Designing and Building High-Performance Crystal Ladder Filters Designing crystal filters for SSB is made easier using readily available software. cepite the several excellent ar- ticles about crystal filters _ that have been published in amateur magazines over the years, building high-quality erystal filters is still een by many amateurs as either black magic or asa complicated proce- dure beyond the reach of the average builder. ‘A crystal filter, being the beart of « superhetorodyne receiver, has a pro- found effect on its selectivity. A low- quality crystal filter in even a high priced commercial transceiver can degrade ite selectivity and dynamic range. On the other hand, a good erys- tal filter can significantly enhance receiver performance, whether in a ple “weekend” project or in a com- t Canyon Dr Burbank, CA 91504 -By Jacob Makhinson, NSNWP- Commercially available crystal fil- tera are uaually expensive and often discourage constraction-minded ama- teurs from pursuing projects that include crystal filters. In addition, studies conducted in recent years con- clude that in a high-performance re- ceiver, a erystal filter may become the “hottleneck” restricting the receiver's dynamic range. So, the goal of this article is to provide design and build- ing methods that can be used to con- struct erystal filters that rival or ex- ‘coed the quality of commercially avail- able filters. I will describe a simple, practical step-by-step procedure to ign, construct and align erystal fil- ters using equipment available to ‘most construction-minded amateurs. ‘The resulting filters achieve top-qual- ity performance at a fraction of the filters Most of the crystal filters described Page 3, QEX, January 1995, published by The American Radio Relay League, inc. in amateur projects and those being sold commercially are lattice, half tice or cascaded half-lattice filters like those shown in Fig 1. A two- or four- crystal filter ofthis type can provide a symmetrical response with reason- ably steep skirts, But the bandwidth of such filters is a function of the fre- quency separation of the crystals. Ifa steeper response is desired, designing alf-lattice filter with more than four crystals becomen more complex, re- quiring matched pairs of erystals and several adjustments. While it. is reasonably easy to obtain matched crystal pairs for CW filters, it becomes considerably more problematic to ob- tain pairs of crystals separated by @ couple of thousand hertz for usein SSB filters. In addition, the coile used for lattice filter alignment often use small cores, which can result in the degra- dation of dynamic range because of ‘core saturation at high signal levels. January 1995 3Another form of filter—which is the subject of this article—is the lad filter shown in Fig 2. It typically has an asymmetrical response and is sometimes called the “lower-sideband ladder” configuration. But as we'll see, with a eufficient number of poles this asymmetry is significantly reduced. Ladder filters offer soveral advan- tages to the amateur experimente: ‘© there is no need to pick crystals for proper frequency separation ‘and no need for matched crystal pairs: * the inherently simpler filter topology results in simple con- struction methods; required after alignment is com- pleted; ‘the absence of coils allows a com- pact assembly and reduces the possibility of dynamic range degradation; ‘the simple topology is conducive toa high number of poles, which allows very steep skirts; and Hy lye 1) Lattice Goyette bm ave] 2) Halt-Latice CryattFiter i ©) Comcde HatLatt IK 0 cotter Fig 1—tattice eryst er circu 4 QeEx ‘+e computer program is available that eliminates the need for em- pirical approaches or cut-and-try ‘methods and allows the designer to shape the filter response with great accuracy. ‘This work was inpired by an article by Bill Carver, K6OLG/7.1 Carver's work is quite ‘remarkable; firat, it proves that itis possible to build high- quality CW and SSB crystal filters with a predetermined frequency re- sponse “without black magic,” and sec- cond (but of no less importance), it proves that the performance of filters built in a home lab using home-built ‘equipment successfully rivals that of filters built using sophisticated pro- ‘refines the crystal filter design crite: ria and methodology, walks the reader through a complete design cxample, provides the results of measurements, ‘on several crystal ladder filters and ‘analyzes the result ‘The scope of this study has been lim- applicable to CW filters ‘The computer-design stage is based on a collection of computer programs _ designed by Wes Hayward, W7ZOI. ‘The ARRL has just republished Wes Hayward’s textbook Introduction to Radio Frequency Design, now includ- ing the software as part of the pack- age.? The computer programs (which 1 will refer to as IRD) run on an IBM PC or compatible computer. The com- puter requirements are minimal, since IRFD fits on a single floppy disk and the computer's speed is of no concern. A VGA card is required for graphic display, hows ‘The Design Procedure Design and construction of these ladder crystal filters are performed ‘using these steps: ‘© election of the filter center fre- quency; + measurement of erystal param- ‘Notes appear on page 17 eer tk Fig 2—Circult of a ladder eryatat fiter. Page 4, QEX, January 1995, published by The American Raclo Relay League, Inc. lection of the shape of the response; ‘+ computer design of the filter; and ‘© construction and alignment. Frequency Selection Ifthe required filter frequency is not already defined, you can select an IF to suit your needa. In doing #0, con- sider that certain frequencies may re- sult in in-band intermodulation prod- ucts. Tables and charts have been developed to help designers avoid these frequencies. ! Practical consider- ations also impose some limitationson IF selection. ‘The crystals used in color-burst gen- eratorsat 3.579 and 4.433 MHzare the is ve cryetale around and the required termination resistances of filters built with such erystale may excoed 10 ka, which necessitates an impedance transformation with a very high ratio (for a 50-0 system). As a result, very high voltage levels may be developed atthe filter input, which may cause an overload condition. In addition, the required values of the coupling capaci- tors may be under 6 pF, making con- struction difficult due to stray eapaci- tances. For these reasons, crystal filters with center frequencies under 6 MHz are not recommended. ‘The useful upper frequency limit is determined by the influence of stray ‘capacitances at frequencies above 10 MHz and by the limitations im- posed on the VFO circuit for multiband HF operation. Consequently, the ree- ‘ommended frequency range for an HF SSB crystal filter is between 6 and 12MHz. The remaining criteria for the crystal frequency selection are the crystal Q and the price. Microproces- sor crystals in HC18/U or HC49/U cages are reasonably inexpensive, but, being manufactured in large quanti- ties, they are optimized for param- eters other than Q. Q is typically not specified by the manufacturer, and it varies signifi- cantly from batch to batch and from ice to device within a batch. There- fore, the only way to find the Q of ecific type of erystal isto obtain eral samples and to measure the pa- rameters. This should be done before buying a large batch of crystals, Toriginally intended to build erystal filters at 9 MHz, which is a popular IF within the amateur community, but it turned out that all the 9-MHz eryetals T obtained (from different vendors)had Qs below 80,000. On the other hand, I found 8-MHz crystals with much higher Qs, so all of the crystal filters described in this article are built using 8-MHz (series resonance) ceryst Crystal Parameters ‘The equivalent cireuit of a quartz crystal is shown in Fig3. The computer software we will use to design the lad- der filter requires entry of the crystal parameters. These parameters are easily measured with the use of a lab- quality impedance analyzer, but they also can be measured quite accurately using home-built equipment. The test equipment required to measure crys tal parameters has been described previously and is beyond the scope of this article.!45 ‘The parameters needed for the design process are: AF, the frequency offset or deviation from the specified center frequency; r, the series resistance of the crystal; fr and fu, the 3-€B points required for the Q calculation; and Lq, the motional inductance, which is derived from the Qand r. C,, the parallel or “holder” capacitance, can be measured, but an assumption that C, is 5 pF (which 1 verified for several crystals in HIC49/U cases) appears tobe adequate in most cases. ‘There are several practical consid- erations in selecting and handling the you will use. For ane thing, the design process is easier ifthe crys tals tobe used ina particular filterare selected from a large pool of crystals. Although it's not essential, the ery tals can be matched for Q, Lig and AF. Buying a large batch of crystals may providea volume discount and furnish you with several sets of filter erysta. To illustrate the point, I bought 100 &-MHz crystals for a total of $60. Out ofthat batch Lused 14 erystals with an average Q of 145,000 for a L4-pole fil- ter, 10 crystals with an average Q of 123,000 fora 10-polefilter and l2crys- tals with an average Q of 110,000 for a 12-pole filter. Irejected the remaining G4 crystals, Three high-quality crystal filters for $60—not bad! ‘The crystals should be tagged before measurement, and the measurement results should be logged for future use Invest sufficient time in this initial stage ofthe design since the accuracy ofthe data will affect the shape of the frequency response. Take care to avoid heat transfer from your hands to the erystal cases, and allow 40 to 60 seconds between handling the crystals and performing the measurements to let the resonant frequency stabilize. ‘Once the measurements are com- pleted, a preliminary group or groups oferystals with sufficiently high Qcan be identified (grouped by Q within a certain range). Caleulate the average motional inductance (Ling) and aver- age Q (Qu. for each group of erysta Filter Selection ‘An important part of the design pro- cess is selection of the desired filter parameters. The parameters you need to select are: ‘the filter response type— Chebyshev, Butterworth, Gaussian, ete, the number of poles, * the filter bandwidth, and ‘the value of terminating resi tances, ‘The Chebyshev response with 0.14B of ripple is the most commonly used response type for SSB HF filters. It may be advantageous in the final fil- ter design to deviate slightly from the 0.1-dB ripple value to obtain mare con- venient values for the end coupling capacitors. Decreasing the ripple level value will resultin aslightly smoother frequeney response but will degrade the shape factor; an increase in the ripple value causes the opposite effect. Several factors have a significant influence on the number of poles cho- sen for the crystal ladder filter: ‘© the desired shape factor, ‘© the insertion loss, the degree of asymmetry of the frequency response, construction considerations, and tthe size and weight (for portable use). ‘The shape factor is defined as the ratio of the filter bandwidth at a level of -80 or ~60 dB to its bandwidth at the ~6-dB level. In this article I'll use the -B0/-6 dB shape factor, SFys0 = Bf s/f 6 The required shape factor depends ‘on the complexity of the receiver, its architecture and its specifications. Filters with more poles have better shape factors. (For example, the XF-9B10, a 10-pole, 9-MHz SSB filter manufactured by’ KVG Inc, has SF eso = 1:1.8.) Fig 4 may help you select the needed number of poles, Chabyater 10 ‘iter insertion Lowe warnun est Ty PROSPER SERERRRA Fig 4—Graph of ladder filter shape factor (6:00) versus the ‘number of poles for ‘Chebyshev respon ‘pple. with 0.1 dB of Page 5, QEX, January 1895, published by The American Radio Relay League, Ine. Fig 5—Graphs of fiter Insertion loss versus crystal Q. January 1995 5based on the desired shape factor. Receivers having filters with good shape factors exhibit improved selec- tivity and a distinct “crisp” quality when tuning from one station to another. Insertion loss, while strongly influ- enced by the Q of the crystals, also depends on the number of poles. The curves in Fig 5 show these effects. All other factors aside, the availability of ‘good-quality crystals may be the de- ciding or limiting factor in selecting the number of poles, As shown in Fig5, a 14-pole filter made of crystals, having Qs of 160,000 has the same in- sertion loss as a 10-pole filter made of crystals having Qs of 110,000, The curves in Fig 5 were generated using IRFD and represent theoretical caleu- lations performed by the program. Due toseveral limitations I'l discuss later, the practical results differ somewhat from the predicted values. But it is possible with a good degree of accuracy to make an estimate of the practical value of the insertion loss by adding 0.8 dB to the value obtained from Figs. ‘The asymmetry of the frequency re- sponse is inherent in crystal ladder filters, but by increasing the number Considering all the factors listed above, I recommend keeping the num- ber of poles between 10 and 14. There is an advantage in having an even number of poles—it results in a sym- metrical design, minimizing the aum- ber of different capacitor values needed. Several factors influence the choice of bandwidth of the crystal iter: + the desired selectivity—nar- rower filters may be preforred for contest work while wider filters may be more appropriate for casual rag-chewing, + receiver sensitivity and dynamic All of the filters discussed in this article have a bandwidth of 2500 Hz— mostly due to the last factor. ‘The value of the terminating resis- tance should be as low as possible to ‘minimize the transformation ratio of the impedance-matching transform- ‘ers, and the value should exceed the IRFD recommended value by an amount that ensures that the end cou- plingeapacitorsareat least 16 pF. You should choose the lowest impedance value consistent with these two crite- ria from Table 1. (The transformer ratios assume you want to match the filter to 60-0 source and load imped- ances.) Computer Design Before starting the computer de- sign, make sure you have calculated the average Q (@.y) and motional in- ductance (Ley) of the crystals and have selected the desired bandwidth, number of poles and the ripple value. ‘Note that if the actual desired band- width is used directly in the crystal filter design, the filter designed by the IRFD program will have a bandwidth narrower than predicted. (This occurs because of simplifying assumptions used in the equations and the use of & ‘Table 1—Termination Resistances and Transformer Design Impedance Termination Primary —_‘Sacondary of pole it is possble wo overcome this cto” eistance (a) tums ‘ume dbwvious ina 1Ospoe iter, it becomes 75 ‘ 5 stmost unnotieablein a d-pae fiver {0 5 ; Construction considerstions wil 450 a : undoubtedly differ from one builder ta 200 ‘ 8 another, but one observation is worth 250 a 8 mentioning noticed during the con- a2 ‘ ‘0 struction of several ladder filters that 378 4 u to achieve good ultimate attenuation 450 4 12 {more than 120 dB), the requirements 00 ‘ is for shielding between filter sections ore ‘ is tre considetly move stringent fora 00 i 3 Iospole ter than fora 12 oF poe coo ‘ is fee I 2 awe oe Lo Bigg Oraphetencuaon bandwith verun crystal fora Fig 7—Graph of clculaton bandwith vraue crystal fore Hann poet Sebasin Tepole ie 6 QEX Page 6, QEX, January 1995, pubfshed by The American Radio Relay League, Ine.tals.4)I've developed correction factors to compensate for this discrepancy. ‘The bandwidth used for the computer calculations will be referred to as the calculation bandwidth (BWe), which canbe obtained from the charts in Figs 6 through 8 for a desired bandwidth (BWa) of 2500 Hz. To illustrate the design process, let's walk through the steps of an actual filter design. We'll begin by listing the design parameters: ‘Number of poles (N):_ 12 Desired bandwidth (BWa): 2500 Hz Filter response type: Chebyshev, 0.1.48 ripple Crystal (Quy): 110,000 ‘Motional inductance (Lin.): 0.0155 H Parallel eapacitance (C,): 5 pF Nominal crystal frequency: 8 MHz Four programs from the IRFD soft- ware package will be used to accom. plish the design: GPLA, the general purpose ladder analysis program; the low-pass filter design program; X, the ladder erystal filter design pro- gram; and MESHTUNE, a utility for tuning meshes in a crystal filter. In each program, the menu selections are selected by typing the appropriate letter key. Entry of numeric valu is done by typing the value, then pressing the Enter key. (See the IRFD2MAN.TXT file supplied with IRFD for details of program op- eration.) To start the design process, use the L program to generate k and q values forthe filter. Run L, then select K from ‘the menu. Type 12 for the number of poles, then .1 for the ripple, in dB. (The progeam calculates the needed values and stores them in a disk file for the X program to use.) Press the Enter key until the L program exits Next, run the X program to perform the ladder filter design. Select K from the menu to load the & and g values from the disk file. Type 12 for the num- ber of meshes, 8 for the nominal erys- tal frequency, 0.0186 for the motional inductance in henries, 1 for the over- tone, § for the parallel capacitance in pF (assumed), and 110 for the crystal Q in thousands, X now wants you to enter thek andg values. Since they are on disk, you can just press the Enter kkey repeatedly to load the data gener- ated by the L program, until X stops asking for the values and instead asks for the bandwidth in hertz. The band- width you enter, 2700, is the band- ‘width obtained from the chartin Fig 7, fora 12-polefilter witha desired band- width of 2500 Hz. Press Enter to get to the souree termination resistance prompt. Although the end resistance value given by the X program in this step is 206 Q, an attempt to use the 250 Q given in Table 1 fails—the ac- ‘tual minimum possible termination is 2740. But the use of that termination value would make for very small (2.2pF)end coupling capacitors, so the next highest termination value is selected from Table 1. Enter $12. X will next ask for the termination value for the load end of the filter. Just press Enter, as 312 is now the default. At this point the values of the coupling capacitors are displayed on the screen. You can print a hard copy by using your computer's print-screen key. The ‘values of all the capacitors are practi- cal and can be easily realized with either a single capacitor or two capaci- tors in parallel. The values of the coupling capacitors will not be altered during the alignment stage and should be considered final. Pressing Enter displays the second set of data vital to the dosign: the mesh offset frequen- EEPDGGE mma x 000 Fig 6—Gray of calculation bandwidth versus crystal Q for a 2500-Hz, 14-pote filter, Page 7, QEX, January 1995, published by The American Radio Relay League, Ino. cies. You should print this screen as well, as this data will be used later in the design procedure. Another Enter displays the menu. It’s useful at this point to closely examine the coupling capacitor values. Ifthey appea factory, you are done with this stage of the design and can save the designed filter to disk by pressing the D key. If ‘you want tochange the design, youcan alter the termination resistances with the R key or change the calculation bandwidth with the W key. Altering the termination resis tances affects only the values of the end coupling capacitors; it is easy to vary these values and see theeffect,s0 feel free to experiment. Altering the bandwidth affects all of the coupling capacitors. You shouldn't change the bandwidth by more than about 30 Hz from the initial value since it will in- validate the later stages of the design procedure, Changing the ripple value is another option, but the design will have to be repeated from scratch since the k and to be changed by the Z ible values for the filter ripple may be between 0.07 and 0.15 aB. Be sure to store the final design in a file using the D key. For this example, ‘we'll call the file 18POLI. (The X pro- gram will automatically add .CIR to the end of the file name.) The stored file can be viewed or edited with a text editor. Make a hard copy of the file for future reference. We can investigate the response of the filter we've designed using the GPLA program. If you have « VGA display, GPLA can plot the frequency response of your screen. Run GPLA and read in the saved circuit using the R key. Once the file has been read in, press the P key to select plotting, then press the H key to get the filter gain plot. The filter response should be dis- played. After viewing the plot, press the Esc key to return to the menu. To find the theoretical insertion lose of the filter, we need to adjust the sweep parameters to geta close-in look at the pass-band response. Press 8 to set up the sweep, type 400 for the beginning frequency, 3500 for the end frequency, 80 for the frequency step, 310 for the iid spacing, 10 for the sereen bottom, and 1 for the dB/division. Then press P followed by H. The theoretical inser- tion loss of the filter is the distance from the top of the plot to the highest point on the response curve, at 1 dB per division, The loss appears to be vanuary 1995 7slightly over 2 4B. A practical value for the lose may be estimated by adding 0.8 dB to the theoretical loss, for @ result of about 3 4B. Toexit this sereen press Ese. Iferystal resonators with frequency offsets equal to the mesh offset fre- quencies listed by the X program were available, the computer design would be finished at this point. Since that unlikely, additional stepsare required in order to tune the crystals to the required offset frequencies. making the use of crystals with random (but known) offset frequencies possible To accomplish this, tuning capaci- tors are inserted in series with the crystals. These capacitors allow con- siderable design flexibility and tune every eryatal to the same “loop fre- quency." This modification results in the schematic diagram shown in Fig 9. Our next step is to “build” this new filter circuit for analysis using a util- ity program and the IRFD software. We will do this by creating a modified copy of the 12POLL.CIR circuit file. Our new cireuit file adds series tun- ing capacitances to the circuit and changes the original crystal offset fre- ‘quencies calculated by the X program tothe actual measured offset frequen- cies (AF) of the crystals we will use in the filter. All of the values we need to deal with are shown in Table 2. If you investigate the coupling c pacitor values and the mesh offset fre- quencies shown in Table 2 (or on your hard copy from the X program), you'll find that the values are symmetrical around the middle. Although it’s not ‘ossential, it is helpful to arrange the an attempt to pre~ serve the symmetry. Columns A and B in Table 2 help to illustrate the ar- rangement. The two crystals with the highest positive frequency offset are placed at the edges (X2 and X35). The twoerystals with the highest negative offset are placed next (X5 and X32). ‘The remaining crystals are arranged ‘monotonically and symmetrically around the middl ‘The CLFMOD program (see Listing 1) is a simple utility program written in BASIC by Jon Bloom, KE9Z, of ARRL HQ-*It reads the original filter circuit file output by the X program ‘and writes the modified circuit file we will work with (call it 12POL2.CIR). This file adds the tuning eapacitors to the original filter circuit, setting the value of each tuning capacitor to 200 pF. It will ask you to enter the mea- sured frequency offsets of the crystals, as shown in column B of Table 2, and it will calculate and display the com- bined offset shown in eolumn D of Table2. These values will alsobe writ- ten to the OFFSETS.CLF file. (It's helpful to make a table like Table 2.20 ‘you don't get lost.) Note that if you arranged the crystals as recom- mended, the spread of the offset frequencies in the middle of column D is minimized, which leads to a narrow spread of the eventual tuning capaci- tor values, as shown incolumn F.(The coupling capacitor values obtained ‘earlier are presented in column J.) Now the values ofthe tuning capaci- tors should be calculated and entered into the circuit, We will use the oo See eee. ] ceca aye me tT Fig Schematic diagram of the example 12-pole crystal filter. Fig 10-1 luction graph. The needed frequency offset can be obtained ‘once the desired bandwidth reduction is found. 8 OEX Page 8, QEX, January 1996, published by The American Radio Relay League, Inc. “* HaIH ., HR o Fig 11—Filter alignment circuits. The resonant frequency of the reference oop is measured using the circuit at A. For the other loops, the required tuning capacitance ia found by adjuating the Variable capacitor at B to get the same resonant frequency.MESHTUNE program to find the required tuning capacitor values. Run MESHTUNE and press B to start a ‘new mesh. Types for the nominal crys- tal frequency and press Enter for the crystal offset frequency. (Do not enter ‘an offset frequency!) Type 0.0155 for ‘the motional inductance and 0 for the crystal parallel capacitance. Press Enter to get lo the interior/exterior prompt, then press I. Enter 10000000000 1 followed by 10 zeroes) for the left-hand coupling capacitor and the same value for the right-hand coupling capacitor. Enter 800 for the initial value for the tuning capacitor. For the target offset frequency, enter the target offset frequeney calculated by CLFMOD for this mesh (column D of Table 2). The actual mesh offset fre- ‘quency and the target offeet frequency will be displayed. Note whether the ‘mesh is tuned too low or too high. (It’s toolow in thiscase.) Press Enter to get back to the menu. If the mesh was tuned too'low, press the down arrow key to reduce the tuning capacitor; if the mesh is tuned high, press the up arrow key instead. The new tuning eapacitor value will be displayed, as will the target and new actual mesh frequencies. Continue pressing the same key until the mesh frequency crosses over the target frequency. To fine-tune the capacitor value, press the 8 key to set the tuning ca- pacitor step. Enter 0.1 for the step ize. Now, using the arrow keys, tune back across the target frequeney until it again changes from too high to too low, or vice versa. At the point where it changes, the displayed tuning ea- pacitor value will be within 0.1 pF of the perfect value—close enough! Enter the final capacitor value into column, F of the table. To obtain the next tuning capacitor value, start over by pressing B. Note now that most of the values you en- tered previously default to the correct ‘numbers, so you can just press Enter for those parameters. Remember to enter the correct target offset fre- quency for the crystal you are tuning. Repeat the tuning process. For some crystals, the required value of tuning capacitance may be quite large. In such cases, you may want to increas the tuning step size using the S com- mand. It's easy to watch the actual ‘mesh froquency as you tune und get a feel for whether you need to raise or lower the step size foreffective tuning. ‘The above tuning procedure may seem cumbersome, but once you get the hang ofit, things go pretty quickly. ‘We now need to create the final fil- ter circuit file so we can check its re- sponse using GPLA. Use a text editor such as DOS EDIT or Windows Notepad to open the 12POL2.CIR file. Each tuning capacitor entry consists of three line: cap 200 where is the reference designa- tor of the component (e.g., C3). For each tuning capacitor, C3, C6,...C36, change the line reading 200 to the value for that capacitor from column F of Table 2. For C3, the result is: cap 8 279.8 Do not change any of the other lines in the file. When all of the tuning eapaci- tor values have been changed, save the file as 12POL3.CIR. ‘Next, run GPLA and press R to read in the 12POLS.CIR file. Then press 8 to set up the sweeps for a beginning frequency of 0, an end frequency of 4000 a frequency step of 40, a fre- quency grid spacing of 200, a screen bottom of 10 and 1 dB per division. ‘Then press P followed by H to see the filter response curve. Make an esti- ‘mate of the insertion lossin the middle of the pass-band. Press Esc to return to the menu. Determine the -8 4B points by changing the sweep settings to make the bottom of the sereen 3 dB below the insertion loss. The insertion loss in this example is about 2.3 dB, soset the bottom to 2.9 + 3 = 5.3. When you redisplay the gain plot, the -3 dB level willbe at the bottom of the screen, and arough estimate of the bandwidth can ‘be done by determining the two points fat which the filter response curve meets the bottom of the chart. A more accurate estimate can be performed by modifying the sweep such that the gain curve originates at the beginning of the sweep interval and finishes at the end of the sweep interval. This sweep modification may require sev- eral tries to get it just right. In this example, you should end up with a sweep range of about 620 to 3225, The -3-4B bandwidth ean be easily caleu- ‘Table 2—Design Parameters for the Example 12-Pole Filter A 8 c D E F 6 4 ' J Crystal Mesh Combined Tuning New New tuning Coupling Crystal offset —oftset_—oftsat capacitor offset capacitor capacitor + (Hz) (Hz) (Ha) cr (oF) (He) (pF) oe (pF) 2 +17 981.99 364.99 3 279.8 534.99 190.9 1 23.84 5 -67 0 67.0 6 1624 237 431.0 4 77.20 8 60 © 920.66 380.66 8 268.3 55066 © 185.5 7 103.26 " -85 «383.96 438.96 «= 12S 232.7 © G0R.96 167-7 10 109.54 4 “15 405.96 420.96 «= 152426 = 590.968 172.8 13 111.79 7 -4 41386 4178818 24a4 587.86 173.7 18 11270 20 0 «413.868.413.862 2468 © 583.86 174.9 19 112.95 23 9 405.96 414.9624 248.1 584.98 ©1746 «= 22,112.70 28 34 383984179627 2464 587.96 173.7 0 2511.79 290-85 -820.66 «375.66 = 30 2719 645.66 «187.2 «= 2B (108.54 3200 -6t 0 61.0 33, 1674 231 442.1 31 103.26, 35 +260 «381.99 355.99 36 286.9 52599 19424 7.20 37 23184 Page 9, QEX, January 1995, published by The American Radio Relay League, Inc. January 1995 9lated: BW, = 3225 — 520 = 2705 He ‘This bandwidth is tao high for the final design, but that was done inten: tionally. The final step in the com- puter design involves adding an addi tional frequency offset to reduce the indwidth to a predetermined value. The purpose of this procedure is two- fold: first, it reduces the value of the tuning capacitors, making it more practical to use two paralle} capaci tors; second, it further narrows the range of capacitor values (see column Hof Table 2), possibly minimizing the number of required capacitor values. ‘The target handwidth for the com: puter design is 6.5% higher than the desired bandwidth: BW, = 1.065 x BW, = 1.085 x 2500 = 2662 Hz ‘The bandwidth reduction: BW’ = RW, — BW, = 2705 - 2662 additional frequency offset required to accomplish this bandwidth reduction can be obtained from Fig 10. From the chart, a 43-Hz reduction requires an additional frequency offset, Fo = 170 Hz, Offset Fis added to the value of the combined offset in eslumn D of Table 2, and the new value is entered in calumn G. (The CLFOFS program, included with the CLPMOD program: in be used to perform these caleula- tions.) By using the MESHTUNE pro: gram and the tuning procedure de- scribed earlier new values for the tun- ing capacitors are obtained and entered in column H of Table 2. File 12POLS.CIR should be modified one more time to include the updated capacitor values and saved as 12POLA.CIR. Finally, the filter bandwidth should be checked again. If the -8 dB band- width (BWy) is within 15 Hz of the target bandwidth (BW, = 2662 Hz), the computer design is completed. Other. wise modify the value of offset Fusing the chart in Fig 10 as a guideline und calculate new tuning capacitor values to obtain the required target band- width, Construction and Alignment ‘The construction method deseribed this section is an alternative to a printed-circuit board. The filter com- ponents are mounted on a piece of Vector board (Vector part no. 8007) The crystals, matehing transformers and the attenuator are mounted on the copper side, which is used as a ground plane. All ground connections are made directly tothe ground plane. The capacitors and the intersection shields are mounted on the pad side of the board. The interface with other stages is done through BNC connectors which can be soldered directly to the bourd, for via coaxial eables which can be soldered to Vector pins (Vector part no. T44), Photo 1 shows the filter assembly The construction involves several steps. In the first step, only the crys tals are mounted on the board. The crystals identified previously should be arranged on the board in the sequence determined in column A of Table 2. The separation between the leads af adjacent crystals is 0.2". The Teads should be soldered to the pads while making sure that the erystals ‘mounted firmly against the graund plane. Ido not recommend you ground cach erystal case via a wire ground strap; the risk af altering the crystal parameters during soldering is too rat totake. The crystal leads should beclippedto Vsinch. It ishelpfal at this, point to label the crystals on the pad side of the board for easy identi- fication, The second step is the formation of ‘the coupling eapacitances (see column Jof Table 2). These are formed by par- alleling capacitors, taking into ac- count the approximately 3 pF af stray capacitance present: C1, C97: 10 pk + 10 pk (4.3.44 pF of stray capacitance) (C4, C4: 47 pF + 27 pF (+ 3.20 pF of stray capacitance! C7, C31: 100 pF (+ 9.26 pF of stray capacitance) C10, C28: 68 pF +39 pF 1s 2.54 pF of stray capacitance C13, C25: 82 pF + 27 pF (+ 2.79 pF of stray capacitance) C16, C22: 100 pF + 10 pF + 2.70 pP of stray enpacitance) C19: 100 pF 4 10 pP (+ 2.85 pF of stray capacitance) The selection process ix greatly facilitated if the eapacitors are picked froma 5% or 10% stock using a eapaci- tance meter with at least 0.1-pF reso: lution. The parallel combinations of the selected capacitors should be formed by soldering them together and trimming the leads to %i ineh. The capacitors should then be tagged with their component number for easy iden: tification, To complete the design. the final values of the tuning capacitors must be determined. The values calculated in column H of Table 2 are preliminary but indicative of the capacitance range required for tuning. The tuning proce- dure is based on the fact, from filter theory, that each loop in the ladder filter should be resonant at the same frequency." In ther words, each erystal-and-tuning-capacitor combi nation, when put in series with the coupling capacitors on each side of it, should resonate at. the same loop fre~ quency, Photo 1 10 QEX Photo 2 Page 10, QEX, January 1985, published by The American Radio Relay League, IncTo find a starting point, refer to column H of Table 2. A convenient value from the middle of the column is selected: C21 = 174.93 pF = 150 pF + 22 pF (+2.93 pF of stray capacitance). Connect this parallel combination of two capacitors in series with crystal X20 (see Fig 9) and the two coupling ‘capacitors on either side of the crystal, C19 and C22, as shown in Fig 1A. The series-resonant frequency of this cir- cuit ean be found with the aid of crys- tal tester. This becomes the loop fre- quency, and all remaining loops have tobe tuned to this frequency. Note that the predicted values of the tuning capacitors in the center section of the filter vary within a narrow range. ‘To determine the value of the next tuning capacitor, connect a parallel ‘combination of a 150-pF capacitor and a 40-pF variable capacitor in series with crystal X17 (see Fig 10) and the two coupling capacitors on either side of the crystal, C16 and C19 (see 11B). Use the crystal tester tomeasure the resonant frequency of this series combination, tuning the variable ca- pacitor until the circuit is resonant at the loop frequency. The proper value for C18is then found by measuring the value of the parallel combination of the 180-pF capacitor and the variable ‘capacitor. Form this capacitance using paralleled fixed-value capacitors and tag it as C18. This procedure is re- peated for all of the tuning capacitors except those at either end of the filter. During this procedure, each coupling capacitor is used twico: once while selecting a tuning capacitor on its left and again while selecting a tuning capacitor on its right. ‘The procedure for determining the values of the tuning capacitors at the ends of the filter differs only slightly: A 316-0 resistor (the termination resistance) is placed across the end coupling capacitor, C1, when deter- ining the value of C3 and across capacitor C37 when determining the value of C36. The terminating resis- tors lower the Q of the circuit, so the tuning of C3 and C36 is not very criti. cal. Remember that when you perform filter measurements the scope probe capacitance (approximately 10 pF) and the generator output capacitance (approximately 5 pF) will be added to the respective terminals. Therefore, after the tuning process has been accomplished the values of capacitors C1 and C37 have tobe modified to com- pensate for the source and load capaci- tances, Once the values of all of the tuning capacitors have been determined, filter construction can continue. Con: struet the ground bars and mount them on the pad side of the board before installing the capacitors. Each ground bar is made from #16 copper wire (stripped of insulation), cut to @ Tongth of (0.2xNV) +0.5) inches, where ‘Nis the number oferystals. Place each ground bar symmetrically, 0.35-inch away from the pins of the crystals (see Fig 12) and solder it to the ground plane every 0.2 inch (every other hole) using jumpers made out of small- diameter hook-up wire. Next, solder the tuning capacitors in place follow- ing the placement diagram shown in Fig 12. Take care not to averheat the crystals during soldering. The load coupling capacitor, C37, and tuning capacitor, C36, are soldered to a Vector pin, Install the coupling capacitors following the diagram of Fig 12. While the tuning capacitors are mounted horizontally, the coupling capacitors are placed vertically toconserve space. Make the capacitor leads as short practical, keeping in mind that longer leads can cause excessive crosstalk between sections, and very short leads can lead to capacitor cracking during the soldering process. Provide suffi- cient clearance between adjacent sec- tions to accommodate the shield ribs, ‘These are made from %-inch-wide strips of metal. Brass, copperor tinned 0.02.inch steel will do, providing itean be easily soldered to the ground plane. ‘The shield is arranged in a“fish-bone” fashion. The center section is cut to a length of (0.2% N) + %«l inches and is Fig 12—Layout diagram ofthe eryatal Page 11, QEX, January 1995, published by The American Radio Relay League, Inc placed vertically exactly in the middle between the pins of the crystals (see Photo 2). Keep it in place by soldering it tof shaped posts. These are made from #18 copper wire (stripped of in- sulation), with the short end soldered to the ground plane. Inert the inter- section ribs between adjacent sections of the filter and solder them to the center rib on one side (at the top only) and to the ground bars on the other side. More comprehensive shielding can be accomplished by fully eneapsu- lating the filter components in « metal enclosure, but I've obtained adequate performance without this precaution. ‘Matching transformers are used to present a 50-Q input and output im- pedance to the outside world. The winding information is taken from Table 1. The transformers are wound with #32 enameled wire on two-hole ferrite balun cores (see Fig 13). In Table 1, “primary turns” refers to the number of turns from the ground, and “secondary turns to the number of turns of the entire ‘winding. Mount both transformers on the ground-plane side of the board in close proximity to the crystals, Solder the tap leads to Vector pins on the ground-plane side of the board, The Vector pin on the input side is also conneeted to the output of a 3-dB at- tenuator (see Fig 14). Place the attenu- ator components on the ground-plane side of the board as well, making the connections on the pad side. Then con- Ula ~**® © Fig 13—The matching transformers are tapped avtotranstormers (A), wound a shown at 8. January 1995 11nect the attenuator input to another ‘Vector pin, which serves as the filter input terminal. The board area taken by the entire 12-pole filter cireuit is 3% x 1% inches. Measurements Now that the filter is finished, we need to measure its response, which should match the design objectives: shape of the response, -3 dB points, bandwidth, amount of ripple, inser- tion loss and the aymmetry of the re- sponse. Making the measurements is a trivial task when a spectrum ani i at hand, or at least a good syn- ed generator and a scope. Since fow amateurs have the luxury of using lab-quality equipment, T suggest. a simple method that should be within reach of the average experimenter. It requiresthe construction of variable- frequency erystal oscillator (VXO). ‘The VXO should tune from 8,000 MHz to at least 8.005 MH. (One of the re- jected filter erystals can be used in the ‘VXO—use the one with the most nega tive offset.) The VXO should have @ level adjustment control and adequate buffering at the output to prevent fre- quency pulling when being loaded by ey 20 Fig 14—Schematle diageam of the 3-d8 Input auendator used with the Inder 12 QeX the erystal filter. Many variations of ‘VXO design have been covered in the amateur literature.$758.10 A fre- quency counter with 1-Hz resolution and an oscilloscope are also needed to perform the measurement. ‘The block diagram of the measure- ‘ment setup is presented in Fig 15. The ‘measurement procedure is to vary the frequency of the VXO in small incre- ‘ments (20 to 80 Hz) while monitoring and recording the signal level at the ‘output of the filter at each frequency. ‘Hold the VXO output level constant at every measurement point. Tused lab-quality test equipment (a Hewlett-Packard HP3585A spectrum analyzer) to measure three ladder fil ters—10-pole, 12-pole and 14-pole designed and constructed using the procedures outlined in this article. The measured response curves of these filters are shown in Fig 17, with the ripple level for the 14-pole filter shown in Fig 18. The measured band- width of all three filters is within 2% of the desired bandwidth, and the shape factor is within 8% of the caleu- lated value. The ripple level is slightly higher than the projected value but significantly better than several com- ‘mercial units I tested. ‘To compare the performance of my home-built ladder filters to that of commereislly available crystal filters, Tmade a series of measurements us- ing laboratory-grade measurement equipment. The block diagram of the measurement setup is given in Fig 16. ‘The amplifier used in the measure- ment has an output third-order inter- cept point (OIP) of +48 dBm and a =1 dB output compression point of +27 dBm. One of my objectives was to study each crystal filter’s dynamic range and its effect on the output in- tercept point of the driving amplifier. Four crystal filters have been evalu ated: a2.2-kHz, 8-pole SSB filter from Fox-Tango Corp (part no. 2809) and the three home-built ladder filters mentioned previously. The Fox-Tango filter requires a 500-0 impedance ter- mination, so it was coupled to the amplifier via a 1:9 transformer. The test results were very similar for all three ladder filters, so they will be referred to collectively as the “ladder crystal filter.” The input impedance of the filters was examined using an HP impedance analyzer and re-examined after in- serting an attenuator between the amplifier and the erystal filter. For the ladder filter, a 3-4B attenuator was used; a 6-dB attenuator was used with the Fox-Tango filter. Figs 19 through 22 show the results of these measure- ments. ‘To measure the IMD of the filters, I used two equal-level tones placed out- side of the filter pass-band. The tone spacing was varied between 2 and 20 kHz without an appreciable effect on the third-order products measured ‘at the output of the amplifier. These measurements were made at two different signal lovels (+15 dBm and +10 dBm) in order to gauge the linear ity of the system. The measurements were made while driving the filter directly by the amplifier and again Fig 16—Block dlagram of the filter IMD measurement tes! setup. Page 12, QEX, January 1985, published by The American Radio Relay League, Inc.Fig 18—A close-in look at the response of the 14-pole ladder filter shows the pass-band ripple. Vertical divisions ‘rlzontal divisions are 300 Fig 19—The measured impedance of a 14-pole crystal tadder filter. The vertical divisions are 20 2, with zero at the bottom. The horizontal divisions are 4 kHz each. 1a (A), 12-pole bout using the techniques Fig 20—The measured impedance of a Fox-Tango lattice fitter. The vertical givisions ae 20 1, with zero at tho horizontal aivisions are 1 KHz bottom. The horizontal dvisione are 4 kez each. January 1995 13 Page 13, QEX, January 1995, published by The American Rado Relay League, nowhile driving the filter through a re- sistive pad. The measurement resulta are given in Table 3, The output inter- cept point of the amplifier was de- graded by 5 dB by the ladder filter and by 12 dB by the Fox-Tango filter. J tried using resistive pads with differ- ent attenuation values in an attempt Fig 21—The measured impedance of a to “smooth out” the impedance pre- sented to the amplifier. A 6-dB pad is ‘required in the case of the Fox-Tango filter to improve the OIP by 5 to 6 dB, and a3-dB padis sufficient in the ease of the ladder filter to improve the OIP by 3 to 5 dB. In the next experiment, one of the pole crystal ladder fitter with a 3-dB attenuator at the Input. The verticel divisions are'20 0, with zero at the bottom. ‘The Norizontel dvisiona are & kHz each. Fig 22—The measured impedance ot a Fox: atfonuator at the input. The vertical divisor orlzontal divisions are 4 kHz each. Table 3—Measurement results rango crystal iattice filter with « 6-38 ‘ore 20 0, with Zero atthe bottom. The Third-order output intercept point of the amplifier (+d8m) tones was placed in the center of the passrband and the second tone was "20 kifs away. In this measurement, he OTP ofthe amplifier was degraded by 4 dB in the case of the ladder filter and by 11 dB in the case of the Fox- Tango filter, A'S resistive pad in front of the ladder filter reduced the dogeadation by 1 t0 348, and a 6B pad in front of the Fox-Tango filter Feduced the degradation by 6to 7 dB, Finally, one af the tonce was placed atthe lending edge ofthe filter pass- hand (notice the impedance dips in Figs 19 and 20), Due to the severe impedance mismatch presented othe two tones at both frequencies the O1P of the amplifier is degraded by 11 4B fn the cane of the ladder filter and by 22 dB inthe case ofthe Fox-Tango fil- ter, Resistive pads have » profound effect on the performance inthis eate: f.3-dB pad in front of the ladder filter reduces the degradation by 5 to 6 4B, tnd 2 6-4B ped in front of the Fox Tango filter reduces the degradation by up to 17 dB! "The following conclusions can be drawn from these measurement re- sults * Because of the highly reactive nature of ite input impedance, 2 crystal filter has a significant loading effect on the preceding stage, This may become the lime iting factor when ealeulating the overall dynamic range of a re- «The experiments suggest that it ia more meaningful to evaluate the effect of the filter on the pre- ceding stage than to attempt to measure the OIP of the crystal filter itself. Some non-linear behavior was abserved (the degradation of the OlP of the preceding stage de pends on the signal tevel: The filter behavior ts more predict. Tone location relative Ladder fiter Fox-Tango filter Tone level fothe passband — Nopad 3-dBpad Nopad 6-dB pad —(+dBm) Both tones outside 45 47 38 43 15 ‘of passband 43 45 36 42 10 (One tone inside 44 468 37 44 15 of passband 45 46 37 43 10 (One tone at the 38 42 30 43 15 edge of passband 37 42 26 43 10 14 QeEx Page 14, QEX, January 1996, pubished by The American Radio Relay League, ncing t Page 1: BASIC Modify IRFD circuit files contining crystal-ladder filters to add tuning capacitors for offeet crystal frequencies From "Designing and Building High-Performance Crystal Ladder ' Filters," QEX, January, 1995 ' J. Bloom, KE3Z 12/08/94 Note: Little checking of the input file is performed. only files written by the IRFD program “X" ghould be used as input files. cus PRINT "CLFMOD -- Modifies IRFD ladder crystal filter circuit files for tuning" PRINT INPUT “Enter name of original circuit file: ", £18 OPEN £1§ FOR INPUT AS #1 INPUT “Enter name of output circuit file: OPEN £2§ FOR OUTPUT AS #2 OPEN “OFFSETS.CLE" FOR OUTPUT AS #3 ' Get the number of circuit elements INPUT #1, x N= (x= 2) /2 + Number of poles ‘ Weite the new number of circuit elements, including tuning caps PRINT #2, N34 2 " Copy the unchanging parametera FOR i= 170 14 INPUT #1, x PRINT #2, x NEXT 4 ' Leop through each of the meshes, annotating the elements with ‘ component designators as in Pig 9, adding the tuning capacitors, * and replacing the crystal offset frequencies with those of the £28 ' Coupling capacitor INPUT #1, x$ PRINT #2, x$, LEPTS(x$, 1) + MIDS(STRS(i * 3 +1), 2) INPUT #1, x$ PRINT #2, x$ INPUT #1, x PRINT #2, x ' Ceystal, INPUT #1, x¢ y$ = LEFTS(x$, 1) + MIDS(STRS(i * 3 + 2), 2) PRINT #2, x$, ys INPUT #1, x$ PRINT #2, x$ INPUT #1, x PRINT "Crysta: INPUT ") delta F. PRINT #2, y PRINT "Target offset frequency for mesh"; i + 1; "="; x- y PRINT #3, x - y "OG ys January 1995 15 Page 15, QEX, January 1995, published by The American Radio Relay League, ine.isting 1 (continued) Page 2: BASIC ‘tuning capacitor PRINT #2, “capt, "c" + MIDS(STRE(A * 3 4 3), 2) PRINT #2, "ser" PRINT #2, 200 NEXT i * Load end coupling capacitor INPUT #1, x PRINT #2, xS, LEPT$(x$, 1) + MIDS(STRS(N + 3 +2), 2) INPUT #1, x8 PRINT #2, x$ INPUT #1, x PRINT #2, x * Should only be one line left, but "just in case," copy to * end of file WHILE NOT EOF (1) LINE INPUT #2, x$ PRINT #2, x$ WaND "Clean up and exit cLosE SYSTEM CLFOFS -- Generates final offsets for ladder crystal filter design From "Designing and Building High-Performance Crystal Ladder Filters," QEX, January, 2995 + 3. Bloom, KE3Z + 32/08/94 cus PRINT "CLFOPS -- Calculates final design offsets for ladder crystal filters” PRINT INPUT “Frequency offset for bandwidth reduction: ", £ OPEN "OFFSETS.CLF" FOR INPUT AS #1 WHILE NOT BOF (2) INPUT #1, x PRINT x + £ wend CLOSE SYSTEM able at the flat portions of the the Fox-Tango filter is replaced * The level ofthe third-order prod- frequency response and less pre- by a ladder filter ucts at the output of the Fox- dictable when extreme imped- «A 6-4B rocistive pad is required Tango filter deviates from the ance changes are encountered to significantly reduce the degra- calculated value by more than Resistive pads tend to improve dation of the amplifier’s OIP in 6 dB, even with a 6-dB resistive the linearity. the ease of the Fox-Tango filter. pad. In the case of the ladder fil + Ino resistive pad is used, the __4'3-dB pad is sufficient to pro. ter, this deviation is reduced to a degradation of the amplifier's duce thesameeffectinthecase of value under 1 4B if'a 3-dB re OIP is reduced by at least 74Bif the ladder filter. tive pad is employed. 16 QeEX Page 16, QEX, January 1995, publshed by The American Radio Relay League, Ine* Examination of the plot of input impedance of the two types af fl- ters Figs 19 and 20) reveals that, the ladder filter has a much smoother response in the pass- band. This must be one ofthe rea- sons for the ladder filter's supe- rrior performance. Summary I'veshown that homeconstruction of high-performance crystal filters is quite practical, and that laboratory- grade equipment, although helpful, is not required. Home-built ladder filters, can exhibit performance superior to that of commercially available filters at reasonable cost. The design and construction procedure outlined above enables the amateur te tailor the fre: quency response of the filter to fit the aneeds of the project. Acknowledgments I wish to express my appreciation to Wes Hayward, W7Z0I, Bill Carver, K60LG/?, Colin Horrabin, G3SBI and Peter Chadwick, G3RZP for the help- fal discussions that enabled me to bet- ter understand this subject and the ‘encouragement to pursue this project. Notes ‘Carver, B., KBOLGI7, “High-Performance ‘Crystal Fite Design,” Communications uarterly, Winter 1993. pp 11-18. "Hayward, W.. WTZOI, Introduction to Rado Frequency Design, ARAL, Newington, Connecticut, 1994, chapters 2 and 3, "Orentea. C., WB3I20. Radio Communic: tions Receivers, Tab Books Ine, Blue Ridge Summit. 1982, pp 69-75. ‘Hayward, W . W7Z0l, “A Unified Approach to the Design of Crystal Ladder Firs, OST. May 1982, pp 21-27 “DeMaw, ., WIFB, “A Tostor for Crystal F. ‘Qand R, QST, January 1990, 2 the source code and execulable versions of CLEMOD and CLFOFS are in the QEXOLE ZIP tile, available for download from the ARRL BBS (203-686-0576) or via the Intemet by anonymous FTP to ftp.cs.butlalo.ecu, in the /pubsham-radio |, seciory ‘Wayward, W., W7Z0! and DeMaw. D., WIFB, Sotd State Design for tne Radio Amateur, ARAL, Newington, Connecticut, 4977p i? "The ‘ARAL 1986 Handbook, ARAL. Newington, Connecticut, 1985, p 10-3 “Noble, F., WIMT, “A Variable Frequency Crystal Oscilator* QST, March 1981, pp 34:37 "DeMaw, D., W1FB, "Some Practical As- pects of VXO Design’, OST, May 1972, bit cx Finding Parts Crystals: The most consistent results were obtained using crystals trom Fox Electronics (Tel: 813-693-0096), part number: FOXOBO. These are series: resonance microprocessor crystals in an HO4@/L case, Ask for list of oistibu- tors in your area Capacitors: Use monolithic ceramic capacitors rom Panasonic or equivalent auality capacitors. Panasonic capacitors are available rom Dig Key Corpora- tton (Tel: 800-344-4538), P4800 series. COG ceramic parts are recommended {or good temperature stabiliy. Low loss is an important requirement. 5% or 10% tolerance is acceptable. Transformer cores: Use two-hole balun cores, part number: BN-43-2402 from Amidon Associates (Tel 310-763-5770) Board: The project board, with a ground plane on one side, is part number 8007 trom Vector (Tet: 800-423-5659; 800-426-4652 Inside California), The Vector pins are part number T44, (Digi-Key is one of the nationwide distibu- tors for Vector.) January 1995 17 Page 17, QEX, January 1985, published by The American Radio Relay League, Inc.