JITTER

Slide 1

JITTERIMPACT ON CLOCK DISTRIBUTION IN LHC EXPERIMENTS

S. BARON - TWEPP 2012AIM OF this talkUnderstand the contributions of all the systems to the bunch clock jitterRF system Long distance transmission Digital electronics within experiments (TTC)Compare orders of magnitude ofJitter of the beamJitter of the Bunch ClockPut them in perspective to jitter sensitivity of systems in LHC detectorsMany thanks to Themistoklis Mastoridis (BE/RF), Philippe Baudrenghien (BE/RF)and Jan Troska (PH/ESE)Sophie Baron - CERN2TWEPP2012 - Jitter On Clock Distribution - 19.09.12Two words about JitterQuantifying JitterJitter TypesJitter sensitivity of electronics in LHC experimentsTraditional ElectronicsLHC detector exotic susceptibilityJitter sourcesBeam JitterBunch Clock JitterRF systemLong distance transmissionElectronic components within experimentsConclusion

OUTLINESophie Baron - CERN3TWEPP2012 - Jitter On Clock Distribution - 19.09.12Two WORDS ABOUT JITTERMeasuring, Quantifying and Representing JitterJitter TypesSophie Baron - CERN4TWEPP2012 - Jitter On Clock Distribution - 19.09.12Time Domain (oscilloscope or SDA traditional views)Probability Density Function (PDF)Frequency Domain (spectrum, phase noise)

= std deviation = rms jitterpkpk jitterTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONMeasuring, Quantifying, Representing JitterJitter TypesSophie Baron - CERN5TWEPP2012 - Jitter On Clock Distribution - 19.09.12Cycle-to-cycle jitter: short term variation in clock period between adjacent clock cycles. Contains highest frequency components of jitter. Period jitter:short term variation in clock period over all measured clock cycles, compared to the average clock period. Contains relatively high frequency components of jitter. TIE jitter (Time Interval Error or accumulated/phase Jitter):Actual deviation from the ideal clock period over all clock periods.includes jitter at all modulation frequencies. Calculation on its pdf gives many information on jitter sources.Skew jitter:Deviation from reference signal (source, adjacent channel) over all clock periods.Phase noise:Plots the Phase Spectral Density (S(F)=|F((t)|) in dBc/Hz, over a frequency domain of the offset frequency f=f-fc. Very useful to have a full picture of the jitter contributions. Full integration of the plot gives the phase jitter.Wander:jitter less than 10Hz from the carrier. Any of the above jitters, when very slow is considered as wander.Pi=Pi-Pi+1Pi+1= Pi+1-Pi+2Pi Pi+1 Pi+2 TIE(n)=T(n)-nT0TIE(n+1)=T(n+1)-(n+1)T0S(n-1)=T(n-1)-T0(n-1)S(n)=T(n)-T0(n)S(n+1)=T(n+1)-T0(n+1)S(n+2)=T(n+2)-T0(n+2)

TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONMeasuring, Quantifying, Representing JitterJitter TypesSophie Baron - CERN6TWEPP2012 - Jitter On Clock Distribution - 19.09.12JITTER SENSITIVITY OF ELECTRONICS IN EXPERIMENTS Traditional ElectronicsLHC Detectors Exotic SusceptibilitySophie Baron - CERN7TWEPP2012 - Jitter On Clock Distribution - 19.09.12Digital Systems (flip-flops):Very sensitive to setup and hold timebasically related to pkpk cy2cy and period jitter.

PLLs:Track the slow variations of the clocks, and filter out the high frequency components. Can not deal with sudden jumps which unlock them. pkpk cy2cy jitterWander can also be a problem when it means frequency drifting out of the locking range.

ADCs: Very sensitive to timing errors as they directly convert into sampling errors, and SNR. Unregular sampling edges can distort of the shape of digitized pulses. This is thus more about pkpk cy2cy and period jitter than about TIE.

Serial Data Links:Need to combine low Bit Error Rate (BER) and good Clock Recovery for further uses.BER is related to the quality of the clockTransmitter is very sensitive to any clock jitter (because of clock multiplication).On the channel, data jitter is correlated to Duty Cycle Distortion of the clock (DCD)Receiver and CDR are highly sensitive to high frequency jitterQuality of the Clock Recovery is a trade off between low BER (requires high bandwidth) and noise rejection (requires narrow bandwidth) Serial Data Links understanding requires TIE decomposition and often frequency domain analysis

TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONTraditional ElectronicsLHC Detectors Exotic SusceptibilitySophie Baron - CERN8TWEPP2012 - Jitter On Clock Distribution - 19.09.12Event reconstruction over a huge system1000s of Bunch Clock destinations spread all over the detectorsskew jitter between all clock signals to guaranty channel-to-channel consistencyTrade off for the PLLs in the clock tree Narrow bandwidth to clean the clock as much as possibleNot too narrow bandwidth to be sure they do not drift too much from each other.stable phase between Bunch Clock and BeamLow skew jitter between clock and Bunches (RF) over a fill (obvious)Deterministic phase from fill to fill and between power cycles not obvious at all, as this requirement is almost never required by industry. Almost impossible to get from commercial components.

The Reference of our system is movingBunch Clock comes from the RF system which is not stable (ramps, blow-ups, trimms, feedback loops)The beam driven by the RF has its own jitter (bunch-to-bunch distance, phase noise wrt RF, bunch profile variation over a fill, etc)TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONTraditional ElectronicsLHC Detectors Exotic SusceptibilitySophie Baron - CERN9TWEPP2012 - Jitter On Clock Distribution - 19.09.12JITTER SOURCESBEAM JITTERBUNCH CLOCK JITTERRF System (Analog)Long Distance TransmissionTTC Electronics (Digital)Sophie Baron - CERN10TWEPP2012 - Jitter On Clock Distribution - 19.09.12Bunch position jitter The data over a turn show a peak-to-peak variation of ~7 ps. This is a single data set (no averaging involved). Each Bunch position is measured once versus 400Mhz RF.

Jitter for each bunch (Standard deviation over 73 turns) is less than 2ps rms

Courtesy of Themistoklis Mastoridis, BE/RFTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN11TWEPP2012 - Jitter On Clock Distribution - 19.09.12Bunch position jitter over a fillConsistent 0.5-1 ps rms for each bunch. Increases to about 2-3 ps during the ramp due to the accelerationAt most 6 ps peak-to-peak over a turn due to beam loading (very reproducible from turn to turn).Flat BottomMiddle of RampFlat TopStable BeamStandard Deviation (rms) of each bunch over 73 turnsMean Bunch Position over 73 turns

12ps 6ps 2ps 1.3ps 20ps 20ps 8ps 6ps Courtesy of Themistoklis Mastoridis, BE/RFTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN12TWEPP2012 - Jitter On Clock Distribution - 19.09.12Almost NO Variation of the bunch profile over a fillSlight distortion from gaussian shape because of the blow-up during rampBlow-up is a noise injected into the RF during the ramp to shake the beam and spread the particules within bunches

ProtonsIonsCourtesy of Philippe Baudrenghien, BE/RFTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERNTWEPP2012 - Jitter On Clock Distribution - 19.09.1213New Bunch Position Plot after LS1After LS1, some modulation of the cavity phase will probably be implemented to help lower the RF power requirementsThe modulation of the cavity phase changes the bunch spacing and therefore the collision point. However the 65 ps displacement is small compared to the 1.2 ns 4-sigma bunch lengthAs the filling pattern of the two rings is very similar, the phase modulations will cancel out in IP1 and IP5 and the resulting displacement of the collision vertex will be much smaller than the above 65 ps

Modulation of the cavity phase by the transient beam loading in physics for nominal conditions. 2835 bunches, 1.7E11 p/bunch, 25ns spacingCurrent Bunch position (1440 bunches). No Cavity phase modulation.Courtesy of Themistoklis Mastoridis and Philippe Baudrenghien BE/RFTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN14TWEPP2012 - Jitter On Clock Distribution - 19.09.12

Modulation of the cavity phase by the transient beam loading in physics for nominal conditions. 2835 bunches, 1.7E11 p/bunch, 25ns spacingCurrent Bunch position (1440 bunches). No Cavity phase modulation.Courtesy of Themistoklis Mastoridis and Philippe Baudrenghien BE/RFReproducible from fill to fillTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN15TWEPP2012 - Jitter On Clock Distribution - 19.09.12Many loops are managed by the low level loop processorBeam control loop (low frequency, < 11kHz)Phase loopSynchro loopRadial loopCavity control loops (high frequency)

Phase Noise Power Spectral Density Plot:Fully integrated phase noise 1Hz-1MHz is 3 ps rms.Keep in mind the propagation time (>50us) between RF and detectorsJitter on RF signal adds to bunch jitter at the level of detectors

Beam Control LoopsCavities Control LoopsTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN16TWEPP2012 - Jitter On Clock Distribution - 19.09.12Up to 14km of burried fiber from SR4 to ALICE, ATLAS, CMS (1m deep)The fibre length changes with core temperature by 7ppm/degC, which induces variation of the propagation delay by about 0.5ns/degC/14kmObviously very slow variation of the phase between beam and clock (wander)

5ns/full year/9km3ns per season for 14kmTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN17TWEPP2012 - Jitter On Clock Distribution - 19.09.12TxRxRFRF2TTCTTCvi/exTTCrq

RF Phase Jitter=2ps rmsTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN18TWEPP2012 - Jitter On Clock Distribution - 19.09.12TxRxRFRF2TTCTTCvi/exTTCrq

TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN19TWEPP2012 - Jitter On Clock Distribution - 19.09.12RF Phase Jitter=2ps rmsRx Phase Jitter=1.9ps rmsTxRxRFRF2TTCTTCvi/exTTCrq

TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN20TWEPP2012 - Jitter On Clock Distribution - 19.09.12RF Phase Jitter=2ps rmsRx Phase Jitter=1.9ps rmsRF2TTC Phase Jitter=10ps rms

TxRxRFRF2TTCTTCvi/exTTCrq

TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN21TWEPP2012 - Jitter On Clock Distribution - 19.09.12RF Phase Jitter=2ps rmsRx Phase Jitter=1.9ps rmsRF2TTC Phase Jitter=10ps rmsTTCex Phase Jitter=4.9ps rmsTxRxRFRF2TTCTTCvi/exTTCrq

TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN22TWEPP2012 - Jitter On Clock Distribution - 19.09.12RF Phase Jitter=2ps rmsRx Phase Jitter=1.9ps rmsRF2TTC Phase Jitter=10ps rmsTTCex Phase Jitter=4.9ps rmsTTCrq Phase Jitter=8ps rmsOFF siteTxRxRFRF2TTCTTCvi/exTTCrqJitter MeasurementTTCrq 40 MHz outputrmspkpkTIE974Cy2Cy13119Per1866TJ (BER 10-12)116RJ7PJ720Phase146mdeg1.1degPhase noise rms jitter8 ps rmsTWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN23TWEPP2012 - Jitter On Clock Distribution - 19.09.12New Electronics for TTC UpgradeTTCrq 40 MHz outputTTC PON very basic proof of conceptTTC FMC board (no PLL)GBT Serdes PrototypeMeasurement typermsPkpkrmspkpkrmspkpkrmspkpkTIE252391179418974Cy2cy Agilent242068731211013119TJ11684116RJ457PJ1511.558210720Skew368mdeg3.3deg190mdeg1.5deg146mdeg1.1degPhase Noise Jitter23.512.7158 TWO WORDSJITTER SENSITIVITYJITTER SOURCESCONCLUSIONBEAM JITTERBUNCH CLOCK JITTER RF JITTER LONG DISTANCETTC JITTERSophie Baron - CERN24TWEPP2012 - Jitter On Clock Distribution - 19.09.12conclusionSophie Baron - CERN25TWEPP2012 - Jitter On Clock Distribution - 19.09.12JITTER VictimsJITTER SourcesJitter VictimQuality CriteriaIdeal referenceJitter type sensitivityTypical requirements at the input

ElectronicsBeam RF phase noiseTemperature driftUsual RJ contributionUsual PJ contributionADCSampling qualityIntrinsiccy2cy