Pre Divider (:10) up to 5GHz - WorldRadioHistory.Com · 2019. 11. 11. · mat 1206 or 0805: 1x 12Ω 4 x 100Ω 1 x ... 1 x 18kΩ 1 x 27kΩ 1 x 2.7pF 1 x 12pF 2 x 15pF 1 x 47pF 1 x

A Publication for theRadio Amateur Worldwide

Especially Covering VHF,UHF and Microwaves

Volume No.34 . Spring . 2002-Q1 . £5.00

Pre Divider (:10) up to 5GHzAlexander Meier, DG6RBP

Contents

K M Publications, 63 Ringwood Road Luton, Beds, LU2 7BG, UKTelephone / Fax +44 (0)1582 581051, email : [email protected]

web : http://www.vhfcomm.co.uk

Wolfgang Schneider Frequency Generator (Wobbler) 2 - 16DJ8ES to 4GHz

Dipl. Ing. Detlef Burchard A Simple Procedure for 17 - 21Nairobi Measurement

Alexander Meier Pre Divider (:10) up to 5GHz 22 - 26

Gunthard Kraus Internet Treasure Trove 28 - 29DG8GB

Index of Volume 33 (2001) 31 - 33

Gunthard Kraus Modern Design of Stripline 35 - 62DG8GB Low Pass Filters

Very interesting articles again with some good constructional projects. Fromfeedback with this years subscriptions most readers prefer the constructionalprojects.One of the most popular projects over the past few years has been the spectrumanalyser by Matjaz Vidmar. This is a very ambitious project with only a set of printedcircuit boards available. I am interested to hear from anyone who has successfullybuilt this project since it would be good to do an article with hints and tips for theconstruction and testing of the unit. I have also been asked to start a web pagedevoted to the project so that "would be" constructors can locate some of the moredifficult to obtain parts.Thanks to UKW Berichte the very popular CAD software PUFF is available again, Iwill be holding a small stock, if you want a copy please use the web site or fax toplace an order.

73s - Andy

VHF COMMUNICATIONS 1/2002

1

The following article describes a fre-quency generator with a wobble func-tion for the frequency range from 10MHz to 4 GHz. The output is approxi-mately +10dBm (10 mW) with a rippleof less than ±1.5 dB over the wholerange. It is split into two overlappingranges. Automatic switching betweenthe two frequency ranges is not abso-lutely necessary for use in amateurradio and was therefore not includedin this design on cost grounds. Thisalso applies to automatic level correc-tion.

1.The Design

The frequency generator/wobbler for upto 4 GHz is a useful aid in measurementand calibration work in the amateur radiofield. The technical requirements for thedesign have been restricted to the abso-lutely necessary. The block diagram (Fig.2) shows the functional blocks used at aglance, together with their interaction.The core of the circuit is a YIG oscilla-tor. The maximum tuning range of themodule used here lies between 1.6 and 4GHz. The lower frequency range, be-tween 10MHz and 1.8GHz, is coveredusing a mixer. This has an associatedoscillator, a low pass filter and an ampli-fier. In both frequency ranges, the output

is approximately +10dBm, with a rippleof less than ±1.5dB. Better values can notbe obtained without a regulated outputamplifier, but this was deliberately dis-counted with for the reasons explainedabove.The wobbler is controlled using a microcontroller. This sets YIG oscillator ac-cording to the frequency range, andgenerates the saw tooth tuning voltagerequired for the wobble. A two line by 16character liquid crystal display is used todisplay the current mean frequency andthe span. A 10dB directional coupler isinserted at the radio frequency output ofthe YIG oscillator. This additional outputis for measurement purposes (e.g. fre-quency counters for calibration) and ifapplicable for the connecting up of a PLLcontrol circuit for narrow band applica-tions. This is, however, optional and isnot covered in this article.

2.The YIG Oscillator

The great advantage of YIG oscillators isobvious: a YIG oscillator can be elec-tronically tuned over a very wide fre-quency range, with a very linear relation-ship between the tuning voltage (and / orcurrent) and the frequency. The responseshown in Fig. 3 shows the frequency

Wolfgang Schneider, DJ8ES

Frequency Generator (Wobbler)to 4GHz


2

response of the YIG oscillator in theprototype over a range from 1.6 to 4GHz.Just a brief note on the theory behind aYIG oscillator: the abbreviation “YIG”stands for Yttrium Iron Garnet. This is asynthetic crystalline ferrite material, con-sisting of yttrium and iron (Y3Fe6O16).If such a crystal is surrounded by amagnetic field and in addition radio

frequency energy is fed in through amagnetic loop, the element begins toresonate in the radio frequency range.The frequency of resonance is directlyproportional to the strength of the mag-netic field.For the practical construction of a YIGoscillator, the magnetic field is generatedthrough an electromagnet. The frequencyof resonance is determined by the current

Fig 1: Frequency generator (wobbler) to 4GHz

Fig 2: Block diagram of frequency generator.


3

in the magnet coil, and displays linearbehaviour. The coils are responsible forthe relatively high weight of YIG oscilla-tors. YIG oscillators are available cover-ing the frequency range between 0.5 and50GHz, in segments. The tuning rangeusually stretches over one or more oc-taves. Equipment of this type, in therange of interest for amateur radio be-tween 2 and 18GHz, can be obtained atreasonable prices in surplus stores.The YIG oscillator used here originates

from the HP141T HP spectrum analyser(radio frequency assembly HP8355) andhas the following operational data:

• Operating voltage +20 V / 70 mA-10 V / 7 mA

• Frequency range 1.6 - 4 GHz

• Output 25 mW ±1 dB

• Coil current 50 - 120 mAThe frequency is set using a micro

Fig 3: Frequency response of YIG oscillator.

Fig 4: The driverfor the YIGoscillator.


4

controller with an integrated D/A con-verter. Since this can not generate therelatively high current required for thetuning coil (main coil), the transistorcircuit with a 2N3055, acts as a driver(Fig. 4). It is important that both thetransistor and the subsequent series re-sistance (47Ω) should be thermally sta-ble. Therefore the 2N3055 is mounted ona cooling surface of generous dimen-sions. A 25 W type was selected for the47Ω resistor, which was mounted di-rectly on the mounting plate (base plate).A 3dB attenuator is fitted directly to theoutput socket of the YIG oscillator withSMA connections (Fig. 5). This isolatesthe oscillator from the subsequent cir-cuits. The output, including all losses inthe attenuator, the directional coupler, thechange over relay and the cable, isalmost exactly 10mW.

3.Mixer With Oscillator, LowPass Filter and Amplifier

3.1. Mixer assemblyFor the lower frequency range up to1.8GHz, the output frequency of the YIGoscillator must be mixed down. Themixer used is an ADE-42MH from MiniCircuits, which is operating in the speci-fied frequency range. In this SMD mod-ule the LO and RF ports are specified forthe frequency range 5 to 4,200MHz andthe intermediate frequency port for 5 to3,500MHz.The mixed signal is fed through a 21 polelow pass filter using stripline technology.This suppresses both the image fre-quency and the signal from the localoscillator at 2,106MHz. Fig. 7 shows thevery good frequency response of the lowpass filter. Next comes a single stageamplifier using an ERA-1 MMIC whichprovides a high degree of amplificationof approximately 19dB over a wide fre-quency range. The output level of themixer assembly is approximately+10dBm (10 mW).

3.1.1. Assembly tips for the mixermoduleA double sided copper coated circuitboard made from epoxy material withdimensions 53.5 mm. x 146 mm, was

Fig 5: Picture of YIG oscillator.

Fig 6: Circuit of mixer with oscillator and low pass filter.


5

designed for the mixer assembly. Firstthe printed circuit board, still without thecomponents, is soldered into a suitablestandard tinplate housing (55 mm x 148mm x 30 mm). Then the ADE-42MHring mixer, ERA-5 MMIC and the two100pF coupling capacitors are solderedin on the copper side. The 120Ω resistorand the 1nF blocking capacitor are sol-dered in on the component side.The L11 choke consists of 2 turns woundonto a 2mm former being one lead of the120Ω resistor. The position of the chokeis thus at a distance of approximately1mm above the printed circuit board.The supply voltage (+12V) is fed througha feedthrough capacitor (1nF, solderable)in the housing wall. At this supplyvoltage, the current consumption of the

circuit is approximately 60mA. ThreeSMA connectors are used for radio fre-quency connections, input (LO and RF)and / or output.

3.1.2. Component list for mixerassemblyIC1 ADE-42MH, ring mixer

(Mini-Circuits)IC2 ERA-5, MMIC (Mini-Circuits)L1-L10 Stripline, printedL11 2 turns on 2mm - see textC1-C11 Capacitor, printedR1 120Ω / 0.6 W, RM 10 mmC12,C13 100pF, SMD 0805C14 1nF, EGPU, RM 2.5 mm1 x DJ8ES 053 PCB1 x Tinplate housing, 55 mm x

148 mm x 30 mm1 x 1nF, feedthrough, solderable3 x SMA flanged socket

3.2. Oscillator for 2,106MHzA tried and tested design from MichaelKuhne (DB6NT) acts as a local oscilla-tor. The circuit (Fig. 10) has been slightlymodified from the original (oscillator fortransverter for the 13cms band).The quartz oscillator runs at 117.0MHzusing a U310 FET. This signal is fedthrough various frequency multiplierstages to reach the desired frequency of2,106MHz. A tripler (BFR93A) to351MHz, a further tripler (BFG93A) to

Fig 7: Frequency response of 21 polefilter.

Fig 8: Printed circuit board for mixer DJ8ES-053.


6

1,053MHz and a frequency doubler(BFG93A) to 2,106MHz. The outputlevel here is approximately 2mW.

3.2.1. Assembly tips for local oscillatorequipmentThe oscillator for 2,106MHz is built on adouble sided copper coated epoxy printedcircuit board with the dimensions 34 mmx 53.5 mm. The printed circuit board issoldered into a suitable tinplate housing(37 mm x 55 mm x 30 mm).The DJ8ES 054 printed circuit board isassembled with the mainly SMD compo-nent on the track side (Fig. 13). Only thewired components such as the U310FET, the 117.0MHz crystal and the fil-ters, are inserted in the usual way fromcomponents side (fully coated side, Fig.

14). The component drawings show allthe details with regard to the position ofthe components, etc.The supply voltage (+12V) is fed througha feedthrough capacitor (1nF, soldera-ble). An SMA socket acts as the outputfor the radio frequency.

3.2.2. Components list for localoscillatorT1 U310, FETT2 BFR93A, SMD transistorT3,T4 BFG93A, SMD transistorIC1 78L09, SMD voltage regulatorQ1 Crystal 117.0MHz, HC-18U,

Series resonance 7th overtoneL1 BV5061, 0,1µH, NeosidL2 5HW-35045A-365, helix filterL3 5HW-100090A-1010, helix

Fig 9: Photograph of the prototype mixer.

Fig 10: Circuit diagram of 2106MHz oscillator.


7

filterL4,L5 Stripline filter, printedC18,C19 Trimmer 5pF, green, Sky1 x DJ8ES 054 PCB1 x Tinplate housing, 37 mm x

55 mm x 30 mm1 x 1nF, solderable, feedthrough1 x SMA flanged socketall remaining components in SMD for-mat 1206 or 0805:1 x 12Ω4 x 100Ω1 x 270Ω1 x 1kΩ1 x 2.2kΩ1 x 2.7kΩ1 x 10kΩ1 x 18kΩ1 x 27kΩ1 x 2.7pF1 x 12pF2 x 15pF1 x 47pF

1 x 82pF3 x 40pF3 x 1nF1 x 1µF, tantalum1 x 10µF, tantalum

3.2.3. Putting oscillator assembly intooperationAfter a final visual check of the fullyassembled oscillator assembly for place-ment errors or solder breaks, it can be putinto operation for the first time. Theoutput is determined using a power metersuitable for this frequency range. Themaximum current consumption at +12Vis 110mA. However, this value is notreached until setup is complete.There must be + 9 V at the voltageregulator output (IC1). The voltage dropthrough resistor R2 should be approxi-mately 0.7V. The crystal oscillator canbe tuned, the core of the oscillator coilL1 is slowly rotated while the voltage

Fig 11: Oscillatorprinted circuitboard, groundplane side.

Fig 12: Oscillatorprinted circuitboard, track sideDJ8ES-054.


8

drop is measured across resistor R5. Assoon as the oscillator starts to, thisvoltage drop increases rapidly, reaching amaximum value of 3.4V. The optimalsetting for the coil core, is slightly belowthe maximum on the slow rising slope.Next the tripler (T2, BFR93A) is tuned to351MHz. To do this, the voltage drop

across R9 is used. Using reciprocal tun-ing of the two circuit helix filter L2, anunambiguous maximum reading (ap-proximately 3.4V) can be obtained. Thesame applies for the tripler to 1,053MHz.Here the voltage drop across R13 (3.4V)acts as an indicator. This time L3 shouldbe tuned (again reciprocally). The dou-bler to 2,106MHz should be tuned by

Fig 13:Componentlayout ofoscillator printedcircuit board,track side.

Fig 14:Componentlayout ofoscillator printedcircuit board,ground planeside.

Fig 15: Prototypeof the 2106MHzoscillator.


9

Fig 16: Circuit diagram of micro controller.


10

means of the two 5pF Sky trimmers(C18, C19). Here the output of approxi-mately 2mW should be obtained.

4.Micro controller control

The micro controller circuit (Fig. 16) isdivided into two functional parts. A µPAT90S2313 (IC5) micro processor gen-erates the saw form tuning voltage forwobbling. This processor also generatesthe blanking signal during the flyback onthe display. The second micro processor,a µP AT90S8515 (IC8), sets the meanfrequency through a shaft encoder andcontrols the LC display.The circuit requires a single positivesupply voltage of +12V. Other voltagesrequired are generated on the circuitboard:

• ±15 V for the operational amplifierwith IC1 (NMA1215S),

• ±5 V for the analogue/digitalconverter with IC2 (78L05) and IC3(79L05) and

• +5 V for the micro-controller andthe LC display with IC4 (7805).

Each micro controller generates a 16 bitdigital word that is converted by thedigital to analogue converter to give therequired voltages for the frequency set-ting. Downstream amplifiers providelevel adjustment to standard values.Thus, for example, the output for thehorizontal deflection is defined on adisplay screen (e.g. oscilloscope) as ±2.5V. For precise calibration, 3 precisiontrimming capacitors are provided. In thelast operational amplifier, the two ana-logue signals are combined and used todrive the YIG oscillator.The amplification of the operational am-plifier IC7b can be switched in stages toset the span. The values given for theresistors in the feedback loop allow aspan of 2GHz, 1GHz, 500MHz, ... rightdown to 20MHz. The CW switch posi-

Fig 17: Printed circuit board for micro controller, top side DJ8ES-055.


11

tion makes standard signal generator op-eration possible at a frequency that canbe set by means of the shaft encoder. Thesweep width is not used in this mode.The frequency stability is better than100kHz over the entire frequency range.Any improvement for genuine narrowband measurements (quartz filters or thelike.), can be made using an additionalfrequency control circuit with a PLL.To set the wobble speed, 5 connectionsare taken from IC5 to a socket strip (K5).For faster wobble speeds, one of theseconnections can be switched to earth.Inverted blanking signals are availablefor flyback with K7 and K8. K28, pin 1makes it possible to switch between theupper and lower frequency ranges on theliquid crystal display and pin 3 makes itpossible to switch between normal opera-tion and frequency calibration.For the settings lower frequency rangeand frequency calibration, the connectionpin in question should be switched toearth. Pin 2 is not used in the currentsoftware.

4.1. Assembly instructions for microcontroller moduleThe micro controller module occupies adouble sided coated epoxy printed circuitboard, with the dimensions 100 mm x160 mm (European standard size pcboard) (Figs. 17,18). In contrast to theoscillator assembly, only wired compo-nents are used, and no SMD parts areused. The components layout is shown inFig. 19. The micro processors and A/Dconverter should be mounted using ICsockets. It is more sensible for all inputsand outputs to use plug connections. TheLC display is connected to K24 as perlist:Pin 1: 11, DB4Pin 2: 12, DB5Pin 3: 13, DB6Pin 4: 14, DB7Pin 5: 6, ENPin 6: 4, RSPin 7: 3, VEE (LCD drive)Pin 8: 5, R/WPin 9: 2, VDD (+5V)Pin 10: 1, VSS (GND)

Fig 18: Printed circuit board for micro controller, bottom side.


12

The 5kΩ trimming potentiometer (R30)adjusts the contrast of the liquid cryataldisplay. For illuminated displays, a stabi-lised +5V voltage is available on K32.The connection usually goes through anexternal 12Ω resistor.

4.2. Component list for microcontroller moduleIC1 NMA1215S, voltage converterIC2 78L05, fixed voltage regulatorIC3 79L05, fixed voltage regulatorIC4 7805, fixed voltage regulatorIC5 AT90S2313-10, micro

controllerIC6,IC9 AD1851, D/A converterIC8 AT90S8515, micro controllerIC7,IC10 LM324, operational amplifierQ1 Crystal, 12 MHz, HC18-U,Q2 Crystal, 8 MHz, HC18-U,R6,R17 2kΩ precision spindle trimmerR27,R30 5kΩ precision spindle trimmerCapacitors:1 x 1000µF, RM 5mm,

electrolytic capacitor5 x 10µF, tantalum electrolytic

capacitor2 x 4.7µF, tantalum electrolytic

capacitor4 x 22pF, RM 2,5 mm, ceramic2 x 10 nF, RM 2,5 mm, ceramic13 x 100nF, RM 2,5 mm, ceramicResistors, ¼ W, RM 10 mm:2 x 100Ω1 x 2.2kΩ1 x 6.8kΩ1 x 8.2kΩ10 x 10kΩ1 x 15kΩPrecision resistor, ¼ W, RM 10 mm:1 x 100Ω1 x 250Ω1 x 500Ω1 x 1kΩ1 x 2.5kΩ1 x 5 kΩ1 x DJ8ES 055 PCB

4.3. Putting into operation with setupof micro controller moduleWhen the micro controller module is

Fig 19: Component layout for micro controller printed circuit board.


13

switched on, the tuning voltage generatedthrough the FC AT90S2313 (IC5) startsautomatically. The waveform and voltagelevel can be monitored at connectionK23. Positive or negative blankingpulses, each with a pulse length of 10ms,are fed to K7 and K8. A low level atK6/pin 3 is used to switch the microcontroller to the lower value. Usingpotentiometer R6, this minimum value isset to precisely 2.50V. A low level at K3(Reset) re starts the sawtooth.A low level at K6/pin 1 is now used toswitch the micro controller to the uppervalue, +2.50V can be monitored at K23.In the event of slight discrepancies ineither of these values, the operationalvoltage, either ±15V for the operationalamplifier (IC1, NMA1215S) or ±5V forthe analogue/digital converter, is notsymmetrical from IC2 (78L05) and IC3(79L05). In this case, the same voltagelevel, positive and negative, is set at K23using R6. The testing and adjustment ofthe tuning voltage for the mean fre-quency of 5V to +5V at IC10/pin 14 is

carried out in a similar way. This time,the micro controller 90AT8515 (IC8) isswitched to the upper or lower maximumvalue using K27 and calibrated with theprecision trimmer R27. The reset connec-tion is available on K21.To monitor the mean frequency of theYIG oscillator, the sawtooth is switchedto CW, i.e. no span, by connecting K9 toK17. The micro controller is switchedinto calibration mode (K28/pin 3 toearth). Following the reset of the microcontroller, the mean frequency is ad-justed and programmed. In the prototype,this is the setting value was 22446 for theD/A converter, which corresponds to afrequency of 2,800MHz for the YIGoscillator. 6.85V is now applied at pin18, or precisely half the voltage, 3.425 Vat IC10/pin 14. Fine adjustments can stillbe made to the frequency using the shaftencoder.All voltage values are taken from thecharacteristic of the YIG oscillatorshown in Fig. 3. The setting value for themean frequency is calculated as follows:

Fig 20: Prototype of micro controller board.


14

Setting value = 32768 * Tuning voltage /5.0 V / 2

= 32768 * 6.85 V / 50 V/ 2

= 22446To set the span, the shaft encoder re-mains set to the mean frequency. Thespan is set to 2GHz (connect K16 withK1). The mean frequencies must be setthrough the micro controller IC5 and thetrimming potentiometer R17; with K6 /pin 3 low minus 1GHz ,and with K6/pin1 low plus 1GHz , for the mean fre-quency of the upper maximum fre-quency. In the specimen apparatus, thefollowing values are valid for this set-ting:

• 1,800 MHz = 4.55 V

• 2,800 MHz = 6.85 V

• 3,800 MHz = 9.14 VThis gives a voltage range of 9.14V4.55V = 4.59V for a maximum sweepwidth of 2GHz. For the final program-ming of the micro controller IC8(AT90S8515) by Frank Peter Richter, thesettings just obtained are important. Thecorrect display for the frequency can nowbe programmed for the upper and lowerfrequency range in the micro controller.The values for the specimen apparatusare:

• Oscillator frequency: 2,106 MHz

• Mean frequency YIG: 2,800 MHz

• Shaft encoder setting:22464 (rounded off, must be divisibleby 64)

• Frequency interval per step: 0.13333MHz per shaft encoder step

If this design is copied using a YIGoscillator which differs considerablyfrom the one in the prototype, then therate of rise of the tuning voltage and thusthe amplification of IC7c must be ad-justed, if applicable. This results in a recalculation of resistor R16. In the proto-

type, the following values apply to thisresistor:R16 = (R15 * dispersion / 5 V) -

(R17 / 2)= (10000Ω * 4.59 V / 5 V) -

(2000Ω / 2)= 8180Ω

Thus for R16 a standard value is selected(E24 series) of 8.2 kΩ.

5.Operational experience andprospects

This frequency generator (wobbler) up to4GHz has already proved its worth manytimes both as a test transmitter and as awobbler (e.g. calibration of band filters)in the VHF/UHF and SHF frequencyranges. The principle of this apparatus isalready being used in a further develop-ment (spectrum analyser up to 1.8GHz oras slimmed down version up to500MHz).Suitable detectors must be used for thisfrequency. DIY products, even in coaxialformats, can frequently be used only upto frequencies of 2GHz. Beyond this, theinput impedance of the detector is farbeyond 50Ω and does not allow anyreliable measurements to be carried out.If a standard oscilloscope is used as adisplay screen, the blanking of the fly-back, if applicable, can not be accom-plished without modification. One con-ceivable plan is to short circuit thedetector DC voltage with an FET (e.g.BS170) using a small additional circuit.Alternatively, a micro controller with adelta voltage is also available instead ofthe sawtooth tuning voltage. However,because of the electrical behaviour of theYIG tuning coil due to the back electro-motive force, this only works for slowwobble speeds.At this point, I would like to express my


15

sincere gratitude to Frank Peter Richter(DL5HAT) for the development of thesoftware and for volunteering to prepareindividually programmed micro control-lers. In this connection it is conceivablethat additional functions could be imple-mented into the software. Any sugges-tions along these lines will be particu-larly welcome.

6.Literature

[1] Bert Kehren, WB5MZJ Microwavecomponents, Proceedings of 45th Wein-heim VHF Congress[2] Michael Kuhne, DB6NT, 13cm lineartransverter; DUBUS 3/1993


16

In the field of RF measurement tech-nology, there are some very good andvery accurate procedures, althoughthey are not necessarily what everyonecan achieve.However, the properties of compo-nents used in radio technology can bemeasured using relatively simplemeans, as is described in the examplebelow.

1.Introduction

There has already been an article in thismagazine by Jirmann [2] on the goodcharacteristics of the AD 606 analoguedevice (Fig 1 shows the pin connectionsfor the AD 606), with evidence obtainedfrom the use of an RF shielded, highpower signal generator and a precisioncalibration measurement circuit. Notevery radio amateur has such equipmentavailable. However, anyone who has, asa minimum, an oscilloscope, can use itfor methods described in [1] (The articlewill be placed on the VHF Communica-tions web site to assist understanding ofthis text - Ed.), as is demonstrated below.It proved not to be very easy to get holdof an AD 606, but one was obtained,

though it took a certain amount of timeand trouble.

2.Measurement procedure

The physical principles have alreadybeen described in [1]. The full measure-ment circuit is shown in Fig. 2. A fewspecial features have to be kept in mindif the measurement is to be carried outsuccessfully at first go.

• The logarithmic AD 606 intermediatefrequency amplifier IC has a rela-tively wide input voltage range, as isspecified in the data sheet. In order toinvestigate it and make the most of it,the generator should be able to sup-ply an initial output of +15dBm. Toobtain this, the voltage on the switch-ing transistor must be increased toapproximately 50V. Not every BF199 can withstand this voltage, andthere may be some background noise.So in case of emergency try anotherunit or use another type of transistor.

• The limiter outputs of the AD 606may not be left open circuit. Nor maythey be connected directly to +5V. Inboth cases there are feedback effects,which can distort the shape of the

Dipl. Ing. Detlef Burchard, Nairobi

A Simple Procedure forMeasurement

An example of the AD606 as a logarithmic amplifier


17

VLOG output curve below -60dBm.A 51 load resistor reduces thesefeedback effects to a minimum.

• Ringing prevention resistors areneeded on the output pin 6 and on thesupply voltage pins 13/14.

• The circuit should be kept as com-pact as possible to ensure wide band-

width and gain of the IC is preserved.

• The Amidon toroidal core is mountedbehind a screen made from printedcircuit board material or tinplate, sothat the coupling to the AD 606 isminimal.

• The probe on the input pin 16 is verysensitive and might act as an antenna

Fig 1: Pinconnections forthe AD 606.

Fig 2: Circuit diagram of the measurement system.


18

for interference. So miniature resis-tors are used (4.95kΩ = 3.65kΩ +1.30kΩ from the E96 range), with50Ω coax cable and a 50Ω termina-tion gives a scale factor of 100:1.This does not need any capacitivecompensation up to and above100MHz.

• There should be no fall off in thefrequency response of the oscillo-scope below 10 MHz.

Let us now look at the details of theprocedure and at the curves that areobtained. We need only to adjust theinitial voltage by varying the supplyvoltage of the BF199 to +15dBm = 1.25Veff = 3.5 Vss.

3.Understanding the outputsobtained

The logarithmic output voltage, VLOGof the AD606, goes through a low passfilter. As Fig. 3 shows, the time taken forthe voltage to build up to the correctvalue is 400ns, which requires four oscil-lations of the decay process. It can beestimated that the first decibel of thedecay process is thus incorrectly repre-sented. Consequently, if the decline takesplace one hundred times more slowly, thedistortion for the remaining curveamounts to only an additional 1/100 dB.

Fig 3: Output display with Y1 andY2: 500mV/div. X: 200nS/div.

Fig 4: Timebase extended to give 1/10amplitude (-20dB), X: 2dB/div.

Fig 5: Timebase five times slower, X:10dB/div.

Fig 6: Y1 gain adjusted to centre -45degree slope.


19

Now the time base sweep of the oscillo-scope is reduced slowly in uncalibratedmode, until the decaying oscillation atthe end of the screen has decreased toone tenth of the initial amplitude. Thetime axis now shows 2dB/section (Fig.4). It is switched to being five timesslower, using the step switch, and weobtain Fig. 5 with 10dB per section. Itcan be seen at a glance that the VLOGcurve displays the desired relationshipabove approximately 90dB, although theresolution is not yet high enough tocheck the manufacturers data.The Y1 amplification is now adjusted inuncalibrated mode, in such a way that thecurve goes down to below -45 degrees inthe central section and we obtain Fig. 6.Subsequently, the X and Y amplifica-tions can be further increased in knownstages, and the section of the curve that isof interest can be brought onto thescreen. For Fig. 7, both axes have beenextended by a factor of 5 and every smalldeparture from the ideal 45 degree linebecomes visible. In Fig. 7a, the line athigh levels can be seen. The manufac-turer gives no details on the response,+15bis + 5dBm. It can be seen that thecurve runs at an angle slightly greaterthan 45 degrees, but remains far belowthe guaranteed deviation of 1.5dB. Theslight ripple effect can also be seen. Inthe central section (Fig. 7b), the line isvery flat. A slight sagging of perhaps0.2dB can be recognised. Finally, in Fig.7c, we can see that the curve at -75dBmis approximately 0.5dB high, i.e. here itslightly deviates from the typical accu-racy limit of 0.4 dB.

4.Summary

As a whole, the AD606 has excellentcharacteristics. In practical applications,people will be happy to use the additionalunspecified input voltage range between

Fig 6a: Amplification of X and Ychannels increased and displaycentred on top part of curve, +15bis-5dBm.

Fig 6b: Display of -25bis -45dBm

Fig 6c: Display of -55bis -75dBm


20

+5 and +15dBm. If you work within a50Ω system, then a 10dB voltage ampli-fication can be brought about with a 1:3step up transformer, which is mounted infront of the AD606 and connected up at450Ω. The input resistance of the compo-nent lies between 500 and 2500, thesecondary load resistor must be individu-ally matched to the individual AD606.However, that is certainly reasonable fora dynamic range amplified to 90dB.

5.Literature

[1] D. Burchard (1993): Logarithmicconverters and measurement of theircharacteristics; VHF Reports, issue 3/1993, Pp. 140-154; Verlag UKW-Berichte,Baiersdorf and VHF Communications3/1994 Pp 174 - 190[2] J. Jirmann (1995): A high-precisionlogarithmic intermediate-frequency am-plifier. VHF Reports, issue 2/1995 Pp.97-101; Ver lag UKW-Ber ich te ,Baiersdorf and VHF Communications3/1996 Pp 185 - 188


21

A frequency counter is part of thestandard equipment in almost any ra-dio-frequency laboratory, but the fre-quency range usually goes up to nofurther than 1.3GHz. Although almostall measurements are carried outwithin this range, we nevertheless of-ten wish we could measure higherfrequencies as well. As an alternativeto purchasing an expensive microwavecounter, there is the option of expand-ing the range of an existing piece ofapparatus with an external pre-di-vider.

1.Circuit description

There are only a few components in thecircuit of the pre-divider for frequencycounters, Fig. 1 shows the wiring dia-gram. A similar pre-divider was pre-sented a few years ago in [1]. Since thePlessey SP 8910 divider IC used has notbeen obtainable for some time, theproject has tended to be forgotten. In themeantime, this IC has been brought back,and is currently available (once again) ina modern SMD housing from Zarlink [2].So what could be more obvious than todevelop a new frequency divider usingthis IC?At the pre-selector input of the circuit

there is a very broad band ERA-1 ampli-fier from Mini Circuits [3]. It amplifiesthe input signal up to 5GHz with ap-proximately 11 to 12dB, before the signalis fed to the actual divider, U2, at PIN 2.Like other dividers, this one also oscil-lates without an input signal, in whichcase approximately 550MHz can bemeasured at the output. On some divid-ers, this oscillation can be suppressed bymeans of a resistance between the inputpin and earth, but this was not successfulhere.The resistor R3 provides for an outputimpedance of approximately 50 Ohms,and the output level is approximately-10dBm. But the required input levelrepresents a greater problem. The curvein Fig. 2 shows how high the minimumlevel must be for the divider to functionsatisfactorily. We can also see that thedivider can still be used over 5GHz. Becareful the input levels are not too low!Fig. 3 shows what happens at the output,with an input frequency of 1GHz, if theinput level (27dBm) is too low, an inputfrequency of 2GHz is faked! Figs. 4 and5 in contrast show the spectrum at theoutput with the correct input levels. Inputlevels that are too high should likewisebe avoided.The supply voltage for the divider isstabilised with a fixed voltage regulator(U3). The one selected here was in aTO-220 housing.

Alexander Meier, DG6RBP

Pre Divider (:10) up to 5GHz


22

2.Circuit assembly

The circuit is assembled on a 45mm x30mm Teflon printed circuit board (Fig.6). This has a gap at one corner for thevoltage regulator, U3.When building the printed circuit boardin accordance with the component layout(Fig. 7), pay particular attention tomounting the amplifier U1. Its earth

connections must be connected to theearth side of the printed circuit board bythe shortest path, or it will have atendency to oscillate due to the feedinductances arising.Earth connections using through platingare not successful unless there is suffi-cient through connection. So anothermethod is used here, which is alwayssuccessful. The MMIC is sunk into ahole (∅ 2.3mm) in the board itself. Theearth connections are bent downwardsand soldered flush with the earth surface.

Fig 1: Circuit diagram of 5GHz pre divider.

Fig 2: Minimuminput level fordivider tofunctioncorrectly.


23

In the same way, the inputs and outputsare bent upwards and soldered to thetracks.When the board has been fully populated(except for the voltage regulator), theflux residues are cleaned off the earthside and it is screwed into the milledaluminium housing. Then the voltageregulator and the connectors can also befixed and soldered on.Before the housing is screwed down, thetop of the board should be cleaned againand the circuit should be tested. Youshould also make sure that U1 is notoscillating!

2.1. Parts listU1 ERA-1 (Mini-Circuits)U2 SP8910 (SO8, Zarlink)U3 7805C1 100pF, 0805C2,C3,C4,C7 1nF, 0805C5,C6 10nF, 0805C8,C9 100nF, 0805C10 4.7µF/25 V, SMDC11 1nF DF capacitor, M3

Actipass AFC-102P-10AL1,L2 10µH, SMD

Fig 3: Output ofdivider wheninput level is toolow (1GHz,-27dBm).

Fig 4: Output ofdivider wheninput level iscorrect (1GHz,-12dBm).


24

R1,R2 680R, 1206R3 100R, 12062 x N-flanged sockets,

small 4-hole flange1 x Teflon PCB, DG6RBP

002, through plated1 x Aluminium housing1 x Soldering lug 3.2 mm for

C11 (earth connection)

3.Technical data

Input freq. range 1 to 5 GHzOutput range 100 to 500 MHzDivider factor 10Input level approx. 13 to 7 dBmOutput level approx. 10 dBmInput/output N socketSupply +15 V, 120 mA

Fig 5: Output ofdivider wheninput level iscorrect (5GHz,-20dBm).

Fig 6: Printedcircuit boardlayout for 5GHzpre divider.


25

4.Literature

[1] Dr.-Ing. J. Jirmann und MichaelKuhne: Measurement aids for the UHFamateur, VHF Reports 1/93, Verlag UK-W-Berichte, Baiersdorf and VHF Com-munications 4/1993 Pp 207 - 213[2] Data sheet SP8910, Zarlink,

www.zarlink.com[3] Data sheet ERA-1, Mini-Circuits,ww.mini-circuits.com

Fig 7: Compo-nent layout for5GHz predivider.

Fig 8: Photo-graph ofcompleted 5GHzpre divider.


26

With over 1000 members world-wide, the UK Six Metre Group is the world’slargest organisation devoted to 50MHz. The ambition of the group, throughthe medium of its 60-page quarterly newsletter ‘Six News’ and through it’sweb site www.uksmg.org, is to provide the best information available on allaspects of the band: including DX news and reports,beacon news, propaga-tion & technical articles, six-metre equipment reviews, DXpedition news andtechnical articles.

Why not join the UKSMG and give us a try? For more information, contactthe secretary Iain Philipps G0RDI, 24 Acres End, Amersham, Buckingham-shire HP7 9DZ, UK or visit the web site.

The UK Six Metre Group

www.uksmg.org


27

Stanford Microdevices / SirenzaMicrodevices

Here you have to make a quick re-adjustment, the well known firm of Stan-ford has suddenly changed its name.Nevertheless, its activities remain un-changed. They still make new low noiseamplifiers, mixers, gain blocks, etc forthe frequency range between 0 and10GHz.And there are naturally lots of applica-tion notes and other interesting data todownload.Address:http://www.stanfordmicro.com

Flex-PDE

The Internet has some new EM simula-tion software, known as Flex-PDE for 2Dand 3D analyses in the microwave range.This is especially suitable for the ex-tremely high frequency range between 10and 100GHz. Those interested can down-load the free test version.The program is so universal that it canalso be used to investigate many otherphenomena, e.g. currents in fluids, heatconduction and distribution, chemicalprocesses, etc. The homepage also lists

some other interesting documents. Thetextbooks, in particular, are first class,and worth the price (e.g. $12 for Fieldsof Physics).Address:http://www.pdesolutions.com

Linmic

Again, something similar, but centring onwhat is (according to the publicity) acombination of EM simulators and layoutorientated CAD software packages whichis unique in the world. Here too there is ademo version for testing.Address:http://www.linmic.com

Metelics

Here you can, find out all about micro-wave diodes, i.e. Schottky diodes, PINdiodes, tunnel diodes, varactor diodes,etc. Naturally, there are also comprehen-sive catalogues, data sheets and applica-tion notes.Address:http://www.metelics.com

Gunthard Kraus, DG8GB

Internet Treasure Trove


28

Marki Microwave

Never heard of them? Well, anyone whoslooking for fast doublers, mixers, multi-pliers or converters at reasonable pricesfor the frequency range between 0 and40GHz in SMD format should take aglance at this page.Address:http://www.MarkiMicrowave.com

Rf Nitro

This site has nothing to do with artificialfertilisers or explosives. It deals with thecutting edge of MMIC development inGaAs or GaN technology. A wide rangeof products for the microwave range andsome excellent application notes practi-cally compel you to visit this site.Address:http://www.rfnitro.com

National Instruments

A company manufacturing and market-ing measurement technology hardwareand software on such a large scale (thinkof LabView, for example) is naturally areal Treasure Trove for those with rel-evant interests. Here we find not onlydata sheets and test software CDs butalso separate document packages for us-ers and developers. Also on offer aretutorials on various subjects and productgroups.You could spend hours finding more andmore items of interest. The documenta-tion on FFT, in particular, is outstanding!Address:http://www.ni.com

Mini Circuits

Well known manufacturer of active radiofrequency components including, amongothers, mixers and MMICs such as, forexample, the Era range.Many comprehensive data sheets.Address:http://www.Mini-Circuits.com

NoteOwing to the fact that the content ofInternet sites can change rapidly, andInternet addresses and the sub categoriesof homepages can change without warn-ing at any time, it is not always possibleto keep information up to date.We therefore apologise for any inconven-ience if Internet addresses listed in issuesof Internet Treasure Trove cease to beaccessible, or if they are altered at shortnotice by their operators.However, the editors and the author willbe happy to help in discovering the newaddress for the site in question and / orobtaining any documents referred to.We would also like to take this occasionto point out that the author and thepublishers are in no way responsible forthe correctness or otherwise of any infor-mation listed here.


29

VHF CommunicationsBack Issues

• Most back issues available from 1969 onwards, an up to datelist is maintained on the VHF Communications web site or seepage 63 of this issue. Some difficult to obtain issues can besupplied as photocopies.

• Locate interesting articles by searching the full index on theweb site, then order the magazine using the secure form on theweb. You can also order by post or fax.

• If you are new to VHF Communications magazine or havemissed some volumes or issues, choose one of the back issuesets or individual magazines to make your collectioncomplete.

• Keep your magazines in good condition with Blue Bindersthat hold 12 issues, £6.50 each + P&P

• Single issue from 1969 to 2000, £1.00 each + P&P

• Single issues from 2001 volume, £4.70 each + P&P

• Complete 2001 volume, £18.50 + P&P

• Back issue set 1972 to 1999 (51 magazines), £45.00 + P&P

• Back issue set 1972 to 2001 (59 magazines), £65.00 + P&P

K M Publications, 63 Ringwood Road, Luton, Beds, LU2 7BG, UK

Tel / Fax +44 1582 581051, web site www.vhfcomm.co.uk


30

Index of Volume 33 (2001)

Article Author Edition Pages

Antenna TechnologyModern Patch Antenna Gunthard Kraus 2001/1 49 - 63Design Part I DG8GB

Modern Patch Antenna Gunthard Kraus 2001/2 66 - 86Design Part II DG8GB

Designing Long Yagis Richard A Formato 2001/3 139 - 155with YGO3 WW1RF

An Intersesting Program Gunthard Kraus 2001/3 166 - 175PCAAD21 - An Antenna DG8GBAnalysis Program

Optimising Yagi Antennas Joop van Sundert 2001/4 206 - 212With YGO3 Genetic PD1APOOptimiser

The Fractal Antenna Angel Viliaseca 2001/4 213 - 226HB9SLV

Audio Frequency TechnologyDigital Speech Store Wolfgang Schneider 2001/2 87 - 91

DJ8ES

Digital Speech Store Wolfgang Schneider 2001/3 156 - 157Instructions and DJ8ESImprovements to the articlein Issue 2/2001

FiltersModern Design for Band Gunthard Kraus 2001/4 228 - 251Pass Filters Made From DG8GBCoupled Lines


31

FundamentalsHigh Precission Frequency Wolfgang Schneider 2001/1 2 - 8Standard for 10MHz Part II DJ8ESFrequency Control via GPS

GMSK, The Modulation Prof. Gisbert 2001/1 9 - 19Used For Mobile GlasmachersCommunications

Shielding Technology Using Dipl. Ing. Hermann L 2001/1 20 - 27Metalised Non Wovens Aichele

Modern Patch Antenna Gunthard Kraus 2001/1 49 - 63Design Part I DG8GB

Modern Patch Antenna Gunthard Kraus 2001/2 66 - 86Design Part II DG8GB

The Noble Art Of Carl G Lodstrom 2001/2 114 - 118De-Coupling SM6MOM, KQ6AX

Is Silver Plating Worth Wolfgang Borschel 2001/2 119 - 122While in RF Applications DK2DO

Line Sections as Cpacitances Michael Hein 2001/3 159 - 165or Inductances in MicrowaveRange

An Interesting Program Gunthar Kraus 2001/4 199 - 205TRL85.exe DG8GB

Measuring TechnologyHigh Precission Frequency Wolfgang Schneider 2001/1 2 - 8Standard for 10MHz Part II DJ8ESFrequency Control via GPS

Tracking Generator from Carsten Kraus 2001/1 35 - 481MHz to 13GHz DJ4GCFor Spectrum Analysers

A Sinadmeter E Chicken 2001/2 92 - 101G3BIK

70MHz Preamplifier for Wolfgang Schneider 2001/4 194 - 198Frequency Counter DJ8ES

Micro Computer TechnologyUniversal Micro Controller Bernd Kaa 2001/2 108 - 113Board, Uniboard C501 DG4RBF


32

MiscellaneousInternet Treasure Trove Gunthard Kraus 2001/1 28 - 29

DG8GB

Radio Astronomy Terms Hermann Hagn 2001/2 102 - 104Explained DK8CI

Astronomical Observations Hermann Hagn 2001/2 105 - 107At The German MuseumMunich

Internet Treasure Trove Gunthard Kraus 2001/2 125 - 127DG8GB

MIMP, Motorolas Impedance Henning Christof 2001/3 130 - 138Matching Program Wedding, DK5LV

An Improved BSPK Matjaz Vidmar 2001/3 176 - 188Demodulator for the S53MV1.2MBit/s Packet Radio RTX



2m BandSupplement to Article on Gerhard Schmitt 2001/2 123 - 125Low Pass Filter for 2m DJ5APand 70cm

A complete index for VHF Communications from 1969 to the currentissue is available on the VHF Communications Web site. This index canbe searched on the web site. There is versions of the index in pdf andExcel format that can be downloaded from the web site so that it can besearched on your own PC.


33

PUFF version 2.1Microwave CAD Software

•••• Available again

•••• Complete with full English handbook

•••• Software supplied on 3.5 inch floppy disc

Price £23.50 + shipping

Shipping - UK £1.50, Surface mail £3.00, Air mail £5.00

As used in article on Stripline Low Pass Filters in this issue

Visit the VHFWeb site www.vhfcomm.co.uk•••• Text of some past articles

•••• List of all overseas agents

•••• Secure form to subscribe and order back issues or kits

•••• Full index of VHF Communications from 1969 to the current issue,this can be searched on line or downloaded

•••• Up to date list of back issues available

•••• Links to other site including all of those from Internet Treasure Trovearticles

•••• Downloads from some article including YGO3 the Yagi designprogram


34

The subject of low pass filters is cov-ered through a comparison of designs,with the assistance of contemporarydevelopment tools based on PC pro-grams. The special features of MicroStripline filters are also referred to.

1.Introduction

The current situation and recent progressin relation to the design of stripline bandpass filters have already been discussedhere The present article is devoted to lowpass filters. In the first section, the workis done using the tried and tested micro-wave CAD program PUFF, while in thesecond section the design work is re-peated using the most recent studentversion of ANSOFT Serenade. The re-sults can thus be compared with oneanother and the progress made in micro-wave CADs and in circuit simulation canbe evaluated Finally, this subject isrounded off with measurements on theassembled specimen circuits.

2.Filter design using PUFF

2.1. Design principles for striplinefiltersThe filter type chosen is the Chebyschevlow pass (for more detailed informationon the selection of filters of a suitabletype and grade, see the article referred toin [1]).The following values are given:Filter grade n Number of poles = number

of components = 5ππππ format (Shut capacitors with series

inductors)Impedance Z 50ΩΩΩΩMaximum ripple 0.1dBof S21 intransmission rangeCut off frequency 1,700 MHz

The components required are calculatedin accordance with these parameters,using the tried and tested filter program“fds.zip” (Fig. 1). We therefore need twocapacitors, each with 2.147pF, a capaci-tor with 3.67pF and two coils, each with6.42nH.Now we use the well known method forimplementation (Fig. 2):The capacitors are created using short but

Gunthard Kraus, DG8GB

Modern Design of Stripline LowPass Filters


35

very wide stripline sections with an im-pedance of less than 20Ω. This ensuresthat the inductive fraction of the lineplays no great role. Very thin, short lines(with Z = at least 100Ω) are used toproduce the inductances, provided thecapacitive fractions are still sufficientlylow.For successful design, stick to the follow-ing rules, which originate from practicalexperience in filter construction:The electrical lengths of the line sectionsshould lie between 10 and 30 degreeswhen the cut off frequency of the filter isused as the design frequency of the line.For the “capacitors”, the impedance levelof the line sections used for a 50Ωsystem should not exceed 20Ω. For the“coils”, select lines with at least Z =100Ω.For the “wide capacitor lines”, the widthto length ratio should not exceed 8.For the “thin inductance lines”, stick asclose as possible to a minimum conduc-tor width of approximately 0.2 mm as thelower limit. This results in a high induct-ance and thus short line lengths. Unfortu-nately such extremely thin tracks arenotable for their high losses; the linestherefore have to be made slightly wider,which makes it easier to manufacture theprinted circuit board.But we must be clear about one thing:Replacing genuine capacitors and coilswith line sections does not immediatelylead to a perfect solution, since it is wellknown that any line consists of inductiveand capacitive parts. In addition, trans-

formed lines also have resistances, thesource resistance and the load resistance.Consequently discrepancies are possibleon the first simulation run, thus optimisa-tion is indispensable.

2.2. Determination of line dataUse a text editor to open the SETUP filefor PUFF, in order to enter and then storethe design frequency, fd, together withthe data for the board material used(Rogers R04003, with a thickness of 32MIL). They are shown in bold in thefollowing file extract:\b{oard} .puf file for PUFF, version 2.1d}d 0 {display: 0 VGA or PUFF chooses, 1

EGA}o 1 {artwork output format: 0 dot-matrix, 1

Laser Jet, 2 HPGL file}t 0 {type: 0 for microstrip, 1 for stripline, 2

for Manhattan}zd 50.000 Ohms {normalizing impedance. 00}s 100.000 mm {circuit-board side length. s>0}c 100.000 mm {connector separation. c>=0}r 0.010 mm {circuit resolution, r>0, use Um for

micrometers}a 0.000 mm {artwork width correction.}mt 35 Um {metal thickness, use Um for

micrometers.}sr 5.000 Um {metal surface roughness, use Um

for micrometers.}lt 1.0E-0003 {dielectric loss tangent.}cd 5.8E+0007 {conductivity of metal in

mhos/meter.}p 2.000 {photographic reduction ratio.

p

Next use a width of 0.25 mm for the“inductance line section”, thus there is nogreat problems in producing the printedcircuit board.Then PUFF is started to determine theimpedance levels for the conductorwidths. This is done in the followingmanner. A real model transmission line isentered in field F3, with an impedancelevel of, e.g., 100Ω and an electricallength of 90 degrees. The “real modellingmode”, as is well known, is invokedusing the exclamation mark after theletter “t”. If the equals sign is nowpressed, then the actual impedance leveland the actual electrical length appear inthe dialogue field. The entry value isaltered until an actual impedance level of120Ω appears in the dialogue field. AsFig. 3 shows, it is necessary to increasethe input value to 126.337Ω? to achievethis. If we now delete the exclamationmark and press the equals sign again, wesee the mechanical data for this line. Fig.4 shows that the width of 0.25 mm iswell attained.In order to add the two outer 2.147pF

capacitors, we use lines with an imped-ance of Z = 17Ω, whilst for the centralcapacitor (3.67pF) a wider line with Z =10Ω makes better sense. If the line is toolong, this unfortunately makes the attenu-ation worse at very high frequencies,which also accounts for the recommenda-tion for an electrical line length of 10 to30 degrees. We use PUFF for these twolines as well, to determine the necessaryconductor widths. We should also in-clude the 50Ω feed lines to connect thefilter structure with the connection sock-ets.If we keep strictly to the methods re-ferred to, we finally obtain the entryvalues in accordance with Fig. 5. If wefinally remove the exclamation mark ineach line and enter the equals sign, wecan compile the first little table:Line 50 ΩΩΩΩ 10 ΩΩΩΩ 17 ΩΩΩΩ 120 ΩΩΩΩConductor width (mm) 1.84 14.62 7.94 0.25

We now continue using simple approxi-mation formulae to determine the re-quired line lengths. Here we start withthe narrow lines serving as coils (w =

Fig 3: Adjustingthe input toobtain therequiredimpedance of120ΩΩΩΩ.


37

0.25 mm), where:

ZL is the selected impedance for the“inductive line”, f is the actual frequencyand L is the required inductance.For f = 1700MHz, L = 6.419nH and ZL=120Ω we obtain the following standard-ised length:

Since a complete wavelength corre-sponds to an angle of 360 degrees, thisline section will have an electrical lengthof 0.090935 x 360 degrees = 32.74degrees. We obtain the associated me-chanical length through the well knownprocedure using PUFF:Put an exclamation mark behind “t”,enter the length, and continue to correctuntil this value is finally obtained whenthe equals sign is entered (Fig. 6). Nowdelete the exclamation mark and pressthe equals sign again to obtain the figuresfor the mechanical dimensions. Thisgives a physical length of 10.53 mm.Summary of complete line data for in-ductances with a design frequency of1,700 MHz:Z 120 ΩΩΩΩElectrical length 32,74 DegreeMechanical length 10.53 mmConductor width 0.25 mm

The capacitor line calculation requiresthe following approximation formula:

Fig 4: Removethe equals sign tocalculate themechanical datafor the line.

L

LineL

ZLfI ⋅=−

λ

090935.0120

41.61700 =⋅=−λLineLL

Fig 5: Real inputs for 120ΩΩΩΩ, 17ΩΩΩΩ, 10ΩΩΩΩand 50ΩΩΩΩ lines.


38

where Zc is the impedance of the capaci-tor line.For the first and third capacitors with C =2.147pF, at 1,700MHz we obtain:

That gives an electrical length for the lineof 0.062 x 360 degrees = 22.32 degreesand in addition PUFF supplies a conduc-tor length of 6.31 mm with a conductorwidth of 7.94 mm.Summary of full line data for first andthird capacities for a design frequency of1,700MHz:Z 17 ΩΩΩΩElectrical length 22.32 degreesMechanical length 6.31 mmConductor width 7.94 mm

It is frequently necessary to repeat thisrun for the central capacitor of 3.698 pF:

That gives an electrical length for the lineof 0.062866 x 360 degrees = 22.63degrees and a conductor length of 6.29mm with a conductor width of 14.62 mm.Summary of full line data for centralcapacity for a design frequency of1,700MHz:Z 10 ΩΩΩΩElectrical length 22.63 degreesPhysical length 6.29 mmConductor width 14.62 mm

2.3. Determination of substitute datafor stripline stepsIn the existing circuit, a repeated change(Impedance Step) is carried out betweenwide and narrow stripline sections. Buteach individual step acts as an irregular-ity, the approximate effect of these canbe determined by using a substitute cir-cuit made from a series inductance and acase capacitance (like a simple low passfilter!).

Fig 6: Calcula-tion of theelectrical lengthof the lines.

CLineC ZCfI ⋅⋅=−

λ

062.017147.21700 =⋅⋅=−λLineCI

082866.010698.31700 =⋅⋅=−λLineCI


39

If the simulated circuit is to correspondprecisely to the parameters (as far aspossible) without a lot of reworking, andat the same time be capable of use, wecan not avoid the need for these circuitextensions. But they are already in exist-ence in a modern microwave CAD pro-gram as modules and the correction isthus relatively simple. PUFF is unfortu-nately not that well developed yet, so thecalculation formulae from [4] are used.In order to specify these substitute com-ponents, we require not only the imped-ance values and conductor widths of eachline section, but also the associated effec-tive dielectric constants. These constantsare determined and used by PUFF duringits calculations, but unfortunately theyare not output. So here we have to outwitthe program:First step:Use a pocket calculator to work out thefree space wavelength for the frequencyof 1,700MHz

Second step:Increase the inputs on all four lines infield F3, one after another, until the reallengths are precisely 360 degrees whenthe equals sign is pressed (i.e. one wave-length). Then delete the exclamationmark in each line and enter the equalssign instead. This will give you thecorresponding physical lengths. Theyare:Line 50 ΩΩΩΩ 10 ΩΩΩΩ 17 ΩΩΩΩ 120 ΩΩΩΩConductor length 108.496 99.998 102.004 115.785

Third step:Using the following simple rearrangedshort cut formula,

All the effective dielectric constants cannow be determined using the pocketcalculator:Line 50 ΩΩΩΩ 10 ΩΩΩΩ 17 ΩΩΩΩ 120 ΩΩΩΩConductor width 1.84 14.62 7.94 0.25εεεεeff 2,646 3,114 2,993 2 ,323

All the step components can now bespecified using this data. The seriesinductance at each step can be deter-

Fig 7: Circuit diagram of low pass filter produced using ANSOFT Seranade,showing impedance steps.

mmfc 47.176

170010.3 8 ===λ

2

=

Line

aireff λ

λε


40

mined using the following approximationformula:

where: h is the board thickness in µm, Z1the impedance level of the wide line, Z2the impedance level of the narrow line,ε1eff the effective dielectric constant ofof the wide line on the board, ε2eff theeffective dielectric constant of the narrowline on the board. The inductance is innH.Thus we can once again compile a table:Transition from 50 ΩΩΩΩ to 17 ΩΩΩΩ to 120 ΩΩΩΩ to

17 ΩΩΩΩ 120 ΩΩΩΩ 10 ΩΩΩΩ

Inductance 1.84 nH14.62 nH 7.94 nH

It is somewhat more laborious to calcu-late the case capacitance arising:

where: Z1 = Impedance of the wide line,ε1eff = effective dielectric constant ofwide line, W1 = conductor width of wideline, W2 = conductor width of narrowline, h = board thickness in µm (here:0.813 mm = 813 µm). The capacity is inpF.With the above data, we obtain a secondtable using the pocket calculator:Transition from 50 ΩΩΩΩ to 17 ΩΩΩΩ to 120 ΩΩΩΩ to

17 ΩΩΩΩ 120 ΩΩΩΩ 10 ΩΩΩΩ

Capacity 0,1 pF 0,125 pF 0,227 pF

When drawing the simulation wiringdiagram in the next chapter, please bearin mind that:series inductances are always connecteddirectly to the narrower line section,whilst the shunt capacitances are alwaysmounted parallel to the wider line sec-tion!The circuit diagram that applies to thesimulation, with all the irregularities isshown in Fig. 7. It was drawn using thecircuit editor of ANSOFT Serenade and

⋅−⋅⋅=

eff

eff

ZZhL

2

1

2

11000987.0ε

ε

+

+⋅

−

+⋅⋅

−⋅⋅=

8.0

264.0

258.03.0

100137.01

1

1

1

1

2

1

1

hWhW

hWW

ZC

eff

effeff

εεε

Fig 8: Results ofsimulating idealfilter.


41

immediately gives an impression of thelayout of this modern program. But thenumber of components required is notvery far off the maximum of 25 compo-nents that can be used with the freestudent version of ANSOFT Serenade.

2.4. Simulation and optimisation ofcircuitFirst we have to simulate the ideal lowpass filter made from coils and capaci-tors, in order to have the optimisationtarget visible. The result can be seen inFig. 8, where special attention should bepaid to the precise heights of the two“camels humps” of S11. S11 rises rightup to -16.44 dB since it is associated withthe S21 ripple of -0.1 dB). If this value isobtained again following the optimisa-tion of the stripline circuit, the ripple isalso tuned, being approximately -0.1 dBin the transmission range. It is increasedmerely by the attenuation, that rises withthe frequency, due to the printed circuitboard losses and the skin effect, etc.The ideal curve for S21 is shown in Fig.9 on a greatly expanded scale. From now

on, these two sheets should be kept nextto the PC for the work that follows.We now enter all the components re-quired for the circuit in field F3. Makesure you dont forget the exclamationmark for all line sections. See Fig. 6 forthe precise data for the lines. Onceeverything has been entered, the circuit iscorrectly assembled in F1, then switchedto F2 and simulated. The simulationresult from S11 and S21 in the frequencyrange from 0 to 5GHz and the amplituderange from 0 to 1dB for monitoring theripple is shown in Fig. 10. In Fig. 11 theamplitude range between 0 and -50dBhas been selected in order to check thetwo S11 humps. It can be seen that theimpedance leaps have the most importanteffect, and also that the cut off frequencyhas fallen by almost 400MHz.But dont worry: optimisation is not espe-cially difficult, for you merely have tobalance out the influence of the addi-tional inductances:

• The camel hump at the bottom righthand corner can be lifted by means

Fig 9: Magnifiedview of S21curve.


42

Fig 10:Simulationresults from 0 to5GHz.

Fig 11:Expanded viewof 0 to 5GHzresults.


43

Fig 12: Resultsshowing S21curve afteroptimisation.

Fig 13: Resultsshowing S11curve afteroptimisation.


44

of an extension of the two external17Ω line sections.

• Then, if necessary, the coils (the twothin 120Ω line sections) are slightlyshortened, and in this way the twohumps become the same height(target = -16.44dB).

Once this has been done, the displayrange for S21 and S11 is restricted to 0 to1dB and simulated again. The pre setripple value of 0.1dB can then be veryeasily recognised on the fundamentalattenuation, which rises as the frequencyincreases. Yet the cut off frequency ofthe filter is still much too low. Wetherefore simply increase the design fre-quency in field F4 until, at 1,700MHz,we can observe the sharp transition fromthe transmission range into the filterattenuation band. This is at fd =2,100MHz in this case. Once again, weconvert this to the display range between0 and 50dB using a linear plot diagramand the precise height of the two humpsis checked with the help of the cursor.Any small discrepancies present areeliminated in accordance with methods a)and b). The results can be seen in the twodiagrams, Figs. 12 and 13.Finally, another table can be compiled, inwhich the definitive mechanical and elec-trical data for the individual line sectionsare entered after optimisation. These arenot needed for the draft layout, but theycan be used as a way in to monitor thisPUFF design in the second part, usingANSOFT Serenade.Make sure that:

• The design frequency has been

increased to 2,100MHz and that thematter of the open end extension(the increase in the line lengthcaused by projecting field lines) hasalso been dealt with automaticallythrough the insertion of theimpedance steps.

• The pre set mechanical lengths andwidths for the line sections maytherefore be transferred directly intothe printed circuit board layoutwithout any further correction! SeeTable 1 below.

2.5. PUFF versus HARMONICAUsing PUFF we can switch, by means ofa simple exclamation mark, from idealmodelling (entering impedance level andelectrical length at design frequency) toreal modelling (with physical length andwidth, together with all the line losses).ANSOFT unfortunately separates thesetwo options completely and even usesdifferent graphic symbols in the twocases. So separate wiring diagrams andprojects must also be drawn up for it!Here we can go straight to the simulationof physical dimensions in the first run.The original PUFF simulation is investi-gated (housing cover not taken into ac-count). In the second run a space of13mm between the board and the cover isused which simulates typical apparatusand its influence determined.

2.5.1 Simulation check of PUFF designwith SERENADESerenade is started up, a new project isset up and then the circuit is drawn.Assistance should be obtained from Fig.

Width 1.84mm 14.62mm 7.94mm 0.25mmLength 5.09mm 6.15mm 8.20mmElectrical length in deg. 22.63 26.8 31.51at 2100MHz

50Ω line 10Ω line 17Ω line 120Ω line

Table 1: Pre set mechanical dimensions.


45

14 here. It shows the buttons behindwhich the various components and ele-ments are hidden. The correct entries inthe substrate screen can be seen in Fig.15 (board and housing data), where theunit mm should not be forgotten, for alldimensions. Fig. 16 shows the circuitready for simulation, including the fre-quency module for the sweep from 0 to5GHz in 5MHz steps.As soon as all this has been done, theanalysis button (button with the littlegear on top) can be pressed. At firstnothing can be seen, until the turquoise /grey Report Editor button is pressed, thisgives the option of outputting the resultsin diagram form. S11 and S21 are shownin dB as set up by the method shown inFig. 17. Fig. 18 shows, on a greatlyenlarged scale, the result of the simula-tion. Both the shape of the S21 curve andthe ripple cut off frequency entered (withattention being paid to the fundamental

filter attenuation!) tally completely withFig. 12, the simulation using PUFF. Thiscan also be seen in Fig. 19, here theresults are represented without magnifi-cation and thus the shape of the S11curve can be checked. A comparisonwith Fig. 13 gives a big boost to our trustin PUFF.

2.5.2. Simulation using housing dataTo do this we first call upon the StriplineCalculator TRL85 and calculate the cor-rected physical line data for an interval of13 mm between board and cover. Thefirst step in doing this is naturally toprepare the line and board targets re-quired:First use the line data for the simulationusing PUFF at a design frequency of2,100MHz (see Table 1) and then thetechnical data for the printed circuitboard. The interval referred to between

Fig 14: The opening window of Seranade showing main items of interest.


46

the cover and the board is added here aswell (= marked in bold):Material: Rogers R04003Board thickness: H = 32 MIL = 0.813 mmDielectric constant: er = 3.38Loss factor: TAND = 0.001Copper coating: TH = 35 µmSurface roughness: RGH = 5 µmDistance betweencover and board: HU = 13 mm

The determining the new values for thefirst line section (Z = 17Ω, electricallength = 26.8 degrees at 2,100MHz) and,once again, the steps required to get thereare shown in Fig. 20. If we repeat that forthe other lines, then we finally obtain the

results table (Table 2).Anyone who compares the two tables(with and without housing) is immedi-ately struck by the fact that including thecover distance referred to affects only thewide line sections, their conductor widthis reduced. This is easy to understand,since some additional electrical fieldlines are pulled up by these large plates.This increases the capacity and must becorrected, whereas for the short “fips”,showing the widths of 120Ω and 0.25mm, scarcely anything is noticeable.Now the previous HARMONICA projectis re started and every physical line valuein it is checked and / or corrected in

Fig 15: Entry ofsubstrate data.

Fig 16: Circuit diagram ready for simulation.


47

accordance with the table.The new entry “HU = 13 mm” for thesubstrate control block is very importanthere, and should not be forgotten!If we now start the simulation and extractand magnify S21 (Fig. 21), all commentbecomes superfluous: see Fig. 12, seeFig. 18.One further tip: in diagrammatic repre-sentations, the right hand mouse buttonconceals not only the zoom functions, butalso the options for the insertion of fixed

or changing data markers, together withthe output of the associated curve valueat the selected frequency in small addi-tional windows. You should try out theseoptions yourself!

Fig 17: Setting up reports editor.

Electrical length in deg. 22.63 26.8 31.51at 2100MhzWidth 1.83mm 14.53mm 7.90mm 0.25mmLength 5.07mm 6.13mm 8.20mm


Table 2:


48

3.Repetition of design usingANSOFT Serenade

3.1. Assembly of filter circuitFor this, go right back to the beginningand begin as if you knew absolutelynothing about the preceding design usingPUFF. Just avoid duplicating work bytaking all the data already determinedwhich are valid for both designs fromSections 2.1 to 2.3.First step:Use a filter program to determine thecoils and capacitors required for thecorresponding LC low pass. Fig. 1 givesthe following values for this: 2.147pF,6.419nH, 3.698pF, 6.419nH, 2.147pFSecond step:Use the approximation formulae pro-vided to determine the data for the line

sections for a design frequency of1,700MHz. (Do you remember ? Wedidnt get involved with optimisation withPUFF until we reached a frequency of2,100MHz, so we ignore that here too).In accordance with the selected indi-vidual impedance level, the followingvalues were obtained from Table 3.Third step:Use the TRL85 calculator to determinethe physical dimensions of the individualline sections, and also include the coverdistance of 13 mm in the list of givenprinted circuit board data:Material: Rogers R04003Board thickness: H = 32 MIL = 0.813 mmDielectric constant: εεεεr = 3.38Loss factorr: TAND = 0.001Copper coating: TH = 35 µmSurface roughness: RGH = 5 µmDistance betweencover and PCB: HU = 13 mm

The associated simulation (Synthesis) forthe first line section, with 17Ω at the

Fig 18: Resultsof simulationshowing S21curve.

Electrical length at 22.63 22.32 32.741700MHz


Table 3:


49


Fig 20: Setting up data for simulation taking distance to cover into account.


50

frequency of 1,700MHz, shows Fig. 22.The remaining results have been entereddirectly into Table 4.Fourth step:Now we must tackle the irregularities byaltering the conductor widths (ImpedanceSteps). Unfortunately ANSOFT was a bitmean here and blocked this option in thestudent version (and who just happens tohave $ 20,000 to spare for the fullversion?). So we have to go back to theformulae in Section 2.3 and determinethe substitute components using thepocket calculator. Fortunately, the con-ductor widths with and without the hous-ing cover do not differ very much, so thecomponent values can be transferredfrom the section referred to (quite liter-ally, its only in the third decimal placethat there is a small difference!):

Transition from 50 ΩΩΩΩ to 17 ΩΩΩΩ to 120 ΩΩΩΩ to17 ΩΩΩΩ 120 ΩΩΩΩ 10 ΩΩΩΩ

Capacity 0.1pF 0.125pF 0.227pFInductance 0.327nH 0.565 nH 0,655 nH

Fifth step:A new project is set up and then thecircuit is assembled, using the physicalline model. But for the length specifica-tions of the lines we enter not values, butvariables, to make optimisation easier.The following classification applies here:

• 17Ω line: Line1

• 120Ω line: „Line2

• 10Ω line: Line3Here we access the variables block

Fig 21: Resultsof simulationtaking intoaccount distanceto cover, showingS21 curve.

Electrical length in deg 22.63 26.8 31.51at 1700MHzWidth 1.83mm 14.53mm 7.89mm 0.25mmLength 6.27mm 6.31mm 10.52mm


Table 4:


51

through the “Parts” button, followed by“Control Blocks” and finally “Vari-ables”. Double click on this block toshow the entries required. The lowestlimiting value permitted, the target valueand the upper limiting value for optimi-sation must always be entered betweenquestion marks (Fig. 23). In Fig. 24 wecan finally see the complete circuit, readyfor simulation, together with a substratecontrol block, a frequency control block,a variables block and an optimisationblock.Sixth step:Now the simulation is carried out andthen the results are analysed (Fig. 25 andFig. 26). The result is exactly the same asthe simulation with PUFF from Figs.11and 12. Because of the components caus-ing irregularities, the ripple cut off fre-quency has fallen to 1,300MHz and the

two camels humps in S11 are of differentheights.3.2. Adjusting components foroptimisation of designHere we basically proceed as for thedesign with PUFF. In the first run, wevary the length of the first and last linesections, together with that of the narrowlines acting as inductances. The tuningmode is available for such purposes. Thefollowing conditions apply:

• Extending Line1 lifts the right hump,S11, in particular.

• Shortening Line2 lifts the left hump,S11, in particular.

In this manner, we can bring the maxi-mum values of the two S11 camelshumps to a value of -16.44 dB. Finally,we increase the cut off frequency by

Fig 22: Setting up simulation for repitition of PUFF design.


52

means of simultaneous reduction of allline lengths by the same factor to therequired 1,700MHz.Now we begin, for example, with the twoouter 17Ω line sections and their lengthLine1. First we call up the circuit dia-gram onto the screen and execute thefollowing checklist exactly (Fig. 27):

• Check that the Accumulate buttonhas been activated.

• Click on the Tune button, whichcalls up the Tune menu.

• Double click on Variables ControlBlock in the circuit diagram to

ensure that the three variables (Line1to Line3) are transferred into thetuning list. Then click on Line1 toselect it.

• Switch over to Sweep

• Delete the tick for Step Width in %

• The start value ( 6.5mm selected) ,the final value ( 8mm selected) andthe step width of 0.5mm can beentered in the corresponding fields.

• We are ready, we can press Tune.But first, you should call up thewindow showing the representationof S11 and S22 to the foreground,

Fig 23: Settingup variables forsimulation.

Fig 24: Completed circuit ready for simulation.


53

this is very useful for looking atindividual simulations andclassifying them.

We can see immediately from Fig. 28that a value of Line1 = 7.5 correspondsalmost exactly to the required specifica-tions. It is thus entered into the list as anew target value by means of doubleclicking on Variables Control Block andthen the simulation is repeated. This timewe are dealing with the precise value ofthe present limiting ripple value, whichcan be taken from Fig. 29 as S21 = -0.26dB at 1,380 MHz. This is 320MHz toolow and so we use our pocket calculator

in order to shorten all line lengths by thesame factor. The relationship which ap-plies is:

and thus we obtain the following values,which are immediately entered into thevariables control block again:

• Line1 = 6.31 mm

• Line2 = 8.85 mm

• Line3 = 5.28 mm



( )32017001700+⋅= OLDNEW

II


54

A new simulation immediately providesinformation on the effect of thesechanges. It can be seen that here we havea cut off frequency of 1,620MHz and thata further correction of 80MHz is re-quired. (One striking factor here is that

the design frequency of the lines is nowobtained at 2,100MHz, which corre-sponds exactly to the design usingPUFF).So now we have the following line

Fig 27: Setting up for tuning process.

Fig 28: Results of tuning showing S11curve.

Fig 29: Results of tuning showing S21curve.


55

lengths for the next stage:

• Line1 = 6.07 mm

• Line2 = 8.52 mm

• Line3 = 5.08 mmIf we use these for a simulation, wealready have a cut off frequency of1,670MHz. Yet we should look at theshape of S11 again, for these correctionshave changed the hump amplitudesagain, by 0.5dB. So Its clear what mustbe done next:First correct the shape of S11 throughLine1 and (if necessary) Line2, until youarrive at the desired hump amplitude ofapproximately -16.5dB. Then check themagnified shape of S21 to find thecurrent value of the limiting ripple fre-quency (rated value : -0.26dB at1,700MHz) and change all the lengthlines again, if necessary, by the samecorrection factor. If you are not satisfied,repeat this procedure again. Here, as acheck and as an example, is the optimisa-tion result for the printed circuit boarddesign (Table 5).Fig. 30 shows the associated final simu-lation of S21 as a proof of the successful

design. In Fig. 31, on the other hand, wecan see a sweep from 0 to 10GHz. Itshows the required shape for S11 anddemonstrates what should never be for-gotten in relation to these microstrip linefilters, the attenuation rises to a maxi-mum and then falls again (the technicalexpression is scatter resonance). In prin-ciple, this is unavoidable when line sec-tions are used in filters, and can, forexample, be improved by inserting afurther low pass filters with a higher cutoff frequency in series.Finally, it is very interesting to comparethis design with the one using PUFFfrom the first part of the article. It canactually be seen that the differences inthe line lengths for all designs amount toonly a few hundredths of a millimetre,and it is therefore certainly the time toprepare some test boards. Their measureddata (with and without housing) will thendecide on any further steps necessary tomake the design ready for production.Normally, an accuracy of approximately2% can be obtained in the first stage byapplying the design principles of thisarticle.

3.3. The Optimisation Option usingHARMONICAThere is also the option of carrying outthis optimisation entirely by a program.To this end, for specific frequencies orfrequency ranges, a minimum value or amaximum value can be entered for sev-eral parameters.It would also be entirely conceivable toforecast as many points as possible onthe S11 curve and / or the S21 curve forthis filter. Then carry out about 1,000tests, and thus to get the computer to sortout the problem, which in certain circum-

Fig 30: Final simulation resultsshowing S21 curve.

Width 1.83mm 14.53mm 7.89mm 0.25mmLength 5.08mm 6.14mm 8.20mm


Table 5:


56

stances can be done within hours.Unfortunately a maximum of 3 optimisa-tion object are possible using the studentversion, and that just isnt enough. Aprogram with so few basic points to workon quite often bends the filter curves inunpredictable or impermissible ways dur-ing the optimisation. So there would haveto be a lot more objects for this to worksuccessfully. If anyone would like toknow more about this all the same, this ishow to start on your way:First step:This would be to enter the minimumvalue, the target value and the maximumvalue for the three line lengths in thevariables block. They must be enteredbetween question marks. Luckily, thathas already been done some time ago.Second step:

Double click on the Linear Optimisationbutton (=black yellow red target) andopen the sheet with the target objectives(Fig. 32) and enter the following condi-tions correctly:Objective 1:Make sure that in the frequency rangebetween 1MHz and 1,700MHz the ampli-tude of S21 does not fall below -0.26dB.Objective 2:Make sure that in the frequency rangebetween 1MHz and 1,500MHz the ampli-tude of S11 always remains below -16.5dB.Objective 3:Make sure that at 1,700MHz precisely anamplitude of -0.26dB can be measured atS21.Tip: Please avoid the value Zero MHz

Fig 31: Sweep from 0 to 10GHz showing scatter resonance.


57

when entering frequencies, since this willlead to an error message!Then close the Objective menu and pressthe button to start the optimisation.Third step:Please follow the sequence listed in Fig.33 exactly and to do this call up the(blue) diagram with the simulation re-sults for S11 and S21 into the foregroundfor observation. Then close the windowagain with Close. Hopefully everythinghas gone all right as described, and ausable result has been obtained.

4.Measurement results

Both design methods produce a prototypebuilt into a milled aluminium housing,fitted with SMA flanged sockets, andfinally measured with a modern networkanalyser. The results are presented in theoriginal untouched condition, for we areactually trying to make sure that thesimulation can produce something thatagrees with reality.Fig. 34 shows the measured transmissionrange for the design using PUFF (ampli-tude range between 0 and -1 dB). Severalthings immediately strike one by com-parison with the simulation:The first Chebyschev ripple has a higheramplitude than the second

Fig 32: Setting up objectives for optimisation.


58

The cut off frequency, at 1.76GHz, is agood 3 % higher than forecastThe fundamental filter attenuation ismarkedly lower than was expectedNow lets look at these points in order:In Fig. 35 we can immediately see thereasons for the unequal ripples observedin a). The two camels humps of S11 arenot the same height! (If we look at theliterature again, we see where the rootcause lies> The formulae for the replace-ment components for the impedancesteps are no longer accurate enoughwhen there are extreme differences in thewidths, such as 0.25mm to 14.6mm.Observant readers will realise at oncethat the lengths of the two externalcapacitor lines must be increased some-what for correction purposes. This auto-matically lowers the cut off frequency, sothat both objections, a) and b), have

disappeared afterwards.The low fundamental filter attenuation isalso interesting. It has its origins in thefact that the loss factor decreases mark-edly as the frequency falls, which is whya comprehensive data sheet was re-quested from the German importer. Thiscontains an interesting diagram, showingthe insertion attenuation of a microstripline for various frequencies, in whichvarious ROGERS materials are used oneafter another (Fig. 36). If we analysethese curves very precisely, for theR04003 material, then we can come tothe following conclusions:The loss factor tanδ, that is 0.002 at10GHz, will fall by a factor of approxi-mately 5.3 if the operating frequency isreduced to 1.5 to 2GHz.If we now repeat the simulation with thenew value of tanδ = 0.002 / 5.3 =

Fig 33: Final set up to satrt optimisation.


59

0.00038, then we obtain, very accurately,the attenuation observed during themeasurements. (From now on, this mustnaturally be taken into account in futuredesigns).There is thus almost nothing more to saywith regard to Fig. 37 and Fig. 38, for theANSOFT design which supplies only aslightly lower ripple in the transmission

range and smaller discrepancies for thecamels humps of S11. Both the essentialshape and the fundamental attenuationand the cut off frequency show scarcelyany difference from the design usingPUFF, so that what has just been said isalso true for any further procedures usingthe ANSOFT printed circuit board.The old adage from developers experi-

Fig 34:Measurementresults on PUFFdesign.

Fig 35:Measurementresults on PUFFdesign showingunequal humps.


60

ence is also confirmed in this case, thatits not until the third board design thatyoure ready for full production.

5.Summary and Evaluation

PUFF and ANSOFT HARMONICA areequally suitable for filter design. Thosewho prefer to trust their instincts and

their experience as regards the optimisa-tion which is required, or who want totest out spontaneous ideas as fast aspossible, or who work with such pro-grams only occasionally, are usually go-ing to achieve their objectives fasterusing PUFF.However, as soon as housing influencesare brought in, or if many parameters arevaried, or if various zoom functions areused, if markers are superimposed or ifall WINDOWS options are to be used,

Fig 36: Data sheet on Rogers material.

Fig 37:Measurementresults onANSOFT design.


61

you are naturally much better off with thestudent version of Serenade. It is simplya very modern development tool, accu-rately tailored to the practical needs ofradio frequency developers. If only thoseall important options currently blockedwere also accessible, then there would beno question at all about what to use.

6.Literature

[1] Gunthard Kraus: Modern Design ofStripline Band-passes VHF Reports, is-sue 2/2001[2] English. Original Manual for PUFF[3] Manual for ANSOFT-Serenade, (in-cluded in program for download)[4] K. C. Gupta, Ramesh Garg, InderBahl, Prakash Bhartia: Microstrip Linesand Slotlines. Artech House PublishersBoston and London 1996. ISBN-Nr.:0-89006-766-X[5] David Pozar: Microwave Engineer-ing. John Wiley and Sons, New York1998. ISBN Nr.: 0-471-17096-8

Fig 38:Measurementresults on PUFFdesign.


62

Year Q1 Q2 Q3 Q42001 Y Y Y Y2000 Y Y Y Y1999 Y Y Y Y1998 Y Y Y Y1997 Y Y Y Y1996 Y Y Y Y1995 Y Y Y Y1994 Y Y Y Y1993 Y P Y P1992 Y Y Y Y1991 P P P P1990 P P Y P1989 P L P L1988 N P P L1987 N N P P1986 Y Y Y Y1985 Y Y Y Y1984 Y Y Y Y1983 Y Y Y Y1982 N N N N1981 Y Y P N1980 N N N P1979 L N L N1978 L L N N1977 L N N L1976 N N P N1975 L L L N1974 Y N N N1973 N L N N1972 L L N Y1971 P P P P1970 P P P P1969 P P P P

VHF CommunicationsBack Issues Available

KeyY = Available L = Low quantity P = Photocopy N = Not availableThe back issue sets contain issues marked Y, issues marked L or P must beordered separately.


63

AUSTRALIA - W.I.A. South Australia Division, GPO Box 1234,Adelaide, SA 5001, Australia. Tel/Fax: 8 8522 4859BELGIUM - UKW-BERICHTE, POB 80, D-91081 BAIERSDORF,Germany. Tel: 09133-77980. Fax: 09133-779833 .Postgiro Nbg. 30445-858.DENMARK - KM PUBLICATIONS , 63 Ringwood Road,LUTON, LU2 7BG, UK. Tel: +44 1582 581051.

Fax: +44 1582 581051. Email: [email protected] - Christiane Michel F5SM, Les Pillets, 89240 PARLY,FranceFax: (33) 03 86 44 08 82 Tel: (33) 03 86 44 06 91FINLAND - KM PUBLICATIONS , 63 Ringwood Road, LUTON,LU2 7BG, UK. Tel: +44 1582 581051.Fax: +44 1582 581051. Email: [email protected] - UKW-BERICHTE, POB 80, D-91081BAIERSDORF, Germany. Tel: 09133 7798-0. Fax: 09133 779833.GREECE - KM PUBLICATIONS , 63 Ringwood Road, LUTON,LU2 7BG, UK. Tel: +44 1582 581051.Fax: +44 1582 581051. Email: [email protected] - KM PUBLICATIONS , 63 Ringwood Road, LUTON,LU2 7BG, UK. Tel: +44 1582 581051.Fax: +44 1582 581051. Email: [email protected] - R.F. Elettronica di Rota Franco, Via Dante 5 - 20030Senago, MI, ItalyFax 0299 48 92 76 Tel. 02 99 48 75 15NEW ZEALAND - KM PUBLICATIONS , 63 Ringwood Road,LUTON, LU2 7BG, UK. Tel: +44 1582 581051.Fax: +44 1582 581051. Email: [email protected] - WAVELINE AB, Box 60224, S-216 09 MALMÖ,Sweden. Tel: +46 40 16 42 66. Fax: +46 40 15 05 07.GSM: 0705 16 42 66e-mail: [email protected] http://www.algonet.se/~wavelineSOUTH AFRICA - K

Documents

Pre Divider (:10) up to 5GHz - WorldRadioHistory.Com · 2019. 11. 11. · mat 1206 or 0805: 1x 12Ω 4 x 100Ω 1 x ... 1 x 18kΩ 1 x 27kΩ 1 x 2.7pF 1 x 12pF 2 x 15pF 1 x 47pF 1 x