bob burns - the hewlett packard archive · 2018. 4. 15. · j(.(fm)fm dfm o.3khz i i 1'-i o.3khz...
TRANSCRIPT
-
MEASURING PHASE NOISE AND ADJACENTCHANNEL POWER WITH THE 8901B
MODULATION ANALYZER.
Bob Burns
RF ~ MicrowaveMeasurementSymposiumandExhibition
FliOW HEWLETTa:~ PACKARD
www.HPARCHIVE.com
-
MEASURING PHASE NOISE AND ADJACENTCHANNEL POWER WITH THE 8901BMODULATION ANALYZER.
This paper describes a method for measuringthe phase noise of RF and microwave signalsusing a new version of the 8901B modulationanalyzer. This approach is an extension of theadjacent channel power measurementscommonly made on communications systemssignals, and actually measures total SSB noise atoffsets greater than 5 kHz away from thecarrier. Advantages and limitations of thistechnique are discussed and compared withother methods of measuring noise around RFand microwave signals.
OUTLINE:
1. Why measure phase noise and/or adjacentchannel power?• What is phase noise? Adjacent channel
power?• Why is this type of noise undesirable?• Why are these measurements required?
2. Existing methods for making thesemeasurements (advantages anddisadvantages):• Spectrum analyzers• Phase noise measurement systems• Modulation analyzers
3. How to make this measurement using the8901B:• Block diagram with noise/adjacent
channel power option• Spectral representation• Measurement algorithm
• adjacent channel power• phase noise
• Using an external L.O.• Measuring microwave signals• Advantages of this technique
4. Limitations• Accuracy and dynamic range
• amplifier and detector linearity and noisefloors
• L.O. noise and settability: internal vs. external• filter response
• Image response• Spurious
5. Automating noise and adjacent channelpower measurements:• System block diagram• Applications.
www.HPARCHIVE.com
-
As the communications spectrum becomes morecrowded, noise performance of communicationssignals becomes increasingly important as a wayto ensure overall system performance. Theintent of this paper is to describe a method formeasuring this performance. This method iscompared to other measurement techniques, andthe relative advantages and disadvantages ofthese methods are discussed. An automatedsystem for making carrier noise measurements isdescribed, and examples given.
Noise around a signal can be represented asfluctuations of either the amplitude or phase ofthe carrier. This noise is commonly referred toas phase noise, since fluctuations in phase aretypically the dominant source of noise close tothe carrier.
733
734
AGENDA
• Why measure noise and/oradjacent channel power
• Techniques for measuringsignal nOise
• Measuring carrier nOiseWi th the aQ01 B
• Limitations of thiS technique
• Rutomating the measurement
SIGNAL REPRESENTATION
V(f> = Vs [l+t,V(t)] cos [21Tfo T +6,~(1)]
= 5i,gnal with ampli1uoe anD phase fluctuations,tly(t) and tl0m
www.HPARCHIVE.com
-
t.Um) DEFINITION
t (fm)=55B nOiS.,·to·catriet,atio at ~m offs..t f,om catt ie,
in dBc/Hz
t (fm)=PssB/Hz
Ps
Ps
735
e:t(fm) defines the ratio of power in a 1 Hzbandwidth fm Hz away from the carrier to thecarrier power. This is SSB power which can becaused by fluctuations in the amplitude orphase of the source, as well as additivebroadband noise. e:t(fm) is usually the importantmeasure of noise around communicationssignals.
INTERFERENCE
RECIEVER LO.
736
Noise around communications LOs can limit thecommunication system performance whenreceiving weak signals. Noise at adjacentchannel offsets on a locally transmitted signalcan prevent clear reception of a distant signal inthe adjacent channel. Noise on receiver LOs atoffsets equal to the IF frequency can limit thereceiver's overall sensitivity. Fast and accuratemeasurement of this noise is necessary foridentifying these causes and improving systemperformance.
ADJACENT CHANNEL POWERUNMODUlATED CARRIER
MODULATED CARRIER
>70 dB
Regulatory agencies are recognizing the need torequire minimum adjacent channel interferenceperformance on equipment used incommunications systems to ensure that all userscan make full use of allocated channels.Measurement of transmitted noise is required toguarantee conformance to regulations that arelikely to become increasingly stringent with time.
CHANNEL Nt CHANNEL N
737
www.HPARCHIVE.com
-
There are many types of commercially availabletest equipment that can be used to measurenoise of communications signals. The spectrumaround a signal can be measured directly on aspectrum analyzer. Systems of spectrumanalyzers and interfaces providing low noisereferences and sensitive detectors are available.Modulation analyzers also can be used tomeasure residual FM, and this can be related tophase noise.
Spectrum analyzers are valuable for aqualitative quick evaluation of signals if thenoise of interest is not too low. The phase noiseof the spectrum analyzer LO limits many RFspectrum analyzers to performance in theneighborhood of -100 dBc/Hz for et(f)measurements. Care must be used in correctingfor attenuator and IF linearity errors, and forusing the proper equivalent noise bandwidthcorrection factor for the IF filter in use. If theanalyzer logs the signal prior to detection,additional correction factors need to be used fornoise measurement. These limitations canprevent the use of a spectrum analyzer formaking fast, calibrated measurements of lownoise signals.
This plot shows the noise performance of theLO in a high-performance spectrum analyzer. Atmany offsets this noise is too large to allowmeasurement of some typical communicationssignals.
NOISE MEASUREMENT USING:
Spectrum Rnalyzers
Phase NOise Measurement Systems
Modulati.on Rnalyzers
738
DIRECT SPECTRUM MEASUREMENT
Rdvantages:
Fasl and Easy
Shows "8 i9 Piclute"
FleKibl~
Disadvantages:
Limiled DynamiC Ranqe
No Ph ase Infotmalion
739
ANALYZER L.O. NOISE
-ClO
-100
t(fm) -110
dBdHz -120
-130
-140
-150
,,~ SPECTRUM R JRIXlER
"- r-"-
""-~WBSI~ RLGEN.i:P 500 MHz'
r-
740
www.HPARCHIVE.com
im(kHz) 1 10 100
-
,I
NOISE MEASUREMENT SYSTEMS
Rdvantages:
V~ty Hi,9h P'erfotmance, Lcrw NOise
Accutate, Calibtated Output in d Bc IH zCan Disti"9Uish AM Ftom PM Noise
Disadvantages:
Requites Two Sources Ot Rpptoptiat~ Reference
Can be Lal"9~, Expensive, Slow
Many Pitfalls to Watch Out Fot
741
Noise measurement systems such as the HP3047A can be used to obtain accurate anddetailed noise information of very low noise RFand microwave signals. The expense of suchsystems has kept them from enjoyingwidespread popularity in all but the mostcritical testing applications. Additionally, thereare many possible pitfalls that can beencountered in using such systems, leavingoperation to relatively sophisticated users. Whena phase-locked signal is required, tuningcharacteristics of the oscillator must beconsidered, as well as its ability to beinjection-locked.
742
RESIDUALMODULATION MEASUREMENTS
Rdvantages:
QUi ck, Easy Qual itati ve Check of 5pecltal Pu ti ty
AM and FOt Can Be Disti"9Uished
Disadvantages:
Results Depen
-
HfmJdBc
Residual FM does not tell the whole story. Thisplot shows the noise performance of the internalLO in the 8901 at 1 GHz. Two other noise plotsare also shown. Although the noise at mostoffsets is different for these plots, the residualFM from 0.3 kHz to 3 kHz is roughly the samefor all three sources.
744
·70
-80
-ClO
-100
-110
-120
RESIDURL FMRII three curves have 2.5 Hz
ReSidual FM,o.3 to 3.0KHz
""- " 8'1D1 L.Q,[gl 11:;1-~... '" ....... --
1"- -~3KHZ 2 '""-b.ftef../2. J(.(fm)fm dfm 1'-O.3KHz I II O.3KHz 3,OKHz
0,1 1.0 10fm(KHz)
Options to the HP 8901B Modulation Analyzerhave been designed to allow using thisinstrument to measure et(f) at adjacent channeloffsets directly as well as residual modulationmeasurements. The 8901 is a heterodynereceiver which down-converts signals up to 1300MHz to one of two IFs: 455 kHz or 1.5 MHz.This IF signal is then fed to AM, FM, or leveldetectors to allow demodulation and analysis.
SIMPLIFIED BLOCK DIRGRRMSQ01B MODULRTION ANRLYZER
AM DEMOO
745
LOCAL 05CILLRTOR DVM
SIMPLIFIED BLOCK DIRGRRM8Q01B MODULRTION RNRLYZERWITH NOISE MERSUREMENT OPTION
The 8901B options for measuring adjacentchannel power and noise consist of an RFswitch for an external LO input, bufferamplifiers in the IF to allow splitting the IFsignal, added narrowband filters, variable gainprecision amplifiers, and an RMS detector. TheRF switch allows use of an external LO whenlower noise and/or finer resolution frequencysettability is required than is provided by theinternal local oscillator. The narrowband filtersprovide rejection of the large down-convertedsignal when noise near this signal is beingmeasured. The precision variable-gain amplifiersprovide a method for making calibratedmeasurements with the rms detector over awide dynamic range.
IN
RF SWITCH
8Cl018 MOD ANRLYZER
-c:::J--BPF #:L
746
www.HPARCHIVE.com
-
•
FILTER ASSEMBLY
TOI.F.AMPS
747
A MPLIFIER/DET ASSYFROM I.F. FILTERS
455 KHzFILTERS
RMS DETEClOR
Anyone of three filter combinations areavailable with the 8901B for making noisemeasurements. One combination is equippedwith two standard filters which allowrecommended adjacent channel powermeasurements to be made for any of threechannel spacings: 25 kHz, 20 kHz, and 12.5 kHz.Additionally, a narrow filter of approximately2.5 kHz bandwidth may be substituted for eitherof the standard channel filters. This narrowfilter allows measurements as close as 5 kHz tothe carrier, and is useful for making generalpurpose noise measurements. The effective noisebandwidth of approximately 34 dB relative to a1 Hz bandwidth can be compensated for by theinstrument firmware, permitting direct readingsin dBc/Hz. Amplifiers are provided tocompensate for insertion loss in these filters,and to keep system SIN high.
95 dB of auto-ranging IF gain, variable in 5 dBsteps, is provided between the narrowband IFfilters and the rms detector. The detector canoperate accurately over approximately 20 dB ofdynamic range. Summing the 95 dB of availablevariable gain, 20 dB of detector range, and 34 dBof noise bandwidth yields 149 dB formeasurement range of noise in a 1 Hzbandwidth relative to the carrier power level.Additional filtering is provided between theamplifiers and detector to reject broadbandamplifier noise and to provide additional carrierrejection.
aQ01B DOWN - CONVERSION
II\P CONTROL
DOWNCONVERTED
NOISE
IMAGENOISE
TO 8'101 VOLTMETER
748
im
~~455 kHz!.ARROW8At~D
When the La of the 8901B is tuned to a givenfrequency, it converts signals separated from theLa by the IF frequency into the IF passband.This can be thought of as two IF filters at theLa frequency + and - the IF frequencies. Thesignal that appears at the detector is the sum ofthe signals in these two passbands. If the La istuned such that the desired signal is in one ofthese passbands, the peak signal power isdetected. If the La is then moved 20 kHz, noise20 kHz to one side of the signal being measuredis then detected by the IF detector.
749
www.HPARCHIVE.com
-
MEASUREMEN T/A LG 0 RITHM
To measure noise, first set the instrument to aknown state, such as automatic operation infrequency measuring mode. Feed a signal to theinput port, and allow the 8901B to automaticallytune to this signal. Alternately, the frequency ofthe 8901B can be manually tuned to the desiredsignal. Use the special function keys to enterthe noise measurement mode and to select theappropriate narrowband IF filter. The instrumentwill read the peak signal strength at the IF rmsdetector in volts. The amplifiers and inputattenuator will autorange to keep this levelbetween 1.5 and 3.0 Volts. This number is thenused as the reference level for making relativenoise measurements. The LO can now be movedto the offset of interest by entering this offset inkHz and pushing the kHz UP key. Enter thespecial function for the filter to be used tomeasure noise. The IF amplifiers will rangeautomatically to maintain the power in the filterbandwidth in the linear range of the detector.The display will show this level in decibelsrelative to the peak signal power reference levelpreviously measured. A separate special functionmay be used to display this level in dBc/Hz ifthe narrowband filter is used.
Command
Ruro/Fteque ncy
24.0 Spcl
24.1/3/5 Spcl
KHz UP Ot KHz Down
24.2/41S/7 Spcl
750
www.HPARCHIVE.com
Function
Sef insftument fo know n sfate
Ini ti al izes instrum~nt fot noise measutement
Measutes Signal refetence powe t Wi fh one off htee Ii lfe,s
Tune Bqo1B L.O. fo desited offsef
Displays noise in one of fhtee filletS BWs
Ot in 1Hz BW telafive fo SiGnal in dB
-
USING AN EXTERNRL L.a.
• Especially Useful Rbove 100MHz
• RII ows Better Frequency Settabil ityResolution
• Improves Measurement NOise Floor
• Can Extend Range to 2.1 GHz
751
The noise measurement option for the 8901Bincludes a programmable RF switch for selectingan external source to be used as the localoscillator. This switch, LO amplifiers, andinternal mixer will all operate to 2.1 GHz, thusoffering one method of extending the 8901Bmeasurement range above the 1300 MHz limit ofthe internal LO. The 8901B internal LOfrequency resolution can be as coarse as 2 kHzat 1300 MHz, decreasing linearly with frequency.When using the 2.5 kHz wide filters formeasuring phase noise, frequency settingresolution better than 1 kHz is desirable to peakthe signal in the IF passband. An external LOwith fine frequency resolution allows this forhigher measurement frequencies.
To use an external LO, select 23.1 SPCL prior toentering noise measurement mode. Tune theexternal LO 455 kHz above the desired signalfrequency. Set the reference level as before,then increase the frequency of the LO by thedesired offset frequency.
www.HPARCHIVE.com
-
EXTERNAL L.O. NOISE
im(kHz) 1
"- 5~Hz'" i'..' ,.,,"""., .............
'-- ""-""- ................
~ .........·• "'-..... ...............·
The phase noise on the La internal to the8901B limits the instruments dynamic rangebelow the capability allowed by the othercritical components in the measurement path.Use of an external low noise La and carefulattention to the RF attenuator and IF detectorsettings during reference level setting allowsmeasurement floors approaching -150 dBc to beobtained. Since the external La is amplified andused as the large signal input to the mixer, AMnoise is suppressed. There are manyhigh-quality signal sources available with UHFnoise performance that is superior to the 8901La noise performance at offsets addressed bythis technique, as is illustrated in this plot.
.BG01 LO.
.8656-qO
·100
£ (fm)@!I -110500 MHzdBc ·120
-130
·1'10
-150
752
.86oW• NewS.G.
.866213
10 100
SWITCH CONTROLPOWER SUPPLY
Extending this measurement technique tomicrowave frequencies simply requires amicrowave mixer and a stable, microwavesource with phase noise lower than thesource-under-test. The difference frequency canbe tuned to the frequency range of the 8901Band analyzed directly. These components areincluded in the 11793A microwavedown-converter. When coupled with amicrowave La such as the 8673A, the 11793Aextends the range of all 8901B measurementcapabilities to 26.5 GHz.
117q3A BLOCK DIAGRAM
LOIN
-L]r-__-iTOk8673AFREGI OFFSET
L..----------------INPUT FROtABG02A
753
INTERNAL L.O. NOISE
When down-converting a microwave signal fornoise analysis by the 8901B, it's best to keep thedifference frequency low (below 10 MHz) to takeadvantage of the lower noise of the 8901B's Laat the lower frequencies. Frequency settingresolution also becomes finer as the 8901 La istuned to lower frequencies. Keeping thedifference frequency low also minimizes in-bandmixer spurious, and makes the filtering job easier.
-100
-110
£ ·120
(10kHz) -130dBc
-140
-150
8673L
HGHz~ 1
....-".., 8901B-LO.-
~,,/
i,MHz 10
754
www.HPARCHIVE.com
100 1000
-
•
MEASURING SIGNAL SSBNOISE WITH THE SQ01B
• Fast/Easy
• Moderat~ Cost
• EasilyRJtomated
• Standard RQjacent Channel Pow~rTests BUilt-In
• Measures Total SSB NOisePower,Not Just RM or PM
755
The noise measurement options on the 8901Ballow extremely easy and fast measurement ofSSB noise on RF and microwave signals. Themeasurement is relatively insensitive to manycommon pitfalls encountered in making noisemeasurements. There is no need to phase-locksignals. The data is displayed conveniently,either as -dBc in a 1 Hz bandwidth, or as totalpower in a standard channel test filter. Noise ata given offset is measured and displayedreal-time, allowing adjustment of the devicebeing tested. Although there are times when itis advantageous to separate the AM and PMcomponents of noise, total SSB noise, regardlessof type, is what is usually important whenanalyzing the limitations in communicationssystems. This is precisely what the 8901B noisemeasurement option measures.
www.HPARCHIVE.com
-
There are several limitations involved in thistechnique which can render it inapplicable forsome test situations. The bandwidth of the IFfilters prevents this instrument from measuringnoise closer than 5 kHz to the carrier before IFfeedthrough begins to saturate the detector.Dynamic range limitations in the RF and IFpaths prevent measurement of noise below-150 dBc even with a noiseless LO. Since theLO and signal under test are not locked to eachother, both must have good frequency stability.Otherwise, the offset frequency being measuredwill drift as the signal drifts, and results will beinaccurate. (However, many non-phase lockedsources are stable enough with adequatewarm-up.) This technique rejects neither AMnor PM noise, thus the measured result is thesum of both of these types of noise. If neithertype of noise is dominant, and only one type isof interest, a different technique must be used.Also, if noise at the image frequency (910 kHzaway) is comparable to noise at the offset ofinterest, the image noise will sum into the IFand the result will be affected. This typicallyhappens at larger offsets where white noisepredominates, and the image noise adds 3 dB tothe result. Finally, any discrete signals, such asspurious sidebands, may be interpretedinaccurately by this method. This is because thefilter noise bandwidth correction factor (34 dBfor 2.5 kHz filter) applies only to evenlydistributed noise throughout the filter passband.Discrete signals therefore will be measured as34 dB lower than the actual value.
SQ01B LIMITATIONS• Offsets must be greater than 5KHz
• Dynamic range limited to less than 150dB
• Source must be stable to within 1KHz
• lOtaI nOise measured, not Just phase nOise
• NOise at image is also measured
• Poor identification and resolution of spurious
756
www.HPARCHIVE.com
-
MEASUAEMENTACCUARCY
:l:'3'l-dB±0.3dBtmlSE BRNDWIDTHCORRECTION
;;;:Q5d6±O.2d6$'.vlTCHED I.F. GRIN
:':'20dB±O.2 10 ±idSRMS DETEL10R RANGE
TYPICRL UNCERTRINTY
-
The equivalent input noise figure of the 8901Bis approximately 24 dB. This is limited by theinsertion losses encountered throughout the RFand IF paths, the maximum signal level thatmay be applied to the RF mixer, and theequivalent noise in the amplifiers. This requiresthat input signal levels should be greater thanodBm to take full advantage of the dynamicrange available in the IF. The RF inputattenuator should be manually adjusted (throughspecial function keys) to ensure that themaximum signal level is being applied to the IF(no IF gain when setting a reference level). Thisis not normally necessary when making astandard adjacent channel power test, but isdesirable when measuring phase noise. The8901B will display an error message when thereference signal level is beyond the upper limitof the rms detector (error 17). The inputattenuator should be adjusted just below thelevel needed to cause this error message.
Once a reference level has been set, thedynamic range can be quickly checked byremoving the signal. The instrument shouldthen display greater than 145 dB relative to thesignal in a 1 Hz bandwidth. This number willthen be approximately the measurement floor.As noise being measured approaches this level,this noise will sum into the result and willcause errors as this floor is approached. Thischeck will not take into account the noise ofthe La, as this noise is not down-converted intothe IF when no input signal is present.
SIMPLIFIED BLOCK DIAGRAM8Q01B MODULRTION RNRLYZER
WITH NOISE MERSUREMENT OPTION
8"1016 MOD ANALYZER
IN
746
www.HPARCHIVE.com
-
EFFECT OF SPURIOUS
-60
-70
-80
-qO
-100
-110
......!t FILTER A1SSBANlJ -3+d BeF' Eq. NBW
READING = ..q4dBe 1Hz
"- SPUR (gI -60dBe
r---"",,~_
2.5 kHz 1= ~NOISE -l00dBe/Hz
45S kHz
759
The noise measurement option of the 8901Breceiver measures total power passed throughthe IF filter to the rms detector. When using thenoise bandwidth correction factor for this filter,the assumption is made that the noise in thisfilter is evenly distributed, and therefore thenoise power in a 1 Hz bandwidth at the centerof the filter is equal to the total power in thefilter passband minus the noise bandwidthcorrection factor. This is not always the case.Phase noise from 5 kHz to 100 kHz from a UHFsignal will typically vary from l/f to l/r.This causes very little measurement error if thefilter bandwidth is small compared to the offsetbeing measured. Even if the noise changes froma 9 dB/octave slope to 6 dB/octave in the filterpassband, the error introduced will be less than0.15 dB.
Of more concern is the effect of a noise "hump"or spurious sidebands. If a spur falls into thefilter passband that is well above the noisepower in the filter passband, the rms detectorwill accurately measure the peak spur power,i.e., the noise bandwidth correction factor is nolonger valid. Therefore, when measuring aknown discrete spurious signal that is wellabove the noise power in the IF filterbandwidth, the noise bandwidth correctionfactor should not be applied. Many times spurscan be identified, as they will cause the resultto increase suddenly as the frequency offset isincreased. In the absence of spurs, noisereadings will generally decrease smoothly orremain constant with increasing offset. Spursless than 30 dB above noise in a 1 Hz bandwidthwill not be detected.
www.HPARCHIVE.com
-
In summary, the measurement floor of thistechnique is limited by the noise of the LO andby the equivalent input noise of the 8901B. Theinternal LO of the 8901B is a good choice atlower center frequencies. However, an externalLO will improve the measurement floor at UHF.The input level and input attenuators should becarefully adjusted to maximize dynamic rangewhen setting the reference level.
The accuracy of the IF detection scheme variesfrom typically 0.5 dB when measuring highernoise levels to approximately 2 dB at the lowerlevels.
Attention must be paid to the effect of discretespurious and very steep or non-monotonic noiseslopes. The noise bandwidth correction may notapply in some of these cases. Spurs less than30 dB above noise in a 1 Hz bandwidth will notbe seen at all.
oise at the image frequency (910 kHz away)also sums into the IF filter. If white noise isbeing measured, this will add 3 dB to the result.Also spurious signals at the image frequencywill be detected as if they were at the offset ofinterest.
Finally, the accuracy of the offset frequencybeing measured is only as good as the accuracyof the difference frequency between the LO andsource under test. Both must remain stableduring the time between reference setting andmeasuring at the desired offset.
ERROR SUMMARY
• L.a. Noise
• 8Q01B EqUivalent Input NOise Figure
• Rccuracy and Linearity of Filter BWIf Gai n, RMS Detector
• Noise Slope and SpuriOUS Signals
• Image NOise
• Frequency Settability and Stabilityof L.a. and Source
760
www.HPARCHIVE.com
-
oAPPLICATIONS
• Transmitter testi ng
• Production testing.of sources
• Signal generat.or calibration .orcharacterization
• R&D evaluation
AUTOMATEDNOISE MEASUREMENT
With this measurement capability, we can nowmake adjacent channel power measurements ontransmitters. Specifically, CEPT standards can bemet with the required filters. Phase noise canbe easily checked on the production line fordifferent types of sources. Because of HP-IB, theag01B can be put into an automatic system.Because of the capability to make themeasurement real time, the ag01B can be usedfor calibration and design evaluation.
761
Here is a block diagram for making an adjacentchannel power measurement or a phase noisemeasurement with an external LO. If theinternal LO of the ag01B is used, then there isno special connection. To automate the system acontroller is connected through HP-IB.
CONTROLLER
HP-IB
RFOUTPUT LD.
762
www.HPARCHIVE.com
-
1000-1066
r/i~ HEWLETT~~ PACKARD
www.HPARCHIVE.com
Printed in U.S.A.
doc-115-PAGE0102Adoc-115-PAGE0103Adoc-115-PAGE0103Bdoc-115-PAGE0104Adoc-115-PAGE0104Bdoc-115-PAGE0105Adoc-115-PAGE0105Bdoc-115-PAGE0106Adoc-115-PAGE0106Bdoc-115-PAGE9894Cdoc-115-PAGE9894Ddoc-115-PAGE9895Cdoc-115-PAGE9895Ddoc-115-PAGE9896Cdoc-115-PAGE9896Ddoc-115-PAGE9897Cdoc-115-PAGE9897Ddoc-115-PAGE9898Cdoc-115-PAGE9898D