Docstoc

Cellular Phone Filtering_ RF experience

Document Sample
Cellular Phone Filtering_ RF experience Powered By Docstoc
					Cellular Phone Filtering: RF experience

                  By

         Kyriakos Sourounis

                  &

          Alexandros Pavlos

        ECE345 Senior Design

             Julio Urbina

            April 29, 2001

             Project #49
                                                                                                             ii
                                               ABSTRACT



       With such a limited understanding of RF engineering, little is known about how the signals of

desired frequencies are selected from the enormous range of possible choices. But lately, RF high

frequency filtering has become a field with increasingly growing interest. The following report will

explain one of the most common challenges of the field, building a Butterworth bandpass filter that can

perform predictably at high frequencies. The process consisted of three general steps. First, a series of

mathematical equations had to be applied in order to get a starting point for the component values and

circuit layout. Next, the data had to entered into a computer simulator for design verification. Finally

the circuit had to be built and tuned to meet the specifications. The challenges of this particular filter

design did not lie in the complexity of the circuit. Instead, the simulation, building and tuning of the

filter, provided the greatest challenge.
                                                              iii
                             Table of Contents




1. Introduction ………………………………………………………1
  1.1.    Functionality ……………………………………………………………...1
  1.2.    Project Goals ……………………………………………………………...2
  1.3.    Block Diagrams …………………………………………………………..2
  1.4.    Design Stages……………………………………………………………..4
         1.4.1. Calculations ………………………………………………………..4
         1.4.2. Software Simulations ……………………………………………...4
         1.4.3. Build Process ………………………………………………………4

2. Design Procedure ………………………………………………...5
   2.1. General Design Alternatives ……………………………………………...5
       2.1.1. Chebechev …………………………………………………………..5
       2.1.2. Bessel ……………………………………………………………….5
       2.1.3. Elliptical …………………………………………………………….5
       2.1.4. Butterworth …………………………………………………………5
   2.2. Calculations……………………………………………………………….6
   2.3. Software Simulations ………..……………………………………………8
   2.4. Building Process ………………………………………………………...10

3. Design Verification.………………………………………….….12
  3.1. Testing Procedure ……………………………………………………….12
      3.1.1. First Attempt / Debugging…………………………………………12
      3.1.2. Second Attempt / Debugging ……………………………………...12
      3.1.3. Final Testing……………………………………………………….13

4. Cost ………………………………………………………………15

5. Conclusion ………………………………………………………16

6. Appendix ……………………………………………...…………17
  6.1. Appendix A ……………………………………………………………...17
  6.2. Appendix B ……………………………………………………………...18
  6.3. Appendix C ……………………………………………………………...19

7. Reference ………………………………………………………..20
                                                                                                                1

                                              1. Introduction

        Many filters, whether they are of the lowpass, highpass, or bandpass types, are essential in the

operation of modern electronic circuits[1]. An example of such a filter application, is the cellular phone.

Due to the popularity of cellular phones, and because RF engineering is becoming an expanding field

with far reaching applications, a design of a filter was a project of great interest for us. Filter design is

often an unfamiliar aspect of engineering. This may be attributed to the fact that the mathematics

required are often quite involved, or that students may be unwilling to tackle such an unfamiliar part of

electrical engineering. This project is and attempt at providing an easy to follow design process for

creating a very high frequency (VHF) RF filter that can be used as a technical teaching tool for students

interested in getting started in the field.



1.1     Functionality

        A Butterworth filter designed at 100 MHz can be used as a 1st Intermediate Frequency (IF)

bandpass filter for modern cell phones, GPS receivers and even bluetooth receivers. It all depends on

the frequency scheme that the designer has chosen. A center frequency of 100 MHz is common, but

other frequencies such as 120 MHz or 60 MHz can be used. The choice depends on the application.

The most important concept to understand when dealing with the application of a 1st IF filter, is that such

a device doesn’t filter the original incoming signal from the antenna. Such a filter is used to filter out

undesired frequencies after the incoming signal has already been filtered at a higher frequency and then

mixed down to 100 MHz using a local oscillator. Hence the name, 1st IF filter.



1.2     Project Goals

               Build a Butterworth bandpass filter with a center frequency of 100 MHz

               Produce a passband of 10%(5% on each side of the center frequency),

                generating a 3 dB bandwidth between 95 MHz – 105 MHz
                                                                                                              2
                Obtain a maximally flat passband

                Achieve good characteristics at cuttoff

                Keep insertion loss to a minimum

                Attain a 40 dB rejection bandwidth of less than 40% from the center frequency



1.3    Block Diagrams

       The location of an IF Butterworth filter in a general radio receiver can be seen in Figure 1.1. As

the signal is passed through the mixer, it is filtered again by the Intermediate Frequency filter. Since the

signal is amplified right after the filtering stage, the insertion loss at passband frequencies is not of great

concern.




               Figure 1.1. The location of our Butterworth Bandpass Filter
                                                                                                         3
        Figure 1.2 describes the steps taken for creating and testing the high frequency filter.

Mathematical equations were first applied for setting the specifications and generating a general

schematic and values for the capacitors and inductors. Next, the data was inputted into Advanced

Design System (ADS), a software simulator. The filter was then built, tested, and tuned using a network

analyzer. Finally, it was confirmed that the circuit worked in the desired frequency ranges by passing

signals from a signal generator through the filter and displaying the output on a spectrum analyzer.




                                         Network Analyzer



                                                                               Spectrum
 Signal Generator                               Filter                         Analyzer




                Zverev                           ADS




Figure 1.2. Block diagram of Testing Procedure for Butterworth Filter
                                                                                                          4
1.4    Design Stages

       The approach taken for the design of the Butterworth filter can be described in three parts:

               1.      Mathematical Analysis

               2.      Computer Simulation

               3.      Build Process.

1.4.1 Calculations

       This stage involved the basic “number crunching” method. Using graphs and tables from

Zverev, predictions could be made for the amount of resonators needed, values of capacitors, inductors,

and Q values for a specific filter response [1].



1.4.2 Software Simulations

       Software Simulations involved the use of the software program called Advance Design System

1.3 (ADS). Applying s-parameter measurements using the program generated graphs that showed how

the filter handled ranges of signals sent through. With this information, the values of the components

were able to be “tuned” to produce the desired plot.



1.4.3 Building Process

       The building process involved the search for high quality components and the minimization of

the parasitic effects. Inductors needed to be shielded, and required a Q value of around 87. In addition,

it was crucial that they be adjustable for tuning purposes. Capacitors needed to be silver mica with

tolerances of 1-2%. The techniques used in the actual assembling of the circuit, minimized parasitic

effects as much as possible.
                                                                                                         5
                                         2. Design Procedure


2.1 General Design Alternatives

      In his book, Zverev describes numerous philosophies on filter design. He explains the

      characteristics, the advantages, and the disadvantages of various types of bandpass filters [1]:

2.1.1 Chebychev

      Advantages                                    Disadvantages
      - Best characteristics near cutoff            - More involved mathematics
      - Predictable ripple factor                   - Presence of ripple factor 
      - Minimizes the maximal deviation             - Insertion loss function
        from ideal response                         - Increased ripple results from
                                                              increased number of poles


2.1.2 Bessel

      Advantages                                    Disadvantages
      - More linear phase response than             - Poor attenuation characteristics
        equal order Gaussian filter                 - Wide passband beyond desired
      - Freedom from ringing or overshoots            frequency range
      - Smooth passband                             - Gentle slope at cuttoff



2.1.3 Elliptical

      Advantages                                    Dissadvantages
      - Steep stopband slope                        - Many components
      - Equal ripple in passband                    - Element tolerances extremely critical
        and stopband                                - Components with insufficient
                                                      Q values have drastic effect


2.1.4 Butterworth

      Advantages                                    Disadvantages
      - Good phase response                         - Poorer characteristics around cutoff
      - Straight forward math                       - Danger of spikes occurring in
      - No insertion loss function                    frequencies far beyond cutoff
      - Smooth passband
      - Emphasis on fc characteristics
                                                                                                            6
2.2    Calculations

       After determining the type of filter that would be designed, the specifications of the filter were

obtained so that the procedure for calculating the component values with specified Q value could take

place. With the set specifications, the Q needed was 90 with inductor values of 75 nH [1]. This gave an

approximate insertion loss of 3.5 dB at a center frequency of 100 MHz. The response form factor was

calculated using Eq. (2.1) to show how many resonators were needed to achieve the set parameters.


                                      bw40 dB / bw1dB  36 / 8.4  4.29                              (2.1)

       Goals were set at 10%, or 10 MHz for the 3 dB bandwidth, and 36 MHz for the 40 dB

bandwidth. The 1 dB bandwidth was set at 8.4 MHz for the creation of a maximally flat passband. The

minimum power, Amin equaled 1 dB and the maximum power equaled 40 dB. With these values, the

number of resonators was determined from the data in Appendix A. Using the above information, the

number of resonators required for this filter was 4.

       With the known bw1dB, and through the use of the graph of Appendix B, the 3 dB bandwidth was

determined. In our case it came out to be 10 MHz. This technique also illustrated an attenuation of

approximately 42 dB in the stopband. Using Eqs. (2.2), (2.3) and Appendix C, the k and q values were

selected.

                                  bw1dB
                                         .84  since bw1dB  8.4 MHz
                                  bw3db                                                              (2.2)
                                                   bw3dB  10.0MHz


                                 bw401dB                                                             (2.3)
                                          3.6  for n  4
                                  bw3db
                                                  attenuation  40dB
                                                                                                             7
       The number of resonators corresponded to the Qmin value that was used for the internal

resonators through Eq. (2.4).


                                                        fm           100                             (2.4)
                                          Qm in  q0          8.710      90
                                                       bw3dB          10


       Next, the value for q0 was selected and used to transverse across the chart shown in Appendix C.

The values from the chart were:

       q1=.5417
       q4=1.8605
       k12=1.0411
       k23=.5373
       k34=.6992

       Using the above values and Eqs. (2.5) through (2.11), the values for coupling and shunt

capacitors were calculated. The generated values were:
                         f                     10 
               C12  k12 3dB C I C II  1.0411      33 .8  3.5189 pF                              (2.5)
                           f0                   100 
               C a  C I  C12  33 .8  3.5189  30 .281 pF                                         (2.6)
                             f 3dB                     10                                         (2.7)
               C 23  k 23            C I C II  .5373      33 .8  1.816074 pF
                              f0                        100 

               Cb  C I  C12  C 23  33 .8  3.5189  1.816074  28 .465 pF                        (2.8)
                             f 3dB                     10                                         (2.9)
               C 34  k 34            C I C II  .6992      33 .8  2.363296 pF
                              f0                        100 
               C c  C I  C12  C 23  33 .8  3.5189  1.816074  28 .465 pF                       (2.10)

               C d  C I  C34  33 .8  2.36296  31 .43704 pF                                      (2.11)
       The resulting circuit with the calculated values for the inductors and capacitors, along with their

respective locations is shown in Fig. 2.1.




                               Figure 2.1. 4-pole butterworth bandpass filter
                                                                                                           8

         The generator impedance on the first and last resonator had to be transformed in order to produce

the desired quality factor. This was accomplished through Eq. (2.12).


                                               C1C2                                                   (2.12)
                                                       Ca
                                              C1  C2
         By arbitrarily setting C1=100 pF and using the previously calculated value of Ca, the value for C2

came out to be 43.43 pF. This created a circuit that was 50  matched on either side.

         After the completion of the design stage, the next part of the process involved the simulation.

Using the calculated values and a partial design, ADS simulated the results for the intended frequency

range.



2.3 Software Simulations

   After entering the calculated values, ADS was used to plot insertion loss graphs, and phase plots.

Figure 2.2 shows the circuit during an S-parameter analysis.




            Figure 2.2. Circuit design as seen in ADS


         ADS has an array of options that allows the user to specify exactly what the requirements for a

filter are. It also has the capability of changing component values on the fly and simultaneously

showing the effects of the changes on the plots. After entering the values obtained from our
                                                                                                             9
calculations, the graph displayed by ADS was not optimal. The circuit had to be “tuned” by varying

the values of the capacitors and inductors until an optimal insertion loss curve, with a very linear phase

graph was produced. The optimum simulations can be seen in Figs. 2.3 and 2.4.




                Figure 2.3. Optimal Insertion Loss graph seen in ADS




                                           Linear




                      Figure 2.4. Phase Graph from ADS showing linearity

       After varying the values of the capacitors to approximate real world values, the data was taken

and the building process was ready to commence. ADS is capable of adjusting the values so as to

observe the effects of the filter’s response with slight changes in the values of the components. This

helped in getting closer to the desired response in the fewest amount of building attempts.
                                                                                                            10


2.4 Building Process

         When dealing with a high frequency filter, or any other circuit for that matter, it is important that

it be built with frequency response in mind. First of all, the circuit must be built on a RF board, not a

proto-board which is commonly used in ECE classes. A RF board contains 50  matched transmission

lines, whereas connections on the proto-board must be made using wires, which become capacitive at

high frequencies. The components chosen must be of the highest quality obtainable for the filter’s

characteristics. Capacitors for example, must be made of silver mica as opposed to the more common

ceramic disk material. Even when dealing with only silver mica capacitors, the tolerances must still be

monitored. For our application, we tried to keep capacitor tolerances at 2% whenever possible.

Techniques to offset of effects of poorer tolerance components are explained in the testing and tuning

section. When choosing the inductors, it was important to keep in mind the Q value obtained from our

calculations [1]. The value calculated was about 87. Another important decision we made, was getting

grounded/shielded inductors, which allowed us to place them literally touching each other on the RF

board.

         The layout of the circuit on the RF board was another important factor in the building process.

In order to minimize parasitic effect, we needed to minimize the amount of wire on the circuit. This was

successfully accomplished by placing the inductors right next to each other. We were therefore able to

place the coupling capacitors right on the inductor leads and the shunt capacitors straight from those

same leads to ground, effectively minimizing the amount of wire present in the circuit due to lead

lengths. The inductors we chose not only had to be shielded and grounded, but they had to be variable

as well. The four “tunable” inductors allowed us to adjust the values in order to align the poles or the

filter for the best response.
                                                                                                         11
                                         3. Design Verification

3.1    Testing Procedure

3.1.1 First Attempt / Debugging

       Once the circuit was finally completed, the testing process began. A network analyzer was used

to give the response of the filter through the desired frequency range asked of it. For the actual tuning of

the circuit, the range was kept between 75 MHz and 125 MHz. The first concern was obtaining a good

response shape for the filter around the center frequency and the lower/upper stopbands. The second

concern was to focus on the insertion loss at the center frequency and the stopbands. The first attempt at

building the circuit was not a great success. The response had an insertion loss at a center frequency of

around 17 dB. Although for a 1st order IF filter, insertion loss is not of the greatest concern, this number

was unacceptable. The poles were off, and despite a great deal of time spent on tuning, the response of

the filter was quite poor.

       When looking at the circuit, numerous areas seemed to be contributing to the problem. The first

thing that was underestimated on the first attempt was the lead lengths of the capacitors. The lengths

were around 8-10 mm long, this was deemed to be acceptable. With a lot of patience however, the

circuit was re-built with even shorter lead lengths. In addition to the decreased lengths of the capacitors,

the amount of solder used for the connections was also reduced. Whenever possible, 50  transmission

lines were used instead of wires.

3.1.2 Second Attempt / Debugging

       The second attempt at building the filter produced a much better response. The insertion loss at

center frequency was now at an acceptable level of 4-8 dB. The stopband slope was relatively steep,

and the passband was much smoother than the one from the previous attempt. Still, the filter seemed to

provide the closest to ideal shape of a Butterworth filter when tuned for a center frequency of 110 MHz.

Further research, along with manipulating values for ADS showed that the “real world values” signed by

ADS for our capacitors were consistently lower than the “optimal values.” Using ADS with actual
                                                                                                           12
values limited to in the circuit also shifted the simulation to the right by 10 MHz. In an attempt to

remedy the problem, variable capacitors or “trimmers” were introduced into the circuit.

       The trimmers were used in place of some of the smaller capacitors of 2-4 pF. And placed in

parallel with the larger capacitors. In both cases they allowed for gradual increases in capacitance. The

response could now be moved closer to the desired center frequency of 100 MHz by varying the

trimmers. The new success was not without tradeoffs however. Previously, the circuit had only four

variable components in the variable inductors. Now 5 trimmers were added into the equation. This

made the filter more sensitive to changes due to the fact that 9 components had to be continuously varied

and kept track of. In essence, it produced a much more complicated tuning situation.

       When it was finally correctly tuned, the response of the filter was a excellent. A choice could

have been made between a lower insertion loss at the center frequency, and a higher insertion loss at the

edges of the passband, or a more flat passband as in the choice that was made. The final insertion loss

was quite low at only 5.8 dB across the band. The insertion loss at the cutoff frequencies was below the

maximum allowed limit of 40 dB, and the curve looked very similar to the one simulated on ADS.

Through the use of the variable capacitors and inductors, many of the parasitic effects were overcome.

3.1.3 Final Testing

       When the desired response was obtained, the filter was tested to see if the correct frequencies

would actually pass through it. A signal was sent from the signal generator to a power meter in order to

calculate the loss of the connectors and lines. The signal from the signal generator was then passed

through the filter and the output was displayed on the spectrum analyzer. Using the data from the

network analyzer, at any given frequency the insertion loss was known. When this number was added to

the loss of the cables and connectors, the number coincided, with the loss shown on the spectrum

analyzer. As hoped, the filter passed the desired frequencies and filtered out the rest. The response of

the filter in ADS can be seen in the graphs of Figs. 3.1 and 3.2:
                                                                   13


              -5.8499dB     -5.2159dB       -5.8516dB
                              10 MHz




Figure 3.1. Insertion loss Graph from Network Analyzer with data




              Linear




   Figure 3.2. Phase Graph from Network Analyzer with emphasis
               on the linearity through the passband region
                                                                                                   14
                                              4. Cost


Description                                                  Quantity    Price/Unit   Total

Parts
  Software
             Advance Design System 1.3(estimated)                1         $2000      Available

   Reference Books
             Handbook of Filter Synthesis                        1         $350       Available

   Hardware Parts
             “Unicoil” 10mm tuneable inductor                    4         $2.60       $10.40
             2 pF silver mica capacitor(estimated)               2         $0.20        $0.40
             4 pF silver mica capacitor (estimated)              1         $0.20        $0.20
             39 pF silver mica capacitor(estimated)              1         $0.20        $0.20
             51pF silver mica capacitor(estimated                2         $0.20        $0.40
             100 pF silver mica capacitor(estimated)             2         $0.20        $0.40
             solder                                            1 roll    Available    Available
             copper tape(estimated)                              1       Available    Available
             trimmer capacitor                                   4       Available    Available
             RF board (estimated)                                          $3.00        $3.00

   Hardware Equipment
             Hewlett Packard Network Analyzer(estimated)         1      $19,000.00    Available
             Hewlett Packard Spectrum Analyzer(estimated)        1       $8,000.00    Available
             Signal Generator(estimated)                         1       $9,000.00    Available

Labor (13 weeks)
   Alexandros Pavlos                                            100        $100/hr      $25,000
   Kyriakos Sourounis                                           100        $100/hr      $25,000

Parts Subtotal                                                                           $15.00

Labor Subtotal                                                                          $52,000

Project Total                                                                         $52,015.00
                                                                                                             15
                                                 5. Conclusion

       The design for the 100 MHz filter outlined above may seem to be at the lower end of the

frequency range of what many in the industry may be working on. In reality, because it is an IF filter, it

can be used in applications such as cell phones in the UHF range. This paper can be used as an

extremely useful tool for students in college level courses to better understand the intricacies of filter

design at high frequencies. The parasitics that were involved with this design played a major roll in not

obtaining a better insertion loss. If it were to be built again, a distributed approach would produce a

better result and the specifications could be set at a higher level than what was set going with a lumped

approach. Along the design path, certain areas of improvement that could have also been taken

advantage of became apparent. Such as the effects of Electromagnetic Forces (EMFs) involved at times

with the circuit could have been greatly reduced by placing a metallic shield around the circuit to block

them out. Also if it were possible to obtain higher Q value components tuning would have been much

easier. Other options that could have apparently helped would be to improve the transmission lines

between the components and possibly laying out the circuit in a different manner that would have been

more compact and help shorten the leads even more, thus further minimizing parasitic effects. Still,

even without these options the circuit successfully stayed within the parameters that were set out in the

beginning, and even outperformed in some aspects as in the stopband.

       There are many modifications that can be applied to filter design. The process outlined here is a

method that would be an excellent way of showing students how filter design is at its most raw form.

Ways of applying this project would be to take it into a classroom setting with minor details changed,

and possibly adding to the project by applying a distributed approach as part of the project to be

completed after going through the lumped component method. This way a much more detailed

understanding can be presented to students who are interested in the field of filter design as a possible

career path, as most of the aspects in this project are still used by engineers in the work place.

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:15
posted:3/16/2010
language:English
pages:18