enough for ultimate rejection of 62 db Required isolation on PCB is ~65 db min Page 16 The PCB structure including the via holes was simulated again with Momentum. Unfortunately, the simulation results showed little improvement. The Antenna to Rx isolation is still 58 db and has not improved.
Simulation Results For Layout (II) Duplexer Model & Momentum 15.3 db Page 17 Again, the overall characteristics of the duplexer are simulated, including the PCB and via hole effects. The Rx rejection at the Tx band of the duplexer improved slightly, but the value is negligible. For confirmation of these results, the design is exported to Agilent HFSS for further analysis. Agilent HFSS is a full 3-D EM simulator based on the finite element method.
3D EM-Simulation: Agilent HFSS ADS/Momentum Layout Translation to HFSS Extra microstrip lines are de-embedded in the post-processor Page 18 The ADS layout is translated to the Agilent HFSS drawing environment. After importing the ADS layout, extra microstrip lines are added to construct the port. After the simulation, these additional microstrip lines can be de-embedded in the post processor. Voltage sources can also be used instead of ports; in Agilent HFSS, these are analogous to internal ports in Momentum.
Isolation for PCB Layout (II) -- 3D EM Modeling Ant to Rx Ant Ant to Tx Rx Tx to Rx Tx Ant to Rx isolation is still 60 db Still not enough for an ultimate rejection of 62 db Required isolation on PCB is ~65 db min Page 19 The above graph shows that the HFSS simulation results agree with the Momentum results. The Antenna-to-Rx isolation simulated in HFSS is about 60 db, and this is just 2 db different from the Momentum simulation results. The similarity between the Momentum and HFSS simulation results indicates that the added via holes did not help in preventing field coupling. Field coupling is now suspected of occuring elsewhere on the board.
Simulation Results for Layout (II) Duplexer Model & HFSS 15.4 db Page 20 This slide shows that the overall results for the duplexer mounted on the PCB with via holes have not improved. This result is just a confirmation of the previous Momentum simulation. If so, where is the field coupling taking place?
EM Coupling Line coupling through air causes the degradation in port isolation Page 21 In the Agilent HFSS post-processor, a field animation shows the EM field around the Antenna line and the Rx line. A small amount of field coupling is detected between the Antenna and the Rx lines. This could be the cause of the degradation in port isolation. Now that the problem has been located, the best way to eliminate the field coupling through the air is to use a shield cover.
PCB Layout around the Duplexer (III) The metal pattern is left unchanged Shielding is inserted Wall will prevent air coupling Blocking wall can be incorporated in the shielding case Page 22 As a solution to this problem, a blocking wall can be used for the prevention of air field coupling; this can be done without any change to the PCB metal pattern. This is easily implemented in the duplexer case or handset case. This is a simple but effective idea, and the additional cost is negligible.
Isolation for PCB Layout (III)-HFSS Ant to Rx Ant to Tx Ant Rx Tx to Rx Tx 80 db isolation on the PCB is enough for an ultimate rejection of 62 db Page 23 The above graph shows significant improvement of the Antenna to Rx board isolation when using a blocking wall. The simulated isolation is about -80 db. The improvement is above 20 db, and this value is enough to obtain -62 db of Rx rejection at the Tx band of the mounted duplexer.
Simulation Results for Layout (III) Duplexer Model & HFSS Very close to the original duplexer specs Page 24 A schematic with the 3-port duplexer model and the S-parameter data from Agilent HFSS is created. The simulation results for this modified design are nearly identical to the original unmounted duplexer specifications. Both results show more than 60 db of Rx rejection at the Tx band.
Measurement Results ~18dB Without Blocking Wall With Blocking Wall Page 25 This slide shows a plot with two sets of measured results. The first set shows the duplexer mounted on the PCB without the blocking wall. In this case, the Rx rejection at the Tx band is about 50 db. The second set shows the duplexer mounted on the PCB with a blocking wall. In this case, the Rx rejection at the Tx band is about 68 db. The difference between the two is a substantial 18 db. More significantly, this performance improvement was achieved without changing the metallization on the compact handset board.
Summary Duplexer layout is one of the hot issues in handset design. Trade-off between limited space and performance. Accurate field simulation is required to predict some coupling phenomena in compact PCB layout. Measurement-based 3-port duplexer model can provide mixed EM and circuit simulation of the duplexer. Eliminates the need for multiple board turns and saves time and money. Page 26 Duplexers are some of the most difficult devices to lay out. They are physically large, and their performance is critical to the overall handset performance. This implies that the EM characterization of this type of small, dense board is very important. A measurement-based 3-port duplexer model derived from circuit and EM simulations can enable the accurate prediction of overall design behavior. Such accurate prediction can significantly reduce the number of board turns, and ultimately shorten the time to market.