2004 vectron working 5 - all-electronics€¦ · perature compensated voltage controlled crystal...

VEC

TRO

N Section11Application Notes

152

1. Clock Oscillators and Non-Compensated Sinewave Crystal Oscillators

2. Voltage Controlled Crystal Oscillators and VCOs - Voltage Controlled Oscillators

3. Oven Controlled Crystal Oscillators (OCXO’s)

4. Temperature Compensated Crystal Oscillators (TCXO’s)

5. Phase Noise

6. Jitter in Clock Sources

7. rms Jitter Calculations

8. Absolute Pull Range

9. SAW-Based Frequency Control Product Applications

10. Evacuated Miniature Crystal Oscillators (EMXO™)

153


Application Notes

Vectron International • 267 Lowell Road, Hudson, NH 03051 • Tel: 1-88-VECTRON-1 • Web: www.vectron.com

For the moderate stability crystal controlled oscillatorwhere neither temperature compensation nor oven operationare required, there are three primary parameters: output waveshape, frequency and accuracy/stability.

A. Output. The vast majority of systems require a crys-tal oscillator output which is TTL compatible, ECLcompatible, CMOS compatible or sinusoidal. Any ofthese outputs can be simply generated by circuitswhich follow the crystal oscillator stage. These areillustrated in Figure 1 with a dashed line in eachcase representing supply voltage input and the solidline showing the output.

B. Accuracy/Stability. The most basic element in anoscillator specification is the output frequency. Atany given time however, the oscillator's output fre-quency will differ from the desired specified frequen-cy resulting in a frequency error. This error is com-prised of three primary components:

1. Initial accuracy. This is generally defined as the dif-ference between the oscillator output frequencyand the specified frequency at 25°C at the time ofshipment by the oscillator manufacturer. Whenspecifying accuracy it is assumed that the user hasno provisions to adjust the oscillator's output fre-quency. When a frequency tuning control is includ-ed, accuracy no longer needs to be specified;instead the range and settability of the tuningadjustment become more consequential.

2. Temperature Stability. Figure 2 shows a typicalcharacteristic of crystal frequency vs. temperature.It is one of a family of curves illustrated in Figure 3.Figure 3 shows that one extreme, curve A, has arelatively flat slope (good temperature stability) nearroom temperature, but is very frequency sensitiveat high and low temperatures. The other extreme,curve B, shows greater sensitivity near room ambi-ent but also provides the overall best temperaturestability over wide temperature ranges. The angle atwhich the quartz crystal is cut determines the tem-

perature characteristic of a specific crystal. Theproper characteristic from this family of curves isselect, for each individual crystal oscillator require-ment In a well designed oscillator the stability vs.temperature is determined primarily by the tempera-ture characteristic of the crystal, and the oscillatormanufacturer must select the crystal characteristicswhich conform with the oscillator circuit to insure thatthe intrinsic stability of the crystal is not degraded.

A temperature stability of, for example, ± 10 ppmover 0°C to +50°C means a peak-to-peak frequencychange of 20 ppm over the specified temperaturerange, not referenced to the frequency at any spe-cific temperature. This is the generally accepted def-inition of temperature stability which, in MIL-0-55310, is called "frequency-temperature stability".

Clock & Sinewave Oscillators

If a reference temperature is desired with a maxi-mum allowable frequency change from that refer-ence, it should be specified, for example, as "±10ppm over O°C to +50°C referenced to the frequen-cy at +25°C."

While Vectron segregates initial accuracy and tem-perature stability, the two may be combined in spec-ifying an overall allowable error for oscillators withno frequency tuning adjustment. The appropriateterm is "frequency-temperature accuracy" and it isthe maximum allowable deviation from the specifiednominal frequency over a given temperature range.

154

Application Notes


3. Aging (Long-term stability). Aging refers to the con-tinuous change in crystal operating frequency withtime, all other parameters (temperature, supply volt-age, etc.) held constant. The better the processingof the crystal, the lower the aging rate (that is, thehigher the long-term stability). Figure 4 shows a typ-ical aging curve. It illustrates that when a crystaloscillator is initially turned on by the manufacturer,the crystal ages rapidly but its stability improveswith time. While the aging rate will typically contin-ue to improve with time, most crystals achieve closeto their lowest aging rate within several monthsafter turn-on. As long as crystal current is moderate,solder sealed or resistance welded AT-cut crystalsused in most clock oscillators provide typical agingof 5 ppm during the first year and 3 ppm per yearthereafter (5 ppm = .0005% = 5x10-6). If the errorintroduced by this degree of crystal aging exceedsthat allowed in the user's system, this can be over-come by

(1) specifying the inclusion of a frequency tuning adjust-ment in the oscillator to permit periodic recalibrationand/or, (2) using a higher quality crystal. Improvedaging to 1 x 10-6 per year can be achieved by employinga specially processed crystal housed in an evacuatedglass or coldweld sealed holder. Because aging gener-ally introduces a small part of the overall error in mod-erate stability clock oscillators, it is often ignored inspecifying these devices.

There are numerous other factors contributing to crystaloscillator instability, such as affect of supply voltagevariation, load variation and physical orientation; how-ever, they are not significant with respect to the majorerrors already detailed and are therefore excluded fromdiscussion in this clock oscillator section. In summary, for most clock oscillator requirements, aspecification will be sufficiently complete if it includesthe following electrical elements: frequency, outputlevel (wave shape), supply voltage, initial accuracyand temperature stability.

Following are examples of specifications for clock oscil-lators illustrating the considerations described in thisdiscussion.

Note that the CO-233T-3 which provides a high degreeof temperature stability (± 0.0003%) includes a tuningadjustment to permit accurate settability. While occa-sionally a need arises for high temperature stability con-current with a loose initial accuracy tolerance, general-ly stability and accuracy go hand in hand for a "bal-anced" specification. Similarly, it is generally illogical toinclude a tuning adjustment for precise settabililty whenthe stability requirements are very loose.

When the overall accuracy/stability specificationbecomes too stringent to be met with a simple clockoscillator, improvements can be accomplished in threeareas:

1. The initial accuracy error may be essentially elimi-nated and aging periodically compensated for byincorporating into the oscillator a frequency adjust-ment as previously discussed.

2. Aging may be improved by using a higher grade crys-tal, as previously discussed, and with a higher gradecircuit maintaining low constant crystal current.

3. The temperature stability may be improved by usingtemperature compensation techniques or housingthe oscillator in an oven.

Vectron CO-402A-OX XO-400-DFC-C CO233T-3ModelOutput 10.24 MHz 155.52 MHz 122.560 MHz

Output TTL PECL sinewave,0.5Level Compatible Complementary vrms minimum

(drives 10 into 50 ohmsloads)

Input 5Vdc ±5% 3.3 Vdc ±5% 15 Vdc ±5%VoltageAccuracy ±.005% Included in 10 Tuning adjust

year aging in settable totemp. stability ±.0001%budget

Temp 0°C to +70°C 0°C to +70°C 0°C to +50 °CStability ±.0025% ±.0020% ±0.0003%

Size 0.5” x0.8” x.20” 0.5” x0.8” x.22” 1.5”x1.5”x0.625”Mounting: 14 pin DIP 14 pin DIP Printed circuit

Compatible Compatible board pins


155

App Notes

A VCXO (voltage controlled crystal oscillator) is a

crystal oscillator which includes a varactor diode andassociated circuitry allowing the frequency to bechanged by application of a voltage across that diode.This can be accomplished in a simple clock or sinewavecrystal oscillator, a TCXO (resulting in a TC/VCXO-tem-perature compensated voltage controlled crystal oscilla-tor), or an oven controlled type (resulting in anOC/VCXO-oven controlled voltage crystal oscillator).

There are several characteristics peculiar to VCXOs. Ingenerating a VCXO specification these apply in additionto the characteristics which define fixed frequency crys-tal oscillators. Primary among the specifications whichare peculiar to VCXOs are the following:

Control Voltage - This is the varying voltage which isapplied to the VCXO input terminal causing a change infrequency. It is sometimes referred to as ModulationVoltage, especially if the input is an AC signal.

Deviation - This is the amount of frequency changewhich results from changes in control voltage. Forexample. a 5 volt control voltage might result in a devi-ation of 100 ppm, or a 0 to + 5 volt control voltage mightresult in total deviation of 150 ppm.

Transfer function (sometimes referred to as SlopePolarity) - This denotes the direction of frequencychange vs control voltage. A positive transfer functiondenotes an increase in frequency for an increasing pos-itive control voltage, as in Figure 1 A. Conversely, if thefrequency decreases with a more positive (or less neg-

ative) control voltage, as in Figure 1 B, the transfer func-tion is negative.

Linearity - The generally accepted definition of lineari-ty is that specified in MIL-PRF-55310. It is the ratiobetween frequency error and total deviation, expressedin percent, where frequency error is the maximum fre-quency excursion from the best straight line drawnthrough a plot of output frequency vs control voltage. Ifthe specification for an oscillator requires a linearity of±5% and the actual deviation is 20 kHz total as shownin Figure 2, the curve of output frequency vs controlvoltage input could vary ± 1 kHz (20 kHz ±5%) from theBest Straight Line "A". These limits are shown by lines"B" and "C". "D" represents the typical curve of a VCXOexhibiting a linearity within ±5%. In Figure 3, the maximum deviation from Best StraightLine "A" is - 14 ppm and the total deviation is 100 ppm,so the linearity is ±14ppm/100 ppm = ±14%.

Figure 2

2. VCXO’s Voltage Controlled Crystal OscillatorsVCO’s Voltage Controlled Oscillators. Non-Crystals

Application Notes

Vectron International • 267 Lowell Road, Hudson, NH 03051 • Tel: 1-88-VECTRON-1 • Web: www.vectron.com156

Application Notes


The VCXO which produces the characteristic indicatedin Figure 2 uses a hyper-abrupt junction varactor diode,biased to accommodate a bipolar (±) control voltage.The VCXO which produces the characteristic in Figure3 uses an abrupt junction varactor diode with an appliedunipolar control voltage (positive in this case).

Good VCXO design dictates that the voltage to fre-quency curve be smooth (no discontinuities) andmonotonic. All Vectron VCXOs exhibit these character-istics.

Modulation rate (sometimes referred to as DeviationRate or Frequency Response) - This is the rate at whichthe control voltage can change resulting in a corre-sponding frequency change. It is measured by applyinga sinewave signal with a peak value equal to the spec-ified control voltage, demodulating the VCXO's outputsignal, and comparing the output level of the demodu-lated signal at different modulation rates. The modula-tion rate is defined by Vectron as the maximum modu-lation frequency which produces a demodulated signalwithin 3 dB of that which is present with a 100 Hz mod-ulating signal. While non-crystal controlled VCOs canbe modulated at very high rates (greater than 1 MHz foroutput frequencies greater than 10 MHz), the modula-tion rate of VCXOs is restricted by the physical charac-teristics of the crystal. While the VCXO's modulationinput network can be broadened to produce a 3 dBresponse above 100 kHz, the demodulated signal mayexhibit amplitude variations of 5-15 dB at modulationfrequencies greater than 20 kHz due to the crystal.

Slope/Slope Linearity/Incremental Sensitivity - Thiscan be a confusing area as these terms are often mis-applied. Slope should be really called average slope if itis intended to define the total deviation divided by thetotal control voltage swing. For the VCXO depicted inFigure 2, the average slope is -20 kHz * 10 volts = -2kHz/volt. Incremental sensitivity, often misnomerredSlope Linearity means the incremental change in thefrequency vs control voltage. Thus, while the averageslope in this example is -2 kHz per volt, the slope forany segment of the curve may be considerably differentfrom -2 kHz/volt. In fact, for VCXOs with Best StraightLine linearity of ±1% to ±5%, the Incremental Sensitivityis approximately (very approximately) 10 times as greatas the Best Straight Line linearity. Thus a VCXO with±5% Best Straight Line linearity can exhibit a slopechange of ±50% on a per volt basis. Therefore, a spec-ification which reads "Slope: 2 kHz/volt ± 10%" requiresclarification as it could mean either Average Slope orIncremental Sensitivity. If it were intended to defineaverage slope, it simply specifies a total deviation of 18kHz to 22 kHz and would more properly have stated,"Total Deviation: 20 kHz - 10%." However, if it wereintended that the frequency change for each incremen-tal volt must fall between 1.8 kHz and 2.2 kHz, a highlylinear VCXO is being specified as a ±10% IncrementalSensitivity relates approximately to a ±1% Best StraightLine linearity. That element of the specification shouldread, "Incremental Sensitivity: 2 kHz ± 10% per volt."

Other Design Considerations

Stability - A quartz crystal is a high Q device which is thecrystal oscillator's stability determining element. It inher-ently resists being "pulled" (deviated) from its designedfrequency. In order to produce a VCXO with significantdeviation, the oscillator circuit must be "de-Q'd". Thisresults in degrading the inherent stability of the crystalin terms of its frequency vs. temperature characteristic,its aging characteristic, and its short-term stability (andassociated phase noise) characteristic. Therefore, it isin the user's best interest not to specify a wider devia-tion than that absolutely required.


157

App Notes

Application Notes


Phase Locking - When a VCXO is used in phase lockloop application, the deviation should always be at leastas great as the combined instability of the VCXO itselfand the reference or signal onto which it is being locked.Vectron produces a line of VCXOs especially intendedfor use in phase lock loop applications (described onthe pages which follow). However, if the open loop sta-bility requirements of a system are more stringent thanavailable in this product line, a TC/VCXO may berequired. For the highest stability open loop require-ments, the appropriate oscillators may be thosedescribed in the TCXO or OCXO sections of this cata-log, incorporating a narrow deviation VCXO option,rather than those described in the VCXO section.

Basic Oscillator Frequencies - Fundamental mode crys-tals (generally 10-25 MHz) permit the widest deviation,while 3rd overtone crystals (generally 20-70 MHz) allowdeviation approximately 1/9th of that which applies tofundamentals. Therefore, all wide deviation VCXOs

(greater than ±100 to ±200 ppm deviation) use funda-mental crystals; narrower deviation VCXOs can usefundamental mode or 3rd overtone crystals, the selec-tion of which often depends upon such specifications aslinearity and stability. It is rare that higher overtone, andtherefore higher frequency crystals find application inVCXOs. Thus, VCXOs with output frequencies higher orlower than available from the appropriate crystalfrequencies include frequency multipliers or dividers.

General Note - While it is true of any type of crystaloscillator, it is especially important with VCXOs that theuser not over-specify the product. The particular prob-lem with VCXOs is that increased deviation results indegraded stability which can result in the need for stillwider deviation, further degrading stability, resulting in aspiraling increase in the required deviation.


158

3. OCXO’s Oven Controlled Crystal Oscillators

Application Notes


If stability requirements are too stringent to be met by a

basic crystal oscillator or TCXO, the crystal and criticalcircuits may be temperature controlled by an oven. Theblock diagram for a Vectron oven controlled crystaloscillator is similar to that for a Vectron TCXO exceptthat the varactor diode and associated thermistor com-pensation network are deleted and the oscillator isinstead temperature controlled by a proportionallycontrolled oven.

Proportional Oven Controlled

A proportional control is an electronic servo systemwhich continuously supplies power to the oven; it variesthe amount of oven power, continuously compensatingfor the ambient temperature changes. In many Vectronoven controlled oscillators, a thermistor is heat sunk tothe oven's metal shell to sense temperature. The ther-mistor is one leg of a resistance bridge, as shown in thefollowing diagram.

The bridge operates such that if the temperature at theoven decreases due to an ambient temperaturechange, the change in thermistor resistance causes thebridge to unbalance, developing an increase in bridgeoutput voltage. This voltage is amplified in a high-gaindifferential amplifier. The output of the differential ampli-fier is further amplified in a power amplifier which drivesdirectly into the oven winding. Thus, the small voltageincrease resulting from bridge unbalance generates alarge voltage increase across the oven winding. Thisincrease in power to the oven generates more heat,compensating for the temperature decrease which wasinitially sensed by the thermistor. Similarly, an increasein temperature at the oven causes a reduction in bridgeoutput voltage, which results in reduced power into theoven and a compensating temperature decrease.

An alternative to this design, used in some Vectronoven shell as the heat transfer mechanism, in lieu ofhaving a heater winding. The concept is the same, theonly difference being the vehicle by which the heat isapplied to the oven.

Employing a proportionally controlled oven can improve

oscillator temperature stability relative to the crystal'sinherent stability by more than 5000 times (from ±1x10-5 to ±1x10-9 over 0-50°C, for example). However, theoven control system is not perfect because (a) the openloop gain is not infinite, (b) there are internal tempera-ture gradients within the oven and (c) circuitry outsidethe oven which is subjected to ambient temperaturechanges can "pull" the frequency. Therefore, a changein ambient temperature will result in small changes inoven temperature.

Setting Oven Temperature

As shown above, the actual temperature to which theoven is set is critical in minimizing the effect of ambienttemperature change.

Referring to Figure 2, if the oven temperature were setto the point designated as (1), and a change in ambienttemperature caused a change in oven temperature fromA to B, a frequency change of magnitude X would result.However, if the oven temperature were set to the upperturnover point (2), an equal temperature change (C toD) would result in a significantly reduced change in fre-quency (magnitude Y). Therefore, each Vectron oven isindividually set to the turnover temperature of the crys-tal which it houses. This is accomplished by adjustingthe potentiometer shown as one leg of the bridge inFigure 1.

Warmup with AT cut crystals

When an oscillator is initially turned on at room temper-ature the frequency is extremely high relative to the out-put frequency after the oven stabilizes, typically by

30xl0-6. This is simply due to the fact that the frequency

of an AT cut crystal is considerably higher at room tem-perature than at its upper turnover temperature. As the

159

App Notes


Application Notes


perature than at its upper turnover temperature. As theoven warms up, the crystal frequency rapidly decreas-es. In standard Vectron oscillators, the oven balances in

10-15

minutes but the crystal displays a rubberbandeffect and overshoots its final frequency per Figure 3,prior to stabilizing. Typically, relatively high degree ofstability is achieved within 30 minutes after turn-on; thistime can be reduced to less than 5 minutes in specialfast warm-up designs.

Turnover Temperature

The oven operating temperature (crystal turnovertemperature) must be several degrees higher than thehighest ambient temperature in which the oscillator is tooperate in order that the oven may maintain good con-trol (considering the internal heat rise generated by theoscillator itself).

However, there are disadvantages associated with highoven temperature operation. First, the crystal's frequen-cy vs. temperature characteristic is sharper with higherturnover crystals resulting in more sensitivity to minutechanges in oven temperature as shown in Figure 4.

Second, and more important, crystal aging (discussedbelow) degrades with an increasing temperature.Therefore, in designing an oven controlled crystal oscil-lator, one is faced with a compromise in determining thedesired oven operating temperature; it should be low aspracticable, but it must be high enough to provide goodcontrol at the maximum ambient operating temperature.

Stability

A. Aging - Aging refers to the continuous change incrystal oscillator frequency with time, all other parame-ters held constant. Prior to delivery, each Vectron ovencontrolled oscillator is pre-aged until it achieve itsspecified aging rate. Aging rate is often used synony-mously with the word stability; thus, an oscillator with an

aging rate of one part in 10-8

per day (1x10-8/day) is

sometimes referred to as one part in 108 oscillator. Thisis incorrect terminology, as aging rate (long term stabil-ity) must be referred to time, and represents only onefacet of oscillator stability.

B. Temperature Stability - As previously noted,because no oven control system is perfect, a change inambient temperature causes a small change outputfrequency. The frequency shift is an offset from theoscillator's aging curve. This deviation from the normalaging characteristic is not related to time, but is a fixedoffset. Thus, the frequency offset vs. temperature (tem-perature stability), for a given temperature change is,

for example, 5x10-9, not 5x10

-9/day. This characteristic is

shown below.

Ambient temperature changes do not produce hystere-sis effects; that is, if there is a change in ambient tem-perature followed by a return to the original tempera-ture, the final frequency will be essentially that whichwould have resulted had there been no ambient tem-perature change.

When the required temperature stability is beyond thatwhich can be achieved with a standard proportionallycontrolled oven, a double oven system can beemployed in which the standard oven is housed withina second oven. The outer oven then buffers the ambi-ent temperature changes to the inner oven, which con-tain the oscillator circuit.

160

C. Restabilization And Retrace - When a crystal oscil-lator is turned off for a period of time and then turned onagain (as occurs when the unit is shipped), the crystalrequires a restabilization period. The characteristic issimilar to the initial factory aging characteristic, but highstability is achieved significantly more quickly becausethe crystal has been factory pre-aged.

In most applications, oven-controlled crystal oscillatorsare continuously energized. This being the case, agingis the critical characteristic with turn-off/turn-on charac-teristic being of little or no significance. However, cer-tain applications require that oven controlled crystaloscillators be frequently de-energized and re-energized(a practice which should be avoided whenever possi-ble). When applications require frequent turn-off, anadditional series of characteristics should be consid-ered. In Figure 6, assume that an oscillator is energized untiltime T2 when it is turned off for a period of time and thenturned on again at time T3. Three characteristics maythen be of significance:

Figure 6

1. How close does the oscillator return to the outputfrequency at turn-off, a specified time after turnon.This is called the retrace characteristic. Retraceerror at T4 = fl - f3.

2. How much will the frequency change over moder-ate periods of time (hours) after the oven has stabi-lized. This is called the restabilization, or warmup,characteristic. Restabilization rate from T4 to T5 = (f3 - f2) / (T5 - T4)

3. How long does it take the oscillator to achieve itsspecified aging rate following a specified off period(This is called "reaging").

Many factors affect retrace, restabilization and reagingcharacteristics. Proper circuit design and componentselection minimize their effects, leaving (1) the crystaland, (2) the length of off-period prior to oscillator turn-onas the prime factors. There is significant variation inthese characteristics from crystal to crystal and theyshould only be specified when absolutely required andthen only to the degree needed, as "tight" specificationsin this area can have a major impact upon oscillator costdue to low yield. These characteristics are of little con-sequence in oscillators which are energized continu-ously.

Double Rotated (SC and IT Cut) Crystals

While most high stability crystal oscillators use AT CutCrystals, SC and IT Cut Crystals are often used in thehighest stability models.

An SC Cut Crystal is one of a family of double rotatedcrystals (quartz crystals cut on an angle relative to twoof the three crystallographic axes). Others in the familyinclude the IT Cut and FC Cut. The SC Cut representsthe optimum double rotated design as its particularangle provides maximum stress compensation, but sim-ilar performance is achieved with the IT Cut.

Following is a comparison between double rotated(referred to simply as SC for convenience) and AT Cutcrystals.

Advantage of SC Crystals:1. Improved Aging. For a given frequency and over-

tone (e.g. 10 MHz, third overtone), the SC crystalprovides 2:1 to 3:1 aging improvement relative toAT crystals.

2. Warm-up. In oven controlled oscillators with a givenoven design and turn-on power, the SC crystalachieves its "final frequency" in considerably lesstime than does the AT crystal.

3. Phase Noise. For a given oscillator design, crystal

frequency and overtone, the SC crystal provideshigher Q and associated improved phase noisecharacteristics. This improvement applies primarilyclose to the carrier as the noise floor is determinedby circuit design rather than the crystal.

4. High Operating Ambient Temperature. Figure 7shows the relative frequency-temperature charac-teristics of AT, IT and SC crystals. The upper tem-

Application Notes



161

App Notes

perature turnover point of the AT crystal ("A" in Figure7) and lower temperature turnover point of the SCcrystal ("B" in Figure 7) are optimally in the 70°C to 90°C temperature range. Based upon(a) the desired 1O°C difference between the high-est operating ambient temperature and the crystal

turnover temperature, and (b) the manufacturing tolerance of crystal turnovertemperatures, these crystals are best suited formaximum operating ambient temperatures of 50°Cto 75°C. However, the upper temperature turnoverpoint of the IT crystal ("C" in Figure 7) is well suitedto higher temperature operation and thus the ITcrystal is a logical choice for high stability oven con-trolled oscillators having a maximum operating tem-perature in the 85°C to 95°C range. Note that whileSC and IT crystal curves are relatively flat at ele-vated temperatures, their frequency falls off rapidlyat low temperatures. Thus, while they serve well inhigh stability HIF oven controlled oscillators, theyare generally not well suited for other types of sta-ble crystal oscillators.

5. Orientation Sensitivity (tip-over). When the physicalorientation of an oscillator is changed, there is asmall frequency change (typically not more thanseveral parts in 10-9 for any 90 degree rotation), dueto the change in stress on the crystal blank result-ing from the gravitational affect upon the crystalsupports. Tip-over is expressed in 10-9/g where one

g represents one half of a 180° orientation change.The SC crystal is less frequency sensitive to orien-tation change than is the AT. However, the tip-overdifference between AT and SC crystals is not con-sequential for most applications and this character-istic is usually not a specification consideration.

6. Spurious Under Vibration. When a crystal oscillatoris subjected to vibration, spurious frequencies aregenerated, offset from the frequency oscillation bythe frequency of vibration. The amplitude of thesespurious outputs is related to the amplitude of vibra-tion, the mechanical design of the crystal support,and the mechanical design of the oscillator. The SCcrystal produces lower amplitude spurious outputunder vibration than does the AT; however, thischaracteristic is determined more by themechanical designs of the crystal and oscillatorthan by crystal cut.

Disadvantages of SC Crystals:

1. Cost. Because of difficulties associated with tightly-controlled angle rotations around two axes in themanufacture of SC crystals vs one axis for the AT,the SC crystal is significantly higher in cost than thatof an AT of the same frequency and overtone.

2. Pullability. The motional capacitance of an SCcrystal is several times less than that of an AT of thesame frequency and overtone, thus reducing theability to "pull" the crystal frequency. This restrictsthe SC crystal from being used in conventionalTCXOs and VCXOs, or even in oven controlledoscillators requiring the ability to deviate thefrequency of oscillation by any significant degree.

In summary, the suitability of double rotated crystals foruse in crystal oscillators is essentially restricted to thoseoven controlled applications where the improved aging,warm-up, and close-in phase noise characteristicsjustify a significant cost increase.

Application Notes



162

4. TCXO’s Temperature Compensated Crystal Oscillators

Application Notes


A. Temperature Stability. The temperature stability of

a basic crystal oscillator can be improved by incorpo-rating in the oscillator circuit components with tempera-ture characteristics approximately equal to and opposite

from that of the crystal as shown in Figure 1. The actual technique employed in all except the most

simple TCXOs is based upon use of a varactor diode inseries with the crystal as follows: A change in voltage "V" causes a change in the capac-itance of the varactor diode resulting in a change in fre-quency of oscillation. The thermistor network is tailoredto the crystal to cause voltage "V" to vary with tempera-ture in a manner which will compensate for the crystal'sfrequency versus temperature characteristic. As eachindividual TCXO requires that its compensation networkbe matched to its individual crystal, the cost of a TCXOis closely related to the difficulty of the frequency versustemperature specification. The stability requirements ofmost TCXOs dictate compensation by means of a mul-tiple thermistor network with several interdependentvariable components thus making the solution of simul-taneous equations by computer the only practicalapproach to temperature compensation.

When an oscillator manufacturer specifies a stability of

±1x10-6

over -20°C to of + 70°C, this means a total peak

error of 2x10-6

over the temperature range, not refer-enced to the frequency at any specific temperature. If areference, such as room temperature, is desired with a

maximum allowable error of ±1x10-6

from that reference,

the specification should clearly state ±1X10-6

over -20°Cto +70°C referenced to the frequency at +25°C."Further, it should be noted that the frequency versustemperature characteristic of a TCXO is not linear; thus

a 2xl0-7

total error over O°C to +50°C will not produce a

gradient of 2x10-7

÷ 50 = 4x 10-9

per ºC. Perturbations inthe crystal characteristics (activity dips) make it virtuallyimpossible to guarantee exceptional stability on a perdegree basis in TCXOs.

B. Aging. In clock oscillators with moderate tempera-ture stability, aging is usually of little consequence.However, in highly temperature stable TCXOs, crystalaging becomes a significant factor in the oscillator'soverall frequency error. Therefore, TCXOs employ spe-cially processed crystals in evacuated glass or coldweldholders.*

Many TCXO specifications include both moderate and

long term aging requirements such as ±lxlO-6

per year.The latter actually has more rapaning for a TCXObecause the temperature sensitivity of the device

makes it almost impossible to measure ±1 x10-8

per dayaging except under constant environmental conditions;the small day to day changes in even laboratory ambi-ent temperatures will cause greater frequency shiftsthan those resulting from crystal aging over short timeperiods.

C. Other Factors: Figure 3 illustrates the block diagramfor a typical Vectron TCXO.

163

App Notes

Application Notes


It shows those elements generally not required in sim-ple clock oscillators, but included in the proper design ofa highly stable TCXO: (1) a frequency compensationnetwork to minimize temperature sensitivity, (2) a preci-sion crystal coupled with AGC for minimum aging, (3) amultiturn tuning adjustment permitting precise setting offrequency, (4) buffering following the oscillator to mini-mize the effects of external circuit changes and (5) aninternal regulator to minimize the effects of voltage vari-ation. Each of these elements is a factor in properlyspecified TCXO, an example of which follows.

Vectron Type: TC-210-DAB-307AFrequency: 12.8 MHzOutput Level: HCMOSSupply Voltage: 3.3 Vdc ±5%Current Drain: <15 mATemperature ±3 x 10

-7 over 0°C to +50°C

Stability:Supply Variation: ±1 x 10

-8 percent change in supply

Aging: 3.5 x 10-6

per 10 years Electrical ±5 ppm minimum via externalFrequency voltage, 0 to Vdd.Adjustment:

4. TCXO’s Temperature Compensated Crystal Oscillators

164

5. Phase Noise

Application Notes


Signal sources such as crystal oscillators produce a

small fraction of undesirable energy (phase noise) nearthe output frequency. As performance of such systemsas communications and radar advance, the spectralpurity of the crystal oscillators which they employ isincreasingly critical.

Phase noise is measured in the frequency domain, andis expressed as a ratio of signal power to noise powermeasured in a 1 Hz bandwidth at a given offset from thedesired signal. A plot of responses at various offsetsfrom the desired signal is usually comprised of threedistinct slopes corresponding to three primary noisegenerating mechanisms in the oscillator, as shown inFigure 1. Noise relatively close to the carrier (Region A)is called Flicker FM noise; its magnitude is determinedprimarily by the quality of the crystal. Vectron's bestclose-in noise results have been obtained using 5thovertone AT cut crystals or 3rd overtone SC cut crystalsin the the 4-6 MHz range. While not quite as good onaverage, excellent close-in noise performance may alsobe achieved using 3rd overtone crystals in the 10 MHzarea, especially double rotated types (see page 41 for adiscussion of double rotated SC and IT cut crystals).Higher frequency crystals result in higher close-in noise

because of their lower Q and wider bandwidths. Noise in Region B of Figure 1, called "1/F" noise, iscaused by semiconductor activity. Design techniquesemployed in Vectron low noise "L2" crystal oscillatorslimit this to a very low, often insignificant, value.

Region C of Figure 1 is called white noise or broadbandnoise. Special low noise circuits in Vectron "L2" crystaloscillators offer dramatic improvements (15-20 dB) rel-ative to standard designs.

When frequency multiplication is employed to achievethe required output frequency from a lower frequency

crystal, the phase noise of the output signal increasesby 20 log (multiplication factor). This results in noisedegradation of approximately 6 dB across the board forfrequency doubling, 10 dB for frequency tripling and 20

dB for decade multiplication. As shown in Figure 2, the noise floor isalmost independent of the crystal frequency for oscilla-tors which do not employ frequency multiplication. Thusfor low noise floor applications, the highest frequencycrystal which satisfies long-term stability requirementsshould generally be used. However, when a higherfrequency application specifically requires minimumclose-in phase noise, lower frequency crystals mayoften be multiplied to advantage. This is so becauseclose-in phase noise is disproportionately better thanthe noise performance obtained using higher frequencycrystals.

Note that the introduction of a varactor diode and mod-erate Q crystal, used typically in TCXO and VCXOproducts, result in poorer close-in noise performancewhen compared with fixed frequency non-compensatedcrystal oscillators.

Phase Noise Testing

The phase noise test characterizes the output spectralpurity of an oscillator by determining the ratio of desiredenergy being delivered by the oscillator at the specifiedoutput frequency to the amount of undesired energybeing delivered at neighboring frequencies. This ratio isusually expressed as a series of power measurementsperformed at various offset frequencies from the carrier,the power measurements are normalized to a 1 Hzbandwidth basis and expressed with respect to thecarrier power level.

165

App Notes

This is the standard measure of phase fluctuationsdescribed in NIST Technical Note 1337 and is called (f).Figure 3 shows a block diagram of the method sug-gested by NIST, and used by Vectron, to measure (f).Signals from two oscillators at the same nominal fre-

quency are applied to the mixer inputs. Unless the oscil-lators have exceptional stability, one oscillator musthave electronic tuning for phase locking. A very narrowband phase-locked loop (PLL) is used to maintain a 90degree phase difference between these two sources.The mixer operation is such that when the input signalsare 90 degrees out of phase (in quadrature), the outputof the mixer is a small fluctuating voltage proportional tothe phase difference between the two oscillators. Byexamining the spectrum of this error signal on the spec-trum analyzer, the phase noise performance of this pairof oscillators may be measured. If the noise of oneoscillator dominates, its phase noise is measureddirectly. A useful and practical approximation when thetwo test oscillators are electrically similar is that eachoscillator contributes one-half the measured noisepower. When three or more oscillators are available fortest, the phase noise of each oscillator may be accu-rately calculated by solving simultaneous equationsexpressing data measured from the permutations ofoscillator pairs.

Figure 4 shows a practical system of measurement Of(f). The steps taken to measure phase noise with thissystem are:

1. Calibration of the spectrum analyzer screen. 2. Phase lock the oscillators and establish quadrature. 3. Record spectrum analyzer readings and normalize

readings to dBc/Hz SSB per oscillator. These steps

are detailed below.

Step I - Calibration

To avoid saturating the mixer, the signal level of oneoscillator is permanently attenuated by 10 dB pad(Attenuator "A"). During the calibration, the level of thisoscillator is additionally attenuated by 80 dB (Attenuator"B") to improve the dynamic range of the spectrum ana-lyzer. The oscillators are mechanically offset in frequen-cy and the amplitude of the resulting low frequency beatsignal represents a level of -80 dB; it is the reference forall subsequent measurements. When using a sweptspectrum analyzer, this level is adjusted to the top lineof the spectrum analyzer's screen. When using a digital(FFT) spectrum analyzer, the instrument is calibrated toread RMS VOLTS/ relative to this level . When full levelis restored to the mixer, and the oscillators are phaselocked, the phase noise will be measured with respectto this -80 dB level.

Application Notes


5. Phase Noise

166

Application Notes


Step II - Phase Lock

The oscillators are phase locked to quadrature bymechanically adjusting them to the same frequency.The desired 90 degree phase difference between thetwo oscillators is indicated when the mixer output is 0Vdc. An oscilloscope or zero-center voltmeter connect-ed temporarily at the input to the spectrum analyzer is aconvenient way to monitor progress toward quadrature.The operating bandwidth of the PLL must be muchlower than the lowest offset frequency of interestbecause the PLL partially suppresses phase noise in itsbandwidth. A widely used empirical method of estab-lishing an appropriate loop bandwidth is to progressive-ly attenuate the voltage control feedback via Attenuator"C". By comparing successive noise measurements atthe lowest offset frequency of interest while advancingAttenuator "C", an operating point may be found wherethe measured phase noise is unaffected by changes inthe attenuator setting. At this point the loop bandwidth isnot a factor of the measured phase noise.

Step III - Readings

Readings are taken with respect to the -8OdB calibra-tion level previously established in Step 1. Smoothingor averaging is used if the spectrum analyzer is soequipped to avoid measurement variations. Sweptspectrum analyzer readings generally require each ofthe following corrections while digital analyzer read-ings displayed in RMS/ do not require the first two cor-rections. The analyzer's manual should be consultedregarding corrections for analyzer noise response.

Correction 1. Normalize to 1 Hz bandwidth. "BW" is the 10measurement bandwidth. Calculation logassumes the noise is flat within the (1/BW)measurement bandwidth.

2. Video response of swept analyzer to noise signals. +3 dB3. Double sideband to single sideband display. -6 dB4. Contribution of two oscillators assuming they are of equal noise quality. -3 dB

5. Phase Noise

167

App Notes


Continuous advances in high-speed communication

and measurement systems require higher levels of per-formance from system clocks and references.Performance acceptable in the past may not be suffi-cient to support high-speed synchronous equipment.Perhaps the most important and least understood mea-sure of clock performance is jitter.

The purpose of this discussion is fourfold.1. To define jitter intuitively, and discuss it's proper-

ties.2. To explain how jitter degrades system perfor-

mance.3. To describe various practical methods of measur-

ing jitter, including the relevance and ease of eachmethod.

4. Offer guidelines for specifying high-speed clocksand related devices.

Jitter Defined:Jitter: "Short-term variations of the significant instants

of a digital signal from their ideal positions in time"(ITU).The expected edges in a digital datastream neveroccur exactly where desired. Defining and measuringthe timing accuracy of those edges (jitter) is critical tothe performance of synchronous communicationsystems.

Definition of terms:a) Jitter expressed in Unit Intervals: A single

unit interval is one cycle of the clock fre-quency. This is the normalized clockperiod. Jitter expressed in Unit Intervalsdescribes the magnitude of the jitter as adecimal fraction of one unit interval.

b) Jitter expressed in degrees (deg.): Jitterexpressed in degrees describes the mag-nitude of the jitter in units of deg. where

one cycle equals 360 deg.c) Jitter expressed in absolute time: Jitter expressed in

units of time describes the magnitude of the jitter inappropriate orders of magnitude, usually pico-seconds.

d) Jitter expressed as a power measurement isdescribed in units of radians or unit intervalssquared. Often expressed in dB relative to onecycle-squared (From Bellamy) [1]

e) Pattern Jitter: Pattern dependent jitter. Sometimesreferred to as flanging. Not random in nature.Generally a result of subharmonics. When viewed inthe time domain, it is seen as multiple modes of jit-ter. Pattern jitter is deterministic, it is a phenomenonthat may be attributed to a unique source. All otherjitter referred to in this discussion is stochastic innature, and may only be described as a randomvariable with respect to time.

For example:Assume a clock rate of 155.52 MHz. One unit intervalwould be equal to the period of the signal, 1/155.52MHz = 6.43 nsec. = 360 deg.Assume 100 ps Pk-Pk of jitter. 100 Ps of jitter = .01555 unit intervals (UI) of jitter =5.598 deg. of jitter. (All Pk-Pk) All three measurementsdescribe the same amount of jitter.

For jitter power, rms (one sigma, s) measure-ments are used. For the above case, weapproximate Pk-Pk as 7s, or 7 times the RMSvalue, placing the rms. jitter power at.0000049 UI 2 . {(.01555/7) 2 . Expressed indB, relative to one unit interval, jitter power inthis case would be 10log (.0000049)= -53.1dBui. As will be seen later, jitter can bederived from power spectral density (phasenoise) measurements. Table 1 relates variousmeasures of jitter in a 155.52 MHz systemclock

Application Notes


Figure 1

Digital Waveform with Jittered Edges

168

Application Notes


Jitter Bandwidth and SpectralContent

The displacement of edges inFig.1 is a result of noise. Noisehas spectral content as well aspower. Consequently, the edgejitter in Fig.1 also has spectralcontent. The edges in Figure 1vary randomly with time, howeverthe noise that causes the jitter isnot necessarily uniform over allfrequencies. Jitter due to 10 kHz noise could begreater or less than jitter due to 100 kHz noise.Spectral content of clock jitter differs greatly dependingon the technique used to generate the clock.Measured jitter also varies with measurement tech-nique and jitter bandwidth. Improperly specified ormeasured jitter might result in unnecessary costs, orpoor system performance. See references [2,3] foradditional information on defining and specifying jitterin telecom systems. Jitter characteristics of variousclock sources is discussed later in this article.

How Does Jitter Effect System PerformanceThe effects of jitter on communication systems are wellbeyond the scope of this discussion. Refer to refer-ences [1,4] for a more thorough treatment. A simplediscussion may help to understand the deleteriouseffects of jitter in digital systems. Every bit of datatransmitted over synchronous communi-cation systems is sampled for its valueat the receiver. The sampled data canonly have the value of logical one orzero. The optimum point for samplingdata is at the center of each transmitclock cycle. In order to perform thisfunction, the receiver aligns its ownclock with the clock used to transmit thedata. Figs.2, a, b, and c represent ideal,typical, and corrupted datastreamsrespectively. Commonly referred to asan "eye diagram" each graph is acumulative graphical portrait of theedge placement due to noise or jitter.Ideally, sampling occurs at the center of the "eye". Asedge jitter increases, the apparent eye begins to close.As a result, the likelihood of an error, i.e.…. mistakinga logical one for a zero is more likely. Jitter due tooscillator noise is only one source of jitter in a telecomsystem. System designers must consider manysources of noise in telecom systems. The jitter intro-

duced by clock sourcesis one component of

noise, and becomes only one part of an "error budget"that must be weighed against performance require-ments and cost.

Measurement Techniques

Time Domain MeasurementsEdge to Edge Jitter Using a Delay Line

A true measure of clock jitter is the accurate positionof clock edges over time. The most direct method ofexamining the placement of edges would be to look atthe edges using an oscilloscope. Unfortunately, usingstandard oscilloscope techniques it is impossible toidentify individual clock edges in absolute time. Any jit-ter measured with a standard oscilloscope is due to

trigger instability. As a result, direct waveform mea-surements using an oscilloscope (even a very goodoscilloscope) are not valid measurements of jitter. Anadditional technique is used to locate the referenceedge, discriminate with time, and examine the jitter onfollowing edges. Figure 3 illustrates this method with a


169

App Notes

typical configuration.The output of the unit under test is fed into split-ter /delay line. The non-delayed output o thesplitter is fed to the external trigger input of theoscilloscope (a CSA-803 in this case). Thedelayed output of the DL-11 is connected to theinput of the oscilloscope. By examining the clock-stream at a time after the trigger equal to thedelay used (in this case, 47 nsec), the trigger-edge is located. After the triggered edge hasbeen identified, the next edge is examined. A his-togram plot is then produced of the measured jit-ter of the second edge.

A CSA-803 is used for its statistical and his-togram capabilities. This is a useful techniquelimited by the length of the delay line and thespeed/sensitivity of the oscilloscope. For all fre-quencies greater than 1/(2ptd), the measurementis limited by the noise of the oscilloscope. Below1/(2ptd), the sensitivity drops approximately 20 RMSSensitivity in picoseconds dB/decade. For the 47-nsecdelay shown in fig 3, the corner frequency occurs at3.3 MHz. All jitter due to frequencies above 3.3 MHz

can be resolved to approximately 5 ps using the CSA-803. Jitter at 330 kHz can not be resolved below 50ps. In a similar manner, jitter at 33 kHz can not bemeasured below 500 ps. Figure 4 is a plot of RMS jit-ter sensitivity using a 47-nsec delay line. It is critical tounderstand the advantages and limitations of this mea-surement method. For the numerical example given,low frequency jitter below 300 kHz would not be seen.Conversely, jitter due to sidebands 3.0 MHz offset ormore could easily be identified. This test method is

appropriate when measuring oscillators that employ

direct frequency multiplication or where low frequencyjitter is not considered. (See previous description ofPattern Jitter).

Jitter Measurements Using Phase Locked Loops

It was noted that the length of the delay line limits res-olution when measuring edge jitter. In order to mea-sure jitter below 100 Hz offset, one needs merely toorder up about three hundred miles of very low loss

delay line. In lieu of such a device, phase locked loopsare used for a variety of noise measurements.Figure 4 shows the basic elements of a phase lockedloop (PLL) used to measure the noise of a clocksource. Gardner [5], Best [6], and Woolover [7] arethree excellent references for understanding PLLs.Some key loop requirements follow:

· The PLL loop bandwidth is a critical parameter forsuccessful measurements. The system will onlymeasure jitter frequencies outside (higher than)the loop bandwidth. It is recommended that theloop bandwidth be set to a maximum of 1/10 thelowest jitter frequency of interest.

· Loop damping must be set to at least 5 in order toreduce jitter peaking in the PLL. Jitter peaking willincrease measured jitter.

· The measurement filter corresponds to jitter band-widths recommended in Bellcore and ITU specifica-tions. Refer to [2,3] for a list of jitter bandwidths andspecifications. Band limiting needs to be defined asa pre-condition for any valid measurement.

Application Notes



170

· The output of the phase detector (PD) is a varyingDC signal that is proportional to the varying phasedue to jitter. It is necessary to know the gain con-stant (Kd) of the PD in volts/radian in order toquantify the detected jitter. For example, a phasedetector with a Kd equalto 1 millivolt per degreewill have a peak to peakoutput of 10 mV for anoscillator with 10 deg,. pk-pk jitter. It may be neces-sary to inject a knownamount of jitter in order tocalibrate the system foraccurate measurements.

Jitter Measurements UsingPhase Locked Loops:Interpreting the Data.

· In The Time Domain Theoutput signal of the phasedetector in Figure 5 con-tains a wealth of informa-tion about the jitter of the measured clock. Directexamination of the signal using an oscilloscopecan show Pk-Pk jitter. A true RMS voltmeter canbe used to measure RMS (one sigma) jitter. Forthese measurements, it is critical that the mea-surement filters used represent the band of jitterfrequencies of interest. It would make no sense tomeasure noise from dc to 10 MHz when a band-width of 10 kHz to 1 MHz is required.Oscilloscopes with histogram and statistical capa-bilities are useful for characterizing the measuredjitter.

In the Frequency Domain, the spectrumof the output signal from the phasedetector in Figure 4 represents thespectrum and relative amplitude of jitterin the frequency domain. Examining thespectrum with a low frequency or FFTanalyzer gives the most intuitive pictureof clock jitter in terms of spectrum. Byintegrating the signal jitter spectrumover the frequency of interest, it is pos-sible to derive the RMS jitter of theclock. This is the most accurate andunfortunately the most cumbersome method forcharacterizing jitter, requiring specialized testequipment. A numerical example is included in

appendix I.

Specifying Jitter PerformanceGood jitter performance and low cost are not mutually

exclusive as long as:· The system requirements for jitter are defined in

terms of amplitude and spectrum.· The method used to generate the clock output fre-

quency is optimal for the application.

System Requirements:Although it is impossible to address all possible varia-tions, some general recommendations based on yearsof oscillator manufacturing may be helpful. While not acomplete survey of all applications, Table 1 is a start-ing point for specifying oscillator performance. Jitter

above 1 kHz is considered high frequency jitter. Clock Generation:

Application Notes



171

App Notes


Application Notes

6. Jitter in Clock SourcesVarious methods may be employed to generate highfrequency clocks. Performance may vary significantlybased on the technique used. Below 20 MHz, it can beassumed that direct crystal frequency generation is suf-ficient for all but the most critical requirements. LowNoise options should be considered for low jitter appli-cations for 20 MHz and above. Table 2 may be used asa starting point to select a cost-effective solution.Variations and combinations of methods listed in table 2could also be optimal solutions.

CONCLUSIONTo correctly specify performance of frequency sourcesboth jitter frequency and amplitude should be consid-ered. This requires an understanding of jitter, measure-ment techniques and their limitations. Time spent todetermine system needs will result in fewer problems

and less time spent fixing those problems later on. It willalso determine a cost-effective approach for each appli-cation.

In this paper, we discussed the definition of jitter, theunits used to describe it, and why jitter is an importantparameter. We also reviewed techniques used to mea-sure jitter as well as applications and typical perfor-mance based of various kinds of oscillators. The dis-cussion is by no means complete, but should give thereader enough information to understand the issuesinvolved. Industry standards were listed, as well as ref-erences for further reading. It is hoped that this paper isuseful and considered a good starting point for under-standing and specifying jitter.

172

7. rms Jitter Calculation

Application Notes


For applications requiring a 622.080, or 666.5143,

MHz PECL/ECL VCO with the lowest possible jitter inthe 50kHz to 80 MHz bandwidth, VI is offering theVCO600A/VS500 Series. These devices utilize SAWtechnology that resonates at a fundamental frequencyeliminating the need for internal multiplication, which isusually required to achieve high frequencies. The netresult is no harmonics/subharmonics and low phasenoise, or in the time domain, low jitter. The phasenoise was measured on several pairs of the VCO600A@ 622.080 MHz using an HP 3048A Phase NoiseMeasurement System. The typical phase noise wasthen integrated from 50 kHz to 80 MHz and convertedto jitter using the methodology outlined in VI's "Jitter inClock Sources" by Joe Adler, (available for download

at VI's website www.vectron.com). The following are the results:

X: = -54.95 dB Equivalent sideband level or the inte-grated phase noise

(X) Jrms := (360/(2p) 10 20 = 0.1025 rms jitter in

degreesJrms/360 := 0.0002847 Unit Intervals (UI) rms (Jrms/360)(1/622.080MHz) = 0.458 ps rms or 3.206

ps peak to peak

The sideband level of -54.95 dBc for the integration ofthe phase noise over 50 kHz to 80 MHz has an equiva-lent jitter value of 0.458 ps RMS or 3.206 ps peak topeak.

For applications requiring a 622.080, or 666.5143, MHzPECL/ECL VCO with the lowest possible jitter in the 12kHz to 20 MHz bandwidth, VI is offering theVCO600A/VS500 Series. These devices utilize SAWtechnology that resonates at a fundamental frequencyeliminating the need for internal multiplication, which isusually required to achieve high frequencies. The netresult is no harmonics/subharmonics and low phasenoise, or in the time domain, low jitter. The phasenoise was measured on several pairs of the VCO600A@ 622.080 MHz using an HP 3048A Phase NoiseMeasurement System. The typical phase noise wasthen integrated from 12 kHz to 20 MHz and converted to jitter using the methodology outlined inVI's "Jitter in Clock Sources" by Joe Adler, (availablefor download at VI's website www.vectron.com).

The following are the results:X: = -60.98 dB Equivalent sideband level or the inte-

grated phase noise (X)

Jrms := (360/(2p)) 10 20 = 0.0051 rms jitter indegrees

Jrms/360 := 0.0014225 UI rms (Jrms/360)(1/622.080MHz) = 0.230 ps rms or 1.610

pS peak to peak

The sideband level of -60.98 dBc for the integration ofthe phase noise over 12 kHz to 20 MHz has an equiv-alent RMS jitter value of 0.230 ps or 1.610 ps peak topeak.

rms Jitter Calculation in the 12 kHz to 20 MHz Bandwidth for VI’s VCO600A/VS500 @622.08 MHz

For applications requiring a PECL or ECL VCXO havingan RMS jitter of < 1.0 ps within the bandwidth of 12KHzto 20 MHz, VI is offering our CO-400 Series. The phasenoise on two VCXO's of VI's model VC-400-CFC-20SG@ 155.52 MHz were measured on the HP 3048A PhaseNoise Measurement System. The phase noise was thenintegrated from 12 kHz to 20 MHz using the TraceIntegration function of the 3048A. The equivalent side-band level of the integrated phase noise was -74.76dBc. Per the methodology outlined in VI's "Jitter inClock Sources" by Joe Adler, (available for download atVI's website www.vectron.com) the RMS jitter can betreated as small index phase modulation and can beconverted from radians to degrees:

X:=-74.76 dB Equivalent sideband level orthe integrated phase noise

(X)Jrms:= (360/2 ) 20 jitter in degrees.Jrms/360 = 0.0000409 UI rms.(Jrms/360) (1/155.52 MHz) = 0.187 ps rms or 1.309 pspeak to peak.

The sideband level of -74.76 dBc for the integration ofthe phase noise over 12KHz to 20 MHz has an equiv-alent RMS jitter value of 0.187 ps. Therefore, the CO-600V series VCXO can satisfy the performancerequirement of < 1ps RMS jitter for 12kHz to 20 MHz.

rms Jitter Calculation in the 12 kHz to 20 MHz Bandwidth for VI’s VC-400 Series VCXO

173

rms Jitter Calculation in the 12 kHz to 20 MHz Bandwidth for VI’s VCO600A/VS500 @622.08 MHz

App Notes

7. rms Jitter Calculation

Application Notes


For applications requiring a PECL VCXO having anRMS jitter of <1.0 ps within the bandwidth of 12 kHz to20 MHz, VI is offering our J type series. The phasenoise on two VCXO’s of VI’s model JDUGLMEP-155.52MHz were measured on the HP 3048A Phase NoiseMeasurement System. The phase noise was then inte-grated from 12 kHz to 20 MHz using the TraceIntegration function of the 3048A. The equivalent side-band level of the integrated phase noise was -70./dBc.Per the methodology outline in VI’s “Jitter in ClockSources” by Joe Adler, (available for download at VI’swebsite www.vectron.com) the RMS jitter can be treat-ed as small index phase modulation and can be con-verted from radians to degrees:

X = -70.1 dB Equivalent sideband level or the integrat-ed phase noise. (X)

Jrms = (360/2 ) 10 20 = 0.017911 degreesJrms/360 = 0.0000497 IT rms(Jrms/360) (1/155.52 MHz) = 0.319 ps rms or 2.23

The sideband level of -70.1 dBc for the integration ofthe phase noise over 12 kHz to 20 MHz has an equiv-alent RMS jitter value of 0.319 ps. Therefore, the JType series VCXO can satisfy the performancerequirement of <1 ps RMS jitter for 12 kHz to 20 MHz.

For applications requiring a PECL XO having an rms jit-ter of <1.0 ps within the bandwidth of 12 kHz to 20 MHz,VI is offering our XO-500 series. The phase noise ontwo VCXO’s of VI’s model XO-500 Series. The phasenoise on two XO’s of VI’s model XO-500-DFC-20SN -155.52 MHz were measured on the HP 3048A PhaseNoise Measurement System. The phase noise was thenintegrated from 12 kHz to 20 MHz using the TraceIntegration function of the 3048A. The equivalent side-band level of the integrated phase noise was -70.0 dBc.Per the methodology outlined in VI’s “Jitter in ClockSources” by Joe Adler, (available for download at VI’swebsite www.vectron.com) the RMS jitter can be treat-ed as small index phase modulation and can be con-verted from radians to degrees.

X = 70.0 dB Equivalent sideband level orthe integrated phase noise(X)

Jrms = (360) (2 ) 10 20 = 0.0181185 degreesJrms/360 = 0.0000503 UI rms(Jrms/360) (1/155.52 MHz) = 0.323 ps rms or 2.26 ps peak to peak

The sideband level of -70.0 dBc for the integration ofthe phase noise over 12 kHz to 20 MHz has an equiv-alent RMS jitter value of 0.165 ps. Therefore, the XO-500 series XO can satisfy the performance require-ment of <1 ps RMS jitter for 12 kHz to 20 MHz.

For applications requiring a PECL XO having an rms jit-ter of <1.0 ps within the bandwidth of 12 kHz to 20 MHz,VI is offering our XO-400 series. The phase noise ontwo XO’s of VI’s model XO-400 Series. The phasenoise on two XO’s of VI’s model XO-400-DFC-20SN -155.52 MHz were measured on the HP 3048A PhaseNoise Measurement System. The phase noise was thenintegrated from 12 kHz to 20 MHz using the TraceIntegration function of the 3048A. The equivalent side-band level of the integrated phase noise was -69.0 dBc.Per the methodology outlined in VI’s “Jitter in ClockSources” by Joe Adler, (available for download at VI’swebsite www.vectron.com) the RMS jitter can be treat-ed as small index phase modulation and can be con-verted from radians to degrees.

X = 69.0 dB Equivalent sideband level orthe integrated phase noise(X)

Jrms = (360) (2 ) 10 20 = 0.020329 degreesJrms/360 = 0.0000564 UI rms(Jrms/360) (1/155.52 MHz) = 0.363 ps rms or 2.54 ps peak to peak

The sideband level of -69.0 dBc for the integration ofthe phase noise over 12 kHz to 20 MHz has an equiv-alent RMS jitter value of 0.363 ps. Therefore, the XO-400 series XO can satisfy the performance require-ment of <1 ps RMS jitter for 12 kHz to 20 MHz.

rms Jitter Calculation in the 12 kHz to 20 MHz Bandwidth for VI’s XO-400 Series XO

rms Jitter Calculation in the 12 kHz to 20 MHz Bandwidth for VI’s J-Type Series VCXO

rms Jitter Calculation in the 12 kHz to 20 MHz Bandwidth for VI’s XO-500 Series XO

174

8. Absolute Pull Range Definition

Application Notes


Introduction

Vectron International uses Absolute Pull Range todefine the amount of deviation a VCXO can be adjust-ed about the device's center frequency, (fo). This appli-cation note defines APR and compares the advantageswith alternative terms such as total pull range.

DefinitionAPR is the minimum guaranteed amount the VCXO canbe varied, about the center frequency (fo). It accountsfor degradation's including temperature (0 to 70 or -40 to 85°C), aging (10 years, 40°C), power supply vari-ations (10%) and load variations.

APR = (Pull range) - (degradations due toTemperature+aging+power supply+load)

For a VCXO specified in terms of APR, there is noguess work regarding how much frequency deviation isavailable, over all conditions, to track your incoming sig-nal.

For example, one of the most popular APR options is 50ppm, which is defined by "G" in the part number code.A 16.384 MHz VI V-Type VCXO would typically have nomore than 20 ppm of temperature drift, 5 ppm of aging(10 years, 40C), 5 ppm due to power supply variationand 4 ppm due to load variations. A device specified interms of Total Pull would need to have at least 84 ppmof Total Pull tomeet specifications.

In order to offer the APR guarantee, devices are wellcharacterized for temperature, aging and power supplyvariations. Every VCXO is tested over temperature forpull range. This is a fully automated process wheredevices are continuously tested throughout a 0 to 70, or-40 to 85°C temperature cycle, including a soak at theextreme temperatures. Data is automatically stored foranalysis. Aging characterization has been performedand correlated. Variations due to power supply and loadare also well understood.

Devices specified with Total Pull Range can falselyseem superior due to what seems like a higher pullcapability. However, in order to ensure that a VCXOspecified in terms of Total Pull meets your designneeds, the manufacturer must be contacted to definedrifts due to temperature, aging, power supply and load

variations, which must then be subtracted from theirTPR. It can not be assumed that the VCXO perfor-mance is equal to Total Pull Range minus the "stability"unless the stability is clearly defined - this is seldom thecase. APR reduces time spent defining and under-standing VCXO specifications.

ExampleIn a digital communications network, the source clockhas a defined maximum error from the centerfrequency over temperature, aging and power supplyvariations. This error is defined by a variety of factorsdictated by the application. The receiver incorporates aVCXO which must be capable of tracking the sourceerror to recover the clock and/or translate to a higherfrequency. For example, a Stratum 4 level clock dic-tates a worse case error of 32 ppm over allenvironmental conditions. The VCXO must also beable to pull or track this 32 ppm over temperature,aging and power supply variations.

A VCXO specified in terms of APR with a 32 ppm APRwould be the correct choice and will always be capa-ble of locking to the source clock under the mostadverse conditions - specmanship has been eliminat-ed.

Again using a Stratum 4, definition the VCXO can beselected by;

VI's APR Method Other's TPR MethodF= 32ppm APR Total Pull RangeDONE ! minus temperature stability

minus agingminus supply variationsminus output load variations

ConclusionAPR is a superior method for specifying the deviationcapability of a VCXO and is rapidly being adopted byother suppliers in the industry. In addition to testing toa minimum APR, every device is automatically testedfor rise and fall time, duty cycle, current consumptionand start-up time with test data stored for SPCanalysis. As well as offering superior performance andpackaging technology, VCXO's supplied by VI are themost thoroughly tested and have the best guaranteedspecifications in the industry.

175

App Notes

9. Saw-Based Frequency Control Product Applications

Application Notes


Introduction

In the first half of this paper, several SAW-based fre-quency control products which find applications in mod-ern telecommunication systems like SynchronousOptical Network (SONET), Synchronous DigitalHierarchy (SDH), and Asynchronous Transfer Mode(ATM) will be presented. They play the important role offrequency synthesis, frequency translation, data andclock recovery, and clock signal distribution to ensurelow bit error rate transport of signals in high frequencyoptical telecommunication equipment up to 2.5 Gb/s.

In the second half of this paper, the applications and theavailability of low-loss RF and IF SAW filters for existingand emerging wireless systems, the competing tech-nologies, the challenges to enter the market, and theapplications of conventional high-loss and high-selectiv-ity SAW filters in equipment based on the Code DivisionMultiple Access (CDMA) technique will be reviewed.

1.1 SAW-Based Timing Recovery Unit

The SAW-based timing recovery unit (TRU600) regen-erates data and clock signals from corrupted NRZ digi-tal data streams, such as those encountered in fiberop-tic data link and telecommunication applications.Although there are many suppliers providing discreteSAW filters for timing recovery applications, theTRU600 allows an easy drop-in solution for users.SAW-based timing recovery scheme offers the best jit-ter performance in many situations[1]. One example isin a SONET/SDH/ATM network interface card applica-tion situation where TRU600A can be used between theO/E converter and a serial to parallel chip[2]. A summa-ry of the specification follows:

Supply Voltage 5 V Acquisition Time <2 ms Output Clock Random Jitter 10 ps rms Power Consumption 325 mW

The TRU600 features a high-speed bipolar ASIC and aSAW filter in a hermetically sealed, 28-lead ceramic sur-face mountable package (18.5x10.5x3.4 mm3). Toextract a clock signal from the input data, the data is firstpassed through a prefilter and frequency doubler stage.This generates pulses containing significant spectralenergy at the input data rate. A precision narrow-band

SAW filter, centered at the clock frequency, substantial-ly suppresses jitter by rejecting other frequencies. Theextracted clock is then accurately aligned with theincoming data signal at the input of a decision circuitwhich then retimes the data.

In addition to producing outputs with very low jitter, theTRU600 has excellent stability, fast acquisition time,and robust operation. It is available with standardSONET/SDH/ATM frequencies at 155.52, 311.04, and622.08 MHz. Additional frequencies (124.416, 125,139.264, 200, 265.625, and 278.528 MHz) for FDDI,ESCON, Fiber Channel, ISDN (CEPT 4), and otherapplications are also available.

To prepare for the increasing capacity demand in thetele/data communication market, a similar device whichworks up to the STS-48/STM-16 rate (2488.32 MHz) isbeing developed. Such a SAW-based clock and datarecovery module is preferred in the emerging highspeed optical communication receiver application[4].

1.2 Discrete SAW Filters for Timing Recovery

For customers who prefer to build their own timingrecovery path on their SONET/SDH/ATM boards, weoffer discrete SAW filters to perform the clock extractionfunction. They are available at 155.52, 622.08, and2488.32 MHz. A summary of the specification follows:

Frequency (MHz) 155.52 622.08 2488.32 Insertion Loss (dB) 17 15.5 19.5 3-dB Q 420 800 750 Phase Slope (°/KHz)0.72 -0.33 -0.07

The 155.52 MHz SAW filter is available in a standard

14-pin, metal dual-in-line package (20.3x12.7x7.4 mm3).

The 622.08 MHz SAW filter is available in a low-profile,22-pin, metal surface mountable package

(15.9x13.6x3.2 mm3). The 2488.32 MHz SAW filter is

available in a compact, surface mountable microwave

package (11.4x10.7x2.1 mm3). The 155.52 , 622.08,

and 2488.32 Mhz timing recovery SAW filters are alsoavailable in the 9mmx7mm, 9mmx5mm, and9mmx7mm leadless chip carrier surface mountablepackages (LCC SMPs) respectively.

1.3 SAW-Based Voltage-Controlled Oscillator The SAW-based voltage-controlled oscillator (VCO600)is a highly integrated device which uses an ASIC with an

176

Application Notes


on-chip phase shifter for frequency pulling and a SAWdelay line with a typical 3-dB Q of 400. The VCO600 hasan ECL output and is available with standardSONET/SDH/ATM frequencies at 155.52, 311.04, and622.08 MHz. Additional frequencies at 278.528 and368.64 MHz are also available. The VCO600 is housedin a hermetically sealed, 28-lead ceramic surfacemountable package. Typical applications are data retim-ing and synchronization as part of a PLL, as well as fre-quency synthesis and frequency translation. TheVCO600A also has a unique output disable and clockthrough feature which improves board-level testing. Asummary of the specification follows:

Absolute Pull Range ±50 ppm Supply Voltage -5 V Control Voltage -0.5 to -4.5 V Linearity ±3% Spurious Output Suppression -60 dB

2.1 SAW Filters for Wireless Applications

Wireless communication systems available includemobile cellular, cordless phones, paging services,Specialized Mobile Radio (SMR), mobile satellite andWireless Local Area Network (WLAN). In Europe, theanalog cellular system (ETACS) is being displaced bythe new Global System for Mobile digital communica-tion system (GSM). The Personal Handyphone System(PHS) and Personal Digital Cordless system (PDC) aregaining momentum in Japan (Table 1).

In the US, the existing analog cellular systems (AMPS)is being converted gradually into dual mode analog/dig-ital systems (IS-54). Digital cellular system (IS-95) usingthe CDMA is also available in some metropolitan areas.These systems operate at the 800 MHz bands andclaim to support more subscribers and provide betterservices. Non-licensed digital cordless phones usingfrequency hopping spread spectrum (FHSS) techniqueat the ISM-15 902-928 MHz band are in the market.They provide more secure services in the crowded con-sumer market of cordless phones. Paging companiesare now providing two-way paging services. In-flight Airto Ground Telephony (AGT) service operating in the849-851 MHz & 894-896 MHz bands is becoming morepopular with the option to route ground to air calls.Wireless data transfer equipment (e.g. WLAN operatingat the non-licensed ISM-15 2400-2483.5 MHz band) isnow available[6]. SMR is changing into the enhanced

version (ESMR) to support digital data/voice transport.Low/Medium Earth Orbit (LEO/MEO) Mobile SatelliteServices like[5] IRIDIUM (Motorola), ARIES(Constellation Communication, Inc.), GLOBALSTAR(Loral & Qualcomm), ELLIPSAT (Ellipsat Corp.),Odyssey (TRW), and Teledesic (Microsoft et al.) willmake "calling anyone, anytime, and anywhere" a reali-ty. In addition, equipment using the Global PositioningSystem (GPS) technology is now available in the com-mercial (automobiles, aircrafts, ships, etc.) and con-sumer market (handheld receivers). Tremendous effortsare being put into developing low-power front end andbaseband chips sets, longer life batteries, etc. In the RFand IF sections of the portable and stationary equip-ment of these systems, low-loss SAW filters and con-ventional high-loss and high-selectivity SAW filters have

become and will continue to be the vital components[6].

Most European, US, and Japanese manufacturers ofSAW filters are adding equipment and expanding theirfacilities to accommodate the business opportunities.

Table 1. World wide Wireless

Telecommunication Standards

ContinuedTable 1. World wide Wireless Telecommunication Standards

Standard Rx MHZ TX MHz #Users RF BW MHZ IF BW MHzAnalog Cellular (FDMA)AMPS 869-894 824-849 832 25 30ETACS 916-949 871-904 1240 33 25NTACS 860-870 915-925 400 10 12.5NMRT450 463-468 453-458 200 5 25NMT900 935-960 890-915 1999 25 12.5Digital Cellular (TDMA)IS-54/-136 869-894 824-849 832X3 25 30IS-95 869-894 824-849 20X798 25 1250(CDMA)GSM 935-960 890-915 124x8 25 200PDC 810-826 940-956 1600x3 16 25

1429- 1477- 1600x3 24 251453 1501

Digital Cordless/ PCN (TDMA/TDD)CT2 & 864/868 & 40 4 100 CT2+

944/948DECT 1880- 10X12 110 1728

1990PHS 1895 300X4 12 300

1907DCS1800 1805- 1710 750X16 75 200

(FDD) 1880 1785


177

App Notes

Application Notes


2.2 Low-Loss SAW Filters for Front-End RFApplications

The RF bandwidth in Table 1 shows the minimum band-width requirement for the front-end filtering in both thetransmitting and receiving paths of different wirelesssystems. For many years, dielectric resonator filters (2to 3 poles) have been widely used especially for the ter-minals. They are low cost, rugged, and easily imple-mented. In addition, they can handle high power.Though they were bulky, it was not a problem sinceearly wireless terminals were mostly mobile but station-ary (e.g. mounted in cars). However, the current trend ofwireless equipment is moving toward more and moreportable, and component size is one of the many factorsthat designers of terminals are concerned with.

Low-loss RF SAW filters (as low as 2.5 to 4 dB)between 800 and 1500 MHz are now available in LCCs

as small as 3.8x3.8x1.6 mm3

for AMPS, GSM, PDC,and other applications and the trend is moving toward

3.0x3.0x1.0 mm3

for PCS, WAN, and WLAN, and other

applications at 1.8 to 2.5 GHz band[7,8

]. There are manysuppliers of these devices from and Europe. Some ofthese companies offer robust products in this 800 to1500 MHz range and are developing devices toward the1.8 to 2.5 GHz range with the goal to attain even lowerinsertion loss. The popular designs are In-Line CoupledResonator Filter (In-Line CRF, Figure 1) and ImpedanceElement Filter (IEF, Figure 2). The latter does not offeras good ultimate rejection and it can only be slightlyimproved by adjusting the capacitance ratio of the par-allel and series arms. It does hold a good prospect inoffering insertion loss lower than 2 dB with high fre-quency operation. Table 3 compares their applicationsand performance. Almost all low-loss RF SAW filters formobile applications use LiTaO3, LiNbO3, or Li2B4O7 as

the substrate materials in order to provide the wide

bandwidth requirement (up to 6%)[9].

One Japanese supplier recently announced the avail-ability of SAW-based duplexers at the 800 MHz band

has

put itself as the leader of the pack[10]. They have suc-

ceeded in overcoming one major obstacle- power han-dling requirements for the transmission path in duplex-er applications. In North America, only two suppliersmanufacture RF SAW filters for cellular terminalsand/or digital cordless phones (e.g. CT-2) and they areprimarily captive.

Most low-loss RF SAW filter suppliers consider theyhave shrunk the footprints of the devices to smallenough sizes. The trends are to put in efforts toreduce the package height possibly through flip-chip

method[12] and, more importantly, to develop ways to

further reduce the insertion loss to below 2 dB. Thelatter is to compete with the dielectric resonator filterswhich now have 1.3 to 3 dB insertion loss.

Standard Rx MHZ TX MHz #Users RF BW MHZ IF BW MHzWireless Data-WAN/LAN (TDMA)CDPD 869-894 824-849 832 25 30RAM 935-941 896-902 480 6 12.5

403-470 450 67 12.5Ardis 851-869 806-824 720 18 25IEEE 2400-2483FHSS/79 83 1000 802.11

(US/Europe)(CSMA) 2470-2499DSSS/7 29 10000 (Japan)

Emerging Personal Communication System (PCS)1930- 1850 601990 1910

High Tier (Larger Cell)PCS TDMA (based on IS-136 Ericsson, AT&T, Hughes

cellular)PCS CDMA (based on IS095 Qualcomm, AT&T, Nokia

cellular)PCS 1900 (based on GSM Nortel, Ericsson, Nokia, Alcatel,cellular) Motorola, MCI, Pacific TelesisWideband CDMA InterDigital, OKI Low Tier (Small

Cell)PACS (based on PHS Cordless) Motorola, Panasonic, NEC, Hitachi

Hughes, BellecoreDCT-U (based on DECT

cordless)Composite CDMA/TDMA Omnipoint

Figure 1. In-line CRF Design for RF Applications


178

Many suppliers provide dielectric resonator filters, SAWfilters, and chip monolithic LC-type filters to meet thefrequency and bandwidth requirements for RF filtering.These devices are primarily different in insertion loss,attenuation, price, and size. Major progresses havebeen made in the development of chip monolithic LC-type RF filters and dielectric resonator filters[11]. Theformer has comparable size and insertion loss as SAWfilters except they do not offer as good attenuation.Suppliers have also shrunk the size of dielectric res-onator filters significantly in the last couple of years andit is a formidable competitor especially in the >2 GHzRF applications[12]. Table 2 depicts the generic com-parison of these RF filter technologies.

Dielectric Filter SAW Filter LC Multilayer Filter Loss Best Good Good Attenuation Good Best Good Size (cubic mm) Fair Best Good Design Flexibility Good Fair Best

Table 2. Generic Comparison of RF Filter Technology[11]

2.3 High Velocity Longitudinal Leaky SAWs andHigh Velocity SAWs in Piezoelectric Film/DiamondStructures

One way to maintain the physical feature size of trans-ducer fingers while pushing up the operating frequen-cies is to increase the SAW velocity. LSAWs with lowleakage loss are being used extensively in modern low-loss RF SAW filters. Popular LSAW cuts are 36° Y-X

LiTaO3, 41° and 64° Y-X LiNbO3 [9]. LSAW's velocity is in

general higher than that of the Rayleigh wave, and isalways sandwiched in between the slow shear and fastshear velocities. They are attractive because of its highvelocity, low leakage loss, and strong electromechani-

cal coupling. In the last several years[13], we have seen

progresses in the study of longitudinal LSAWs which

have low leakage loss, strong electromechanical cou-pling, and comparable temperature coefficient of delay(TCD). It is foreseeable that wafer cuts using longitudi-nal LSAWs will become commercially available in thefuture to support high frequency SAW devices.

In the past several years, the synthesizing of polycrys-talline diamond films using chemical vapor deposition(CVD) has become quite successful. In 1989,Yamanouchi et al. suggested theoretically that high fre-quency SAW devices (>3 GHz) could be realized in apiezoelectric AlN or ZnO/diamond structure because ofthe hardness of diamond film (Rayleigh wave velocity

could exceed 12,000 m/s)[14]. Extensive experimental

work is being actively pursued in Japan and Russia[15].

It's likely we will see vendors supplying diamond filmcoated SAW wafers in the future.

Table 3. Comparison of Different Low-Less SAW Filter Designs for Current Applications

2.4 Low-Loss IF SAW Filters for Very Narrow BandApplications

Design TCRF SPUDT In-Line CRF IRFApplication IF IF IF & RF RFCurrent <600 MHz <400 MHz <1GHz As high asFrequency 2.4 GHzRangeInsertion >3 dB 6~10 dB >2 dB >1 dBLossBandwidth 0.04 to 0.1% 0.3 to 5% 0.08 to 5% up to 6%Materials Quartz Quartz & Quartz, Quartz,

LiTaO3 LiTaO3, LiTaO3,, &LiNbO3 & LiNbO3

Li2B4O7

Strengths Superior Superior Small size; Small size;near-in out-of-band Wide Widerejection; rejection. bandwidth; bandwidth;Low-loss. Traps placing Matching is

is easy; not needed;Matching is Excellentnot needed. near-in

rejection.Weakness Metallization Matching is Sidelobe on Poor out-of

usually thick generally high bandand needed; long frequency rejectionuniformity is chip size; </8 side due to Varied

is critical; fingers direct transducerMatching is usually transmission. frequencies.generally needed;needed Difficult to

Synthesize/optimize

Application Notes


Figure 2. Impedance Element Filter Design for RF Applications


179

App Notes

Application Notes


The IF bandwidth listed in Table 1 depicts the channelwidth. For many years, 20 to 45 MHz monolithic crystalfilters (MCFs) of 10 to 30 KHz bandwidth (4 to 6 poles)dominated the IF filtering segments of analog cellularsystems like AMPS. Nowadays, the trend is to push upthe IF frequency to help to suppress images and spursdue to mixing and the continuous narrowing of thetransmitting and receiving bands in the RF carriers to

support more channels[16] (e.g. the expansion of AMPS

bands from 20 MHz to 25 MHz several years ago andthe current expansion of GSM to EGSM). MCFs above45 MHz, in addition to being fragile, are costly to make.Only one company from Japan is persistently pushingthe MCFs to higher frequencies using the inverted

mesa quartz resonator technique[17].

Figure 3. 4-Pole TCRF Design for IF Applications

Low-loss IF SAW filters at around 80 MHz for narrow

channel analog systems like AMPS applicationsemploying the 4-pole Transversely-Coupled ResonatorFilter design (TCRF, Figure 3) are now widely available.In addition to being low-loss (6 to 10 dB), these filtershave excellent rejection in suppressing images after themixing stage. It was also used as the RF front-end filterbetween 200 and 300 MHz in earlier narrow bandpagers to allow down conversion directly to 455 KHz IFfrequency without going through a first IF frequency of21.4 MHz. Since the bandwidth is very narrow, quartzSAW substrate is exclusively used to minimize frequen-cy drift due to temperature. Standard frequencies at82.2, 83.16, 85.05, 86.85, and 90 MHz are available.

13.3 x 6.5 x 1.3 mm3, 15.4 x 6.5 x 1.5 mm

3, and other

LCCs are widely used. There are many suppliers ofthese "standard" filters from Japan and Europe. 2.5 Low-Loss IF SAW Filters for Narrow BandApplications

IF filtering in digital cellular systems usually requires

wider bandwidth and stringent delay characteristic.Single-Phase Unidirectional Transversal SAW filter(SPUDT) is one of the few designs used to achieve thebandwidth and delay requirements for the digital sys-tems. The SPUDT design ingeniously directs powertransmission from the electrical port to the forwardacoustical port by setting a certain phase shift (±45° or±135°) between the transduction center and the reflec-tion center. When the matching of the electrical portbegins, the insertion loss will decrease (Figure 4). Thetriple transit echo will also decrease (reflection coeffi-cient decreases) as the acoustical reflection and thepiezoelectric regeneration begin to cancel out.

Figure 4. SPUDT Design for IF Applications

Comparing with the conventional transversal filterdesign which has only a transduction function to work

with, the challenge for the SPUDT design is now onehas to properly account for the reflection function also.The In-Line CRF design described in the low-loss RFSAW filters section can also be used to provide evenwider bandwidth than the SPUDT design in IF applica-tions (Figure 5). Table 3 compares their applicationsand performance. Depending on the requirements,devices in LCCs with footprints anywhere from

5 x 5mm2

to 19 x 6.5mm2

are available.

Similar to the analog systems like AMPS (or IS-54), webegin to see standardized IF frequencies in some pop-ular digital wireless terminals (e.g. 71 MHz for GSM,110.592 MHz for DECT, 130 MHz for PDC, 248.45 MHzfor PHS et al.). Same as the low-loss RF SAW filters,suppliers of these standard low-loss IF

SAW filters are usually more bulky and costly than thelow-loss RF SAW filters. Many designers are seekingways to reduce the size as a means to reduce the


180

cost[18]. In-Line CRF design can usually fit into smaller

packages than the SPUDT design and circuit matchingis in general not needed.

Most IF filterings in emerging systems though, areimplemented differently; very much dependent on thesystem designers. In the US there are tremendousneeds of SAW filters to perform IF filtering for differentnew wireless equipment (terminals and base stations)based on the newly allocated PCS bands. These newsystems will be operating at the 1.8 to 2.0 GHz range. The equipment will very likely need to down convertthe radio signals to IF frequency (50 to 500 MHz) forprocessing.

Figure 5. In-Line CRF Design for IF Applications

2.6 SAW Filters for CDMA Base stationApplications

The technique of using direct sequence spread spec-trum method in providing multiple access to frequencychannels (CDMA) is gaining momentum worldwide. Inthe US, IS-95 regulates the usage of this technologyfor cellular applications in the 900 MHz range. Newsystems using the CDMA will appear in the PCS band.

Figure 6. Frequency Response of a 131.01 MHz SAW Filter forCDMA Base station Applications

Conventional SAW filters, though high in insertion loss,offer unsurpassed selectivity (low shape factor) andhigh rejection. They begin to find extensive applicationsin the terminals and base station of existing and emerg-

ing systems[19]. One example is a 131.01 MHz SAW fil-

ter for IF filtering in CDMA base station (Figure 6).Quartz substrate is used for such a device in order tomeet the 1.25 MHz channel width requirement overtemperature. For many emerging CDMA systems (e.g.W-CDMA), SAW filters will probably continue to be theonly solution which can provide robust filtering functionin the IF segment.

References

1. "Compact-Same-Size 52- and 156 Mb/s SDHOptical Transceiver Modules, " K. Yamashita et al.,J. Lightwave Tech., Vol. 12, No. 9, Sept. 1994, pp.1607-1615

2.1"ATM/SONET/SDH Network Interface CardDesign at 622 Mb/s," I. Verigan & B. Wheeler,Proc. 1995 Hewlett-Packard DigitalCommunication Design Symposium, pp. A4.1-4.14.

3. "GaAs IC Set for 2.4-Gbits/s Optical TransmissionSystems," Y. Hatta et al., Hitachi Review, Vol. 43(1994), No. 2, pp. 91-94.

4. "Complete Chip Set Speeds Design of PCMCIAWLANS," Microwaves & RF, Oct. 1995, pp. 146-147.

5. "The Market & Proposed Systems for SatelliteCommunications." R. Rusche, applied Microwave& Wireless, Fall 1995, pp. 10-34.

6. "Filtering and Frequency Control for the NextGeneration of Mobile Communication Systems," J.Schwartzel, Proc. EFTF, 1994, pp. 175-189.

7. "Surface Mount Type SAW Filter for Hand-HeldTelephones," S. Suma et al., Proc. JapanElectronic Manufacturing Technology Symposium,June 9-11, 1993, pp. 97-100.

8. "1.9 GHz Low-Loss Surface Acoustic Wave Filter:Liftoff Method and 3mm Square Small SizePackage," N. Mishima et al., Japan J. Appl. Phys.,Vol. 33, 1994, pp. 2984-2988.

9. "Current Status of Piezoelectric Substrate andPropagation Characteristics for SAW Devices," Y.Shimizu, Japan J. Appl. Phys., Vol. 33, 1993, pp.2183-2187.

10. Electronic Engineering Times, Jan. 15, 1996.

Application Notes



181

App Notes

Application Notes


11. "RF Filter Technology for WirelessCommunications- The Choice is Yours," EldenGrace and H. Iwatsubo, Wireless Design &Development, June 1996, pp. 8.

12. "Filters Compete for Mobile Phone Niche,"Masaharu Tanaka, Nikkei Electronics Asia,Technology Trend, May 1996.

13. "Longitudinal Leaky Surface Waves for HighFrequency SAW Device Applications," T. Sato andH. Abe, Proc. Ultrasonic Symp., 1995, pp. 305-315.

14. "SAW Propagation Characteristics and FabricationTechnology of Piezoelectric Thin Film/DiamondStructure," K. Yamanouchi, N. Sakurai, and T.Satoch, Proc. Ultrasonics Symp., 1989, pp. 351-354.

15. "SAW Devices on Diamond," H. Nakahata et al.,Proc. Ultrasonics Symp., 1995, pp. 361-370.

16. "SAW Filters Aid Communications SystemPerformance," A.Victor & S. Skiest, Microwave &

RF, Aug. 1991, pp. 104-111. 17. "High Frequency Fundamental Resonators and

Filters Fabricated by Batch Process UsingChemical Etching," O. Ishii et al., Proc. FrequencyControl Symp., 1995, pp. 818-826.

18. "Innovative SPUDT Based Structures for MobileRadio Applications," M. Solal et al., Proc.Ultrasonics Symp., 1994, pp. 17-22.

19. "Frequency Coordination Between CDMA andNon-CDMA Systems," S. Soliman and C.Wheatley, Microwave Journal, Jan. 1996, pp. 86-98.


182

10. Evacuated Miniature Crystal Oscillators (EMXOTM)


Vectron International Announces aTechnological Breakthrough

in Oscillator Design byIntroducing our Evacuated Miniature

Crystal Oscillator (EMXO™)

OCXO’s (Oven Controlled Crystal Oscillators) are usedwhen frequency vs. temperature requirements are toostringent to be met by a basic XO (Crystal Oscillator) orTCXO (Temperature Compensated Crystal Oscillator).With an OCXO, the temperature of the crystal and criti-cal circuits is kept constant as the temperature outsidethe oscillator varies. Controlling the temperature insidethe oscillator with an oven maintains this constanttemperature. In an OCXO, the changes in the ambienttemperature are sensed and then fed back to an ovencontrol that continually maintains a constant optimumtemperature inside the oscillator enclosure. An OCXOcan improve the crystal's inherent stability by more than5000 times. The oven control system is not perfect, theopen loop gain is not infinite, there are internal temper-ature gradients inside the oven (oscillator) and, in aconventional oven, the circuitry outside the oven shell issubjected to ambient temperature changes that can"pull" the frequency.

The improved temperature stability performance of aconventional OCXO over an XO or TCXO comes at asteep price. OCXO power consumption, for instance, isgreater by a factor of over 200. There is also a sizeconsideration. In an ordinary OCXO, a crystal isenclosed in a metal case, which is then placed inside anoven shell together with temperature sensitive circuitry,and then surrounded by thermal insulation. All this,plus any additional circuitry are then placed in a metalhousing making for a bulky package, which becomesvery difficult to miniaturize.

To overcome these obstacles, the EX-380 SeriesEMXO (Evacuated Miniature Oven Controlled Crystal

Oscillator) was specifically developed to achieve OCXOperformance while significantly lowering power con-sumption and reducing package size (See Figure 1).

These characteristics go hand in hand since reducingpackage size makes it easier to improve power con-sumption. First, the volume of the package was madeas small as possible to reduce the volume that the ovenneeds to heat. Secondly, the most effective insulationneeded to be used. In the EX-380 Series, Vectron hasdone both. The package was designed to single DIPdimensions of 0.82" x 0.52" x 0.3", and the oscillatoruses a vacuum as the insulation medium - a dramaticimprovement over conventional polyfoam or fiber basedinsulation material. Another great contributor toVectron’s efforts to reduce size and still provide "ovenoscillator" performance was to eliminate the use of largepackaged crystals which, up until now, had to be usedto achieve good aging. Instead, a way to use an opencrystal blank was found and made practical. Vectronhas succeeded in resolving outgassing and contamina-tion issues, which could degrade performance. To dothis required manufacturing the oscillator with a highinternal vacuum level, with low internal outgassing, toprovide the needed thermal insulation. A high level ofcleanliness was needed to prevent contamination of theopen (un-encased) crystal blank and to ensure excep-tional long term crystal aging.

The key design feature of this package utilized theconcept of integrating the precision crystal in blankform in combination with hybrid microelectronics cir-cuitry. In doing this, obtaining good aging performancewas paramount. For this reason, a cold welded pack-

183

Application Notes

App Notes



age was chosen rather than a more traditional resis-tance welded package. Cold weld sealing provided atrue metallurgical bond between ductile metal surfaceswithout added heat from the sealing process. Underthe high tonnage pressure introduced through theindentation of the welding die, a plasticity flow of mate-rial takes place on the mating surfaces. The endresult is a hermetically sealed enclosure without conta-mination from weld splashes, dust and vapors. And,most important, cold weld sealed enclosures achievesa high level of vacuum integrity.

Mechanically, the hybrid circuit and crystal assembly issuspended directly over a highly insulating structure tominimize heat energy loss through conduction. Inaddition, the entire assembly is thermally insulated tothe enclosure by vacuum at a pressure level of 10-6torr. Based on the steady state thermal conductioncalculation, this package design resulted in a thermalresistance of >300 degrees C/watt.

The EMXO manufacturing process is interesting and aprocess flow diagram is shown below in figure 2.

Substrates are fabricated with thick film screen printingtechniques with each deposition layer subjected tothree different process stages - print, dry and fire.Crystal clips are attached to the gold conductor traceon a substrate with high thermal conductivity. Allactive and passive components are mounted on thesubstrate using a conductive adhesive and thenmoved to a convection oven for curing. After the cureprocess, the hybrid is cleaned to remove organic andnon-organic contaminants. Wires are bonded on thehybrid circuit as interconnects. The hybrid circuits are

then attached to the cold-weld package with adhesive.Finally, blank crystals are mounted onto the clips andtuned to the nominal frequency needed, by an evapo-ration process, to a typical accuracy of 1 ppm. Theunits are then cold weld sealed. The oven itself isheated by direct thermal conduction applied to a heatconductive substrate.

This "Open Blank" EMXO concept is shown and com-pared to conventional OCXO construction, in Figure 3below.

In figure 3 above, 1,2 and 3 reflect conventionalOCXO construction while 4 is the EMXO. It shows themajor elements of the SDIL EMXO. Only one substrateis used and all the elements are heated. The oscillatoris essentially a CMOS gate type with an additional var-actor diode and LC trap for overtone select. The res-onator is a 3rd Overtone, doubly rotated cut asrequired by the application, both of which offer superi-or aging performance when compared to a traditionalfundamental resonator.

It is anticipated that the EX-380 Series will find appli-cations where performance in severe mechanical envi-ronments is equally important to electrical perfor-mance. An additional focus for the EX-380 series,therefore, was to provide robust construction to with-stand high shock and vibration and to yield good G-sensitivity performance. For example, when the physi-cal orientation of an oscillator is changed, there is asmall frequency change (typically not more than sever-al parts in 10-9 for a 90-degree rotation) due tochange in stress on the crystal blank resulting from thegravitational effect upon the crystal supports. Thischaracteristic is known as "tip-over" and is expressedin 10-9/g where one g represents one-half of a 180-degree orientation change. To minimize this change

184

Application Notes



and also to enhance performance under shock andvibration, the crystal blank is mounted in a symmetricalmounting structure, instead of the more traditional 2 or3 points. This helps to achieve a high shock and vibra-tion endurance level, low g-sensitivity performanceand symmetrical heat transfer. Also, when a crystaloscillator is operating and subjected to vibration, spuri-ous frequencies are generated, offset from the fre-quency of oscillation by the frequency of vibration.These are commonly referred to as "vibration inducedside-bands" and these side bands behave similarly tophase noise. The amplitude of these spurious outputsis related to the amplitude of vibration, the mechanicaldesign of the crystal supports, and the mechanicaldesign of the oscillator assembly, including the crystalmounting. Here also the symmetrical crystal mountingstructure helps to reduce unwanted noise.

Even though many types of doubly rotated crystalsproduce lower amplitude spurii under vibration thanAT cut crystals, this characteristic is primarily deter-mined by the mechanical design of the crystal andoscillator rather than the specific crystal cut. Vectronuses a doubly rotated resonator in the EMXO seriesoscillator to provide lower close in phase noise, betteraging rates and reduced acceleration sensitivity. Figures 4 through 10 represent the typical actual testdata on qualification samples for various characteristics.

185

Application Notes

App Notes



Summary:

The model EX-380, low profile 4 Pin DIP EvacuatedMiniature, Oven Controlled Crystal Oscillator (EMXO)is available at selected frequencies from 10 MHz to 80MHz.

The unit provides exceptionally low aging rates andhigh temperature stabilities in an extremely smallpackage over a wide range of environmental condi-tions. The through hole unit measures only 20.8mm x13.2mm x 7.6mm (0.82" x 0.52" x 0.30"), providesaging rates of <1x10-9 /day average, <1x10-7 for thefirst year and <1x10-6 for 10 years with temperaturestabilities to ±1x10-7 over -40ºC to +85ºC. Wider tem-perature ranges are available from -55ºC to +85ºC.This performance is achieved by the application ofnew resonator design concepts and technologicalbreakthroughs. The EMXO bridges the gap betweencurrent large, high precision OCXO’s and smallerTCXO's and becomes the most economical choicewhere there is a need for spectral purity, short andlong term stability, along with small size and dramati-cally reduced power consumption.

Standard supply voltages for the EX-380 series are3.3 Vdc and 5 Vdc, all with an HCMOS,Complementary PECL, LVDS and sinewave output. Asurface mount version of this oscillator is available(EX-381/4/5). Sinewave output of 0 dBm to +3.0 dBm/ 50 ohm is available in the EX-381 surface mountversion. LVDS and complementary PECL are availablein the EX-384 surface mount version and HCMOS isavailable in the EX-385 surface mount version.

These EMXO units are ideal for SONET/SDH, DWDM,FDM, ATM, 3G, Telecom Transmission and SwitchingEquipment, Wireless Communication Equipment andMilitary Airborne and Mobile systems.

186

Application Notes

2004 vectron working 5 - all-electronics€¦ · perature compensated voltage controlled crystal...

Documents