INTERNATIONAL TELECOMMUNICATION UNION

INTERNATIONAL TELECOMMUNICATION UNION ITU-T J.133 TELECOMMUNICATION STANDARDIZATION SECTOR OF ITU (07/2002) SERIES J: CABLE NETWORKS AND TRANSMISSION OF TELEVISION, SOUND PROGRAMME AND OTHER MULTIMEDIA SIGNALS Transport of MPEG-2 signals on packetised networks Measureent of MPEG-2 transport streas in networks ITU-T Recoendation J.133

ITU-T J-SERIES RECOMMENDATIONS CABLE NETWORKS AND TRANSMISSION OF TELEVISION, SOUND PROGRAMME AND OTHER MULTIMEDIA SIGNALS General Recoendations General specifications for analogue sound-prograe transission Perforance characteristics of analogue sound-prograe circuits Equipent and lines used for analogue sound-prograe circuits Digital encoders for analogue sound-prograe signals Digital transission of sound-prograe signals Circuits for analogue television transission Analogue television transission over etallic lines and interconnection with radio-relay links Digital transission of television signals Ancillary digital services for television transission Operational requireents and ethods for television transission Interactive systes for digital television distribution Transport of MPEG-2 signals on packetised networks Measureent of the quality of service Digital television distribution through local subscriber networks IPCableco Miscellaneous Application for Interactive Digital Television J.1 J.9 J.10 J.19 J.20 J.29 J.30 J.39 J.40 J.49 J.50 J.59 J.60 J.69 J.70 J.79 J.80 J.89 J.90 J.99 J.100 J.109 J.110 J.129 J.130 J.139 J.140 J.149 J.150 J.159 J.160 J.179 J.180 J.199 J.200 J.209 For further details, please refer to the list of ITU-T Recoendations.

ITU-T Recoendation J.133 Measureent of MPEG-2 transport streas in networks Suary MPEG-2 Transport Streas that are transitted over any real networks are exposed to certain effects caused by the network coponents which are not ideally transparent. One of the predoinant effects is the acquisition of jitter in relation to the PCR values and their position in the TS. This Recoendation specifies easureents that enable deterination of this jitter. Source ITU-T Recoendation J.133 was prepared by ITU-T Study Group 9 (2001-2004) and approved under the WTSA Resolution 1 procedure on 29 July 2002. ITU-T Rec. J.133 (07/2002) i

FOREWORD The International Telecounication Union (ITU) is the United Nations specialized agency in the field of telecounications. The ITU Telecounication Standardization Sector (ITU-T) is a peranent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recoendations on the with a view to standardizing telecounications on a worldwide basis. The World Telecounication Standardization Assebly (WTSA), which eets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recoendations on these topics. The approval of ITU-T Recoendations is covered by the procedure laid down in WTSA Resolution 1. In soe areas of inforation technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. NOTE In this Recoendation, the expression "Adinistration" is used for conciseness to indicate both a telecounication adinistration and a recognized operating agency. INTELLECTUAL PROPERTY RIGHTS ITU draws attention to the possibility that the practice or ipleentation of this Recoendation ay involve the use of a claied Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claied Intellectual Property Rights, whether asserted by ITU ebers or others outside of the Recoendation developent process. As of the date of approval of this Recoendation, ITU had not received notice of intellectual property, protected by patents, which ay be required to ipleent this Recoendation. However, ipleentors are cautioned that this ay not represent the latest inforation and are therefore strongly urged to consult the TSB patent database. ITU 2002 All rights reserved. No part of this publication ay be reproduced, by any eans whatsoever, without the prior written perission of ITU. ii ITU-T Rec. J.133 (07/2002)

CONTENTS Page 1 Scope... 1 2 References... 1 2.1 Norative references... 1 2.2 Inforative references... 1 3 Abbreviations and acronys... 1 4 Syste clock and PCR easureents... 1 4.1 Reference odel for syste clock and PCR easureents... 1 4.2 Measureent descriptions... 3 4.3 Progra Clock Reference Frequency Offset (PCR_FO)... 3 4.4 Progra Clock Reference Drift Rate (PCR_DR)... 4 4.5 Progra Clock Reference Overall Jitter (PCR_OJ)... 4 4.6 Progra Clock Reference Accuracy (PCR_AC)... 5 Appendix I PCR-related easureents... 6 I.1 Introduction... 6 I.2 Liits... 6 I.3 Equations... 7 I.4 Mask... 8 I.5 Break frequencies... 9 I.6 Further iplicit liitations... 10 I.7 Measureent procedures... 11 I.7.1 PCR_Accuracy (PCR_AC)... 13 I.7.2 PCR_drift_rate (PCR_DR)... 14 I.7.3 PCR_frequency_offset (PCR_FO)... 14 I.7.4 PCR_overall_jitter easureent... 15 I.8 Considerations on perforing PCR easureents... 15 I.9 Choice of filters in PCR easureent... 17 I.9.1 Why is there a choice?... 17 I.9.2 Higher dearcation frequencies... 18 I.9.3 Lower dearcation frequencies... 18 ITU-T Rec. J.133 (07/2002) iii

ITU-T Recoendation J.133 Measureent of MPEG-2 transport streas in networks 1 Scope A MPEG-2 Transport Strea that is transitted over any real network is exposed to certain effects caused by the network coponents which are not ideally transparent. One of the predoinant effects is the acquisition of jitter in relation to the PCR values and their position in the TS. The four easureent paraeters defined in the following describe the various jitter coponents which can be differentiated by dearcation frequencies. 2 References The following ITU-T Recoendations and other references contain provisions which, through reference in this text, constitute provisions of this Recoendation. At the tie of publication, the editions indicated were valid. All Recoendations and other references are subject to revision; users of this Recoendation are therefore encouraged to investigate the possibility of applying the ost recent edition of the Recoendations and other references listed below. A list of the currently valid ITU-T Recoendations is regularly published. The reference to a docuent within this Recoendation does not give it, as a stand-alone docuent, the status of a Recoendation. 2.1 Norative references [1] ITU-T Recoendation H.222.0 (2000) ISO/IEC 13818-1: 2000, Inforation technology Generic coding of oving pictures and associated audio inforation: Systes. [2] ETSI TR 101 290 V1.2.1 (2001), Digital Video Broadcasting (DVB); Measureent guidelines for DVB systes. 2.2 Inforative references [3] ISO/IEC 13818-9:1996, Inforation technology Generic coding of oving pictures and associated audio inforation Part 9: Extension for real tie interface for systes decoders. 3 Abbreviations and acronys This Recoendation uses the following abbreviation: PCR Progra Clock Reference 4 Syste clock and PCR easureents 4.1 Reference odel for syste clock and PCR easureents This clause presents a reference odel (see Figure 1) for any source of a Transport Strea (TS) concerning the generation of PCR values and delivery delays. It odels all the tiing effects visible at the TS interface point. It is not intended to represent all the echaniss by which these tiing effects could arise in real systes. ITU-T Rec. J.133 (07/2002) 1

PCR inaccuracy source M p,i + D + J i N p,i B Delivery tiing delay C A PCR counter J.133_F01 Reference clock f = 27 MHz + f dev (t) Figure 1/J.133 Reference odel Reference points are indicated by dashed lines. This is a odel of an encoder/ultiplexer (up to reference point B) and a physical delivery echanis or counications network (between reference points B and C). The coponents of the odel to the left of reference point B are specific to a single PCR packet identifier (PID). The coponents of the odel to the right of reference point B relate to the whole Transport Strea. Measuring equipent can usually only access the TS at reference point C. The odel consists of a syste clock frequency oscillator with a noinal frequency of 27 MHz, but whose actual frequency deviates fro this by a function f dev (p,t). This function depends on the tie (t) and is specific to a single PCR PID (p). The "Frequency Offset PCR_FO" easures the value of f dev (p,t). The "Drift Rate PCR_DR" is the rate of change with tie of f dev (p,t). The syste clock frequency oscillator drives a PCR counter which generates an idealized PCR count, N p,i. p refers to the specific PCR PID p, and i refers to the bit position in the Transport Strea. To this is added a value fro a PCR inaccuracy source, M p,i, to create the PCR value seen in the strea, P p,i. The siple relationship between these values is: P p, i = N p, i + M p, i (1) M p,i represents the "Accuracy PCR_AC". The physical delivery echanis or counications network beyond point B introduces a variable delay between the departure tie T i and the arrival tie U i of bits: U T = D + J i i In the case of a PCR, U i is the tie of arrival of the last bit of the last byte containing the PCR base (see 2.4.3.5 of ITU-T Rec. H.222.0 ISO/IEC13818-1 [1]). D is a constant representing the ean delay through the counications network. J i represents the jitter in the network delay and its ean value over all tie is defined to be zero. J i + M p,i is easured as the "Overall Jitter PCR_OJ". In the coon case where the Transport Strea is constant bitrate, at reference point B, the Transport Strea is being transitted at a constant bitrate R no. It is iportant to note that in this reference odel this bitrate is accurate and constant; there is no error contribution fro varying bitrate. This gives us an additional equation for the departure tie of packets: i (2) 2 ITU-T Rec. J.133 (07/2002)

T i = T 0 + Rno (3) T 0 is a constant representing the tie of departure of the zeroth bit. Cobining equations (2) and (3) we have for the arrival tie: i U + i i = T 0 + + D Ji Rno (4) 4.2 Measureent descriptions The following easureents require a dearcation frequency for deliiting the range of drift rate and jitter frequencies of the tiing variations of PCRs and/or TSs. The dearcation frequency used should be chosen fro the Table 1 and indicated with the easureent results. A description of the derivation of the dearcation frequencies is included in Appendix I. Table 1/J.133 Profile Dearcation Coents frequency MGF1 10 Hz This profile is provided to give the total coverage of frequency coponents included in the tiing ipairents of PCR-related easureents. This profile provides the ost accurate results in accordance with the liits specified in 2.4.2.1 of ITU-T Rec. H.222.0 ISO/IEC 13818-1. If jitter or drift-rate easureents are found out of specification when using other profiles, it is suggested to use this one for better accuracy. MGF2 100 Hz This profile is accounting for interediate benefits between the profiles MGF1 and MGF3, by giving reasonable easureent response as well as reasonable account for low frequency coponents of the tiing ipairents. MGF3 1 Hz This profile provides faster easureent response by taking in account only the highest frequency coponents of the tiing ipairents. This profile is expected to be sufficient in any applications. MGF4 Manufacturerdefined This profile will provide any benefit that the anufacturer ay consider as useful when it is designed and ipleented in a easureent instruent. The dearcation frequency has to be supplied with the easureent result. Optionally any other data that the anufacturer ay consider to be relevant ay be supplied. For testing against ISO/IEC 13818-9 (±25 s jitter liit) a dearcation frequency of 2 Hz is required. A filter for such dearcation ay be ipleented under this MGF4 profile. 4.3 Progra Clock Reference Frequency Offset (PCR_FO) Definition PCR_FO is defined as the difference between the progra clock frequency and the noinal clock Frequency (easured against a reference which is neither PCR- nor TS-derived). ITU-T Rec. J.133 (07/2002) 3

The units for the paraeter PCR_FO should be in Hz according to: Measured Frequency Noinal Frequency or in pp expressed as: Measured Frequency (in Hz) Noinal Frequency (in Hz) Noinal Frequency (in MHz) Purpose The original frequency of the clock used in the digital video forat before copression (progra clock) is transitted to the final receiver in for of nuerical values in the PCR fields. The tolerance as specified by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] is ±810 Hz or ±30 pp. Interface For exaple, at interface G in Figure I.8. Method Refer to Appendix I for a description of a easureent ethod. 4.4 Progra Clock Reference Drift Rate (PCR_DR) Definition PCR_DR is defined as the first derivative of the frequency and is easured on the low frequency coponents of the difference between the progra clock frequency and the noinal clock frequency (easured against a reference which is not PCR derived, neither TS derived). The forat of the paraeter PCR_DR should be in Hz/s (@ 27 MHz) or pp/hour. Purpose The easureent is designed to verify that the frequency drift, if any, of the progra clock frequency is below the liits set by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1]. This liit is effective only for the low frequency coponents of the variations as indicated by the dearcation frequency described in Appendix I. The tolerance as specified by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] is ±75 Hz/s @ 27 MHz or ±10 pp/hour. Interface For exaple, at interface H in Figure I.8. Method Refer to Appendix I for a description of a easureent ethod. 4.5 Progra Clock Reference Overall Jitter (PCR_OJ) Definition PCR_OJ is defined as the instantaneous easureent of the high frequency coponents of the difference between when a PCR should have arrived at a easureent point (based upon previous PCR values, its own value and a reference which is neither PCR- nor TS-derived) and when it did arrive. The forat of the paraeter PCR_OJ should be in nanoseconds. 4 ITU-T Rec. J.133 (07/2002)

Purpose The PCR_OJ easureent is designed to account for all cuulative errors affecting the PCR values during progra strea generation, ultiplexing, transission, etc. All these effects appear as jitter at the receiver but they are a cobination of PCR inaccuracies and jitter in the transission. This value can be copared against the axiu error specification by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] for PCR Accuracy of ±500 ns only if the jitter in the transission is assued to be zero. Interface For exaple, at interface J in Figure I.8. Method Refer to Appendix I for a description of a easureent ethod. 4.6 Progra Clock Reference Accuracy (PCR_AC) Definition The accuracy of the PCR values PCR_AC is defined as the difference between the actual PCR value and the value it should have in the TS represented by the byte index for its actual position. This can be calculated for constant bitrate TS; the easureent ay NOT produce eaningful results in variable bitrate TS. The units for the paraeter PCR_AC should be in nanoseconds. Purpose This easureent is designed to indicate the total error included in the PCR value with respect to its position in the TS. The tolerance as specified by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] is ±500 ns. This easureent is considered to be valid for both real tie and off-line easureents. Interface For exaple, at interface E in Figure I.6. Method Refer to Appendix I for a description of a easureent ethod. NOTE PCR Accuracy is defined by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1]: "A tolerance is specified for the PCR values. The PCR tolerance is defined as the axiu inaccuracy allowed received PCRs. This inaccuracy ay be due to iprecision in the PCR values or to PCR odification during re-ultiplexing. It does not include errors in packet arrival tie due to network jitter or other causes". ITU-T Rec. J.133 (07/2002) 5

6 ITU-T Rec. J.133 (07/2002) Appendix I PCR-related easureents This appendix provides background inforation on the concept of PCR-related easureents and the reasoning behind the definition of the paraeters. The ai is to gather the inforation which enables different ipleentations of PCR-related easureents to show consistent and coparable results for the sae Transport Strea. I.1 Introduction Recovering the 27 MHz clock at the decoder side of a digital TV transission syste is necessary to re-create the video signal. To allow recovery of the clock, the PCR values are sent within the Transport Strea. It is required that the PCR values are correct at the point of origin and not distorted in the transission chain to the point of creating probles in the process of decoding the copressed signals. Measuring the accuracy of the PCR values and the jitter accuulated on the when transitted in a Transport Strea is necessary to assure the confidence of decodability of such strea. As jitter and drift rate are iportant paraeters for the overall process, a clear definition is needed for what is understood as PCR jitter and a guidance to its easureent ethod. I.2 Liits Fro the specifications set in ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1], it is possible to define a liit ask for the frequency deviation fro the noinal 27 MHz. Frequency offset: the difference between the actual value and the noinal frequency of the clock (27 MHz). The liit is set to ±810 Hz. Converting this value into relative or noralized units results in 810/27 10 6 = 30 10 6. This eans that the frequency of the clock at any oent should be the noinal ±0.003 %, or the noinal ±30 pp. Rating the liit of the frequency offset as relative has the advantage of obtaining a liit valid for any value of frequency for a reference clock used to synthesize the noinal clock of 27 MHz. For exaple, the frequency error in Hz of a 270 MHz serial clock derived fro the 27 MHz syste clock can be divided, or noralized, by 270 MHz to deterine if the frequency offset is within 30 pp. Frequency rate of change, or frequency drift rate: the "speed" at which the frequency of a clock varies with tie. In other words it is the first derivative of the frequency with respect to tie or the second derivative of phase with respect to tie. The liit is set to 75 illihertz (Hz) per second for the 27 MHz clock. It can be converted into relative liit by dividing by 27 MHz which produces a result of (75 10 3 ) / (27 10 6 ) = 2.777... 10 9 /s. It eans that the axiu rate of change allowed for the clock frequency is ±0.000 000 277 7...%/s of the noinal value, or ±0.00277...pp/s of the noinal, or ±2.77...ppb/s of the noinal value of the syste clock frequency. (Note that a billion is taken here as 10 9, in any countries a billion is represented as 10 12 ). This result can also be presented as 0.001 %/hour, or as being 10 pp/h. 27 000 000 810 syste_clock_frequency 27 000 000 + 810 @ 27 MHz Frequency tolerance = ±30 10 6 @ 1 Hz (I-1) Rate of change of syste_clock_frequency = 75 10 3 Hz/s @ 27 MHz Drift tolerance = ±2.7778 10 9 /s @ 1 Hz (I-2)

Phase tolerance = ±500 10 9 s (I-3) This represents the axiu error of a PCR value with respect to its tie position in the Transport Strea. The axiu liit for the phase represented in a PCR value is ±500 ns; this value is an absolute liit at the generation of PCRs and does not include network-induced jitter. ISO/IEC 13818-9 [3] (Extension for real tie interface for systes decoders) specifies in clause 2.5 (Real-Tie Interface for Low Jitter Applications) a liit for t-jitter equal to 50 µs. Low jitter applications tolerance = 25 10 6 s (I-3b) NOTE The liits for frequency offset and drift rate are iposed for the syste clock as it is represented by the values of the corresponding PCR fields. They include the effects of the syste clock and any possible errors in the PCR calculation. The liit of 500 ns is not iposed to the syste clock, but to the accuracy representing the PCR values with respect to their position in the Transport Strea. However, the PCR errors are fully equivalent to phase and jitter errors when the PCRs are used at the decoding point to reconstruct the syste clock. I.3 Equations The wavefor of the phase odulation ay have any shape that can be analyzed as a coposition of sinusoidal wavefors of various aplitudes and phases. Also the clock ay be a pulsed signal. In this case the forulas below apply to the fundaental coponent of such periodic signal. For exaple, the equation for a sinusoidal clock with sinusoidal phase odulation can be written as: where: c Fclk [ ] () t = A sin[ ω t + φ( t )] = A sin ω t + φ sin( ω t) c ω noinal angular frequency of the progra clock, ( = 2π 27 MHz) φ () t phase odulation function; φ p peak phase deviation in radians; c p ω ; c ω phase odulating angular frequency in units of radians/s. The instantaneous phase of the clock has two ters as: () t = ω t + φ() t = ω t + φ sin( ω t) φ i c c p (I-4) The instantaneous angular frequency of the clock is found as the first derivative of the instantaneous phase as: where: ω i i () t = dφ () t d t =ω + φ ω cos( ω t) ω (I-5) i c p instantaneous angular frequency of the clock, ω = φ i i, in units of radians/s. The frequency rate of change, or drift rate, is given by the first derivative of the angular frequency, or the second derivative of the phase as: ri 2 () t dω () t d t = φ ω sin( ω t) i p = (I-6) where: r i instantaneous rate of change of the clock, r φ, in units of radians/s 2. i = i ITU-T Rec. J.133 (07/2002) 7

I.4 Mask A liit ask can be derived as a group of functions representing the liit specifications. Fro the instantaneous phase equation (I-4) it can be seen that the axiu peak value of phase odulation is φ p which can be copared to the liit set by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1]. The phase equation ay be found as: where: 9 p c ax = φ = ω T = 2π 27 MHz 500 10 seconds 84.823 radians (I-7) T ax axiu tie error of clock edge = 500 10 9 s Fro the instantaneous angular frequency equation (I-5) it can be seen that the axiu peak value of angular frequency offset is given by φ p ω, which can be copared to the liit set by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] of 810 Hz. The axiu angular frequency deviation fro the noinal is: φ p ω = 2π 810 radian/s By dividing by ω, the frequency equation for peak phase error as a function of odulation frequency ay be found as: 2π 810 φ p = (I-8) ω Fro the instantaneous drift rate equation (I-6) it can be seen that the axiu peak value of 2 angular frequency drift rate is φ p ω, which can be copared to the liit set by ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] of 75 Hz/s. 2 φ p ω 2 = 2 π 0.075 radian/s By dividing by ω, the drift rate equation for peak phase error as a function of odulation frequency ay be found as: 2 p = 2π 0.075 ω All three equations ay be noralized by dividing by 2π 27 MHz. The phase equation becoes: 2 φ (I-9) φp 84.823 9 Tax = = = 500 10 (seconds) (I-7a) 6 6 2π 27 10 2π 27 10 The frequency equation becoes: The drift rate equation becoes: φ 6 p 2π 810 30 10 Tf ( ) s 6 6 ω = = = (I-8a) 2π 27 10 2π 27 10 ω ω φ 9 p 2π 0.075 2.7778 10 Tr( ) s 6 6 2 2 ω = = = (I-9a) 2π 27 10 2π 27 10 ω ω 8 ITU-T Rec. J.133 (07/2002)

The three equations (I-7a, I-8a and I-9a) can be seen in the graph of Figure I.1. T in seconds 1.E+01 1.E+00 1.E 01 1.E 02 1.E 03 1.E 04 Frequency-offset liit = 810 Hz Tf(ω ) = 30 10 6 / ω (s) 1.E 05 1.E 06 1.E 07 1.E 08 Frequency-drift liit = 0.075 Hz/s Tr(ω ) = 2.7778 10 9 / ω 2 (s) Phase error Liit = 500 ns T ax = 500 10 9 (s) 10 7 10 6 10 5 F 1 F 10 4 10 3 0.01 2 0.1 1 10 100 1000 Frequency of phase variation, in Hz J.133_FI-1 I.5 Break frequencies Figure I.1/J.133 PCR jitter coponents Values for two break frequencies of Figure I.1. F1 can be found by re-arranging the equations for frequency and drift rate (I-8 and I-9 respectively) and solving for the value of ω that provides the sae peak phase error: 2 φ p = 2π 810 ω and φp = 2π 0.075 ω radians ω = 5 ( 2π 0.075) ( 2π 810) = 9.2592 10 radian/s F 1 = ω 2π = 14.736 10 The break frequency F 1 is extreely low to have any practical use. When the frequency offset is to be easured, there is no need to wait about 5 days to have an averaged result appropriated to the period of such a signal. It is not considered here due to its very long-ter significance. It can be seen that the drift liit is enough for practical purposes of jitter analysis. F 2 can be found by re-arranging and solving the equations of phase and drift rate (I-7 and I-9 respectively) for the value of ω that has the sae peak phase error: 6 Hz 2 φ p = 84.823 radians and φp = 2π 0.075 ω radians ω = 0.4712 84.823 = 0.074535 radian/s F 2 = 0.074535 2π = 0.01186 Hz NOTE 1 The sae values ay be obtained by using the noralized equations (I-7a), (I-8a) and (I-9a). This break frequency (F 2 ~ 10 Hz) is recoended as the dearcation frequency for separating the easureents of jitter and drift. It has been defined as filter MGF1 in Table 1. ITU-T Rec. J.133 (07/2002) 9

This value defines the corner frequency to be used in the filters for processing the PCR data. A ask can be drawn fro the two equations used to obtain this value (phase equation I-7a and drift equation I-9a). The ask so defined is represented in Figure I.2. T in seconds 1.E 01 1.E 02 Drift liit = 0.075 Hz/s Tr(ω ) = 2.7778 10 9 / 2 ω 1.E 03 1.E 04 1.E 05 1.E 06 1.E 07 1.E 08 Region where tiing errors exceed phaseerror liit but do not exceed drift-rate liit. Drift liiting region Phase-error liit = 500 ns T ax = 500 10 9 (s) Region where tiing errors exceed drift-rate liit but do not exceed phase-error liit. 10 5 10 4 10 3 1.E 09 0.01 0.1 1 10 100 F 2 = 0.01186 Hz Dearcation frequency 5 Hz 1/2 PCR repetition rate (Miniu Nyquist liit for MPEG-2) Jitter liiting region Frequency of phase variation, in Hz 12.5 Hz 1/2 PCR repetition rate (Miniu Nyquist liit for DVB) J.133_FI-2 10 ITU-T Rec. J.133 (07/2002) Figure I.2/J.133 Mask for PCR jitter coponents It can be seen that the axiu drift of 75 Hz/s ay only be reasonably applied to jitter frequencies lower than the dearcation frequency. Above such frequency it is possible in practice to find drifts uch faster than the liit, when real PCR errors are considered. Above the dearcation frequency, the liit that applies is the absolute 500 ns for any PCR value. NOTE 2 For the Low Jitter Applications (ISO/IEC 13818-9 [3]), the ±25 µs liit yields a dearcation frequency of 1.67 Hz, to be used in place of the 10 Hz. This suggests the use of a filter with about 2 Hz break frequency when checking against this liit. This filter has been luped under MGF4 due to the long tie constant involved, which akes it to provide a very slow response for a practical ipleentation. I.6 Further iplicit liitations Fro Figure I.2 it can be seen that a practical liit is also iposed to the ability to easure jitter frequencies above a certain frequency. For PCR values inserted at the iniu rate of 100 s as per ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1], the saples arrive to the easureent instruent at a 10 Hz rate. The Nyquist value (half the sapling rate) is equal to 5 Hz. For PCR values inserted at the iniu rate of 40 s as per TR 101 154 the saples arrive to the easureent instruent at a 25 Hz rate. The Nyquist value is equal to 12.5 Hz. If higher PCR insertion rates are used in any of the above environents, the corresponding Nyquist frequency increases proportionally. This iplies that any statistics ade by the easureent instruent based in jitter spectral analysis has to easure the actual PCR rate.

Depending on the type of analysis, it is necessary to take in account that the PCR saples do not necessarily arrive at regular intervals. For any practical ipleentation, the designer ay decide what is the preferred way for ipleenting the filters: digital signal processing (DSP) techniques (infinite ipulse response (IIR) or finite ipulse response (FIR) filters), with interpolation (linear, sinx/x, etc.) or without interpolation, analogue circuitry or hybrid technology by ixing analogue and nuerical analysis, etc. It is interesting to note, however, that in ost practical cases the rate of saples will occur at very high frequencies (1000 ties higher) copared to the frequency break points of the proposed filters (MGF1 at 10 Hz). The iniu rate for PCRs is 10 Hz for general MPEG Transport Streas (25 Hz in DVB systes) and at this over-sapled PCR values the transient response shape of filters with bandwidths near 10 Hz are not significantly affected by the non-unifor rate. I.7 Measureent procedures It is possible to do jitter easureents fitting the data with a second-order curve (quadratic regression) liited by drift-rate specification (see Figure I.3). However, this is not necessary if one takes the view of creating separate easureents of jitter and frequency-offset/drift-rate based on the ore failiar ethod of sinusoidal spectral content of the tiing variations. Drift-rate liit = 10 pp/h 1 2 500 ns Jitter spec Dearcation frequency 1/2 sync-byte rate Wander region Jitter region J.133_FI-3 Figure I.3/J.133 Total spectral ask of tiing variations For jitter spectral coponents below the dearcation frequency, the peak sinusoidal coponents of the PCR tiing-error can increase proportional to the square of the period of the spectral coponent without exceeding the drift-rate liit of 10 pp/h (also, equivalently, 2.8 ppb/s and 75 Hz/s @ 27 MHz). Since the decoder phase-locked loop (PLL) and all subsequent video tiing equipent track this error, these coponents can far exceed the peak liit of 500 ns. By inverting the specification ask, a spectrally weighted easureent or easureent filter becoes apparent as follows in Figure I.4. ITU-T Rec. J.133 (07/2002) 11

0 db 2 1 Total spectral weighting response cobining jitter and wander perturbations into one easureent. Dearcation frequency 1/2 sync-byte rate Wander region Jitter region J.133_FI-4 Figure I.4/J.133 Filter by inverting the spectral ask of tiing variations This can be decoposed into two separate easureents such that the su of the jitter and driftrate easured outputs is essentially the sae as the original (see Figure I.5). 0 db 1 3 Jitter HPF 0 db Frequency-offset easureent response 1 1 2 1 1 1 Drift-rate easureent response Dearcation frequency 1/2 sync-byte rate Dearcation frequency 1/2 sync-byte rate Wander region Jitter region Wander region Jitter region J.133_FI-5 Figure I.5/J.133 Third-order high-pass filter for jitter and first-order roll-off for drift easureents Now jitter can be evaluated against given perforance liits soewhat independently of the frequency Drift-rate perforance liits. Note that in Figure I.5 the jitter high-pass filter (HPF) has a third-order response to reject the drift-rate coponents fro the easureent. Also in Figure I.5 right, the drift-rate easureent response has a first-order roll-off to reject the jitter coponents fro its output. Also shown is the preferred frequency-offset easureent response, which also rejects jitter spectral coponents. Note (see Figure I.5 right) that below the dearcation frequency, the frequency-offset is a first-derivative slope and the drift-rate is a second-derivative slope. The tiing error need not be directly easured since its tie-derivative or frequency-offset contains all that is needed to ipleent the easureent filters. This eans that only two saples to copute the tie-delta or first-past-difference of the byte arrival tie are needed. This is equivalent to easuring the instantaneous frequency offset rather than the actual tie-error of the Transport Strea and greatly siplifies the easureent with no loss in inforation. 12 ITU-T Rec. J.133 (07/2002)

I.7.1 PCR_Accuracy (PCR_AC) The result of PCR_AC is obtained at interface E of Figure I.6. 1/F No 500 ns PCR_Accuracy PCR(i ) value PCR(i ) PCR(i ) PCR(i ) + Z 1 PCR(i ) 1/x + 2nd order high-pass filter (Cascade) 500 ns Sliding window Interface E PCR_Accuracy 1/x Noinal TR PCR Byte index i + Z 1 i (i i ) 3rd order low-pass filter Data soothing Interface F Average, PCR-derived, Transport Strea rate J.133_FI-6 PCR(i ) PCR Byte index PCR_base(i ) * 300 + PCR_ext(i ) index of the byte containing the last bit of the iediately following progra_clock_reference_base field applicable to the progra being decoded. Figure I.6/J.133 PCR_Accuracy easureent The PCR_ACs that affect the PLL clock recovery for a specific progra can be easured independently of arrival-tie by extracting the change in adjacent PCR values and the nuber of bytes between PCRs as follows: K(i) = i i, bytes, [PCR(i) PCR(i 1)]/F No K(i)/TR = d(pcr_ac(i))/dt where: TR noinal Transport Strea rate, bytes/s, F No = 27 MHz; K(i) nuber of bytes between current PCR(i) and previous PCR(i 1). All high-pass and low-pass filter bandwidths as MGF1, MGF2, MGF3 and MGF4. Note that this ethod easures PCR_AC independently of arrival-tie. This can only be done for constant bitrate TS. Drift rate and frequency offset are not easured. PCR interval errors are also not easured but can be deterined indirectly fro K(i)/TR. Also note that PCR_AC is easured above the dearcation frequency to be consistent with those spectral coponents that contribute to PLL jitter. The drift coponents of PCR_AC are likely negligible copared to clock drift. The second-order high-pass filter represents a second-order HPF response to the PCR accuracy due to the first-derivative effect of the first-past-difference calculation of the PCRs shown in the diagra. This is best illustrated as a discrete-tie syste operating at the average PCR rate in Figure I.7. ITU-T Rec. J.133 (07/2002) 13

1st-past difference 2nd-order cascade filter HPF 2nd-order (2) J.133_FI-7 Figure I.7/J.133 Second-order HPF In ters of the reference odel presented in 4.1, Figure I-6 easures the difference in two PCR inaccuracies Mp,i Mp,I. A series of these easureents can be processed further to derive the individual PCR inaccuracies Mp,I by assuing that average inaccuracy is zero. I.7.2 PCR_drift_rate (PCR_DR) The result of PCR_DR is obtained at interface H of Figure I.8. Valid PCR PCR(i) PCR(0) Counter Latch + Load counter, PCR(0), at start of easureent and when PCR_discontinuity flag detected. 1/F No + + Integrator H(z) = 1/(z 1) 1st order LPF Type II PLL 1st order Latch Open-loop gain = K/PCR_rate K Kv*SF*1/F No *BW 2 HPF K Digital filter 1 st order LPF Fz = BW/3 Tz = 1/2*PI*Fz Interface J PCR Overall jitter DAC Interface H PCR drift rate VCXO F No = 27 MHz Kv = VCO gain Interface G PCR Frequency offset SF = DAC gain or scale-factor SCF when the loop locks. F No when in free run K*PCR_rate J.133_FI.8 Figure I.8/J.133 Overall PCR jitter easureent cobining the effects of PCR_AC and PCR_arrival_tie_jitter This easureent result is obtained after the cobined action of the second-order HPF represented by the loop (before the integrator represented by the adder and latch), followed by the first-order low-pass filter (LPF). This cobined action provides the response indicated in Figure I.5 for drift rate. I.7.3 PCR_frequency_offset (PCR_FO) The result of PCR_FO is obtained at interface G of Figure I.8. This easureent is obtained after the cobined action of the first-order HPF represented by the loop and the integrator (represented by the adder and latch) followed by the first-order LPF. This cobined action provides the response indicated in Figure I.5 for frequency offset. 14 ITU-T Rec. J.133 (07/2002)

I.7.4 PCR_overall_jitter easureent The result of PCR_OJ is obtained at interface J of Figure I.8. This easureent result is obtained after the cobined action of the second-order HPF represented by the loop (before the integrator represented by the adder and latch), followed by the first-order HPF. This cobined action provides the response indicated in Figure I.5 for jitter (left drawing). Overall jitter includes the coposite effect of PCR accuracy errors and PCR arrival-tie jitter. It is iportant since this relates directly to the effect on the progra-recovered clock jitter and drift. This ethod should also include a easureent of clock drift-rate and frequency-offset. Therefore, the ost practical ethod is to ipleent a SCF recovery PLL like the one in the progra decoder. By carefully controlling the bandwidth and calibrating the VCXO, it is possible to easure, siultaneously, PCR overall jitter, SCF frequency-offset, and syste clock frequency (SCF) driftrate with the frequency responses described before. Explanation Note that the PLL is a Type II control syste with two ideal integrators (digital accuulator shown and VCXO). This creates a second-order high-pass closed-loop response at the output of the phase subtraction. Therefore, below the loop bandwidth, the response is proportional to drift-rate and proportional to jitter above the loop bandwidth. It is necessary to add an additional first-order HPF to the jitter easureent to reove the effects of drift-rate. Conversely, it is necessary to add a first-order LPF to the drift-rate output to reove the effects of jitter fro that easureent. NOTE 1 If the filters are ipleented using DSP techniques on the raw data, and since the PCR_rate is the saple rate, the average PCR_rate should be deterined by easuring the PCR_interval and filtering the result with a 10 Hz LPF or lower. The value of PCR_rate can be used for those values shown in the figure to effect the selected easureent bandwidth, BW, such that it is independent of PCR_rate. NOTE 2 The design shown is a digital/analogue hybrid with a digital-to-analogue (DAC) converter driving the analogue loop filter. For a 14-bit DAC the SF would be 2 14. The voltage-controlled crystal oscillator (VCXO) with gain Kv can be constructed fro a sub-syste consisting of an oven-controlled crystal oscillator (OCXO) and a frequency-locked loop (FLL) locking a VCXO. This can be used to calibrate the Frequency-offset output to the wanted accuracy if desired. Otherwise, the VCXO can be used alone and its frequency error or offset verified by applying a known, accurate frequency, TS and subtracting the error fro subsequent easureents. NOTE 3 Alternatively, a free-running OCXO can be used to deterine the PCR_interval with known ethods and a nuerical voltage-controlled oscillator (VCO) can be constructed. With this ethod a copletely digital or software only version can be constructed using the easured PCR_interval and the PCR values. It can be shown that this ethod can have a bandwidth that is essentially independent of average PCR_rate with the easured jitter values relatively independent of variations in PCR_interval. Although this ethod describes a PLL ipleentation as a hybrid of DSP and analogue signal processing, other ethods that yield the sae filtered responses are possible. I.8 Considerations on perforing PCR easureents The easureent and validation of contributions to jitter and drift rate of a progra syste tie clock (STC) carried by its discrete-tie saples via PCR values of each progra in a TS require certain atheatical analysis of such saples in order to copute the perforance liits for direct coparison to those indicated in this Recoendation. Typical sapled syste analysis relies on a regular sapling rate of the data to be analyzed. This is not generally the case of the discrete-tie saples carried by PCR values which, per their own nature, depend on criteria and priorities at the ultiplexing stage. ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] establishes a axiu interval of 100 s between consecutive PCR values. The DVB recoends that all DVB copliant systes will transit the ITU-T Rec. J.133 (07/2002) 15

PCR values with a axiu interval of 40 s, but all receivers should work properly with intervals as long as 100 s. None of the standards forced that the interval, whatever it is, should be constant. This is because in the ultiplexing process there is a need for an allowance as to the instant the packet containing the PCR field for a given progra is to be inserted into the TS. However, the intention of the designers and operators of ultiplexers is to provide such values at the ost regular rate as possible. At the receiver, the regeneration of the 27 MHz of syste clock for the progra under the decoding process is controlled by a signal that akes use of each of the PCR values corresponding to such progra at the tie of arrival to introduce corrections when needed. It is assued that the stability of the clock regenerator is such that the phase does not unduly drift fro one PCR value to the next over intervals as long as 100 s. However, it is the responsibility of the TS to provide the values of PCR correctly with an error no greater than 500 ns fro the instantaneous phase of the syste clock. The liit of 500 ns ay be exceeded as an accuulated error over any PCR values. However, when the accuulated error spans a sufficiently long duration, it should be considered in ters of its drift contribution and, allowed to exceed the 500 ns liit. What "sufficiently long" eans has been derived in I.5 and is represented graphically by the break points of the graph in Figure I.2. For sinusoidal frequencies lower than 12 Hz, the liit is set by the drift rate specification rather than by the 500 ns liit. If appropriate filters are built into the easureent device to separate the received PCR value spectral coponents around a jitter vs. drift dearcation frequency, then it is possible to copare the errors received against the appropriate liits indicated in this Recoendation. Should the design of the easureent device be built as analogue device with hardware filters, then the designer will use the dearcation frequency as a requireent for the design of the filters with independence of the sapling rate at which the PCRs are actually arriving. This dearcation frequency is derived fro the liits indicated in this Recoendation and does not depend on sapling rate for the PCR values. If the design of the filters is done by DSP techniques, the designer ust take into account the average sapling rate of the PCR values and adapt the filters to aintain a relatively fixed bandwidth for the easureent. This approach iplicitly assues that the sapling rate (average arrival rate of PCR values) is not only known but is relatively constant. A good recoendation is to have the value of the coefficients deterined adaptively by easuring the actual arrival rate of PCR values. In other words, use an adaptive filter with the variable paraeter being the easured PCR rate. This approach has been tested in practice using very strong frequency odulation for the PCR values rate; the results in the easured jitter and drift do have a very close correlation (within the accuracy liits of the easuring device) to the jitter and drift errors inserted by the test generator into the PCR values under test. Generally, sall differences in easureent filter bandwidths do not affect jitter easureent results significantly since the jitter spectral coponents are ost often broad band. In fact, the order of the filter is ost iportant since this deterines the filter output sensitivity to out-of-band coponents, which ay have sall aplitudes but very high first- and second-tie derivatives. Another consideration to have in account is not related to the verification of strea validity but is related to a debugging tool to find the origin of the jitter should it exist and have certain periodicity or resonant frequencies. This tool shall apply Fourier analysis to the received sapled data. Again, for this type of analysis to be valid, it is assued that the sapling rate is known and is regular. Then the sapling rate has to be easured in order to know frequencies analyzed in each frequency bin (the resolution as a function of the nuber of tie doain saples used in the calculation and the relative stability of the sapling rate over the easureent interval). 16 ITU-T Rec. J.133 (07/2002)

The proble of the non-unifority of the sapling rate could be overcoe by careful interpolation before the Fourier technique is applied. In general this interpolation is not necessary due to the fact that as a debugging tool, the need is not to know what is the "exact" value of the frequencies and its aplitudes. What is needed is only to obtain an idea on whether the jitter is just rando or whether it has soe predoinant frequencies ebedded. Generally, when a Fourier analysis is done on regularly sapled signals and there is a stable sinusoidal coponent on the signal, its paraeters can be obtained with great accuracy and a clear spectral line could be displayed with such data represented as in a spectru analyzer. If the sinusoidal coponent were not stable, then a broad spectral line with lowered aplitude would be expected, broader and lower as greater is the frequency odulation (FM) iplicit in such a sinusoid. If a stable sinusoid is present but the sapling rate is FM odulated, as is the case of PCR arrival rate, then a broad and lower spectral line can be expected, just siilar to the previous case described. When a great deal of FM (rando or not) is present in the sapling signal, the spectru becoes broader with less aplitude in each bin. However as a diagnostic tool it ay still be valid. I.9 Choice of filters in PCR easureent I.9.1 Why is there a choice? PCR easureent is a difficult task. The PCR values do not occur very often and when they do, they are rather large (42 bits) nubers. The Clock reference is intended to be very stable, and as such a easureent device ust have at least the sae stability to ake a easureent. It is this long-ter stability (of the order of a few pp change in frequency per hour) in a counter which is increenting very fast (27 MHz), but transitted infrequently (40 s or so) which causes the probles. A "Dearcation" frequency has been defined (Figure I.2) which is able to divide the inaccuracies added to the PCR clock into Drift (low frequency coponent) and Jitter (high frequency coponent). It is based on the liits indicated in ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1] that sets a region below 10 Hz (MGF1) where the drift liit (75 Hz/s) is doinant and a region above 10 Hz (MGF1) where errors are allow to exceed the drift rate but not the phase-error liit (500 ns); that is why MGF1 is the highly recoended dearcation frequency used for accurate copliance to ITU-T Rec. H.222.0 ISO/IEC 13818-1 [1]. For practical easureents, however, three fixed dearcation frequencies have been specified. MGF1-3 and a user- or anufacturer-defined one is also allowed MGF4. The dearcation frequency chosen is a coproise between the desired accuracy of the clock as defined in the MPEG specification, and the practical concerns with perforing the easureent. In order for two easureent devices to give the sae results for a given Transport Strea, they ust use the sae dearcation frequency in the easureent. In addition, any secondary effects due to irregular arrival of the PCR saples ay be reoved so that results ay atch ore closely. The way this is done is beyond the scope of this easureent guideline, but designs should give siilar results when, say, a 10-inute strea has PCRs every 20 s for the first 5 inutes and then 40 s for the next 5 inutes. When the filter profiles MGF1 to MGF4 defined in this Recoendation are ipleented, there will be deviations between the real response of the filters and the desired response of the ideal filters. This will give soe easureent errors between devices. In general, the precision of the filtering is a coercial choice of the equipent anufacturer who is building equipent for a specific arket. ITU-T Rec. J.133 (07/2002) 17

The choice These guidelines are intended to create an environent where siilar achines give siilar results, and users are able to understand the iplications of choosing different easureent paraeters. The errors between different devices will vary depending on a nuber of factors: 1) Are the sae dearcation frequencies being used? This is the ajor factor. If different devices use different dearcation frequencies, then they will give different results. This will be a ajor source of error. A discussion of the nature of the error is given below. 2) Are the dearcation filters of the sae order? This is less iportant. If one device uses a second order filter and another uses a fifth order filter, then the nature of the filter response will be quite different. There is likely to be a sall difference between easureent devices particularly if significant frequency coponents of the errors are close to the chosen dearcation frequency. 3) Is the easureent being ade near the crossover of the offset/drift/jitter frequencies? Near the crossover frequency, the order of the filter and its ipulse response are likely to affect the frequency coponents which are included or rejected fro the easureents. This has uch less of an affect than the choice of dearcation frequency. I.9.2 Higher dearcation frequencies There are several effects of choosing a higher dearcation frequency (e.g. MGF3): 1) Jitter turns into drift or frequency offset. A higher dearcation frequency eans that frequency coponent which would have been classed as jitter will now be classed as frequency offset or drift. This has the effect of reducing the agnitude of the overall jitter frequency coponent. It also akes the syste clock look less stable than it actually is. 2) The easureent settles faster. The settling tie is closely related to 1/frequency. If the frequency is increased by two orders of agnitude, then the settling tie ay be reduced by two orders of agnitude. There are DSP techniques which can be used to iprove settling ties, and the use of these is a coercial choice of the equipent vendor. As a rough rule of thub, a higher dearcation frequency settles faster but gives a less accurate result. Jitter easureents should appear saller and drift easureents should appear larger. I.9.3 Lower dearcation frequencies There are several effects of choosing a lower dearcation frequency (e.g. MGF1): 1) Separation of drift and jitter into ore representative groupings. A lower dearcation frequency eans that frequency coponents are ore accurately classed as jitter, frequency offset or drift. This has the effect of easuring the frequency coponents based on assuptions which are closer to the values in the MPEG2 specification. 2) The easureent takes longer to settle. The settling tie is closely related to 1/frequency. If the frequency is reduced by two orders of agnitude, then the settling tie ay increase by two orders of agnitude. There are DSP techniques which can be used to iprove settling ties, and the use of these is a coercial choice of the equipent vendor. 18 ITU-T Rec. J.133 (07/2002)