Simulation Tools, Real Components and Limitations

By Bryan Ackerly, VK3YNG

Abstract

In recent times there have been a number of software manufacturers who have made affordable, or even free, simulation software available to amateurs. Although many of these free tools are cut down versions of much more powerful packages, they still allow simulation of many real world circuits that would have been well beyond the means of many amateurs only a few years ago.

To make proper use of these tools, an awareness of their limitations is required. This paper outlines some of the tools available and some tips on getting effective use from them.

1. Introduction

In the past high frequency simulation tools were only available to relatively large commercial engineering concerns. These software tools were extremely expensive and in many cases required a large amount of equally expensive training to make effective use of them.

In recent times a number of companies have released cut down evaluation or “student” versions of some very good simulation tools that in most cases are available free of charge from the companys’ web site. These tools are a lot more “user friendly” than their predecessors and provided some understanding of the limitations involved is taken into account, these tools can provide very close approximations to real world designs to the point where “trimming” a final design is unnecessary. This paper concentrates on getting effective use from these “free” tools.

2. Why simulate

Simulation tools are an invaluable design aid which allow concepts to be tried out without having to spend many hours trying to coax a physical circuit into operation. They are also very useful as a learning tool and allow the experimenter to quickly see the effects of changing various circuit components in an otherwise working design.

You can, for example, determine if a circuit is stable and what needs to be done to make it that way rather than finding out when you build it for the first time and let all the smoke out. Existing designs that were done before the availability of these tools can be further optimised for one or all of gain, return loss or noise figure for example using real components.

These modern tools are also a lot more user friendly allowing you to get up and operational without needing several days or weeks of training on how to use the tool. However, the tools still require some fundamental understandings of what you are working with and the old adage of “Garbage in = garbage out” still applies.

3. Types of simulation

3.1 Linear Simulation:

Linear simulation is used to simulate small signal performance of circuits and assumes there are no bounds on signal levels or linearity. Linear analysis allows the modelling and optimisation of input and output return loss (VSWR) amplifier forward and reverse small signal gain, stability and noise figure. Active and passive lumped and distributed circuits can generally be modelled and optimised.

3.2 Non-linear simulation

Non-linear simulation is useful for determining intermodulation performance and compression levels of linear and non-linear amplifiers and for modelling oscillators and mixers. Many free simulation tools offer a small range of non linear models for common active devices. Unlike linear simulation where models are commonly available in the form of S-parameters, non-linear models are usually proprietary and a lot more difficult to obtain.

3.3 System simulation

System simulation is useful for analysing and optimising the performance of cascaded major functional blocks in an Analog or Digital RF communication system. System simulation is beyond the scope of this paper but could prove very useful for those who are interested in the “bigger picture”.

4. The mathematics behind

4.1 Methods of simulation

There are several methods used inside simulation tools. An analysis of these is beyond the scope of this paper, but briefly the common methods used are:

Finite difference – time domain, Finite element method, Method of moments and Method of lines.

4.2 Types of simulation:

Simulation varies in its modelling ability and accuracy due to tradeoffs in models used. Older software had to strike a balance between the usefulness of the tool and modelling accuracy due to processing and memory limitations. With increases in processing capabilities offered by today’s personal computers, the capabilities of modelling tools has been able to increase to the point where very accurate results can be obtained very quickly.

There are effectively three categories to these tools:

4.2.1 2D (infinitely thin tracks):

Some older simulators used 2D modelling to enable equation based simulation which allowed minimal memory requirements and usable results through the UHF range. In many cases however, the results were somewhat limited.

4.2.2 2.5D (equation based)

Models box modes and track thickness. Most current simulators currently fall into this category, including many which claim to be 3D simulators. The modelling is equation based for the third dimension rather than using full EM modelling and the effects of obstacles generally cannot be taken into account.

4.2.3 3D (full space modelling)

This is presently the ultimate simulator which models obstacles in full 3D using Maxwells equations. This type of simulator is useful for waveguide and cavity filters and accurate antenna modelling (Note that most of the tools that amateurs currently use for Yagi design for example are based on interpolated tables and are not true 3D analysis tools).

This type of simulator is the most accurate for higher frequency and microwave use, but is extremely computationally intensive. Generally, full 3D tools are used for analysis only rather than as an optimisation and design tool due to the processing time, power and extremely large memory requirements needed for accurate analysis.

5. Available Tools

5.1 Berkley Spice

This simulator has been around for a number of years and many older implementations are public domain. More recent versions of this simulator have found their way into commercial PCB and schematic design packages like OrCad and Protel.

This simulator is good for modelling audio circuits and is particularly good at modelling OpAmps. Many small signal transistors and basic integrated circuits have spice models available for them. Spice is particularly good for low frequency non-linear simulation and can give usable results through VHF.

Pspice is also good for modelling HF oscillators and some RF circuits and it is apparently possible to extract S-parameters from spice models for devices where they are not supplied.

5.2 Sonnet Lite

This is a 2.5D simulator which is good for modelling rectangular elements, transmission lines, vias in multiplayer boards and other 3D structures. The free version is limited in memory which places some limits on the analysis but it is otherwise a good tool for modelling EM behaviour in more than 2 dimensions.

It has a feature where it generates colour plots of current density and is useful for modelling many microwave components.

5.3 Aplac

This is a very powerful simulator which uses a script based programming language entry. This makes it very useful for performing multiple simulations for comparing topologies since multiple sets of directives can be generated and analysed simultaneously.

Aplac has lots of component and stripline models, handles offset stripline and is very well documented. This simulator may be difficult for a first time user due to its scripting based nature but the package is becoming very commercial and graphical front ends are apparently now available.

5.4 Ansoft Serenade SV

This simulator is one of the more recent simulators to become available and is extremely powerful. It allows Linear and non-linear harmonic balance simulation, Schematic capture, linear and non-linear circuit and system analysis. An excellent overview of this tool was provided recently in QST [4]

In particular it offers a facility called “smith tool” which is very useful for designing impedance matching networks. Noise and stability analysis is also offered and is fairly easy to use for those who have understanding of how noise and stability are modelled using smith charts.

Another very powerful feature of this software is the ability to do tuning and linear optimisation. A set of goals are set up and particular components can be targeted to optimise a circuit to achieve desired results. There are a number of different optimisation methods available.

Serenade includes Models for basic transmission lines, edge coupled transmission lines (useful for microstrip filters), open ended lines and optimally mitred bends.

The student version has limitations which include:

- 25 circuit elements

- 1001 sweep points

- 4 probes

- 2 DC bias sources

- 4 non-linear device ports

- 31 spectral components (useful to 5^th order)

- 2 fundamental tones

Of all of these, the 25 circuit elements is the biggest limiting factor when trying to optimise a circuit with real components. However, there are ways of getting around this problem by selective use of S-parameter or ideal component models.

5.5 Other tools

This list is by no means complete. New tools are being made available and some older versions are becoming freely available via the Internet. Of interest recently has been an extremely powerful tool called Microwave Office 2002. The capabilities of this tool are very impressive. It will be interesting to see if the company releases an entry level version of this tool, but they do offer free trials of their software.

6. Using simulation tools – their limitations

Many simulation tools have limitations and understanding them is crucial to get the most out of the tool. For modelling circuits in the HF through low VHF spectrum the effects of wires and PCB tracks can often be ignored without major impact provided Q’s are low and the design is relatively compact. For modelling high VHF through microwave, transmission lines and PCB tracks must be included.

Generally most simulation tools are usable to around 6 to 10GHz. Above these frequencies, physical structures start to become less than ideal and it becomes increasingly important to model structure geometry, imperfections and proximity effects in full 3D using maxwells equations.

7. Components

In order to gain the best results from VHF and up an understanding of the operation and limitations of components at RF frequencies is paramount.

For most designs in the VHF through low microwave frequencies, if the results didn’t agree closely with the simulation then there is either a mistake somewhere in your design or there is a parameter that has been oversimplified or left out of the simulation. Even millimetre long lengths of track can make a difference as the frequency goes up.

7.1 Why use surface mount components?

Possibly the most important rule of any successful simulation is that the component models are accurate. At HF frequencies it is possible to disregard lead inductances and other parasitic effects as their contributions are generally minimal.

As frequencies increase through VHF and beyond, component leads and termination methods contribute a growing uncertainty to measurement that is difficult to predict and can severely limit the accuracy of results if ignored.

Surface mounted components on the other hand are predictable and consistent and are often supported by manufacturers models. Connections are usually done through PCB tracks which can also be modelled accurately giving accurate results even through microwave frequencies.

7.2 Resistors

Surface mount resistors are accurate to many GHz. The lowest and highest resistors tend to depart from their marked values through parasitic effects. Low resistance values tend to be dominated by series inductance while high values begin to get swamped by parallel capacitance.

A simple resistor model has been determined through measurement and is shown in figure 1.

7.3 Capacitors

For low RF and audio frequencies, capacitors can be used as if they are ideal components provided the ESR is not important. As frequencies increase, series inductance plays an increasing role and the ultimate capacitive reactance is limited by series resonance.

Since some larger ceramic capacitors start to become self resonant in the low megahertz, this parameter cannot even be ignored at HF frequencies.

Table 1 shows some actual measurements of capacitor self resonance and ESR performed on a calibrated Vector Network Analyser. Of interest is that in nearly all cases the equivalent series inductance is around a nanohenry. The surprise was that the very tiny 0402 components possessed similar self inductance to larger 0603 and 0805 case styles. This table shows the claim that smaller components possess lower parasitics appears to be unfounded.

Table 1 - Capacitor Self Resonance

Measured on HP8753ES Network Analyser.

Capacitor Value (size) & type:	ESR (ohms)	Self Resonant Frequency (MHz)	ESL(nH)	Notes:
1pf (0603)	0.65	5300	0.90
2p2 (0603)	0.63	3400	1.00
3p3 (0603)	0.43	2860	0.94
4p7 (0603)	0.46	2160	1.16
5p6 (0603)	0.36	2025	1.10
5p6 (1206)	0.7	2100	1.03
6p8 (0603)	0.46	1820	1.12
6p8Q (0805)	0.3	1940	0.99	High Q
10p (0402)	0.42	1570	1.03
10p (0603)	0.42	1530	1.08
10p (0805)	0.3	1640	0.94
22p (0603)	0.33	980	1.20
22p (0805)	0.28	1080	0.99
33p (0603)	0.52	850	1.06
47p (0603)	0.31	730	1.01
47p (0805)	0.43	776	0.89
100p (0603)	0.14	490	1.05
220p (0603)	0.15	350	0.94
470p (0603) C0G	0.13	240	0.94
470p (0805) C0G	0.11	240	0.94
470p (0603) X7R	0.92	280	0.69
1n (0603) X7R	0.42	173	0.85
1n (1206) NP0	0.1	151	1.11
10n (0805) X7R	0.16	61	0.68
10n (0612) X7R	0.49	168	0.09	Special low ESL capacitor
100n (0805) X7R	0.07	14	1.29
1uF (0805) Y5V	0.07	4	1.91	Measured @ 0V (C=0.83uF)
1uF (0805) Y5V	0.06	7	1.08	Measured @ 5V (C=0.48uF)
68uF/6V (C)	0.165	0.6	1.03	Tantalum Measured @ 5V

The only major change was the special 0612 package which is said to be designed for high frequency bypassing. This particular component exhibited an order of magnitude lower series inductance and a much higher series resonant frequency. The main reason for performing a test on this device was to prove that stray inductance had been calibrated out of the test jig.

ESR varies between about 0.05 to 1.0 ohms with lower values occurring for the larger capacitance values and better dielectric materials. NPO and C0G give much better results than X7R for example. Note how both ESL and capacitance vary substantially with voltage in the Z5U dielectric.

For simulation purposes, typically a capacitor can be replaced with an equivalent series circuit comprising a capacitor of the selected value, a resistor similar to that in table 1 and an inductor of 1nH as shown in figure 2. This allows the capacitor value to be optimised while taking into account its parasitic behaviour. In many cases either or both of these parasitic components can be omitted, particularly for coupling use but it is worth adding them and then replacing them with equivalent short circuits once their contribution, or lack of, has been verified.

7.4 Inductors

Generally we are taught that capacitors can be treated as ideal and inductors are very much a compromise due to mainly to limitations of interwinding capacitance and wire resistance. The reality is that at higher frequencies with smaller values the roles tend to reverse. Typically small value (<100nH) surface mount inductors have self resonant frequencies well into the Gigahertz region.

To accurately model these components with an equivalent linear model is difficult since it involves modelling five components one of which varies with the root of frequency. Equivalent parallel winding capacitance is typically 0.1pf in series with a resistance (R1) that increases with approximately the root of inductance and is typically 15 ohms for a 100nH 1008CS inductor. Note that this is much higher than Rvar (typically = ESR) and ultimately limits the self resonance Q.

Table 2 - Typical inductor ESR based on Coilcraft 1008CS
Inductance:	10nH	33nH	100nH	330nH	1uH
min SRF (MHz)	4100	1600	1000	570	290
Cp(pf)=	0.13	0.145	0.17	0.158	0.217
typ SRF (MHz)	4414	2301	1221	697	342
K=	2.00E-05	5.00E-05	1.35E-04	3.80E-04	1.20E-03
Freq=	ESR=	ESR=	ESR=	ESR=	ESR=
0	0.08	0.14	0.56	1.05	1.75
10	0.11	0.16	0.43	1.20	3.79
50	0.19	0.35	0.95	2.69	8.49
100	0.25	0.50	1.35	3.80	12.00
200	0.33	0.71	1.91	5.37	16.97
400	0.45	1.00	2.70	7.60	*
1000	0.68	1.58	*	*	*
1400	0.80	1.87	*	*	*
2000	0.94	*	*	*	*
2500	1.05	*	*	*	*
* = impedance limited by self resonance

Effective modelling can be done using a discrete ideal inductor to simplify initial design and optimisation and then replace it with an S-parameter equivalent for final analysis and tweaking. If ESR is important table 2 can be used to model this parameter at the frequency of interest. In many cases this parameter can be omitted, but it is wise to include it initially. Note that this technique only works when the inductor is used well below self resonance. Near resonance the S-parameter equivalent model is recommended, but this usually excludes the part from automated optimisation.

Many reputable inductor manufacturers provide S-parameter equivalent circuits via their web sites. For small values, final tweaking can often be performed in simulation by altering the track length to the inductor once the ideal inductor has been replaced with its S-parameter equivalent.

7.5 RF Beads

These devices are often used as a “cure all” and after doing some measurements it appears their effectiveness is somewhat overrated. Figure 5 shows the transmission loss of a typical 500MHz RF bead. Note that the transmission loss is typically in the order of only 5dB in a 50 ohm system, hardly the earmarked 40 plus dB that many users of these devices expect.

These devices can be modelled as a well damped (low Q) RLC parallel circuit. Figure 6 shows a derived equivalent circuit for the Murata BLM31A700 RF Bead. Note that this circuit only holds for zero bias as these devices are also somewhat dependant on the current going through them.

Generally beads with higher impedance have higher Q and the attenuation falls off more rapidly either side of resonance.

Figure 7 shows that even the popular 6 hole “VK200” beads only possess an insertion loss of 19dB at resonance (120MHz) dropping back to about 10dB at 1GHz. Contrast this with the attenuation measured on an AVX 22nF W3F15C2238 surface mount feedthrough capacitor shown here for comparison.

At first look it may appear that these devices are not particularly useful. However, there is one particular application where they can be used to particular benefit and this is to damp the parallel resonance that occurs when two bypass capacitors of different value are added in parallel. Figure 8 shows the effect of paralleling two “real” C0G bypass capacitors. This is often done to so called “broadband” bypass filtering. Note the peak in the attenuation response at 360MHz. This peak is 14dB worse than the attenuation at this frequency of either capacitor on its own! Such a peak can be the source of instability problems if it is not taken into account. Parallel capacitors can be very problematic if their effects are not properly understood.

Figure 8 - parallel bypass capacitors

Note the lower trace on figure 8 which is the combination of the two parallel capacitors separated by a ferrite bead. The series resonance points of each capacitor are still evident but the parallel resonance between the two has been significantly damped. Note also that the overall attenuation is more than 20dB better at all frequencies and is also far better than the bead on its own. Loss at near DC frequencies is minimal. Figure 9 shows the equivalent circuit of these three components used for this simulation.

Figure 9 - bypass capacitor simulation

7.6 Active devices

Simulation tools generally contain a rather large number of well characterised linear and non-linear active device models. Where linear models are not included with the simulator they can usually be obtained from manufacturers web sites or by contacting the device manufacturer and requesting them. Linear models are generally provided in the form of S-parameter (.S2P) files.

Some manufacturers also provide optimal noise match data in these files. If this data is not included the device is assumed noiseless and the modelling will only show the contribution of added noise due to losses in matching components. If you are interested in accurately optimising noise figure of an amplifier check that there are two lines towards the end of the S2P file containing something like the following:

!FREQ Fopt GAMMA OPT RN/Zo

!GHZ dB MAG ANG -

Non-linear models are generally proprietary and the tool designers will charge many thousands of dollars to create them. If you are interested in modelling intermodulation or compression points or designing non-linear amplifiers or oscillators, make sure the device you intend to use exists as part of the tool’s non-linear components library.

7.7 PCB tracks

For a successful design to be properly modelled, information on the PCB tracking should be included. Most simulation tools provide this facility and have facilities to include one or more substrates in Microstrip or symmetrical stripline structures. Some tools may offer limited support for offset stripline.

It is vital to have some information about the dielectric constant and tangent loss characteristics of the PCB material used. Typically standard FR4 has a dielectric constant of around 4.4 at UHF frequencies due to currently used levels of glass resin. The old “standard” of Er=4.7 is only really applicable at very low frequencies and it generally drops as the frequency increases[8]. Tangent loss of FR4 is typically around 0.05. If in doubt consult manufacturers datasheets. FR4 can be used through to microwave frequencies provided careful attention is paid to keeping Q’s low and minimising lengths of unnecessary transmission lines.

Teflon dielectrics have better controlled dielectric constants and tangent loss of 0.005 or better. Newer Composite dielectric materials also offer Teflon like characteristics at much better prices and the manufacturability of FR4.

8. A sample design

A 70cm amplifier was built using the Ansoft Serenade design tool as shown in figure 10. This circuit uses an ATF21186 GaAsFET in a self biased configuration. A 0.8mm thick FR4 laminate was used

Figure 10 - 70cm preamplifier simulation

A schematic was then developed and a PCB using Protel as shown in figures 11 and 12. This particular amplifier was designed to be phantom powered (i.e. power supplied via output coax connection)

Figure 11 – 70cm Preamplifier Schematic

Figure 12 – 70cm Test Preamplifier Figure 13 – measured input match

Figure 14 – Simulated input and output match. Figure 15 – Simulated Gain and Noise Figure.

Actual measured performance was comparable to that simulated with the input impedance measuring 53.5+j3.88 ohms (figure 13) compared to the simulation result of 56.5+j1.7 ohms at 430MHz (figure 14). Measured gain was also slightly higher than predicted measuring at about 19.4dB where the simulation result was 17.48dB (figure 15). This difference in gain has been attributed to the drain source voltage being somewhat higher than the model due to self biasing.

Noise figure was measured at 1.6dB although this is difficult to compare with the simulation as there was no noise data in the model. The simulation tool predicted 0.42dB of added noise figure due to component losses indicating a device noise figure of 1.2dB. It should be noted that the optimisation was set to obtain a target gain of more than 15dB, input and output return loss of better than 13dB and an added noise figure of less than 0.5dB. Although source degeneration was used in an effort to try to obtain simultaneous noise match and input match while sacrificing some gain, the noise figure result is probably not optimal due to the lack of optimal noise match data. In any case the results are quite usable as far as a demonstration of the design tool is concerned. There was no tweaking of any components in the final design.

9. Test equipment

Test equipment can form a very useful method of validating a design. Here are some devices which may prove useful, however a full description of test and verification methods is beyond the scope of this paper.

9.1 The MFJ SWR analyser:

Figure 16 - The MFJ 259 SWR Analyser

This device is claimed to measure impedances between 7 and 650 ohms. But measurements (tested with the B version) become inaccurate near these limits and even properly calibrated 50 ohm loads become somewhat inaccurate as the frequency of measurement increases. Also there is no sign for the reactance so it is difficult to determine if the load is inductive or capacitive. Both open and short circuits tend to read as a real resistance of zero ohms.

This unit has limited ability to measure one port S-parameters at HF frequencies and some self resonance parameters can be determined for larger components provided accurate ESR measurements are not required.

9.2 Vector Network analyser

This is an extremely useful device for characterising RF circuits and devices. It can accurately measure complex impedance from milliohms to tens of kilo ohms and sweeps from tens of KHz through to around 40GHz depending on the instrument. Unfortunately even older base model units on the secondhand market still attract prices starting at well over US$10,000 putting them out of the reach of most amateurs.

Figure 17 the Agilent 8753ES Vector Network Analyser

9.3 A Cheap network analyser?

Recently, Analog Devices released a device called the AD8302 gain and phase detector. This device performs two port measurement of relative gain to within +/-30dB and phase accurate to 1 degree at frequencies up to 2.7GHz. Paired up with a good directional coupler and a suitable swept or even manual frequency source, this device could perform some very useful device characterisations. The author is considering a project based on this device for a future Gippstech presentation.

10. Conclusion

Software simulation tools allow amateurs to test out circuits and scenarios quickly and easily and take a lot of the guesswork out of circuit and system design. When teamed with surface mounted components, which in many cases are accurately modelled by manufacturer’s data, a designer can try out “what if” scenarios quickly and accurately predict the performance of an RF circuit before construction.

With the aid of these new design tools and the availability of well characterised components, the task of designing higher frequency or microwave RF circuits has never been easier. These are empowering tools and many of them are free. Use them!

11. References:

The author would also like to thank Richard Gipps, from Trio Datacom for invaluable guidance in use of simulation tools and their limitations.