Friday, March 29, 2024
Login

Using Component Models to Achieve First Pass Success - A Transmitter Case Study: Part 1, Harmonic Filter Design

Parent Category: 2017 HFE

By Ted Longshore and Larry Dunleavy

Abstract

Utilizing models that include component and PCB parasitics in place of ideal lumped elements in microwave circuit design software produces filter simulation results that are accurate enough to provide first pass success in an example filter design for a transmitter application. The cost of the required models is shown to be easily justified on the basis of cost, schedule and/or improved designer productivity.

Introduction

This two-part article presents a case study in simulation-based RF/Microwave design in the context of an example transmitter design. Our focus will be on an output filter design using advanced linear component models and simulation methods (Part 1) and a 2-stage power amplifier design, optimized with linear component and non-linear device models utilized in harmonic balance simulations (Part 2).

RF/microwave design computer-aided-engineering (CAE) tools have existed for many years and are used extensively by engineers to design linear and non-linear RF/microwave circuits. These design efforts have been supported by linear S-Parameters “models,” for passives, and compact non-linear models and load-pull power device data for active devices, as provided by many component manufacturers. More recently X-parameters models by some manufacturers and model providers.

At application frequencies above 1 GHz, the component parasitics such as series inductance in capacitors and shunt capacitance in inductors as well as substrate-dependent component parasitics significantly impact the actual circuit performance. If these component parasitics are not included in the simulation, the accuracy of the simulation results will be significantly degraded. As a consequence, extensive prototype board level bench-tuning and re-work is typically required to achieve the desired results. Often a second or third board iteration (or spin) is needed to achieve acceptable performance. This is costly, not only in terms of board fabrication costs, but in engineering time, design productivity, and in terms of lost calendar time and missed time-to-market windows for new products.

Simulations

Circuit simulations with accurate component models that include the component parasitics can produce very accurate results. For example, an accurate low pass filter simulation might show steeper roll-off due to the parasitic series inductance in the shunt capacitors, and will certainly show the ‘flybacks’ (frequencies where filter has degraded rejection in stop band) where the series inductors are resonant. Many circuit designers recognize the importance of including component parasitics in their simulations and create their own models for the desired components by measuring S-parameters using a network analyzer and then de-embedding the results. Models created using this approach are typically accurate only for a specific pad size, PCB thickness and dielectric constant. This effort is time consuming and requires specialized test fixtures and measurement expertise.

In part 1 of this article, a low pass harmonic filter is successfully designed in a single pass using NI AWR Design Environment software, specifically Microwave Office circuit simulation, and Modelithics’ Global Microwave Models™ included in the Modelithics® CLR Library1, 2.

These models provide for pad scaling that takes into account the substrate/ground-plane effects, and are measurement validated on numerous substrates over a wide frequency range. Obviously there is a cost required to develop component models, even if it is done in house on a per project basis as described above3, 4. In addition to the technical filter design example, a cost - benefit analysis of using a purchased component model library is provided through a “return on investment” (ROI) calculation5.

Part 2 of this article will illustrate how the use of suitably accurate non-linear power transistor models can facilitate successful design flows for transmitter designs.

Filter Example

In the design of a harmonic filter for a transmitter application, the cutoff frequency and level and frequency of the flybacks is crucial to ensuring the level of the transmit harmonics are below the required limit. Using today’s advanced CAE tools, the design of a low pass harmonic filter appears to be a straightforward effort. For an example 1 W transmitter application, a filter was required to reduce the level of the 2nd and 3rd harmonics at the output to be below -20 dBm, and the level of the higher order harmonics below -80 dBc.

Starting with ideal lumped element capacitors and inductors and including the interconnecting microstrip lines, the following 2.2 to 2.4 GHz harmonic filter of Figure 1 was designed for low S11 in the passband and high rejection at the 2nd harmonic and above. The schematic shows the circuit using Modelithics model symbols but the parasitics have been set to “ideal” so are not included in the simulation results.

1708 HFE models 01

Figure 1 • Filter Schematic: Ideal Harmonic Filter - Including Transmission Line Interconnects.

The simulated response (Figure 2) shows insertion loss less than 0.05 dB, 2nd harmonic rejection of 49 dB, and minimum 47 dB rejection up to 20 GHz. The experienced engineer would recognize that this response is unrealistic and assume the simulation software is somehow incorrect; however, this is due to the fact that the lumped element components utilized in this harmonic filter differ from their ideal models at microwave frequencies.

1708 HFE models 02

Figure 2 • Simulation Results of 2.4 GHz Harmonic Filter - Including Transmission Line Interconnects.

In addition to the component parasitics, including resistance, series inductance, and shunt capacitance, the response is further changed when the shunt capacitance between the component PCB pads and the substrate ground is included in the model. Taking all these parasitics into account, the simulated results look like the simulation in Figure 3, in which the parasitics and shunt pad capacitance were enabled in the Modelithics component models (simmode=0 model setting).

It is apparent, by comparison of Figures 2 and 3, that the filter response is significantly different when the parasitics L-C models are included in the simulation. The 3dB cutoff frequency response is shifted lower by 20 %. Even worse, the level of the 4th and higher harmonics, leaking through the filter, will be significantly higher due to the filter flybacks and degraded rejection. Improving the rejection after fabrication to meet the requirements would require time consuming re-design and an expensive 2nd and probably 3rd spin of the PCB.

1708 HFE models 03

Figure 3 • Simulation of 2.4 GHz Harmonic Filter - With all Parasitic Effects Included in Simulation.

The harmonic filter shown in Figure 4 was designed using Modelithics models for ATC capacitors and Coilcraft inductor families.  Because there was not yet a Modelithics model for the Coilcraft 0403HQ inductors used in this transmitter at the time of this design, the model for the larger 0604HQ inductors was simulated instead. In comparison to the previous circuit, additional shunt capacitors were added to decrease the parasitic shunt inductance in the capacitors in order to reduce the level of the flybacks. The completed PCB is shown in Figure 5.

1708 HFE models 04

Figure 4 • Re-designed 2.4 GHz Harmonic Filter Schematic.

1708 HFE models 05    1708 HFE models 06

Figure 6 shows a comparison between the simulated and two measured S21 responses. The electromagnetic analysis co-simulation (using EM for microstrip lines and solder pads and circuit models for components) displays a -36 dB flyback at 8.4 GHz and only about 15 dB rejection above 13 GHz due to coupling from the microstrip lines at the input to the output around the filter. The upper measured response is as built with no shielding and compares favorably to the simulated measurement through the passband, except for a higher resonance peak at 8.8 GHz The level of this resonance in the simulation results is very sensitive to the layout, and is expected to be substantially lower in the measured performance after the shield is installed. The measured insertion loss is 0.14 dB higher and the 2nd harmonic rejection is 3.6 dB less than simulated. The reason for the differences in the passband is likely due to the smaller, lower Q 0403HQ inductors used in the actual board compared to the 0604HQ inductors used in the simulation.

1708 HFE models 07

Figure 6 • Microwave Office and Modelithics: 2.4 GHz Discrete Harmonic Filter Simulation vs. Measured.

Compared to the simulation using ideal components (Figure 2), the simulation plots in Figure 6 indicate the need for additional filtering to improve the rejection above 8 GHz. For that reason, a radial stub microstrip filter (Figure 7) was appended to the output of discrete filter.

1708 HFE models 08

Figure 7 • Schematic for Radial Stub Filter.

The completed PCB layout for the combined discrete plus distributed low pass harmonic filter is shown in Figure 8 below. The gold outline shows where shielding will be added to reduce flybacks.

1708 HFE models 08ab

Figure 8 • Discrete LC plus Radial Stub Harmonic Filter.

Good Agreement

A comparison of the simulated and measured performance of the above filter shows good agreement as presented in Figure 9. The passband insertion loss and stopband rejection were accurately simulated especially given the difference in inductor model vs actual inductor body size as noted previously. The degraded rejection above 16 GHz is due to coupling around the filter and will improve once the filter is placed in a shielded housing.

1708 HFE models 15

Figure 9 • Microwave Office and Modelithics: 2.4 GHz Harmonic Filter Simulation vs. Measured.

The source of the measured 6.2 GHz flyback was traced to coupling between the discrete inductors and radial stub filter added at the output. Unfortunately, this radiated coupling from the inductors is not included in component models nor S-parameter data on individual components. This type of interaction is best addressed with 3D geometry modeling with a full-wave EM analysis tool, which is beyond our scope here.

Development Effort and Cost

RF circuit development is typically accomplished by prototyping the individual circuit components and individually characterizing and optimizing them via tuning on the bench; this typically results in multiple iterations of the design before acceptable performance is achieved.  Modern RF and Microwave CAE tools such as NI AWR Design Environment and Keysight ADS provide the capability to accurately simulate microwave circuits. If the engineer first accurately measures the optimized component performance on his particular substrate, the resulting component model can be inserted into the Microwave Office simulator. Instead of relying on multiple prototypes, the engineer can utilize the simulator to optimize the combined circuitry for optimal performance. This approach can eliminate one or two PCB spins and shorten the circuit development cycle one or two months.

Accurately characterizing a component and de-embedding the results for inclusion into the simulator requires substantial RF expertise. The alternative approach explored in this article utilizes component models developed by a third party (in this case Modelithics). This approach eliminates the time consuming component characterization and allows the engineer to proceed directly with circuit development, thereby saving several weeks of effort.

1708 HFE models 10

Figure 10 • ROI Analysis Results on Filter Design (completed with ROI calculator on Modelithics website).

It is interesting to explore the savings in project cost and development time resulting from a successful first pass design. The results of that analysis with actual costs of this real-world design are entered into the ROI tool and shown in Figure 10. The assumption that an investment of $5K was made to acquire or develop sufficiently accurate inductor and capacitor models that enabled a first pass design success. An estimate was also made of the costs and schedule impact of proceeding with inaccurate models that resulted in the need for a second design iteration. Under these assumptions the results show that in this case the cost of the required models (which should also be useful for future designs), the return on investment is achieved or justified on this specific design alone. It is also estimated that this project was complete four weeks sooner using the accurate models. One other way of looking at the benefit of investing in the models that enabled first pass success for this example is that it is estimated to result in a 90% productivity improvement, suggesting the same engineer could accomplish nearly twice as many similar designs in a given year.

Conclusion

At frequencies above 1 GHz, simulating with ideal components produces ideal results which can deviate from the actual performance by 20 % and omit critical responses such as flybacks. Simulating circuit performance using component models which include parasitics produces results which are typically accurate enough to realize design goals on a first pass PCB fabrication. This example demonstrates very good agreement in the passband, 2nd harmonic cutoff and reasonable agreement through the stopband. The coupling between the closely spaced inductors resulted in undesirable flybacks which were not included in the component model and not predicted by the simulation. The development time and cost savings of simulating with accurate component models is believed to easily justified the cost of using accurate models as part of this filter design flow.

*X-parameters is a trademark of Keysight Technologies.

About the Authors

1708 HFE models 13Dr. Larry Dunleavy co-founded Modelithics, Inc. in 2001 to provide improved modeling solutions and high quality microwave measurement services for radio and microwave frequency circuit designers.

He is currently serving as President and CEO at Modelithics. He also maintains a position as a Professor within University of South Florida’s Department of Electrical Engineering, where he has been a faculty member since 1990. In this role, he has been teaching in the area of RF & Microwave circuits and measurements for over 26 years. Prior to this he worked as a microwave circuit/MMIC design engineer at Hughes Aircraft and E-Systems companies.

1708 HFE models 12Ted Longshore graduated with a BSEE from the University of Cincinnati, and an MSEE from Illinois Institute of Technology, and has been working in the field of radio communications for nearly 30 years. He started his career working on cellular base stations and handsets at both Motorola and then Ericsson, and more recently on military electronics at Pole/Zero and NuWaves Engineering. In his current position at Quasonix he develops transmitters for telemetry applications.

 

References

1 “Advanced Microwave Chip Capacitor Models,” Microwave Journal, Jan 2002.

2 “Comprehensive Models for RLC Components to Accelerate PCB Designs,” Microwave Journal, May 2004.

3 L. Dunleavy, “Modeling-The Hot Potato In the RF & Microwave Industry,” Microwave Product Digest, April 2002.

4 Mike Golio and Jim Cozzie,  “Who Pays for Characterization?: The Final Dilemma for MESFET Modeling” 48th ARFTG Conference Digest, Fall 1996.

5 L. Dunleavy, “Everyone wants better models, but where’s the ROI?,” High Frequency Electronics, February 2017.

Search

CLICKABLES