A Guide to Digital Television Systems and Measurements

Home , Color television, Communications system, Serial Digital Interface, Television lines

Last active picture sample First active picture sample

COMPOSITE: Full analog line digitized

COMPONENT: Active line digitized, EAV/SAV added

EAV SAV

Y Y Y

Cb Cb

Cr Cr

000 000

3FF 3FF

000 000

XYZ XYZ Data

0 1 2 Word Number 3

1712 1713 1714 1715

A Guide to Digital Television Systems and Measurements

Acknowledgment Authored by David K. Fibush with contributions from: Bob Elkind Kenneth Ainsworth

Some text and figures used in this booklet reprint- ed with permission from Designing Digital Sys- tems published by Grass Valley Products.

i ii Contents 1. Introduction · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 1 2. Digital Basics · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3 Component and Composite (analog)· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 3 Sampling and Quantizing · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 4 Digital Video Standards· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 4 Serial Digital Video · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 8 Rate Conversion – Format conversion · · · · · · · · · · · · · · · · · · · · · · · · · · · 10 3. Digital Audio · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 11 AES/EBU Audio Data Format · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 11 Embedded Audio · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 11 4. System Hardware and Issues · · · · · · · · · · · · · · · · · · · · · · · · · 19 Cable Selection · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 19 Connectors · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 19 Patch Panels · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 19 Terminations and Loop-throughs · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 19 Digital Audio Cable and Connector Types · · · · · · · · · · · · · · · · · · · · · · · · 20 Signal Distribution, Reclocking · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 20 System Timing · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 20 5. Monitoring and Measurements · · · · · · · · · · · · · · · · · · · · · · · · 23 6. Measuring the Serial Signal · · · · · · · · · · · · · · · · · · · · · · · · · · 25 Waveform Measurements · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 25 Measuring Serial Digital Video Signal Timing · · · · · · · · · · · · · · · · · · · · · 26 7. Definition and Detection of Errors · · · · · · · · · · · · · · · · · · · · · 29 Definition of Errors · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 29 Quantifying Errors · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 30 Nature of the Studio Digital Video Transmission Systems · · · · · · · · · · · 30 Measuring Bit Error Rates · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 32 An Error Measurement Method for Television· · · · · · · · · · · · · · · · · · · · · 32 System considerations· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 33 8. Jitter Effects and Measurements · · · · · · · · · · · · · · · · · · · · · · · 35 Jitter in Serial Digital Signals · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 35 Measuring Jitter · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 37 9. System Testing · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 39 Stress Testing · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 39 SDI Check Field · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 39 10. Bibliography· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 42 Articles and Books· · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 42 Standards · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 43 11. Glossary · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · 44

iii

1. Introduction New methods of test and The following major topics 4. Test and measurement of measurement are required to are covered in this booklet: serial digital signals can determine the health of the 1. A definition of digital be characterized in three digital video transmission television including digi- different aspects: usage, system and the signals it car- tal basics, video stan- methods, and operational ries. Analysis methods for dards, conversions environment. Types of the program video and audio between video signal usage include: designer’s signals are well known. forms, and digital audio workbench, manufactur- However, measurements for formats and standards. ing quality assurance, the serial form of the digi- user equipment evalua- 2. Considerations in select- tized signal differ greatly tion, system installation ing transmission system from those used to analyze and acceptance testing, passive components such the familiar baseband video system and equipment as cable, connectors and signal. maintenance, and perhaps related impedance match- most important, opera- This guide addresses many ing concerns are dis- tion. topics of interest to designers cussed as well as contin- and users of digital television ued use of the passive 5. Several measurement systems. You’ll find informa- loop-through concept. methods are covered in tion on embedded audio Much of the transmission detail, including serial (Section 3), system timing system methods and hard- waveform measurements, issues (Section 4), and sys- ware already in place may definition and detection tem timing measurements be used for serial digital. of errors, jitter effects and (Section 6). Readers who are measurements, and sys- 3. New and old systems already familiar with digital tem testing with special issues that are particularly television basics may wish to test signals. start with Section 4, System important for digital video Hardware and Issues. Discus- are discussed, including sion of specific measurement equalization, reclocking, topics begins in Section 6. and system timing.

page 1 page 2 2. Digital Basics Component and Composite ized for 525-line systems, ond, each one with only half (analog) whereas there is an EBU doc- the lines. Every other picture Image origination sources, ument (EBU N-10) for 625 line is represented in the first such as cameras and line systems. Details of com- field and the others are filled telecines, internally produce ponent analog systems are in by the second field. color pictures as three, full fully described in another Another way to look at inter- bandwidth signals – one each booklet from Tektronix, Solv- lace is that it’s bandwidth for Green, Blue and Red. Cap- ing the Component Puzzle. reduction. Display rates of 50 italizing on human vision not It’s important to note that sig- or more presentations per being as acute for color as it nal level values in the color second are required to elimi- is for brightness level, televi- difference domain allow com- nate perceived flicker. By sion signals are generally binations of Y, B-Y, R-Y that using two-field interlace, the transformed into luminance will be out of legal gamut bandwidth for transmission is and color difference signals range when converted to reduced by a factor of two. as shown in Figure 2-1. Y, the GBR. Therefore, there’s a Interlace does produce arti- luminance signal, is derived need for gamut checking of facts in pictures with high from the GBR component col- color difference signals when vertical information content. ors, based on an equation making operational adjust- However, for television view- such as: ments for either the analog or ing, this isn’t generally con- digital form of these signals. Y = 0.59G + 0.30R + 0.11B sidered a problem when con- Most television signals today trasted to the cost of twice the The color difference signals are two field-per-frame bandwidth. In computer dis- operate at reduced band- “interlaced.” A frame con- play applications, the arti- width, typically one-half of tains all the scan lines for a facts of interlace are unac- the luminance bandwidth. In picture, 525 lines in 30 ceptable so progressive scan- some systems, specifically frame/second applications ning is used, often with frame NTSC, the color difference and 625 lines in 25 rates well above 60 Hz. As signals have even lower and frame/second applications. television (especially high unequal bandwidths. Compo- Interlacing increases the tem- definition) and computer ap- nent signal formats and volt- poral resolution by providing plications become more inte- age levels are not standard- twice as many fields per sec- grated there will be a migra- tion from interlace to progressive scanning for all applications. Further bandwidth reduction of the television signal occurs when it’s encoded into composite PAL or NTSC as shown in Figure 2-2. NTSC is defined for studio operation by SMPTE 170M which for- malizes and updates the Figure 2-1. Component signals. never fully approved, RS 170A. Definitions for PAL and NTSC (as well as SECAM) can be found in Y 0-5 MHz Composite Video ITU-R (formally CCIR) Report 624. Where each GBR signal + 0-5 MHz may have a 6 MHz bandwidth, color difference sig- Chroma nals would typically have Y 2-5 MHz at 6 MHz and each color difference signal at 3 MHz; however, a composite signal is R-Y 0-1.5 MHz Quadrature one channel of 6 MHz or less. Modulation The net result is a composite 6 MHz channel carrying color B-Y 0-1.5 MHz at subcarrier frequency fields at a 60 per second rate that in its uncompressed, progressive scan form, would have required three 12 MHz Figure 2-2. Composite encoding. channels for a total band-

page 3 width of 36 MHz. So, data particularly when decoding pling Theorem says that the compression is nothing new; back to component. The interval between successive digital just makes it easier. problem is that it’s relatively samples must be equal to or For composite NTSC signals, easy to add setup; but remov- less than one-half of the there are additional gamut ing it when the amplitudes period of the highest fre- considerations when con- and timing of setup in the quency present in the signal.) verting from the color differ- composite domain are not The second step in digitizing ence domain. NTSC trans- well controlled can result in video is to “quantize” by mitters will not allow 100 black level errors and/or sig- assigning a digital number to percent color amplitude on nal disturbances at the ends the voltage levels of the sam- signals with high luminance of the active line. pled analog signal – 256 lev- level (such as yellow). This Sampling and Quantizing els for 8-bit video, 1024 for is because the transmitter The first step in the digitiz- 10-bit video, and up to sev- carrier will go to zero for sig- ing process is to “sample” eral thousand for audio. nals greater than about 15 the continuous variations of To get the full benefit of digi- percent above 1 volt. Hence, the analog signal as shown in tal, 10-bit processing is there’s a lower gamut limit Figure 2-4. By looking at the required. While most current for some color difference sig- analog signal at discrete time tape machines are 8-bit, nals to be converted to NTSC intervals, a sequence of volt- SMPTE 125M calls out 10-bit for RF transmission. age samples can be stored, as the interface standard. As part of the encoding pro- manipulated and later, Processing at less than 10 cess, sync and burst are reconstructed. bits may cause truncation added as shown in Figure In order to recover the analog and rounding artifacts, par- 2-3. Burst provides a refer- signal accurately, the sample ticularly in electronically ence for decoding back to rate must be rapid enough to generated pictures. Visible components. The phase of avoid missing important in- defects will be revealed in burst with respect to sync formation. Generally, this the picture if the quantiza- edge is called SCH phase, requires the sampling fre- tion levels are too coarse (too which must be carefully con- quency to be at least twice few levels). These defects trolled in most studio appli- the highest analog frequency. can appear as “contouring” cations. For NTSC, 7.5 IRE In the real world, the fre- of the picture. However, the units of setup is added to the quency is a little higher than good news is that random luminance signal. This poses twice. (The Nyquist Sam- noise and picture details pre- some conversion difficulties, sent in most live video signals actually help conceal these contouring effects by adding a natural randomness to them. Sometimes the number of quantizing levels must be reduced; for example, when the output of a 10-bit processing device feeds an 8-bit recorder. In this case, contouring effects are minimized by deliberately adding a small amount of random noise (dither) to the signal. Figure 2-3. Color difference and composite waveforms. This technique is known as randomized rounding. Digital Video Standards Although early experiments with digital technology were based on sampling the composite (NTSC or PAL) signal, it was realized that for the highest quality operation, component processing was necessary. The first digital standards were component. Interest in composite digital was revived when Ampex and Sony announced a composite digital recording for- Figure 2-4. Sampling an analog signal. mat, which became known as

page 4 D-2. Initially, these machines signals CB and CR, which are lier RP-125), and for 625/50 were projected as analog scaled versions of the signals by EBU Tech 3267 (a revi- in/out devices for use in B-Y and R-Y. sion of the earlier EBU Tech existing analog NTSC and The sampling structure 3246). Both of these were PAL environments; digital defined is known as “4:2:2.” adopted by CCIR and are inputs and outputs were pro- This nomenclature is derived included in Recommenda- vided for machine-to-ma- from the days when multi- tion 656, the document chine dubs. However, the ples of NTSC subcarrier were defining the hardware inter- post-production community being considered for the sam- face. recognized that greater pling frequency. This ap- The parallel interface uses advantage could be taken of proach was abandoned, but eleven twisted pairs and 25- the multi-generation capabil- the use of “4” to represent pin “D” connectors. (The ity of these machines if they the luminance sampling fre- early documents specified were used in an all digital quency was retained. The slide locks on the connec- environment. task force mentioned above tors; later revisions changed Rec. 601. Recommendation examined luminance sam- the retention mechanism to ITU-R BT.601 (formerly CCIR pling frequencies from 4/40 screws.) This interface Recommendation 601) is not 12 MHz to 14.3 MHz. They multiplexes the data words a video interface standard, selected 13.5 MHz as a com- in the sequence CB, Y, CR, Y, but a sampling standard. Rec. promise because the sub- CB…, resulting in a data rate 601 evolved out of a joint multiple 2.25 MHz is a factor of 27 Mwords/s. The timing SMPTE/EBU task force to common to both 525- and sequences SAV and EAV determine the parameters for 625-line system. Some were added to each line to digital component video for extended definition TV sys- represent Start of Active the 525/59.94 and 625/50 tems use a higher resolution Video and End of Active television systems. This format called 8:4:4 which has Video. The digital active line work culminated in a series twice the bandwidth of 4:2:2. contains 720 luminance sam- of tests sponsored by SMPTE Parallel component digital. ples, and includes space for in 1981, and resulted in the Rec. 601 described the sam- representing analog blanking well known CCIR Recom- pling of the signal. Electrical within the active line. mendation 601. This docu- interfaces for the data pro- Rec. 601 specified eight bits ment specified the sampling duced by this sampling were of precision for the data mechanism to be used for standardized separately by words representing the vid- both 525 and 625 line sig- SMPTE and the EBU. The eo. At the time the standards nals. It specified orthogonal parallel interface for were written, some partici- sampling at 13.5 MHz for 525/59.94 was defined by pants suggested that this may luminance, and 6.75 MHz SMPTE as SMPTE Standard not be adequate, and provi- for the two color difference 125M (a revision of the ear- sion was made to expand the interface to 10-bit precision. Ten-bit operation has indeed Excluded 766.3 mv 3FF 11 1111 1111 proved beneficial in many circumstances, and the latest 700.0 mv 3AC 11 1010 1100 Peak revisions of the interface standard provide for a 10-bit interface, even if only eight bits are used. Digital-to-analog conversion range is cho- sen to provide headroom above peak white and footroom below black as shown in Figure 2-5. Quantizing levels for black and white are selected so the 8-bit levels

Black 0.0 mv 040 00 0100 0000 with two “0”s added will have the same values as the Excluded -51.1 mv 000 00 0000 0000 10-bit levels. Values 000 to 003 and 3FF to 3FC are Figure 2-5. Luminance quantizing. reserved for synchronizing

page 5 purposes. Similar factors determine the quantizing val- Excluded 399.2 mv 3FF 11 1111 1111 ues for color difference signals as shown in Figure 2-6. Max Positive 350.0 mv 3C0 11 1100 0000 Figure 2-7 shows the location of samples and digital words with respect to an analog horizontal line. Because the timing information is carried by EAV and SAV, there is no Black 0.0 mv 200 10 0000 0000 need for conventional synchronizing signals, and the horizontal intervals (and the active line periods during the vertical interval) may be used to carry ancillary data. Max Negative -350.0 mv 040 00 0100 0000 The most obvious applica-

Excluded -400.0 mv 000 00 0000 0000 tion for this data space is to carry digital audio, and documents are being prepared by SMPTE to standardize the Figure 2-6. Color difference quantizing. format and distribution of the audio data packets. Rec. 601/656 is a well proven technology with a full range of equipment available for production and post production. Generally, the parallel Last active picture sample First active picture sample interface has been super- seded by a serial implementation, which is far more practical in larger installa- COMPOSITE: Full analog line digitized tions. Rec. 601 provides all of the advantages of both dig- COMPONENT: Active line digitized, EAV/SAV added ital and component opera-

V V

SA EA tion. It’s the system of choice for the highest possible quality in 525- or 625-line

Y Y Y

Cb Cb

Cr Cr

000 000

3FF 3FF

000

XYZ XYZ systems. Data

0 1 2 Word Number 3 Parallel composite digital.

1712 1713 1714 1715 The composite video signal is sampled at four times the (NTSC or PAL) subcarrier Figure 2-7. Digital horizontal line. frequency, giving nominal sampling rates of 14.3 MHz for NTSC and 17.7 MHz for Excluded 998.7 mv 3FF 11 1111 1111 PAL. The interface is stan- 100% Chroma 937.7 mv 3CF 11 1100 1111 dardized for NTSC as SMPTE 244M and, at the time of writing, EBU documentation Peak White 714.3 mv 320 11 0010 0000 is in process for PAL. Both interfaces specify ten-bit precision, although D-2 and D-3 machines record only eight bits to tape. Quantizing of the NTSC signal (shown in Figure 2-8) is defined with a Blanking 0.0 mv 0F0 00 1111 0000 modest amount of headroom above 100% bars, a small

Sync Tip -285.7 mv 016 00 0001 0000 footroom below sync tip and Excluded -306.1 mv 000 00 0000 0000 the same excluded values as for component. PAL composite digital has Figure 2-8. NTSC quantizing. been defined to minimize

page 6 quantizing noise by using a sentation of vertical sync and composite. Also, the full maximum amount of the equalizing pulses is also excursion of the composite available digital range. As transmitted over the compos- signal has to be represented can be seen in Figure 2-9, the ite interface. by 256 levels on 8-bit peak analog values actually Composite digital installa- recorders. Nevertheless, com- exceed the digital dynamic tions provide the advantages posite digital provides a range, which might appear to of digital processing and more powerful environment be a mistake. Because of the interface, and particularly than analog for NTSC and specified sampling axis, ref- the multi-generation capabil- PAL installations and is a erence to subcarrier and the ities of digital recording. very cost effective solution phase of the highest lumi- However, there are some lim- for many users. nance level bars (such as yel- itations. The signal is com- As with component digital, low), the samples never posite and does bear the foot- the parallel composite inter- exceed the digital dynamic print of NTSC or PAL encod- face uses a multi-pair cable range. The values involved ing, including the narrow- and 25-pin “D” connectors. are shown in Figure 2-10. band color information Again, this has proved to be Like the component inter- inherent to these coding satisfactory for small and face, the composite digital schemes. Processes such as medium sized installations, active line is long enough to chroma key are generally not but practical implementation accommodate the analog satisfactory for high quality of a large system requires a active line and the analog work, and separate provision serial interface. blanking edges. Unlike the must be made for a compo- 16:9 widescreen component component interface, the nent derived keying signal. digital signals. Many televi- composite interface transmits Some operations, such as sion markets around the a digital representation of digital effects, require that world are seeing the intro- conventional sync and burst the signal be converted to duction of delivery systems during the horizontal blank- component form for process- for widescreen (16:9 aspect ing interval. A digital repre- ing, and then re-encoded to ratio) pictures. Some of these systems, such as MUSE in Excluded 913.1 mv 3FF 11 1111 1111 Japan and the future ATV system in the USA, are intended for high-definition Peak White 700.0 mv 34C 11 0100 1100 images. Others, such as PALplus in Europe will use 525- or 625-line pictures. Domestic receivers with 16:9 displays have been intro- duced in many markets, and Blanking 0.0 mv 100 01 0000 0000 penetration will likely increase significantly over the next few years. For most Sync Tip -300.0 mv 004 00 0000 0100 broadcasters, this creates the Excluded -304.8 mv 000 00 0000 0000 need for 16:9 program material.

Figure 2-9. PAL quantizing. Two approaches have been proposed for the digital representation of 16:9 525- and 625-line video. The first Reference Subcarrier method retains the sampling U-V sample U+V sample frequencies of the Rec. 601 standard for 4:3 images (13.5 MHz for luminance). This “stretches” the pixel -U-V sample -U+V sample represented by each data word by a factor of 1.33 hori- +167 o 0.934 V zontally, and results in a loss of 25 percent of the horizon- Excluded values Highest sample 0.9095 to 0.9131 V tal spatial resolution when 0.620 V Luminance Level 0.886 V compared to Rec. 601 pictures. For some applications, this method still provides ad- 100% Yellow Bar equate resolution, and it has the big advantage that much Figure 2-10. PAL yellow bar sampling.

page 7 existing Rec. 601 equipment 16:9 signals using 13.5 MHz the low-frequency content of can be used. sampling are electrically the transmitted signal, and to A second method maintains indistinguishable from 4:3 spread the transmitted spatial resolution where pix- signals, they can be con- energy spectrum so that els (and data words) are veyed by the SMPTE 259M radio-frequency emission added to represent the addi- serial interface at 270 Mb/s. problems are minimized. tional width of the image. It’s planned that SMPTE In the early 1980s, a serial This approach results in 960 259M (discussed below) will interface for Rec. 601 signals luminance samples for the be revised to provide for was recommended by the digital active line (compared serial transmission of EBU. This interface used 8/9 to 720 samples for a 4:3 pic- 18 MHz sampled signals, block coding and resulted in ture, and to 1920 samples for using the same algorithm, at a bit rate of 243 Mb/s. Its a 16:9 high definition pic- a data rate of 360 Mb/s. interface did not support ten- ture). The resulting sampling Serial Digital Video bit precision signals, and rate for luminance is Parallel connection of digital there were some difficulties 18 MHz. This system pro- equipment is practical only in producing reliable, cost vides the same spatial resolu- for relatively small installa- effective, integrated circuits. tion as Rec. 601 4:3 images; tions, and there’s a clear The block coding based inter- but existing equipment need for transmission over a face was abandoned and has designed only for 13.5 MHz single coaxial cable. This is been replaced by an interface cannot be used. The addi- not simple as the data rate is with channel coding that tional resolution of this high, and if the signal were uses scrambling and conver- method may be advantageous transmitted serially without sion to NRZI. The serial inter- when some quality overhead modification, reliable recov- face has been standardized as is desired for post produc- ery would be very difficult. SMPTE 259M and EBU Tech. tion or when up-conversion The serial signal must be 3267, and is defined for both to a high-definition system is modified prior to transmis- component and composite contemplated. sion to ensure that there are signals including embedded SMPTE 267M provides for sufficient edges for reliable digital audio. both the 13.5 MHz and clock recovery, to minimize Conceptually, the serial digital 18 MHz systems. Because interface is much like a carrier system for studio applications. Baseband audio and VIDEO video signals are digitized and combined on the serial digital ANALOG TO “carrier” as shown in Figure DIGITAL SERIAL DIGITAL 2-11. (It’s not strictly a carrier C ONVERSION VIDEO “C ARRIER” system in that it’s a baseband & digital signal, not a signal DIGITAL DATA 143 Mb/s NTSC modulated on a carrier.) The F ORMATTING bit rate (carrier frequency) is 177 Mb/s PAL AUDIO determined by the clock rate 270 Mb/s CCIR 601 of the digital data: 143 Mb/s 360 Mb/s 16:9 component for NTSC, 177 Mb/s for PAL and 270 Mb/s for Rec. 601 component digital. The widescreen (16:9) component Figure 2-11. The carrier concept. system defined in SMPTE 267 will produce a bit rate of 360 Mb/s.

Parallel In Parallel data representing the MSB LSB Serial Out samples of the analog signal 10 MSB is processed as shown in Fig- NRZ Scrambled ure 2-12 to create the serial Data Shift Encoder LSB Register Scrambler NRZI digital data stream. The parallel clock is used to load 9 4 sample data into a shift regis- Clock Load Data Shift Data G1(x) = x + x + 1 ter, and a times-ten multiple G2(x) = x + 1 of the parallel clock shifts the bits out, LSB (least significant bit) first, for each 10-bit x 10 data word. If only 8 bits of Fclock 10 x Fclock data are available at the input, the serializer places Figure 2-12. Parallel-to-serial conversion.

page 8 zeros in the two LSBs to parallel component interface ing at the serial receiver, complete the 10-bit word. provide unique sequences which should also remove Component signals do not that can be identified in the the TRS from the received need further processing as serial domain. The parallel serial signal. the SAV and EAV signals on composite interface does not The composite parallel inter- the parallel interface provide have such signals, so it is face does not provide for unique sequences that can be necessary to insert a suitable transmission of ancillary identified in the serial timing reference signal (TRS) data, and the transmission of domain to permit word fram- into the parallel signal before sync and burst means that ing. If ancillary data such as serialization. A diagram of less room is available for audio has been inserted into the serial digital NTSC hori- insertion of data. Upon con- the parallel signal, this data zontal interval is shown in version from parallel to will be carried by the serial Figure 2-13. The horizontal serial, the sync tips may be interface. The serial interface interval for serial digital PAL used. However, the data may be used with normal would be similar except that space in NTSC is sufficient video coaxial cable. the sample location is for four channels of AES/EBU The conversion from parallel slightly different on each digital audio. Ancillary data to serial for composite signals line, building up two extra such as audio may be added is somewhat more complex. samples per field. A three- prior to serialization, and As mentioned above, the word TRS is inserted in the this would normally be per- SAV and EAV signals on the sync tip to enable word fram- formed by the same co-pro- cessor that inserts the TRS. Following the serialization of the parallel information the data stream is scrambled by a mathematical algorithm then encoded into NRZI (non- return to zero inverted) by a concatenation of the following two functions: 9 4 G1(X) = X + X + 1

G2(X) = X + 1 At the receiver the inverse of this algorithm is used in the deserializer to recover the correct data. In the serial digital transmission system, the clock is contained in the data as opposed to the parallel Figure 2-13. NTSC horizontal interval. system where there’s a separate clock line. By scram- 1 1 1 1 bling the data, an abundance NRZ of transitions is assured as o o o o o o o required for clock recovery. The mathematics of scram- NRZI bling and descrambling lead to some specialized test signals for serial digital systems NRZ all "0"s that are discussed later in this guide. NRZI all "highs" or "lows" Encoding into NRZI makes the serial data stream polarity insensitive. NRZ (non return to zero, an old digital NRZ all "1"s data tape term) is the familiar circuit board logic level, high NRZI 1/2 clock freq is a “1” and low is a “0.” For either polarity a transmission system it’s convenient to not require a certain polarity of the signal Clock Freq Data detect on at the receiver. As shown in rising edge Figure 2-14, a data transition is used to represent each “1” Figure 2-14. NRZ and NRZI relationship.

page 9 and there’s no transition for a resampling of digital rates. composite-to-component for- data “0.” The result’s that it’s Strictly speaking, rate con- mat converter provides a rea- only necessary to detect tran- version is taking one sample sonable alternative. sitions; this means either rate and making another After post-production work, polarity of the signal may be sample rate out of it. For our component digital often used. Another result of NRZI purposes, we’ll use the term needs to be converted to encoding is that a signal of “format conversion” to mean composite digital. Sources all “1”s now produces a tran- both the encode/decode step that are component digital sition every clock interval and resampling of digital may be converted for input and results in a square wave rates. The format conversion to a composite digital at one-half the clock fre- sequence depends on direc- switcher or effects machine. quency. However, “0”s pro- tion. For component to com- Or a source from a compo- duce no transition, which posite, the usual sequence is nent digital telecine may be leads to the need for scram- rate conversion followed by converted for input to a com- bling. At the receiver, the ris- encoding. For composite to posite digital suite. Further- ing edge of a square wave at component, the sequence is more, tapes produced in the clock frequency would be decoding followed by rate component digital may need used for data detection. conversion. See Figure 2-15. to be distributed or archived Rate Conversion – Format It’s much easier to do pro- in composite digital. Conversion duction in component The two major contributors When moving between com- because it isn’t necessary to to the quality of this process ponent digital and composite wait for every color framing are the encoding or decoding digital in either direction, boundary (four fields for process and the sample rate there are two steps: the NTSC and eight for PAL) to conversion. If either one is actual encoding or decoding, match two pieces of video to- faulty, the quality of the final and the conversion of the gether; instead, they can be product suffers. In order to sample rate from one stan- matched every two fields accurately change the digital dard to the other. The digital (motion frame). This pro- sample rate, computations sample rates for these two vides four times the opportu- must be made between two formats are different: nity. Additionally, compo- different sample rates and in- 13.5 MHz for component dig- nent is a higher quality for- terpolations must be calcu- ital and 14.3 MHz for NTSC mat since luminance and lated between the physical composite digital (17.7 MHz chrominance are handled location of the source pixel for PAL). This second step is separately. To the extent pos- data and the physical loca- called “rate conversion.” sible, origination for compo- tion of destination pixel data. Often the term rate conver- nent environment produc- Of the 709,379 pixel loca- sion is used to mean both tion should be accomplished tions in a PAL composite encoding/decoding and component. A high quality digital frame, all except one must be mapped (calculated). In order to accomplish this computationally intense conversion, an extremely accurate algorithm must be used. If the algorithm is accurate enough to deal with PAL conversions, NTSC conversions using the same algorithm are simple. To ensure quality video, a sophisticated algorithm must be employed and the hardware must produce precise coefficients and minimize rounding errors.

Figure 2-15. Format conversion.

page 10 3. Digital Audio AES/EBU Audio Data Format between 100 and 110 dB. the transmission line, facili- AES digital audio (also Using the above formula for tate clock recovery, and make known as AES/EBU) con- an SNR of 110 dB, this sys- the interface polarity insensi- forms to the specification tem has an equivalent 18.3- tive, the data is channel AES3 (ANSI 4.40) titled “AES bit resolution. coded as biphase-mark. The recommended practice for An AES digital audio signal preambles specifically violate digital audio engineering – always consists of two chan- the biphase-mark rules for Serial transmission format for nels which may be distinctly easy recognition and to two-channel linearly repre- separate audio material or ensure synchronization. sented digital audio data.” stereophonic audio. There’s When digital audio is embed- AES/EBU digital audio is the also a provision for single- ded in the serial digital video result of cooperation between channel operation (mono- data stream, the start of the the Audio Engineering Soci- phonic) where the second 192-frame block is indicated ety and the European Broad- digital data channel is either by the so-called “Z” bit casting Union. identical to the first or has which corresponds to the data set to logical “0.” For- occurrence of the Z-type pre- When discussing digital amble. audio, one of the important matting of the AES data is considerations is the number shown in Figure 3-1. Each The validity bit indicates of bits per sample. Where sample is carried by a sub- whether the audio sample video operates with 8 or 10 frame containing: 20 bits of bits in the sub-frame are suit- bits per sample, audio imple- sample data, 4 bits of auxil- able for conversion to an ana- mentations range from 16 to iary data (which may be used log audio signal. User data is 24 bits to provide the desired to extend the sample to 24 provided to carry other infor- dynamic range and signal-to- bits), 4 other bits of data and mation, such as time code. noise ratio (SNR). The basic a preamble. Two sub-frames Channel status data contains formula for determining the make up a frame which con- information associated with SNR for digital audio is: tains one sample from each of each audio channel. There the two channels. are three levels of implemen- SNR = (6.02 * n) + 1.76 Frames are further grouped tation of the channel status where “n” is the number of into 192-frame blocks which data: minimum, standard, bits per sample define the limits of user data and enhanced. The standard For a 16-bit system, the maxi- and channel status data implementation is recom- mum theoretical SNR would be blocks. A special preamble mended for use in profession- (6.02 * 16) + 1.76 = 98.08 dB; indicates the channel identity al television applications, for an 18-bit system, the SNR for each sample (X or Y hence the channel status data would be 110.2 dB; and for a preamble) and the start of a will contain information 20-bit device, 122.16 dB. A 192-frame block (Z pream- about signal emphasis, sam- well-designed 20-bit ADC ble). To minimize the direct- pling frequency, channel probably offers a value current (DC) component on mode (stereo, mono, etc.), use of auxiliary bits (extend audio data to 24 bits or other use), and a CRC (cyclic 32 bits redundancy code) for error Preamble X, Y or Z, 4 bits checking of the total channel Sample Data status block. 20 bits Embedded Audio Auxiliary Data 4 bits One of the important advan- Audio sample validity tages of SDI (Serial Digital Subframe Structure User bit data Interconnect) is the ability to Audio channel status embed (multiplex) several Subframe parity channels of digital audio in Frame Structure the digital video. This is particularly useful in large sys- Y Channel B Z Channel A Y Channel B X Channel A Y Channel B tems where separate routing of digital audio becomes a cost consideration and the Subframe 1 Subframe 2 assurance that the audio is Frame 191 Frame 0 Frame 1 associated with the appropri- Start of 192 frame block ate video is an advantage. In smaller systems, such as a post production suite, it’s Figure 3-1. AES audio data formatting.

page 11 generally more economical to least four channels of audio using this method for both maintain separate audio thus in the composite serial digi- composite and component eliminating the need for tal signal which has limited signals. As engineers were numerous mux (multiplexer) ancillary data space. A basic using the SMPTE 259M spec- and demux (demultiplexer) form of embedded audio for ifications for their designs, it modules. In developing the composite video is docu- became clear that a more SDI standard, SMPTE 259M, mented in SMPTE 259M and definitive document was one of the key considerations there’s a significant amount needed. A draft proposed was the ability to embed at of equipment in the field standard is being developed for embedded audio which Table 3-1. NTSC Digital Ancillary Data Space includes such things as dis- Pulse Tip Words Per Frame Total tribution of audio samples within the available compos- Horizontal 55 507 27,885 ite digital ancillary data Equalizing 21 24 504 space, the ability to carry 24- Broad 376 12 4,512 bit audio, methods to handle Total 10-bit words/frame 32,901 non-synchronous clocking and clock frequencies other than 48 kHz, and specifications for embedding audio in TRS-ID Ancillary Data Space (not to scale) component digital video. Ancillary data space. Com- posite digital video allows V H Scan ancillary data only in its

V Direction

V ertical serial form and then only in ertical the tips of synchronizing signals. Ancillary data is not

Active Video Blanking Active Video carried in parallel composite

Blanking digital as parallel data is simply a digital representation of the analog signal. Figures 3-2 and 3-3 show the ancillary data space available in composite digital signals which includes horizontal Figure 3-2. Composite ancillary data space. sync tip, vertical broad pulses, and vertical equalizing pulses. A small amount of the horizontal sync tip space is reserved for TRS-ID (timing reference signal and identifi- 790 - 794 795 - 849 cation) which is required for

TRS-ID ANC Data - 55 words digital word framing in the H-sync (all lines except vertical sync) Not to scale deserialization process. Since composite digital ancillary data is only available in the serial domain, it consists of 10-bit words, whereas both composite and component parallel interconnects may be 790 - 794 795 - 260 340 - 715 either 8 or 10 bits. A sum- TRS-ID ANC Data - 376 words ANC Data - 376 words Vertical sync lines only Vertical sync lines only mary of composite digital ancillary data space for NTSC is shown in Table 3-1. Thirty frames/second will provide 9.87 Mb/second of ancillary data which is enough for four channels of embedded audio

790 - 794 795 - 815 340 - 360 with a little capacity left over. PAL has slightly more ancil- TRS-ID ANC Data - 21 words ANC Data - 21 words Vertical equalizing pulses only Vertical equalizing pulses only lary data space due to the higher sample rate; however

Figure 3-3. NTSC available ancillary data words.

page 12 embedded audio for digital HANC (horizontal ancillary channels of embedded audio PAL is rarely used. data) and VANC (vertical are specified for HANC in There’s considerably more ancillary data) with the the proposed standard and ancillary data space available SMPTE defining the use of there’s room for significantly in component digital video HANC and the EBU defining more data. as shown in Figure 3-4. All the use of VANC. Word Not all of the VANC data of the horizontal and vertical lengths are 10-bit for HANC space is available. For in- blanking intervals are avail- specifically so it can carry stance, the luminance sam- able except for the small embedded audio in the same ples on one line per field are amount required for EAV format as for composite digi- reserved for DVITC (digital (end of video) and SAV (start tal and 8-bit for VANC vertical interval time code) of video) synchronizing specifically so it can be car- and the chrominance sam- words. ried by D-1 VTRs (on those ples on that line may be vertical interval lines that are The ancillary data space has devoted to video index. Also, recorded). Total data space is been divided into two types it would be wise to avoid shown in Table 3-2. Up to 16 using the vertical interval switch line and perhaps, the Table 3-2. Component Digital Ancillary Data Space subsequent line where data 525 line Words Per Frame Total Rate Mb/s might be lost due to relock HANC 268 525 140,700 42.2 after the switch occurs. VANC 1440 39 56,160 13.5 Ancillary data formatting. Ancillary data is formatted Total words 198,860 55.7 into packets prior to multi- 625 line Words Per Frame Total Rate Mb/s plexing it into the video data stream as shown in Figure HANC 282 625 176,250 44.1 3-5. Each data block may VANC 1440 49 70.560 14.1 contain up to 255 user data Total words 246,810 58.2 words provided there’s enough total data space available to include five (composite) or seven (component) words of overhead. For composite digital, only the vertical sync (broad) pulses have enough room for the full 255 words. Multiple data packets may be placed in individual ancillary data spaces, thus providing a rather flexible data communications channel. At the beginning of each data packet is a header using word values that are excluded for digital video data and reserved for synchronizing purposes. For composite Figure 3-4. Component ancillary data space. video, a single header word of 3FCh is used whereas for component video, a three- word header 000h 3FFh Data Block Number (1 word) 3FFh is used. Each type of data packet is identified with User Data a different Data ID word. 255 words maximum Several different Data ID words are defined to orga- Data Count (1 word) nize the various data packets Data ID (1 word) Check Sum (1 word) used for embedded audio. The Data Block Number Data Header (DBN) is an optional counter 1 word composite, 3FC that can be used to provide 3 words component, 000 3FF 3FF sequential order to ancillary data packets allowing a Figure 3-5. Ancillary data formatting. receiver to determine if

page 13 there’s missing data. As an sum which is used to detect This basic embedded audio example, with embedded errors in the data packet. corresponds to Level A in the audio the DBN may be used Basic embedded audio. proposed embedded audio to detect the occurrence of a Embedded audio defined in standard. Other levels of vertical interval switch, SMPTE 259M provides four operation provide more thereby allowing the receiver channels of 20-bit audio data channels, other sampling fre- to process the audio data to sampled at 48 kHz with the quencies, and additional remove the likely transient sample clock locked to the information about the audio “click” or “pop.” Just prior to television signal. Although data. Basic embedded audio the data is the Data Count specified in the composite data packet formatting word indicating the amount digital part of the standard, derived from AES audio is of data in the packet. Finally, the same method is also used shown in Figure 3-6. following the data is a check- for component digital video. Two AES channel-pairs are shown as the source; however it’s possible for each of Table 3-3. Embedded Audio Bit Distribution the four embedded audio Bit X X + 1 X + 2 channels to come from a different AES signal (specifi- b9 not b8 not b8 not b8 cally implemented in some b8 aud 5 aud 14 Parity D-1 digital VTRs). The Audio b7 aud 4 aud 13 C Data Packet contains one or more audio samples from up b6 aud 3 aud 12 U to four audio channels. 23 b5 aud 2 aud 11 V bits (20 audio bits plus the C, b4 aud 1 aud 10 aud 19 (msb) U, and V bits) from each AES b3 aud 0 aud 9 aud 18 sub-frame are mapped into three 10-bit video words (X, b2 ch bit-1 aud 8 aud 17 X+1, X+2) as shown in Table b1 ch bit-2 aud 7 aud 16 3-3. b0 Z-bit aud 6 aud 15 Bit-9 is always not Bit-8 to ensure that none of the excluded word values (3FFh- AES Channel-Pair 2 3FCh or 003h-000h) are used. Y Channel B Z Channel A Y Channel B X Channel A Y Channel B The Z-bit is set to “1” corre- sponding to the first frame of Subframe 1 Subframe 2 the 192-frame AES block. Frame 191 Frame 0 Frame 1 Channels of embedded audio AES Channel-Pair 1 are essentially independent (although they,re always Y Channel B Z Channel A Y Channel B X Channel A Y Channel B transmitted in pairs) so the

Subframe 1 Subframe 2 Z-bit is set to a “1” in each

Frame 191 Frame 0 Frame 1 Frame 3 channel even if derived from the same AES source. C, U, and V bits are mapped from the AES signal; however the AES Subframe 32 bits parity bit is not the AES par- Subframe parity ity bit. Bit-8 in word X+2 is Preamble X, Y or Z, 4 bits Audio channel status even parity for bits 0-8 in all Auxiliary Data 4 bits User bit data Audio sample validity three words. There are several restrictions 20 bits Sample Data AES 1 chnl B (subframe 2) regarding distribution of the Embedded Channel 2 audio data packets although there’s a “grandfather clause” 20 + 3 bits of data is in the proposed standard to mapped into 3 ANC words account for older equipment Audio Data Packet that may not observe all the restrictions. Audio data AES 1, Chnl A AES 1, Chnl B AES 2, Chnl A Channel 1 Channel 2 Channel 3 packets are not transmitted

Sum

Data

Count x, x+1, x+2 x, x+1, x+2 x, x+1, x+2

Check

Data ID

Blk Num in the horizontal ancillary data space following the nor- Data Header mal vertical interval switch Composite, 3FC as defined in RP 168. They’re Component, 000 3FF 3FF also not transmitted in the Figure 3-6. Basic embedded audio. ancillary data space desig-

page 14 nated for error detection four audio samples per chan- • Providing audio-to-video checkwords defined in nel in each audio data delay information for each RP 165. For composite digital packet. channel video, audio data packets are Extended embedded audio. • Documenting Data IDs to not transmitted in equalizing Full-featured embedded allow up to 16 channels of pulses. Taking into account audio defined in the pro- audio in component digital these restrictions “data posed standard includes: systems should be distributed as • Carrying the 4 AES auxil- • Counting “audio frames” evenly as possible through- iary bits (which may be for 525 line systems. out the video field.” The rea- used to extend the audio son for this last statement is To provide these features, samples to 24-bits) to minimize receiver buffer two additional data packets size which is an important • Allowing non-synchronous are defined. Extended Data issue for transmitting 24-bit clock operation Packets carry the 4 AES aux- audio in composite digital • Allowing sampling fre- iliary bits formatted such systems. For basic, Level A, quencies other than 48 kHz that one video word contains this results in either three or the auxiliary data for two audio samples as shown in Figure 3-7. Extended data Table 3-4. Data IDs For Up To 16-channel Operation packets must be located in Audio Audio Data Extended Audio Control the same ancillary data space Channels Packet Data Packet Packet as the associated audio data Group 1 1-4 1FF 1FE 1EF packets and must follow the audio data packets. Group 2 5-8 1FD 2FC 2EE The Audio Control Packet Group 3 9-12 1FB 2FA 2ED (shown in Figure 3-8.) is Group 4 13-16 2F9 1F8 1EC transmitted once per field in the second horizontal ancillary data space after the ver- Data ID 2FF (1FD, 1FB, 2F9) Audio Data Packet tical interval switch point. It contains information on AES 1, ch A AES 1, ch B AES 2, ch A Channel 1 Channel 2 Channel 3 audio frame number, sam-

Data x, x+1, x+2 x, x+1, x+2 x, x+1, x+2 Sum

Data

Count

Check pling frequency, active chan- Blk Num nels, and relative audio-to- video delay of each channel. Data Header Composite, 3FC Transmission of audio con- Component, 000 3FF 3FF trol packets is optional for Extended Data Packet 48 kHz synchronous opera- Auxiliary 4 bits tion and required for all other modes of operation

Data

Sum

Data

Count (since it contains the infor-

AUX AUX AUX AUX

Check

AUX

AUX Blk Num mation as to what mode is being used). Data ID 1FE (2FC, 2FA, 1F8) Audio frame numbers are an artifact of 525 line, 29.97 Figure 3-7. Extended embedded audio. frame/second operation. In that system there are exactly Data Header 8008 audio samples in exactly five frames, which Component 000 3FF 3FF +_ Delay (audio-video) Composite 3FC means there’s a non-integer ch 1 (&2) ch 3 (&4) ch 2 =\ 1 ch 4 =\ 3 number of samples per Data ID Data block number frame. An audio frame Data count sequence is the number of frames for an integer number of samples (in this case five)

V V and the audio frame number

DID

ACT

DBN indicates where in the

RSR

AF1-2 AF3-4

RSR

DELA0 DELA1

DELC0 DELC1 DELC2 DELD0 DELD1

DELA2 DELB0 DELB1 DELB2

DELD2 CHK SUM sequence a particular frame belongs. This is important Audio frame Reserved number when switching between Active Words sources because certain chnls 1 & 2 Sampling Channels Checksum equipment (most notably dig- chnls 3 & 4 Frequency ital VTRs) require consistent synchronous operation to Figure 3-8. Audio control packet formatting. prevent buffer over/under

page 15 flow. Where frequent switch- dio ancillary data. The case A receiver sample buffer is ing is planned, receiving is considerably different for needed to store samples nec- equipment can be designed composite digital video due essary to supply the continu- to add or drop a sample fol- to the exclusion of data in ous output required during lowing a switch in the four equalizing pulses and, even the time of the equalizing out of five cases where the more important, the data pulses and other excluded sequence is broken. The packet distribution required lines (the vertical interval challenge in such a system is for extended audio. For this switch line and the subse- to detect that a switch has reason the proposed standard quent line). When only 20-bit occurred. This can be facili- requires a receiver buffer of audio is required the “even tated by use of the data block 64 samples per channel with distribution of samples” dic- number in the ancillary data a grandfather clause of 48 tates that some horizontal format structure and by samples per channel to warn lines carry four samples per including an optional frame designers of the limitations channel resulting in a rather counter with the unused bits in older equipment. In the modest buffer size of 48 sam- in the audio frame number proposed standard, Level A ples or less depending on the word of the audio control defines a sample distribution type of buffer. Using only the packet. allowing use of a 48 sample- first broad pulse of each ver- Audio delay information per-channel receiver buffer tical sync line, four or five contained in the audio con- while other levels generally samples in each broad pulse trol packet uses a default require the use of the speci- will provide a satisfactory, channel-pair mode. That is, fied 64 sample buffer. Deter- even distribution. This is delay-A (DELA0-2) is for mination of buffer size is known as Level A. However, both channel 1 and channel analyzed in this section for if 24-bit audio is used or if 2 unless the delay for chan- digital NTSC systems, how- the sample distribution is to nel 2 is not equal to channel ever the result is much the allow for 24-bit audio even if 1. In that case, the delay for same for digital PAL systems. not used, there can be no channel 2 is located in delay- Synchronous sampling at more than three samples per C. Sampling frequency must 48 kHz provides exactly 8008 horizontal ancillary data be the same for each channel samples in five frames which space for a four-channel sys- in a pair, hence the data in is 8008 ÷ (5 x 525) = 3.051 tem. The first broad pulse of “ACT” provides only two samples per line. Therefore each vertical sync line is values, one for channels 1 most lines may carry three required to carry 16 or 17 and 2 and the other for chan- samples per channel while samples per channel to get nels 3 and 4. some should carry four sam- the best possible “even distribution.” This results in a In order to provide for up to ples per channel. For digital buffer requirement of 80 16 channels of audio in com- NTSC, each horizontal sync samples per channel, or less, ponent digital systems the tip has space for 55 ancillary depending on the type of embedded audio is divided data words. Table 3-5 shows buffer. To meet the 64 sam- into audio groups corre- how those words are used for ple-per-channel buffer, a sponding to the basic four- four-channel, 20-bit and 24- smart buffer is required. channel operation. Each of bit audio (overhead includes the three data packet types header, data ID, data count, There are two common types are assigned four Data IDs as etc.). Since it would require of buffer that are equivalent shown in Table 3-4. 24 more video words to carry for the purpose of under- one additional sample per standing requirements to Receiver buffer size. In com- channel for four channels of meet the proposed standard. ponent digital video, the re- 24-bit audio and that would One is the FIFO (first in first ceiver buffer in an audio exceed the 55-word space, out) buffer and the other is a demultiplexer is not a criti- there can be no more than circular buffer. Smart buffers cal issue since there’s much three samples per line of 24- have the following synchro- ancillary data space available bit audio. nized load abilities: and few lines excluding au- • FIFO Buffer – hold off Table 3-5. Maximum Samples Per NTSC Sync Tip “reads” until a specific number of samples are in 4-Channels 20-bit Audio 24-bit Audio the buffer (requires a 3-samples 4-samples 3-samples counter) AND neither read Audio Data or write until a specified Overhead 5 5 5 time (requires vertical Sample data 36 48 36 sync). Extended Data • Circular Buffer – set the Overhead 5 read address to be a certain Auxiliary data 6 number of samples after the write address at a spec- TOTAL 41 53 47 page 16 ified time (requires vertical audio or for automatic opera- instances where it’s desirable sync). tion of both sample sizes. to break away some of the Using a smart buffer the min- Therefore, a 64 sample-per- audio sources, the digital imum sizes are 24 samples channel smart buffer is audio can be demultiplexed per channel for 20-bit opera- required to meet the specifi- and switched separately via tion and 40 samples per cations of the proposed an AES/EBU digital audio channel for 24-bit operation. standard. routing switcher. Because of the different Systemizing AES/EBU audio. At the receiving end, after requirements for the two Serial digital video and the multiplexed audio has types of sample distribution, audio are becoming com- passed through a serial digi- the minimum smart buffer monplace in production and tal routing switcher, it may size for either number of bits post-production facilities as be necessary to extract the per sample is 57 samples. well as television stations. In audio from the video so that Therefore, a 64 sample-per- many cases, the video and editing, audio sweetening, or channel smart buffer should audio are married sources; other processing can be ac- be adequate. and it may be desirable to complished. This requires a In the case of a not-so-smart keep them together and treat demultiplexer that strips off buffer, read and write them as a single serial digital the AES/EBU audio from the addresses for a circular data stream. This has, for one serial digital video. The out- buffer must be far enough example, the advantage of put of a typical demulti- apart to handle either a build being able to keep the signals plexer has a serial digital up of 40 samples or a deple- in the digital domain and video BNC as well as connection of 40 samples. This switch them together with a tors for the two-stereo-pair means the buffer size must serial digital video routing AES/EBU digital audio be 80 samples for 24-bit switcher. In the occasional signals.

page 17 page 18 4. System Hardware and Issues Cable Selection Connectors designed specifically for The best grades of precision Until recently, all BNC con- serial digital applications. analog video cables exhibit nectors used in television had Terminations and Loop-throughs low losses from very low fre- a characteristic impedance of The serial digital standard quencies (near DC) to around 50 ohms. BNC connectors of specifies transmitter and 10 MHz. In the serial digital the 75-ohm variety were receiver return loss to be world, cable losses in this available, but were not physi- greater than 15 dB up to portion of the spectrum are of cally compatible with 50-ohm 270 MHz. That is, termina- less consequence but still connectors. The impedance tions should be 75 ohms with important. It’s in the higher “mismatch” to coax is of little no significant reactive com- frequencies associated with consequence at analog video ponent to 270 MHz. Clearly transmission rates of 143, frequencies because the this frequency is related to 177, 270, or 360 Mb/s where wavelength of the signals is component digital, hence a losses are considerable. For- many times longer than the lower frequency might be tunately, the robustness of length of the connector. But suitable for NTSC and that the serial digital signal makes with serial digital’s high data may be considered in the it possible to equalize these rate (and attendant short future. As can be seen by the losses quite easily. So when wavelengths), the impedance rather modest 15 dB specifi- converting from analog to of the connector must be con- cation, return loss is not a digital, the use of existing, sidered. In general, the length critical issue in serial digital quality cable runs should of a simple BNC connector video systems. It’s more pose no problem. doesn’t have significant effect important at short cable The most important charac- on the serial digital signal. lengths where the reflection teristic of coaxial cable to be Several inches of 50-ohm could distort the signal rather used for serial digital is its connection, such as a number than at long lengths where loss at 1/2 the clock fre- of barrels or short coax, the reflected signal receives quency of the signal to be would have to be used for more attenuation. transmitted. That value will any noticeable effect. In the specific case of a serial trans- Most serial digital receivers determine the maximum are directly terminated in cable length that can be mitter or receiver, the active devices associated with the order to avoid return loss equalized by a given receiver. problems. Since the receiver It’s also important that the chassis connector need significant impedance matching. active circuits don’t look like frequency response loss in dB 75 ohms, circuitry must be be approximately__ propor- This could easily include √ making a 50-ohm connector included to provide a good tional to 1/ f down to fre- resistive termination and low quencies below 5 MHz. Some look like 75 ohms over the frequency band of interest. return loss. Some of today’s of the common cables in use equipment does not provide are the following: Good engineering practice tells us to avoid impedance return loss values meeting the PSF 2/3 UK mismatches and use 75-ohm specification for the higher Belden 8281 USA and Japan components wherever possi- frequencies. This hasn’t F&G 1.0/6.6 Germany ble and that rule is being fol- caused any known system These cables all have excel- lowed in the development of problems; however, good en- lent performance for serial new equipment for serial gineering practice should pre- digital signals. Cable manu- digital. vail here as well. facturers are seizing the Active loop-throughs are very Patch Panels opportunity to introduce common in serial digital new, low-loss, foam dielectric The same holds true for other equipment because they’re cables specifically designed “passive” elements in the sys- relatively simple and have for serial digital. Examples of tem such as patch panels. In signal regeneration qualities alternative video cables are order to avoid reflections similar to a reclocking distri- Belden 1505A which, caused by impedance discon- bution amplifier. Also, they although thinner, more flexi- tinuities, these elements provide isolation between ble, and less expensive than should also have a character- input and output. However, if 8281, has a higher perfor- istic impedance of 75 ohms. the equipment power goes off mance at the frequencies that Existing 50-ohm patch panels for any reason, the connec- are more critical for serial will probably be adequate in tion is broken. Active loop- digital signals. A recent many instances, but new throughs also have the same development specifically for installations should use 75- need for care in circuit serial digital video is Belden ohm patch panels. Several impedance matching as men- 1694A with lower loss than patch panel manufacturers tioned above. 1505A. now offer 75-ohm versions

page 19 Passive loop-throughs are sional applications. When instead of 75-ohm unbal- both possible and practical. AES/EBU digital audio anced and reduce the 3- to Implementations used in se- evolved, it was natural to 10-volt signal to 1 volt. The rial digital waveform moni- assume that the traditional resulting signal now has the tors have return loss greater analog audio transmission same characteristics as ana- that 25 dB up to the clock cable and connectors could log video, and traditional frequency. This makes it pos- still be used. The scope of analog video distribution sible to monitor the actual AES3 covers digital audio amplifiers, routing switchers, signal being received by the transmission of up to 100 and video patch panels can operational equipment or meters, which can be han- be used. Costs of making up unit-under-test and not sub- dled adequately with the coaxial cable with BNCs stitute the monitor receiver. shielded, balanced twisted- is also less than multicon- The most important use of pair interconnection. ductor cable with XLRs. passive loop-throughs is for Because AES/EBU audio has Tests have indicated that system diagnostics and fault a much wider bandwidth EMI emissions are also finding where it’s necessary than analog audio, cable reduced with coaxial to observe the signal in the must be selected with care. distribution. troubled path. If a loop- The impedance, in order to Signal Distribution, Reclocking through is to be used for meet the AES3 specification, Although a video signal may monitoring, it’s important requires 110-ohm source and be digital, the real world that it be passive loop- load impedances. The stan- through which that signal through for two reasons. dard has no definition for a passes is analog. Conse- First, serial transmitters with bridging load, although loop- quently, it’s important to con- multiple outputs usually through inputs, as used in sider the analog distortions have separate active devices analog audio, are theoreti- that affect a digital signal. on each output, therefore cally possible. Improperly These include frequency monitoring one output terminated cables may cause response rolloff caused by doesn’t necessarily indicate signal reflections and subse- cable attenuation, phase dis- the quality of another output quent data errors. tortion, noise, clock jitter, as it did in the days of resis- The relatively high frequen- and baseline shift due to AC tive fan-out of analog signals. cies of the AES/EBU signals coupling. While a digital sig- Second, when an active loop- cannot travel over twisted- nal will retain the ability to through is used, loss of pair cables as easily as ana- communicate its data despite power in the monitoring log audio. Capacitance and a certain degree of distortion, device will turn off the sig- high-frequency losses cause there’s a point beyond which nal. This would be disastrous high-frequency rolloff. Even- the data will not be recover- in an operational situation. If tually, signal edges become able. Long cable runs are the an available passive loop- so rounded and amplitude so main cause of signal distor- through isn’t being used, it’s low that the receivers can no tion. Most digital equipment important to make sure the longer tell the “1”s from the provides some form of equal- termination is 75 ohms with “0”s. This makes the signals ization and regeneration at all no significant reactive com- undetectable. Typically, inputs in order to compen- ponent to at least the clock cable lengths are limited to a sate for cable runs of varying frequency of the serial signal. few hundred feet. XLR con- lengths. Considering the spe- That old, perhaps “preci- nectors are also specified. cific case of distribution am- sion,” terminator in your tool Since AES/EBU digital audio plifiers and routing switch- box may not do the job for has frequencies up to about ers, there are several ap- serial digital. 6 MHz, there’ve been sugges- proaches that can be used: Digital Audio Cable and tions to use unbalanced wideband amplifier, wide- Connector Types coaxial cable with BNC con- band digital amplifier, and AES/EBU digital audio has nectors for improved perfor- regenerating digital amplifier. raised some interesting ques- mance in existing video Taking the last case first, tions because of its character- installations and for trans- regeneration of the digital istics. In professional appli- missions beyond 100 meters. signal generally means to cations, balanced audio has There are two committees recover data from an incom- traditionally been deemed defining the use of BNC con- ing signal and to retransmit it necessary in order to avoid nectors for AES/EBU digital with a clean waveform using hum and other artifacts. Usu- audio. The Audio Engineer- a stable clock source. Regen- ally twisted, shielded, multi- ing Society has developed eration of a digital signal conductor audio cable is AES3-ID and SMPTE is allows it to be transmitted used. The XLR connector developing a similar recom- farther and sustain more ana- was selected as the connector mended practice. These rec- log degradation than a signal of choice and is used univer- ommendations will use a which already has accumu- sally in almost all profes- 110-ohm balanced signal lated some analog distor- page 20 tions. Regeneration generally several dozen times before a may not, be 6 dB/octave. It’s uses the characteristics of the parallel regeneration is nec- the difference between the incoming signal, such as essary. Excessive jitter build in-band and out-of-band extracted clock, to produce up eventually causes a sys- response that causes the the output. In serial digital tem to fail; jitter acceptable problem. video, there are two kinds of to a serial receiver must still In between these two regeneration: serial and be considered and handled extremes are wideband serial parallel. from a system standpoint digital routers that don’t Serial regeneration is sim- which will be discussed reclock the signal. Such non- plest. It consists of cable later. reclocking routers will gener- equalization (necessary even Regeneration, where a refer- ally perform adequately at all for cable lengths of a few ence clock is used to produce clock frequencies for the meters or less), clock recov- the output, can be performed amount of coax attenuation ery, data recovery, and re- an unlimited number of given in their specifications. transmission of the data us- times and will eliminate all However, that specification ing the recovered clock. A the jitter built up in a series will usually be shorter than phase locked loop (PLL) with of regeneration operations. for reclocking routers. an LC (inductor/capacitor) or This type of regeneration System Timing RC (resistor/capacitor) oscil- happens in major operational lator regenerates the serial equipment such as VTRs, The need to understand, clock frequency, a process production switchers (vision plan, and measure signal called reclocking. mixers) or special effects timing has not been elimi- units that use external refer- nated in digital video, it’s Parallel regeneration is more just moved to a different set complex. It involves three ences to perform their digital processing. So the bottom of values and parameters. In steps: deserialization; paral- most cases, precise timing lel reclocking, usually using line is, regeneration (or reclocking) can be perfect, requirements will be relaxed a crystal-controlled time in a digital environment and base; and serialization. only limited by economic considerations. will be measured in micro- Each form of regeneration seconds, lines, and frames can reduce jitter outside its A completely different rather than nanoseconds. In PLL bandwidth, but jitter approach to signal distribu- many ways, distribution and within the loop bandwidth tion is the use of wideband timing is becoming simpler will be reproduced and may analog routers, but there are with digital processing accumulate significantly many limitations. It’s true because devices that have with each regeneration. (For that wideband analog routers digital inputs and outputs a complete discussion of jit- (100 MHz plus) will pass a can lend themselves to auto- ter effects and measurements 143 Mb/s serial digital signal. matic input timing compen- see Section 8.) A serial regen- However, they’re unlikely to sation. However, mixed ana- erator will have a loop band- have the performance log/digital systems place width on the order of several required for 270 or 360 Mb/s. additional constraints on the hundred kilohertz to a few With routers designed for handling of digital signal megahertz. A parallel regen- wideband__ analog signals, the timing. erator will have a much nar- 1/√ f frequency response characteristics will adversely Relative timing of multiple rower bandwidth on the signals with respect to a ref- order of several Hertz. There- affect the signal. Receivers intended to work with erence is required in most fore, a parallel regenerator systems. At one extreme, can reduce jitter to a greater SMPTE 259M signals expect to see a signal transmitted inputs to a composite analog extent than a serial regenera- production switcher must be tor, at the expense of greater from standard source and attenuated by coaxial cable timed to the nanosecond complexity. In addition, the range so there will be no sub- inherent jitter in a crystal with frequency response losses. Any deviation from carrier phase errors. At the controlled timebase (parallel other extreme, most digital regenerator) is much less 6 dB/octave rolloff over a bandwidth of 1 MHz up to production switchers allow than that of an LC or RC relative timing between oscillator timebase (serial the clock frequency will cause improper operation of input signals in the range of regenerator). Serial regenera- one horizontal line. Signal- tion obviously cannot be per- the automatic equalizer in the serial receiver. In general, to-signal timing requirements formed an unlimited number for inputs to digital video of times because of cumula- this will be a significant problem because the analog can be placed in three cate- tive PLL oscillator jitter and gories: the fact that the clock is router is designed to have extracted from the incoming flat response up to a certain 1. Digital equipment with signal. Serial regeneration bandwidth and then roll off automatic input timing can typically be performed at some slope that may, or compensation.

page 21 2. Digital equipment without the switching area for field 1. will have to be within one or automatic input timing (Field 2 switches on line 273 two nanoseconds so the dis- compensation. for 525 and line 319 for 625.) turbance doesn’t occur. 3. Digital-to-analog convert- The window size is 10 µs Clearly, the downstream ers without automatic which would imply an blanking approach is timing compensation. allowable difference in time preferred. between the two signals of a In the first case, the signals Finally, the more difficult few microseconds. A pre- must be in the range of the case of inputs to DACs with- ferred timing difference is automatic compensation, out timing compensations generally determined by the generally about one horizon- may require the same accura- reaction of downstream tal line. For system installa- cy as today’s analog systems equipment to the vertical tion and maintenance it’s – in the nanosecond range. interval switch. In the case of important to measure the rel- This will be true where fur- analog routing switchers, a ative time of such signals to ther analog processing is very tight tolerance is often ensure they’re within the required, such as used in maintained based on subcar- automatic timing range with analog production switchers rier phase requirements. For enough headroom to allow (vision mixers). Alternately, serial digital signals for any expected changes. if the analog signal is simply microseconds of timing dif- Automatic timing adjust- a studio output, this high ference may be acceptable. ments are a key component accuracy may not be At the switch point, clock for digital systems; but they required. lock of the serial signal will cannot currently be applied Automatic timing adjust- be recovered in one or two to a serial data stream due to ments are a key component digital words (10 to 20 serial the high digital frequencies for digital systems; but they clock periods). However, involved. A conversion back cannot currently be applied word framing will, most to the parallel format for the to a serial data stream due to likely, be lost causing the sig- adjustable delay is necessary the high digital frequencies nal value to assume some due to the high data rates. involved. A conversion back random value until the next Therefore, automatic input to the parallel format for the synchronizing data (EAV) timing is not available in all delay is necessary due to the resets the word framing and equipment. high data rates. There’s a produces full synchroniza- need for a variety of For the second case, signals tion of the signal at the adjustable delay devices to are nominally in-time as receiver. A signal with a ran- address these new timing would be expected at the dom value for part of a line is requirements. These include input to a routing switcher. clearly not suitable on-line to a multiple-line delay device SMPTE recommended prac- a program feed; however and a frame delay device. A tice RP 168, which applies to downstream equipment can frame delay device, of both 525 and 625 line opera- be designed to blank the course, would result in a tion, defines a timing win- undesired portion of the line. video delay that could poten- dow on a specific horizontal In the case where the unde- tially create some annoying line where a vertical interval sired signal is not elimi- problems with regard to switch is to be performed. nated, the serial signal tim- audio timing and the passing See Figure 4-1 which shows ing to the routing switcher of time code through the system. Whenever audio and Switching video signals are processed Area through different paths, a 525-line systems, Field 1 potential exists for differen- 35 µs 25 µs tial audio-to-video delay. The differential delay caused by embedding audio is insig- nificant (under 1 ms). How- 525 1 2 3 4 5 6 7 8 9 10 11 12 19 20 21 ever, potential problems occur where frame delays are Switching used and the video and Area audio follow separate paths.

625-line systems, Field 1 35 µs 25 µs

622 623 624 625 1 2 3 4 5 6 7 8 9 22 23 24

Figure 4-1. Vertical interval switching area.

page 22 5. Monitoring and Measurements Test and measurement of evaluation, system installa- nal with features and accu- serial digital signals can be tion and acceptance testing, racy consistent with today’s characterized in three differ- system and equipment main- analog baseband signal moni- ent aspects: usage, methods, tenance, and perhaps most tors. An operational monitor and operational environment. important, operation. should also include informa- Types of usage would Requirements for a basic tion about the serial digital include: designer’s work- operational monitor include signal itself, such as data bench, manufacturing quality display of the program sig- available, bit errors, and data assurance, user equipment nals carried by the digital sig- formatting errors. A display of the actual serial waveform is not required. Due to the robust nature of digital video signals a reduction in the number of operational monitors may be possible. How- ever, monitoring of the program signal waveform is required at all locations where an operator or equipment has the ability to change program parameters. The Tektronix WFM 601 Series of serial-component monitors is shown in Figure 5-1. Methods for technical evaluation cover several usage areas that have overlapping requirements. In addition to the traditional television system measurements there’s a new dimension for test and measurement – to quantify Figure 5-1. The WFM 601 Series of serial-component monitors. the various parameters associated directly with the serial waveform. The result is several categories of monitoring and measurement methods to be considered: program signal analysis, data analysis, format verification, transmitter/receiver operation, transmission hardware, and fault reporting. An equivalent-time sampling display of the serial waveform is displayed on the Tektronix WFM601M Serial Component Monitor in Fig- ure 5-2. Program signal measurements are essentially the baseband video and audio measurements that have been used for years. An important aspect to these measurements is that the accuracy of signal representation is limited by the number of bits-per-sample. In analog systems there’s been a small amount of digital data present in video sig- Figure 5-2. Eye-pattern display on WFM601M. nals in the form of Vertical

page 23 Interval Time Code (VITC); (coax, patch panels, etc.) is Because of the well known but the serial data stream has similar to that used with crash-point type of failure for much more capacity for data baseband systems except that digital systems, out-of-ser- other than video. Hence, much wider bandwidths vice testing is also especially more complete measurement must be considered. important. In order to know methods are desirable. The third aspect to test and how much headroom is Digital signal waveform anal- measurement is whether it’s available it’s necessary to ysis and its effect on trans- in-service or out-of-service. add stressing parameters to mitter/receiver operation is a All operational monitoring the digital signal in mea- new requirement in test and must be in-service, which sured amounts until it does measurement for television means that monitors must be crash, which is certainly not facilities. Rather than acquire able to give the operator in- acceptable in operational sit- special equipment for inter- formation about the digital uations. preting this high speed wave- signal that contains active The next few sections deal form, appropriate capabili- program material as well as with the details of specific ties are being added to tradi- the program signal itself. If measurement techniques tional television test equip- there are problems to be covering monitoring, mea- ment, helping in the econom- solved, discovering those surements, in-service testing, ic transition to serial digital with an intermittent nature and out-of-service testing. television. Testing of passive will also require in-service transmission components measurements.

page 24 6. Measuring the Serial Signal Waveform Measurements different paths are due to the Figure 6-1 and displayed on When viewed on an appro- fact that the digits in the a monitor as shown on the priate scope or monitor, sev- serial stream vary based on WFM601M Serial Compo- eral time sweeps (overlaid by the data (high or low states, nent Monitor in Figure 5-2). the CRT persistence or digi- with or without change at Analog measurements of the tal sample memory) pro- possible transition times). serial digital waveform start duces a waveform that fol- The waveform that results is with the specifications of the lows a number of different known as an eye pattern transmitter output as shown paths across the screen. The (with two “eyes” shown in in Figure 6-2. Specifications to be measured are amplitude, risetime, and jitter, which are defined in the serial standard, SMPTE 259M. Frequency, or period, is determined by the television sync generator developing the source signal, not the serialization process. A unit interval (UI) is defined as the time between two adjacent signal transitions, which is the reciprocal of clock frequency. The unit interval is 7.0 ns for NTSC, 5.6 ns for PAL, and 3.7 ns for component 525 or 625. Figure 6-1. Eye pattern display. A serial receiver determines if the signal is a “high” or a “low” in the center of each eye, thereby detecting the serial data. As noise and jitter in the signal increase through the transmission Jitter 0.8 Volts channel, certainly the best <0.2 UI p-p, decision point is in the cen- +_ 10% 10 Hz HPF ter of the eye (as shown in 20% to 80% Figure 6-3) although some Risetime receivers select a point at a 0.4 to 1.5 ns fixed time after each transition point. Any effect which Unit closes the eye may reduce Interval the usefulness of the received signal. Figure 6-2. Serial signal specifications. In a communications system with forward error correction, accurate data recovery can be made with the eye nearly closed. With the very low error rates required for correct transmission of serial digital video (see Section 7) a rather large and clean eye opening is required after receiver equalization. This is because the random nature of the processes that close the eye have statistical “tails” that would cause an occasional, but unacceptable error. Jitter effects that close

Figure 6-3. Data recovery.

page 25 the eye are discussed in Sec- bandwidth). Monitoring The factor of 0.5 adjusts for tion 8. quality risetime measure- the fact that the scope rise- Amplitude is important ments can be made with tele- time specification is for the because it affects maximum vision test equipment using 10% to 90% points. As an transmission distance; too equivalent-time sampling at example: with a scope 10 to large can be a problem as the lower bandwidths. Rise- 90% risetime of 1.0 ns, a well as the more obvious too time measurements are made measured risetime of 1.2 ns small. Some receiver equaliz- from the 20% to 80% points would indicate an actual ers depend on amplitude to as appropriate for ECL logic serial waveform 20 to 80% estimate cable length and set- devices. Since the serial sig- risetime of 0.97 ns, and a ting of the equalizer signifi- nal has approximately a 1 ns measured 1.6 ns would indi- cantly affects resulting noise risetime the measurement cate 1.44 ns actual risetime. and jitter. must be adjusted per the fol- Measuring Serial Digital Video lowing formula: Precise measurement of the Signal Timing 2 2 serial transmitter waveform Ta = √ (Tm – 0.5 Ts ) Relative timing between requires a scope with a Where: serial digital video signals 1 GHz bandwidth because of that are within an opera- T = actual risetime the serial signal's 1 ns rise- a tional range for use in studio time. However, amplitude Tm = measured risetime equipment may vary from measurements may be made Ts = risetime of the scope several nanoseconds to a few with lower bandwidth television lines. Measure- scopes (300 to 500 MHz ment of the timing differ- ences in operational signal paths may be accomplished using the Active Picture Tim- ing Test Signal available First full line in an analog field from the TSG 422 Digital Component Generator in conjunction with the timing First and last full amplitude sample cursors and line select of the allowed by nominal analog blanking WFM 601 series of serial component waveform monitors. Figure 6-4 shows a represen- First and last digital active picture samples tation of a picture monitor display of the Active Picture Timing Test Signal (previously known as the Digital Last full line in an analog field Blanking Test Signal). The first and last full active analog lines have a white (luminance-only) bar with nominal blanking edges. Depend- Figure 6-4. Picture monitor display of the Active Picture Timing Test signal.

page 26 ing on the field, these may or mV mV First & last digital sample First & last digital sample may not be the first and last 350 Analog line 350 Analog line 100% amplitude 100% amplitude full digital active lines. Use of the specific full-lines shown below ensure that the Not to scale Not to scale complete signal will be visible after digital-to-analog 0.0 0.0 conversion. It’s the responsi- 0.00 0.30 52.89 53.26 µs 0.00 1.04 52.44 53.26 µs bility of the converter to 000 004 714 719 sample 000 014 708 719 sample 001 009 1429 1439 word 001 029 1417 1439 word make half-lines where appro- 525 Line Luminance 625 Line Luminance priate. The lines with the

mV mV luminance white bar are: First & last digital sample First & last digital sample 525-line Lines 21, 262, 350 Analog line 350 Analog line 100% amplitude 100% amplitude signals: 284, and 525 625-line Lines 24, 310,

Not to scale Not to scale signals: 336, and 622 All other active picture lines 0.0 0.0 have a one-sample, half- 0.00 0.30 52.89 53.19 µs 0.00 1.04 52.44 53.19 µs amplitude word in four loca- 000 002 357 359 Cb sample 000 007 354 359 Cb sample 000 008 1428 1436 word (Cb) 000 028 1416 1436 word (Cb) tions: first and last active 000 002 357 359 Cr sample 000 007 354 359 Cr sample 002 010 1430 1438 word (Cr) 002 030 1418 1438 word (Cr) digital sample, first and last 525 Line Chrominance 625 Line Chrominance sample at the 100% point of a nominal white bar. Figure Figure 6-5. Locations of half-amplitude samples. 6-5 shows the specific location of each half-amplitude word. Note: the last active Digital Signal A T1 DT = 43.22m T2 chrominance sample on each Line 21 line is co-sited with lumi- Digital Signal B nance sample #718 whereas the last luminance sample is #719. Analog Reference To make a measurement of WFM 601, 1-line sweep, Line-select mode large relative timing between two digital signals, set up the Digital Signal A WFM 601 as shown in Figure Line 24 T1 DT = 43.22m T2 6-6. Alternately, select each input in the line select mode Digital Signal B and find the 50% point of the white bar. The line number and delta-T measurement Analog Reference give the total time difference. WFM 601, 1-line sweep, Line-select mode For higher resolution measurement of small time dif- Figure 6-6.Timing measurement, large time difference. ferences, use the same set up to view the time of the dou- ble pulses at the start of the Digital Signal A active line as shown in Fig- T1 DT = 1.05m T2 ure 6-7. Although the figure shows nice clean pulses, Digital Signal B considerable ringing will be seen due to the non-nyquist nature of the single-word Analog Reference excursion. As with other fre- WFM 601, 1-line mag sweep quently used setups, these can be stored as one of the nine WFM 601 presets, to be Digital Signal A T1 DT = 1.05m T2 recalled instantly when needed. Digital Signal B

Analog Reference WFM 601, 1-line mag sweep

Figure 6-7.Timing measurement, small time difference. page 27 page 28 7. Definition and Detection of Errors In our application of comput- values could be considered Considering the possible ers and digital techniques to to be correct. Application of replacement of blanking areas automate many tasks, we’ve this definition to transmis- by some equipment and grown to expect the hard- sion links is straight forward. knowing that the active pic- ware and software to report If any digital data word val- ture sample locations are problems to an operator. This ues change between trans- well defined by the various is particularly important as mitter and receiver there’s an digital standards, it’s possible systems become larger, more error. For operational studio to measure Active Picture complex, and more sophisti- equipment, the situation is a Errors separately from Full cated. It’s reasonable to little more complicated and Field Errors. The Active Pic- expect digital television sys- needs further definition. ture Error concept takes care tems to provide similar fault Equipment such as routing of blanking area replace- reporting capabilities. switchers, distribution ampli- ments. However, there is a Television equipment has tra- fiers and patch panels should, more insidious possibility ditionally had varying levels in general, not change the based on the television of diagnostic capabilities. data carried by the signal; design engineer’s historical With the development of dig- hence, the basic definition right to modify the blanking ital audio and video signals applies. However, when a edges. Most digital standards it’s now possible to add a routing switcher changes the were designed with digital vital tool to system fault source selection there will be blanking narrower than ana- reporting; confirmation of a short disturbance. This log blanking to ensure appro- signal integrity in studio should not be considered an priate edge transitions in the interconnection systems. To error although the length of analog domain. (Large ampli- understand and use this tool, the disturbance is subject to tude edge transitions within it’s necessary to delve into evaluation and measurement. one sample period caused by the technology of digital error SMPTE Recommended Prac- time truncating the blanking detection. The combination tice RP 168 specifies the line waveform can create out-of- of the serial-digital video in the vertical interval where band spectral components interconnect method, the switching is to take place, that can show up as excessive nature of television systems, which allows error measure- ringing after digital-to-analog and typical digital equipment ment equipment to ignore this conversion and filtering.) design lead to the use of error acceptable error. What happens in some equipment is that engineers have measurement methods that There are various types of modified the samples repre- are economical yet special- studio equipment that would senting the analog blanking ized for this application. not normally be expected to edge in order to provide a Serial digital video operates change the signal. Examples desired transition; therefore, in a basically noise-free envi- are frame synchronizers with error measurements through ronment, so traditional meth- no proc amp controls, pro- such equipment even using ods of measuring random duction switchers in a the Active Picture concept error rates are not particu- straight-through mode, and will not work. larly useful. Examination of digital VTRs in the E-to-E the overall system character- mode. All have the capability With VTRs, there’s a further istics leads to the conclusion (and are likely to) replace all limit to measurement of that a specialized burst error of the horizontal and part of errors. The original full bit measurement system will the vertical blanking inter- rate VTR formats (D-1, D-2, provide the television studio vals. In general, replacing and D-3) are designed with engineer with an effective part of the television signal significant amounts of for- tool to monitor and evaluate will destroy the error-free ward error correction and error performance of serial- integrity of the signal. This is work very well. However, it’s digital video systems. true for both component and the nature of the tape recording and reproduction process Definition of Errors composite signals as there could be ancillary data in the that some errors will not be In the testing of in-studio replaced sections of the sig- corrected. Generally the error transmission links and oper- nal. There’s an even stronger correction system identifies ational equipment a single case for composite signals as the uncorrected data and error is defined as one data the sync areas are not a sophisticated error conceal- word whose digital value strictly defined set of digits, ment systems can be applied changes between the signal hence may vary within the to make a virtually perfect source and the measuring limits of the analog signal picture. The systems are so receiver. This is significantly standard. good that tens of generations different than the analog case are possible with no human- where a range of received noticeable defect. Although

page 29 the error concealment is modification of the data. Quantifying Errors excellent, it’s almost certain Therefore, all error measure- Most engineers are familiar that error measurement sys- ment methods external to a with the concept of Bit Error tems (that by definition digital VTR are not meaning- Rate (BER), which is the ratio require perfect data repro- ful because of potential of bits in error to total bits. As duction) would find errors in blanking edge adjustment, an example, the 10-bit digital many completely acceptable error concealment methods component data rate is fields. New VTR formats that and, in some equipment, the 270 Mb/s. If there were one use bit-rate reduction will use of data-rate reduction. error per frame the BER have an additional potential The good news is that vari- would be 30/(270 x 106) = source of small but accept- ous types of error rate and 1.11 x 10-7 for 525 line sys- able data value changes. The concealment rate data are tems or 25/(270 x 106) = 0.93 data compression and available inside the VTR and x 10-7 for 625 line systems. decompression schemes are standards for reporting that Table 7-1 shows the BER for not absolutely lossless, information are being one error over different resulting in some minor developed. lengths of time for various television systems. BER is a Table 7-1. Error Frequency and Bit Error Rates useful measure of system performance where the signal-to- Time Between Errors NTSC PAL Component 143 Mb/s 177 Mb/s 270 Mb/s noise ratio at the receiver is such that noise-produced 1 television frame 2 x 10–7 2 x 10–7 1 x 10–7 random errors occur. 1 second 7 x 10–9 6 x 10–9 4 x 10–9 As part of the serial digital 1 minute 1 x 10–10 9 x 10–11 6 x 10–11 interconnect system, scram- 1 hour 2 x 10–12 2 x 10–12 1 x 10–12 bling is used to lower DC

–14 –14 –14 content of the signal and pro- 1 day 8 x 10 7 x 10 4 x 10 vide sufficient zero crossings 1 week 1 x 10–14 9 x 10–15 6 x 10–15 for reliable clock recovery. 1 month 3 x 10–15 2 x 10–15 1 x 10–15 It’s the nature of the 1 year 2 x 10–16 2 x 10–16 1 x 10–16 descrambler that a single bit error absolutely causes an –17 –17 –17 1 decade 2 x 10 2 x 10 1 x 10 error in two words (samples) 1 century 2 x 10–18 2 x 10–18 1 x 10–18 and has a 50-percent proba- bility of the error in one of Table 7-2. Error Rate as a Function of SNR the words being in the most, for Serial Digital NTSC or next to the most, significant bit. Therefore, an error Time Between Errors BER SNR SNR (dB) (volts ratio) rate of 1 error/frame will be noticeable by a reasonably 1 microsecond 7 x 10–3 10.8 12 patient observer. If it’s 1 millisecond 7 x 10–6 15.8 38 noticeable, it’s unacceptable; 1 television frame 2 x 10–7 17.1 51 but it’s even more unaccept-

–9 able because of what it tells 1 second 7 x 10 18.1 64 us about the operation of the 1 minute 1 x 10–10 19.0 80 serial transmission system. –14 1 day 8 x 10 20.4 109 Nature of Studio Digital Video 1 month 3 x 10–15 20.9 122 Transmission Systems 1 century 2 x 10–18 21.8 150 Specifications for sources of digital video signals are defined by SMPTE 259M. Although the specifications Transmitted Signal do not include a signal-to- noise ratio (SNR) typical values would be 40 dB or greater at the transmitter. Errors will occur if the SNR Signal After Equalization at some location in the system reaches a low enough value, generally in the vicin- ity of 20 dB. Figure 7-1 is a Serial Coax Cable Equalizing Source Amplifier block diagram of the basic 0.8 V 0.03 V 0.8 V serial transmitter and receiver system. An intuitive Figure 7-1. Serial transmission system. method of testing the serial

page 30 system is to add cable – a the proposed serial digital described in the proposed straight-forward method of standard it’s stated that the standard. It’s the 30 dB-of- lowering the SNR. Since expected operational dis- loss point that’s critical.) coax itself is not a significant tance is through a length of This same theoretical data noise source it’s the noise coax that attenuates a fre- can be expressed in a differ- figure of the receiver that quency of 1/2 the clock rate ent manner to show error determines the operating by up to 30 dB. That is, re- rates as a function of cable SNR. Assuming an automatic ceivers may be designed with length as demonstrated in equalizer in the receiver, less or more capability, but Table 7-3 and shown graphi- eventually, as more cable is the 30 dB value is considered cally in Figure 7-2. The added, the signal level due to to be realizable. The data in graph makes it very apparent coax attenuation causes the Table 7-2 for NTSC serial that there’s a sharp knee in SNR in the receiver to be digital transmission shows the cable length vs. error rate such that errors occur. that a 4.7 dB increase in SNR curve. Eighteen additional Based on the scrambled changes the result from 1 meters of cable (5 percent of NRZI channel code used and error-per-frame to 1 error- the total) moves operation assuming gaussian distribut- per-century. For NTSC the from the knee to completely ed noise, a calculation using calibration point for the cal- unacceptable while 50 fewer the error-function gives the culation is 400 meters of meters of cable (12 percent of theoretical values shown in Belden 8281 coax. (Various the total length) moves oper- Table 7-2. The calibration types of coax can be used ation to a reasonably safe, 1 point for this calculation is provided they have a fre- error/month. Similar results based on the capabilities of quency response__ reasonably will be obtained for other the serial-digital interface. In meeting the √ f characteristic standards where the calibration point for the calculation is 360 meters for PAL and Table 7-3. Error Rate as a Function of Cable Length 290 meters for component. Using 8281 Coax in for Serial Digital NTSC Cable length changes Time Between Errors BER Cable Length Attenuation (dB) required to maintain head- (meters) at 1/2 Clock Freq room ‘scales proportionally as shown in Figure 7-3. 1 microsecond 7 x 10–3 484 36.3 Good engineering practice –6 1 millisecond 7 x 10 418 31.3 would suggest a 6-dB margin 1 television frame 2 x 10–7 400 30.0 or 80 meters of cable, hence a 1 second 7 x 10–9 387 29.0 maximum operating length 1 minute 1 x 10–10 374 28.1 of about 320 meters in an NTSC system where the knee 1 day 8 x 10–14 356 26.7 of the curve is at 400 meters. 1 month 3 x 10–15 350 26.2 At that operating level, there 1 century 2 x 10–18 ≤338 25.3 should never be any errors (at least not in our lifetime). Practical systems includes equipment that doesn’t necessarily completely reconsti- tute the signal in terms of SNR. That is, sending the signal through a distribution amplifier or routing switcher may result in a completely useful but not completely standard signal to be sent to a receiving device. The non- standardness could be both jitter and noise, but the sharp-knee characteristic of the system would remain, occurring at a different amount of signal attenuation. Use of properly equalized and re-clocked distribution and routing equipment at intervals with adequate headroom provides virtually Figure 7-2. Calculated NTSC bit-error rate.

page 31 unlimited total transmission the Errored Second. The fol- burst. Errored seconds, and distances. lowing example demon- its inverse, error free sec- Measuring Bit Error Rates (BER) strates the benefits of the onds, do a good job of quan- errored second. Suppose a tifying this. BER measurements may be burst error causes 10,000 made directly using equip- For use in serial-digital tele- errors in two frames of video. vision systems, there are sev- ment specifically designed A BER measurement made for that purpose. Unfortu- eral disadvantages to direct for 1 minute would indicate measurement of BER: nately, in a properly operat- a 1 x 10–6 BER and a mea- ing system, say with 6 dB of surement made for 1 day 1. It must be an out-of-ser- headroom, there’s no BER to would indicate a 8 x 10–9 vice test because tradi- measure. This is because the BER. Whereas an errored sec- tional BER measurements serial digital television sys- ond measurement could use one of several defined tem generally operates in an indicate that there was one pseudo-random sequences environment that is free from second in error at a time 3 at various bit rates. random errors. If errors are hours, 10 minutes and 5 sec- 2. None of the sequences are never going to occur (as onds ago. The errored second particularly similar to the implied by Table 7-2) with method is clearly a more use- serial-digital video bit 6 dB of headroom, what’s ful measurement in this case. stream; hence, some tele- there to measure? A more A significant advantage of vision equipment will not common problem will be errored seconds vs. a straight process the test set bit pat- burst errors due to some sort BER measurement is that it’s terns. of interfering signal such as a a better measure of fitness for 3. Consider a system operat- noise spike that occurs at in- service of links that are sub- ing with a reasonable 6 dB termittent intervals spaced ject to burst errors. The serial of SNR margin with far apart in time. Another digital video system fits this respect to 1 error/frame. source could be crosstalk category because television Errors due to random that might come and go de- images are greatly disturbed noise will occur once pending on what other sig- by momentary loss in syn- every hundred years or so. nals are being used at a par- chronization. A BER mea- A BER measurement will ticular time. Also, there’s the surement could give the not be possible. poor electrical connection at same value for a single, large an interface that would cause 4. BER measurements do not burst as it does for several noise only when it’s mechan- provide meaningful data shorter scattered bursts. But ically disturbed. when the system-under- if several of the shorter test is basically noise free Because of the intermittent bursts each result in momen- but potentially subject to nature of burst errors, data tary sync failure, the subjec- burst errors. recording and communica- tive effect will be more dam- 5. Test sets for BER measure- tions engineers have defined age to the viewed picture another error measurement – ment are expensive con- than caused by the single sidering their limited application in a television system. An Error Measurement Method for Television Tektronix has developed an error detection system for digital television signals and has placed the technical details in the public domain to encourage other manufacturers to use the method. This method has proven to be a sensitive and accurate way to determine if the system is operating correctly and it has been approved for standardization by SMPTE as Recommended Practice RP 165. Briefly, the Error Detection and Handling (EDH) concept is based on making Cyclic Redundancy Code (CRC) calculations for Figure 7-3. Calculated 525/625 component bit-error rate. each field of video at the

page 32 serializer as shown in Figure 1. The CRC data is part of be routed to a central col- 7-4. Separate CRCs for the the serial-digital televi- lection point for overall full field and active picture, sion signal, thereby pro- system diagnostics. along with status flags, are viding a meaningful mea- Composite digital systems then sent with the other sure of system perfor- already must have a copro- serial data through the trans- mance. cessor to handle special tim- mission system. The CRCs 2. RP 165 can be used as an ing signals required in the are recalculated at the deseri- in-service test to automat- serial data and not allowed alizer and, if not identical to ically and electronically in the parallel data. Adding the transmitted values, an pinpoint any system fail- basic CRC capabilities to the error is indicated. Typical ures. coprocessor costs close to error detection data will be 3. For out-of-service testing, nothing. For component sys- presented as errored seconds RP 165 is sufficiently sen- tems that do not have the dif- over a period of time, and sitive to accurately define ferences between serial and time since the last errored the knee of the error rate parallel, there’s a small second. curve during stressing incremental cost which In normal operation of the tests. should be a fraction of a per- serial-digital interface, there cent of the total cost in 4. Where there are errors will be no errors to measure. equipment where it’s appro- present, RP 165 provides What’s of interest to the tele- priate to use this method. the information necessary vision engineer is the to determine errored sec- System Considerations amount of headroom that’s onds, which is more use- available in the system. That In a large serial digital instal- ful than bit-error rate. is, how much stressing could lation the need for traditional be added to the system 5. Facility is provided for waveform monitoring can be before the knee of the error measuring both full-field reduced by the systematic rate curve or crash point is and active-picture errors. application of error detection reached. As an out-of-service Optional status flags are methods. At signal source test, this can be determined also available to facilitate locations, such as cameras, by adding cable or other error reporting. DVEs and VTRs, where oper- stressing method until the 6. CRC calculation can be ational controls can affect the onset of errors. Since it’s an built into all serial trans- program signal, it will con- out-of-service test, either a mitters and receivers for a tinue to be important to ver- BER test set or RP 165 could very small incremental ify the key program signal be used. There are, however, cost. With error informa- parameters using waveform many advantages to using the tion available from a vari- monitors with serial digital RP 165 system: ety of television equip- input capabilities. The results ment, the results can then of certain technical operations, such as embedded audio mux/demux, serial link transmission, and routing switcher I/O, can be moni- tored with less sophisticated digital-only equipment provided the signal data integrity is verified. It’s the function of the RP 165 system to provide that verification. For equipment where the digital signal remains in the serial domain, such as routing switchers or distribution amplifiers, it’s not economical to provide the CRC calculation function. Therefore, a basic error detection system will be implemented as

Figure 7-4. Error detection concept.

page 33 shown in Figure 7-5. Equip- domain processing equip- the monitor’s receiver that’s ment that processes the sig- ment, consider the block dia- being used to process the signal in the parallel domain, gram in Figure 7-6. Since the nal, not the actual destina- such as VTRs, DVEs or pro- signal source doesn’t have an tion equipment. It’s possible duction switchers, should internal CRC data generator that different receiver charac- provide the transmit and one might consider a black teristics of the two pieces of receive CRC calculation nec- box to provide that function. equipment would provide essary for error detection. Unfortunately, the cost of misleading results. If the That equipment can then deserializing and reserializ- receiver being tested has an report errors locally and/or ing makes the cost pro- active loop-through, it’s use- to a central diagnostics com- hibitively high so such a sys- ful to place the error detect- puter. Routing switchers and tem is not recommended. At ing monitor at that output. other equipment operating in a receiver that doesn’t calcu- Errors detected would most the serial domain will have late the CRC, it’s possible, likely be due to the first their signal path integrity and not unreasonable, to pro- receiver, that’s the unit being verified by the error detec- vide a monitor with either tested, eliminating the differ- tion system. passive or active loop- ence in receiver sensitivities. To emphasize the impor- through that determines the To extend the idea of system tance of including the CRC signal integrity. The draw- diagnostics beyond simple calculation in parallel- back of this system is that it’s digital data error detection, it would be useful to provide an equipment fault reporting system. SMPTE 269M documents a single contact clo- sure fault reporting system and discussions have begun on computer data communications protocols for more sophisticated fault reporting. Error detection is an important part of system diagnostics as shown in Figure 7-7. Information relating to signal Figure 7-5. Using error detection (basic). transmission errors are combined with other equipment internal diagnostics to be Error Reporting sent to a central computer. A large routing switcher can economically support internal diagnostics. Using two Signal Processing Signal Routing Rx Signal Processing serial receivers a polling of and/or Tx Equipment and/or serial-data errors could be Storage Equipment (serial domain) Storage Equipment implemented. The normal (parallel domain) (parallel domain) internal bus structure could be used to send serial data to a receiver to detect input Tx = Deserialize, CRC calculation, Reserialize errors and a special output Rx = Monitor with active or passive loop-thru, CRC calculation and compare bus could be used to ensure that no errors were created Figure 7-6. Using error detection (alternate). within the large serial domain routing switcher.

Figure 7-7. System fault reporting.

page 34 8. Jitter Effects and Measurements Digital transmission systems tion of the standard in Febru- that is a multiplex of samples can operate with a consider- ary 1992, there’s a note to the from two audio channels and able amount of jitter; in fact, specification which states some ancillary data (see digi- more jitter than would be tol- “this specification is tenta- tal audio formats in Section erated in the program signal tive with further work in 3). For television, the pre- represented by the digits. progress to determine the ferred audio sampling rate is This section looks at the rea- measurement method.” In 48 kHz, clock-locked to sons behind this separation fact, both the specification video as required for digital of jitter effects – the amount and its measurement method VTRs. There are 32 bits of of jitter that may be found in are the subject of SMPTE en- data associated with each different digital transmission gineering committee work. audio sample, which pro- systems and how to measure In parallel component trans- duces a 2-channel (one serial the jitter in serial transmis- mission, jitter specifications stream) data rate of sion systems. for the 27 MHz clock signal 3.07 Mb/s. Since the channel The jitter specification in the are stated as “the peak-to- coding scheme (biphase standard for the serial-digital peak jitter between rising mark) inserts a transition for video signal is “the timing of edges shall be within 3 ns of each clock, the resulting the rising edges of the data the average time of the rising transition rate of the serial signal shall be within edge computed over at least signal can be 6.14 MHz, giv- +0.25 ns of the average tim- one field.” For NTSC compos- ing an eye pattern unit inter- ing of rising edges, as deter- ite, the specification is 5 ns. val of 163 ns. For AES audio, the jitter specification is mined over a period of one Digital audio per the AES “data transitions shall occur line.” As of the first publica- standard is a serial signal within +20 ns of an ideal jitter-free clock.” Jitter In Serial Digital Signals Jitter, noise, amplitude changes, and other distortions to the serial-digital signal can occur as it’s processed by distribution amplifiers, routing switchers, and other equipment that operate on the signal exclusively in its serial form. Figure 8-1 shows how the recovered signal can be as perfect as the original if the data is detected with a jitter-free clock. As long as the noise and jitter don’t exceed the Figure 8-1. Data recovery with a noise-free clock. threshold of the detection circuits (that’s, the eye is sufficiently open) the data will be perfectly reconstructed. 2 unit-intervals In a practical system, the clock is extracted from the serial bit stream and contains some of the jitter present in Jitter (Clock Unit Intervals) the signal. Jitter in the clock This jitter appears This jitter does not appear in the extracted clock in the extracted clock can be a desirable feature to the extent that the jitter helps position the clock edge in the middle of the eye. Large amounts of low-frequency jit- 0.25 unit-intervals ter can be tolerated in serial receivers if the clock follows Xtal oscillator 10 Hz 10 kHz Frequency the eye opening location as its LC oscillator 250 kHz 2 MHz position varies with time. An example of receiver jitter sensitivity is shown in Figure 8-2 Figure 8-2. Receiver jitter sensitivity.

page 35 where the solid line repre- tolerance decreases to a value Based on the acceptance of sents the amount of jitter that of about 0.25 unit intervals. some jitter in the extracted would cause data errors at the The break points for this toler- clock used to recover the receiver. For low jitter fre- ance curve depend on the data, two types of jitter are quencies, eye location varia- type of phase-locked loop cir- defined: tions of up to several unit cuitry used in clock extrac- 1. Timing Jitter is defined as intervals will not defeat data tion. the variation in time of extraction. As the jitter fre- the significant instants quency increases, the receiver (such as, zero crossings) of a digital signal relative to a clock with no jitter above some low frequency (about 10 Hz). 2. Alignment Jitter (or relative jitter) is defined as the variation in time of the significant instants (such as, zero crossings) of a digital signal relative to a clock recovered from the signal itself. (This clock will have jitter components above 10 Hz but none above a higher frequency in the 1 kHz to 10 kHz range.) Figure 8-3. Data recovery with extracted clock. Since the data will generally be recovered using a clock with jitter, the resulting digital information may have jitter on its transition edges as shown in Figure 8-3. This is still completely valid information since digital signal processing will only consider the high/low value in the middle of the clock period. However, if the same clock with jitter (or simple sub- multiple thereof) is used to convert the digital information to analog, errors may occur as shown in Figure 8-4. Sample values converted to analog at exactly the correct Figure 8-4. Signal errors due to clock jitter. times produce a straight line whereas use of a clock with jitter produces an incorrect analog waveform. The relative effect on the analog waveform produced using the same clock for digital-to-analog conversion and for data extraction from the serial signal depends on the amount of jitter (in unit intervals) and the number of quantizing levels in the digitized signal. Where reduced jitter is required for the digital-to- analog converter (DAC) clock, a circuit such as shown in Figure 8-5 may be used Figure 8-5. DAC with a low-jitter clock. where a high Q (crystal)

page 36 phase-locked loop produces a the clock jitter to an accept- well to follow this tradition clock with much lower jitter. able level. and not produce parallel Typical jitter specifications The situation for parallel dig- clock jitter with values in the for serial audio and seri- ital video is similar; how- region of the published spec- al/parallel video signals are ever, practical implementa- ification. For serial digital shown in Table 8-1. The tions have generally avoided systems, the engineering AES/EBU standard for serial the whole issue. That is, community is still consider- digital audio allows +20 ns most or all parallel digital ing whether low-frequency of jitter, which is appropriate video clocks have jitter in the jitter greater than the nomi- as the peak-to-peak value of 1 ns or smaller region, and nal 0.5 ns should be 40 ns is about 1/4 of a unit most or all parallel receivers allowed. That is, should interval. As explained above, do not allow or account for clocks extracted from all the DAC clock jitter require- clocks with more than that serial-digital signals be ments are considerably small amount of jitter. For appropriate as inputs to D/A tighter. A draft AES/EBU 525/625 video signals, input converters without using jit- standard specifies the DAC to a 10-bit full-range D/A ter reduction or should D/A clock at 1 ns jitter. However, converter, the allowable converters be expected to a theoretical value for 16-bit clock jitter is based on a provide the jitter reduction audio could be as small as burst frequency, 25% ampli- circuitry. 0.1 ns. The important point tude signal (approximately Measuring Jitter is that receivers for AES/EBU burst amplitude) causing less There are three oscilloscope- serial audio should be than one LSB of error. For related methods of measur- designed to handle up to high-definition signals, a ing jitter in a serial digital 40 ns of jitter, whereas the 20 MHz signal is used. In the signal. Measurement of tim- internal DAC circuitry of author's opinion, designers of ing jitter requires a jitter-free such receivers must reduce new equipment would do reference clock as shown in the top part of Figure 8-6. Table 8-1. Typical Jitter Specifications Alignment (relative) jitter is measured using a clock Standard Clock Jitter % of DAC extracted from the serial sig- Period Spec Clock Requirement nal being evaluated as shown AES/EBU Audio 163 ns 40.0 ns 25% 1.0 ns spec in the bottom part of this fig- 0.1 ns 16-bit ure. Timing jitter measure- Serial NTSC 7 ns 0.5 ns 7% 0.5 ns ments include essentially all Serial PAL 6 ns 0.5 ns 9% 0.5 ns frequency components of the jitter as indicated by the Serial Component 4 ns 0.5 ns 14% 0.5 ns dashed line leading into the Parallel NTSC 70 ns 10.0 ns 15% 0.5 ns solid line in Figure 8-7. The Parallel PAL 56 ns 10.0 ns 20% 0.5 ns jitter frequency components Parallel Component 37 ns 6.0 ns 16% 0.5 ns measured using the alignment-jitter method will Parallel HD 7 ns 1.0 ns 14% 0.1 ns depend on the bandwidth the clock extraction circuit.

Serial Low-frequency components Signal V in Oscilloscope Presentation of jitter will not be included because the extracted clock Jitter Reference follows the serial signal jit- Clock H trigger ter. These low-frequency jitter components are not sig- Timing jitter measurement nificant for data recovery provided the measurement clock extraction system has the same bandwidth as the Serial Signal V in Oscilloscope Presentation data recovery clock extraction system. All jitter fre-

Jitter quencies above a certain H trigger value will be measured with break points between the two areas at frequencies appro- Clock Extraction and Phase Locked Loop priate for the bandwidth of the clock extraction system Alignment jitter measurement used. When deriving the scope trigger it’s important to consider the frequency divi- Figure 8-6. Measurements using a clock reference.

page 37 sion that usually takes place scopes, long delays may be filter is much different than in the clock extractor. If it’s a obtained with extremely that obtained with an divide-by-ten, then word- small amounts of jitter extracted clock measure- synchronized jitter will not attributable to the delay cir- ment; hence, the results may be observed depending on cuits in the scope. At first not be a good indication of whether the sweep time of glance, this looks like a the ability of a receiver to the display covers exactly 10 straight forward way to eval- recover the data from the zero crossings. Division by a uate the eye pattern since a serial bit stream. number other than the word clock extraction circuit is not As an example, component length will ensure that all jit- required. The problem is that serial digital video has a unit ter is measured. the components of jitter fre- interval of 3.7 ns. If the The third method of observ- quency that are being mea- sweep delay is 37 ns the ing, but not measuring, jitter sured are a function of the tenth zero crossing will be is the simple but deceptive sweep delay used and that displayed. In many of today’s self-triggered scope measure- function is a comb filter. As a serializers there’s a strong jit- ment depicted in Figure 8-8. result, there are some jitter ter component at one tenth of By internally triggering the frequencies that will not be the clock frequency (due in scope on the rising (or measured and others that part to the times 10 multipli- falling) edge of the waveform will indicate twice as high a er in the serializer). A self- it’s possible to display an eye value as would be expected. triggered measurement at the pattern at a delayed time Although there’s a low-pass tenth zero crossing will not after the trigger point. Using effect with this type of mea- show that jitter. Alternately, modern digital sampling surement, the shape of the a measurement at other zero crossings may show as much as two times too high a value

Total jitter is measured of jitter. To further compli- using a reference clock cate the matter, if the one- Full Value tenth frequency jitter were exactly symmetrical (such as, a sine wave) the fifth zero Jitter crossing would also have no (Relative Value) jitter. In practice, the fifth This jitter is measured using an extracted clock crossing also shows a lot of jitter; only the tenth, twenti- eth, thirtieth, etc., crossings appear to be nearly jitter free. In troubleshooting a system, the self-triggered method can Zero

Xtal oscillator 10 Hz 10 kHz be useful in tracking down a Frequency LC oscillator 250 kHz 2 MHz source of jitter. However, proper jitter measurements to meet system specifications Figure 8-7. Measured jitter components. should be made with a clock extraction method. A SMPTE ad-hoc group is developing methods to measure and specify jitter for serial digital video, most likely incorpo- rating a defined-clock extraction system.

Figure 8-8. Self-triggered jitter observations.

page 38 9. System Testing Stress Testing important in operational test- form. The characteristic of Unlike analog systems that ing as discussed in Sections the SDI Check Field is that it tend to degrade gracefully, 6 and 8, but not as stress test- has a maximum amount of digital systems tend to work ing.) Addition of noise or low-frequency energy in two without fault until they change in risetime (within separate signals. One signal crash. This characteristic is reasonable bounds) has little tests equalizer operation and most clearly illustrated by effect on digital systems and the other tests phase-locked the discussion of error detec- is not important in stress loop operation. It’s a fully tion in Section 7. However, tests. valid signal for component other stressing functions will Cable-length stress testing digital and was originally de- produce much the same can be done using actual veloped to test D-1 recorders. result. When operating a dig- coax or a cable simulator. In the composite domain, it’s ital system, it’s desirable to Coax is the real world and not a valid signal (hence may know how much headroom most accurate method. The be considered a stress test); is available; that is, how far key parameter to be mea- however, it’s also a good test the system is away from the sured is onset of errors to perform. The SDI Check crash point. To date, there because that defines the Field is defined in proposed are no in-service tests that crash point as described in SMPTE Recommended Prac- will measure the headroom; Section 7. With an error mea- tice RP 178. but research is being pursued surement method in place, Scrambled NRZI. It’s the in this area. Therefore, out- the quality of the measure- mathematics of the scram- of-service stress tests are ment will be determined by bling process that produces required to evaluate system the sharpness of the knee of the pathological signal. Al- operation. the error curve. As an exam- though the quantizing levels Stress testing consists of ple, using 8281 coax, a five- 3FF and 000 are excluded changing one or more param- meter change in length will from the active picture and eters of the digital signal go from no errors in one ancillary data, the ratio of until failure occurs. The minute to more than one “1”s to “0”s may be quite amount of change required to error-per-second. uneven for some values of a produce a failure is a mea- To evaluate a cable simula- flat color field. This is one of sure of the headroom. Start- tor, the natural approach is the main reasons for scram- ing with the specifications in to compare its loss curve bling the signal. Conversion the serial digital video stan- with coax using a network to NRZI eliminates the polar- dard (SMPTE 259M), the analyzer. Certainly that’s an ity sensitivity of the signal most intuitive way to stress appropriate criteria; but the but has little effect on the the system is to add cable most important is still sharp- ratio of “1”s to “0”s. Scram- until the onset of errors. ness of the error curve at var- bling does indeed break up Other tests would be to ious simulated lengths. long runs of “1”s or “0”s and change amplitude or rise- Experiments have shown produces the desired spec- time, or add noise and/or jit- that good cable simulators trum with minimum low-fre- ter to the signal. Each of require a 10- to 15-meter quency content, while pro- these tests are evaluating one change in added 8281 coax viding maximum zero cross- or more aspect of the receiver to produce the no-errors to ings needed for clock extrac- performance, specifically more than one error-per-section at the receiver. However, automatic equalizer range ond change. Simulators that there are certain combina- and accuracy and receiver require a longer change in tions of data input and noise characteristics. Experi- coax (less sharp knee) are scrambler state that will, oc- mental results indicate that still useful for comparison casionally, cause long runs cable-length testing is the testing but should be avoided (20 to 40) of “0”s. These long most meaningful stress test in equipment evaluation. runs of “0”s create no zero crossings and are the source because it represents real SDI Check Field operation. Stress testing of low-frequency content in receiver ability to handle The SDI (serial digital inter- the signal. (All “1”s is not a amplitude changes and face) Check Field (also problem because this pro- added jitter are useful in known as “pathological sig- duces an NRZI signal at one- evaluating and accepting nal”) is not a stress test; but half the clock frequency, equipment, but not too because it’s a full-field test which is right in the middle meaningful in system opera- signal, it must be an out-of- of the frequency band occu- tion. (Measuring the signal service test. It’s a difficult pied by the overall signal.) amplitude at the transmitter signal for the serial digital The scrambler is a nine-cell and measuring jitter at vari- system to handle and is a shift register as shown in Fig- ous points in the system is very important test to per- ure 9-1. For each input bit,

page 39 there is one output bit sent to of this analysis the “1”s are with relative ease because of the one-cell NRZI encoder. high and the “0”s are low as their very short duration. An output bit is determined opposed to the transition, no- Multiple long run events can by the state of the 10 cells transition data definition.) be encouraged with certain and the input bit. A typical The longer runs (16, 19, 22, input signals, which is the distribution of runs of “1”s 32, 33, 34, and 39) occur as basis for the generation of and “0”s is shown in Table single events that the trans- pathological signals. When 9-1 for 100 frames of compo- mission system amplifiers the input words alternate nent black. (For the purpose and receivers can handle between certain digital values the multi-event patholog- Table 9-1. 100 Frames of Component Black ical signal will occur for one Length “1”s “0”s Length “1”s “0”s specific scrambler state, generally all “0”s, hence the 1 112615438 112610564 21 sequence usually starts at an 2 56303576 56305219 22 73759 13579 SAV. The multi-event 3 28155012 28154963 23 sequence ends at the next EAV, which breaks the input 4 14076558 14077338 24 sequence. Since the scram- 5 7037923 7038595 25 bler state is more or less a 6 3518756 3518899 26 random number, the multi- 7 1759428 1760240 27 event sequence of long runs of “0”s happens about once 8 1059392 699666 28 per frame (9-bits, 512 possi- 9 610241 970950 29 ble states, 525 or 625 lines 10 1169 1060 30 per frame). 11 14 156 31 Pathological Signals, the SDI 12 40 32 55 51 Check Field. There are two forms of the pathological sig- 13 156 114 33 49 nal as shown in Figure 9-2. 14 34 3 For testing the automatic 16 13713 73865 36 equalizer, a run of 19 “0”s followed by two “1”s pro- 19 101 94 39 50 duces a signal with large DC content. Remember a “0” is no transition and a “1” is represented by a transition in the NRZI domain. In addition, the switching on and off of this high DC content signal will stress the linearity of the analog capabilities of the serial transmission system and equipment. Poor analog amplifier linearity results in errors at the point of transition where peak signal Figure 9-1. Serial scrambler and encoder. amplitude is the greatest. It’s important to produce both polarities of this signal for full system testing. Phase- 19 bits 19 bits locked loop testing is accom- 1 bit plished with a signal consisting of 20 NRZI “0”s followed by a single NRZI “1”. This Signal for testing automatic cable equalization provides the minimum number of zero crossings for clock extraction. The SDI Check Field consists 20 bits 20 bits of one-half field of each signal as shown in Figure 9-3. Equalizer testing is based on Signal for testing PLL Lock-in the values 300h and 198h

Figure 9-2. Signal forms.

page 40 Table 9-2. 100 Frames of SDI Check Field while phase-locked loop test- Length “1”s “0”s Length “1”s “0”s ing is based on the values 200h and 110h. In the C-Y 1 112595598 112595271 21 order shown, the top half of 2 56291947 56292583 22 9344 9425 the field will be a purple 3 28146431 28147229 23 here and the bottom half a 4 14073813 14072656 24 shade of gray. Some test signal generators use the other 5 7035240 7035841 25 order, Y-C, giving two differ- 6 3518502 3518856 26 ent shades of green. Either 7 1758583 1758969 27 one works, but the C-Y order is specified in the proposed 8 879203 879359 28 recommended practice. One 9 706058 706029 29 word with a single bit of “1” 10 51596 51598 30 is placed in each frame to en- 11 146 154 31 sure both polarities of equalizer stress signal for the de- 12 16858 16817 32 49 49 fined SDI Check Field. The 13 34405 34374 33 35 32 single “1” inverts the polar- 14 17445 17442 34 3 4 ity of the NRZ to NRZI func- 16 9558 9588 36 tion because the field (for component video) would 18 29 27 38 otherwise have an even num- 19 20944 19503 39 13 14 ber of “1”s and the phase 20 19413 19413 40 would never change. Statistics for 100 frames of SDI Check Field are shown in Table 9-2. It’s the runs of 19 and 20 that represent the low-frequency stressing of the system. When they happen (about once per frame) they occur for a full active television line and cause a significant low-frequency disturbance. Since the PLL stress signal is a 20 “1”s, 20 “0”s squarewave, you would expect an even number of each type of run. The equalizer stress signal is based on runs of 19 “1”s followed by a “0” or 19 “0”s followed by a “1”. As described above, an extra “1” in each frame forces the two polarities to happen and both are represented in the data. Figure 9-3. SDI check field.

page 41 10. Bibliography Articles and Books

Ainsworth, Kenneth. “Moni- Fibush, David and Elkind, Mendelson, Philip. Digital toring the Serial Digital Sig- Bob. “Test and Measurement Magic, Santa Monica, CA. nal,” Broadcast Engineer- of Serial Digital Television “Building a Serial Compo- ing, Nov. 1991. Tektronix lit- Systems,” 133rd SMPTE nent Facility,” Broadcast En- erature number 25W-7156. Technical Conference, Los gineering, Mar. 1992, p. 38, Ainsworth, Kenneth and Angeles, CA, Oct. 1991. 42, 44, 48, 50. Andrews, Gary. “Evaluating SMPTE Journal, Sept. 1992, Martin, Bruno. Advanced Serial Digital Video Sys- p. 622-631. Tektronix litera- Audio Visual Systems, Mon- tems,” IEEE 4th Interna- ture number 25W-7154. treuil, France. “A Compre- tional Montreux Sympo- Fibush, David. “Understand- hensive Method of Digital sium, Montreux, Switzer- ing the Serial Digital Video Video Signal Analysis,” land, June 1991. Interface,” In View, Midwest 135th SMPTE Technical Ainsworth, Kenneth and Publishing, Winter/Spring Conference, Los Angeles, Andrews, Gary. “Compre- 1991, p. 16-18. California, Nov. 1993. hending the Video Digital Fibush, David. “Error Detec- Reynolds, Keith. Grass Val- Serial,” Broadcast Engineer- tion in Serial Digital Sys- ley Group, Grass Valley, CA. ing, Spanish language edi- tems,” NAB, Las Vegas, NV, “The Transition Process: Get- tion, Sept. 1992, p. 34-44. Apr. 1993. SMPTE Journal, ting from A to D,” Broadcast Tektronix literature number Aug. 1993, p. 688-692. Engineering, Mar. 1992, 25W-7153. Tektronix literature number p. 86-90. Avis, Richard, Sony Broad- 25W-7198. Robin, Michael and Poulin, cast & Pro-Audio Division, Fibush, David. “Jitter Effects Michel. CBC, Montreal, Basingstoke, U.K. “Serial and Measurements in Serial Canada. “Performance Evalu- Digital Interface,” Booklet. Digital Television Signals,” ation and Acceptance Test- Sony Broadcast & Commu- 26th SMPTE Conference, ing of Bit-Serial Digital Video nications, Basingstoke, Eng- Feb. 1993. SMPTE Journal, Equipment and Systems at land, June 1991. Aug. 1993, p. 693-698. the CBC,” 134th SMPTE Elkind, Bob. Ainsworth, Ken- Tektronix literature number Technical Conference, neth. Fibush, David. Lane, 25W-7179. Toronto, Canada, Nov. 1992. Richard. Overton, Michael. Fibush, David. “Error Mea- Robin, Michael. CBC, Mon- “Error Measurements and surements in Studio Digital treal, Canada. “Bit-serial Dig- Analysis of Serial CCIR-601 Video Systems,” 134th ital Signal Jitter Causes, Equipment,” International SMPTE Technical Confer- Effects, Remedies and Mea- Broadcaster's Convention, ence, Toronto, Canada, Nov. surement,” 135th SMPTE Amsterdam, Holland, June 1992. SMPTE Journal, Aug. Technical Conference, Los 1992. 1993, p. 688-692. Tektronix Angeles, California, Nov. Elkind, Bob and Fibush, Dav- literature number 25W-7185. 1993. id. “Proposal for Error Detec- Fibush, David. “Test and Stremler, Ferrel G. “Introduc- tion and Handling in Studio Measurement of Serial Digi- tion to Communications Sys- Equipment,” 25th SMPTE tal Signals error detection tems,” Addison-Wesley Television Conference, De- and jitter,” SBE Convention, Series in Electrical Engi- troit, MI, Feb. 1991. SMPTE San Jose, CA, Oct. 1992. neering, December 1982. Journal, Dec. 1991. Grass Valley Group, “Digital Smith, David R. Digital Tektronix literature number Television: The Transition,” Transmission Systems, Van 25W-7128-1. Grass Valley Group, Inc., Nostrand Reinhold Com- Eguchi, Takeo. Sony Corpo- Grass Valley, CA. 1992 pany, New York, 1985. ration, Japan. “Pathological Huckfield, Dave. Sony Stone, David. Pro-Bel Ltd. Check Codes for Serial Digi- Broadcast & Communications “Digital Signal Analysis,” In- tal Interface Systems,” 133rd Ltd. “Serial Digital Interfac- ternational Broadcast Engi- SMPTE Technical Confer- ing,” International Broad- neer, Nov. 1992, p. 28-31. ence, Los Angeles, CA, Oct. cast Engineer, Sept. 1990. 1991. SMPTE Journal, Aug. 1992, p. 553-558.

page 42 Tancock, M.B., Sony Broad- Ward, Ron. Utah Scientific, Wonsowicz, John. CBC, cast & Communications, Bas- Salt Lake City, UT. “Avoid- Toronto, Canada. “Design ingstoke, U.K. “The Digital ing the Pitfalls in Serial Digi- Considerations for Television Evolution: The Serial Digital tal Signal Distribution,” Facilities in the CBC Toronto Interface,” Sony, Bas- SMPTE Journal, Jan. 1993, Broadcast Center,” 134th ingstoke, U.K. Booklet p. 14-23. SMPTE Technical Confer- released at IBC Convention, Watkinson, John. The Art of ence, Toronto, Canada, Nov. Brighton, England, Sept. Digital Video, Focal Press, 1992. 1990. London & Boston. Walker, Marc S. BTS, Salt Watkinson, John and Rum- Lake City, UT. “Distributing sey, Francis. The Digital In- Serial Digital Video,” Broad- terface Handbook, Focal cast Engineering, Feb. 1992, Press, London & Boston. p. 46-50.

Standards

AES 3 (ANSI S4.40), “AES Report ITU-R BT.624, “Char- SMPTE 267M, “Bit Parallel Recommended Practice for acteristics of Television Sys- Digital Interface, Component Digital Audio Engineering, tems” Video Signal 4:2:2, 16x9 Serial Transmission Format Report ITU-R BT.1212, Aspect Ratio” for Linearly Represented Dig- “Measurements and Test Sig- SMPTE 269M, “Fault Report- ital Audio Data” nals for Digitally Encoded ing in Television Systems” AES 11 (Draft ANSI S4.44), Color Television Signals” SMPTE RP 154, “Reference “AES Recommended Practice EBU Tech 3246, “EBU Paral- Signals for the Synchroniza- for Digital Audio Engineer- lel Interface for 625-line Digi- tion of 525-line Video Equip- ing, Synchronization of Digi- tal Video Signals” ment” tal Audio Equipment in Stu- EBU N-10, “Parallel Compo- SMPTE RP 168, “Definition dio Operations” nent Video Interface for Non- of Vertical Interval Switching Recommendation ITU-R composite ENG Signals” Point for Synchronous Video BT.601, “Encoding Parame- SMPTE 125M, “Bit-Parallel Switching” ters of Digital Television for Digital Interface, Component SMPTE RP 165, “Error Detec- Studios” Video Signal 4:2:2” tion Checkwords and Status Recommendation ITU-R SMPTE 170M, “Composite Flags for Use in Bit-serial BT.656, “Interfaces for Digi- Analog Video Signal, NTSC Digital Television Interfaces” tal Component Video Signals for Studio Applications” SMPTE RP 178, “Serial Digi- in 525-line and 625-line tal Interface Check Field for Television Systems Operat- SMPTE 244M, “Digital Rep- 10-bit 4:2:2 Component and ing at the 4:2:2 level of Rec- resentation of NTSC Encoded 4ƒsc Composite Digital Sig- ommendation 601” Video Signal” nals” Recommendation ITU-R SMPTE 259M, “Serial Digital BT.711, “Synchronizing Ref- Interface for 10-bit 4:2:2 Note: Former CCIR recommendations and reports are erence Signals for the Com- Component and 4ƒsc NTSC Composite Digital Signals” now ITU-R documents. The ponent Digital Studio” International Telecommuni- cation Union, Radio Com- munication Sector has replaced the CCIR.

page 43 11. Glossary 125M – See SMPTE 125M. bandwidth – 1. The differ- blocking – Occurs in a multi- 4:2:2 – A commonly-used ence between the upper and stage routing system when a term for a component digital lower limits of a frequency, destination requests a source video format. The details of often measured in megahertz and finds that source un- the format are specified in (MHz). 2. The complete available. In a tie line sys- the CCIR-601 standard docu- range of frequencies over tem, this means that a desti- ment. The numerals 4:2:2 which a circuit or electronic nation requests a tie line and denote the ratio of the sam- system can function with receives a “tie line busy” pling frequencies of the sin- less than a 3 dB signal loss. message, indicating that all gle luminance channel to the 3. The information carrying tie lines are in use. two color-difference chan- capability of a particular tele- BNC – Abbreviation of “baby nels. For every four lumi- vision channel. N connector.” A cable con- nance samples, there are two baseline shift – A form of nector used extensively in samples of each color differ- low-frequency distortion television. ence channel. See CCIR-601. resulting in a shift in the DC byte – A complete set of 4ƒsc – Four times subcarrier level of the signal. quantized levels containing sampling rate used in com- bit – A binary representation all of the bits. Bytes consist- posite digital systems. In of 1 or 0. One of the quan- ing of 8 to 10 bits per sample NTSC, this is 14.3 MHz. In tized levels of a pixel. are typical. PAL, this is 17.7 MHz. bit parallel – Byte-wise cable equalization – The pro- AES/EBU – Informal name transmission of digital video cess of altering the frequency for a digital audio standard down a multi-conductor ca- response of a video amplifier established jointly by the ble where each pair of wires to compensate for high-fre- Audio Engineering Society carries a single bit. This stan- quency losses in coaxial and European Broadcasting dard is covered under cable. Union organizations. SMPTE 125M, EBU 3267-E CCIR – International Radio algorithm – A set of rules or and CCIR 656. Consultative Committee, an processes for solving a prob- bit serial – Bit-wise transmis- international standards com- lem in a finite number of sion of digital video down a mittee, now replaced by steps. single conductor such as co- ITU-R. aliasing – Defects in the pic- axial cable. May also be sent CCIR-601 – (See ITU-R ture typically caused by through fiber optics. This BT.601.) standard is covered under insufficient sampling or poor CCIR-656 – The physical par- CCIR 656. filtering of digital video. De- allel and serial interconnect fects are typically seen as jag- bit slippage – 1. Occurs scheme for CCIR-601. CCIR gies on diagonal lines and when word framing is lost in 656 defines the parallel con- twinkling or brightening in a serial signal so the relative nector pinouts as well as the picture detail. value of a bit is incorrect. blanking, sync, and multi- analog – An adjective This is generally reset at the plexing schemes used in describing any signal that next serial signal, TRS-ID for both parallel and serial inter- varies continuously as op- composite and EAV/SAV for faces. Reflects definitions in posed to a digital signal that component. 2. The erroneous EBU Tech 3267 (for 625-line contains discrete levels rep- reading of a serial bit stream signals) and in SMPTE 125M resenting the binary digits 0 when the recovered clock (parallel 525) and SMPTE and 1. phase drifts enough to miss a 259M (serial 525). bit. 3. A phenomenon which asynchronous – A transmis- channel coding – Describes occurs in parallel digital data sion procedure that is not the way in which the 1s and buses when one or more bits synchronized by a clock. 0s of the data stream are rep- gets out of time in relation to resented on the transmission A-to-D Converter (analog-to- the rest. The result is errone- path. digital) – A circuit that uses ous data. Differing cable digital sampling to convert lengths is the most common character generator (CG) – A an analog signal into a digital cause. computer used to generate representation of that signal. text and sometimes graphics bit stream – A continuous for video titles or captions. series of bits transmitted on a line. clock jitter – Timing uncer- tainty of the data cell edges in a digital signal.

page 44 clock recovery – The recon- compression artifacts – Com- differential phase – A change struction of timing informa- pacting of a digital signal, in chrominance phase of a tion from digital data. particularly when a high video signal caused by a coaxial cable – A transmis- compression ratio is used, change in luminance level of sion line with a concentric may result in small errors in the signal. pair of signal carrying con- the decompressed signal. digital word – The number of ductors. There’s an inner These errors are known as bits treated as a single entity conductor and an outer con- “artifacts,” or unwanted by the system. defects. The artifacts may re- ductive metallic sheath. The discrete – Having an individ- semble noise (or edge “busy- sheath aids in preventing ual identity. An individual ness”) or may cause parts of external radiation from af- circuit component. fecting the signal on the the picture, particularly fast dither – Typically a random, inner conductor and mini- moving portions, to be dis- low-level signal (oscillation) mizes signal radiation from played with the movement which may be added to an the transmission line. distorted or missing. analog signal prior to sam- composite digital – A digi- coding – Representing each pling. Often consists of white tally encoded video signal, level of a video signal as a noise of one quantizing level such as NTSC or PAL video, number, usually in binary peak-to-peak amplitude. form. that includes horizontal and vertical synchronizing infor- dither component encoding – coefficients – A number mation. A slight expansion of the (often a constant) that analog signal levels so that contouring – Video picture expresses some property of a the signal comes in contact defect due to quantizing at physical system in a quanti- with more quantizing levels. too coarse a level. tative way. The results are smoother component analog – The D1 – A component digital transitions. This is done by unencoded output of a cam- video recording format that adding white noise (which is era, videotape recorder, etc., uses data conforming to the at the amplitude of one quan- consisting of three primary CCIR-601 standard. Records tizing level) to the analog sig- color signals: green, blue, on 19mm magnetic tape. nal prior to sampling. (Often used incorrectly to and red (GBR) that together drift – Gradual shift or indicate component digital convey all necessary picture change in the output over a video.) information. In some compo- period of time due to change nent video formats, these D2 – A composite digital or aging of circuit compo- three components have been video recording format that nents. Change is often caused translated into a luminance uses data conforming to by thermal instability of signal and two color-differ- SMPTE 244M. Records on components. ence signals, for example, Y, 19mm magnetic tape. (Often D-to-A converter (digital-to- B-Y, R-Y. used incorrectly to indicate analog) – A device that con- composite digital video.) component digital – A digital verts digital signals to analog representation of a compo- D3 – A composite digital signals. nent analog signal set, most video recording format that DVTR – Abbreviation of digi- often Y, B-Y, R-Y. The en- uses data conforming to tal videotape recorder. coding parameters are speci- SMPTE 244M. Records on fied by CCIR 601. The paral- 1/2" magnetic tape. EAV – End of active video in component digital systems. lel interface is specified by delay – The time required for CCIR 656 and SMPTE 125M a signal to pass through a EBU – European Broadcast- (1991). device or conductor. ing Union. An organization of European broadcasters composite analog – An demultiplexer (demux) – A that, among other activities, encoded video signal, such device used to separate two produces technical state- as NTSC or PAL video, that or more signals that were ments and recommendations includes horizontal and ver- previously combined by a for the 625/50 line television tical synchronizing informa- compatible multiplexer and system. tion. transmitted over a single channel. EBU TECH.3267-E – The EBU recommendation for the deserializer – A device that parallel interface of 625-line converts serial digital infor- digital video signal. A revi- mation to parallel. sion of the earlier EBU differential gain – A change Tech.3246-E, which in turn in chrominance amplitude of was derived from CCIR-601 a video signal caused by a and contributed to CCIR-656 change in luminance level of standards. the signal.

page 45 EDH (error detection and gain – Any increase or microsecond (µ) – One mil- handling) – Proposed SMPTE decrease in strength of an lionth of a second: 1 x 10–6 or RP 165 for recognizing inac- electrical signal. Gain is mea- 0.000001 second. curacies in the serial digital sured in terms of decibels or Miller squared coding – A signal. It may be incorpo- number of times of magnifi- DC-free channel coding rated into serial digital cation. scheme used in D-2 VTRs. equipment and employ a group delay – A signal defect MPEG-2 – Motion pictures simple LED error indicator. caused by different frequen- expert group. An interna- Equalization (EQ) – Process cies having differing propa- tional group of industry ex- of altering the frequency gation delays (delay at perts set up to standardize response of a video amplifier 1 MHz is different from delay compressed moving pictures to compensate for high-fre- at 5 MHz). and audio. quency losses in coaxial horizontal interval (horizon- multi-layer effects – A cable. tal blanking interval) – The generic term for a mix/effects embedded audio – Digital time period between lines of system that allows multiple audio is multiplexed onto a active video. video images to be combined serial digital data stream. interpolation – In digital into a composite image. encoder – In video, a device video, the creation of new multiplexer (mux) – Device that forms a single, compos- pixels in the image by some for combining two or more ite color signal from a set of method of mathematically electrical signals into a sin- component signals. manipulating the values of gle, composite signal. neighboring pixels. error concealment – A tech- nanosecond (ns) – One bil- nique used when error cor- I/O – Abbreviation of lionth of a second: 1 x 10–9 or rection fails (see error correc- input/output. Typically 0.000000001 second. tion). Erroneous data is refers to sending information NICAM (near instantaneous replaced by data synthesized or data signals to and from companded audio multiplex) from surrounding pixels. devices. – A digital audio coding sys- error correction – A scheme ITU-R – The International tem originally developed by that adds overhead to the Telecommunication Union, the BBC for point-to-point data to permit a certain level Radio Communication Sector links. A later development, of errors to be detected and (replaces the CCIR). NICAM 728 is used in sev- corrected. ITU-R BT.601 – An interna- eral European countries to eye pattern – A waveform tional standard for compo- provide stereo digital audio used to evaluate channel per- nent digital television from to home television receivers. formance. which was derived SMPTE nonlinear encoding – Rela- field-time (linear) distortion – 125M (was RP-125) and EBU tively more levels of quanti- An unwarranted change in 3246E standards. CCIR zation are assigned to small video signal amplitude that defines the sampling sys- amplitude signals, relatively occurs in a time frame of tems, matrix values, and fil- fewer to the large signal 16 ms. ter characteristics for both Y, peaks. B-Y, R-Y and GBR compo- format conversion – The pro- nonlinearity – Having gain nent digital television. cess of both encoding/decod- vary as a function of signal ing and resampling of digital jaggies – Slang for the stair- amplitude. step aliasing that appears on rates. NRZ – Non return to zero. A diagonal lines. Caused by in- frequency modulation – coding scheme that is polar- sufficient filtering, violation Modulation of a sinewave or ity sensitive. 0 = logic low; of the Nyquist Theory, “carrier” by varying its fre- 1 = logic high. and/or poor interpolation. quency in accordance with NRZI – Non return to zero jitter – An undesirable ran- amplitude variations of the inverse. A video data scram- dom signal variation with modulating signal. bling scheme that is polarity respect to time. frequency response rolloff – insensitive. 0 = no change in A distortion in a transmis- MAC – Multiplexed Analog logic; 1 = a transition from sion system where the higher Component video. This is a one logic level to the other. frequency components are means of time multiplexing not conveyed at their original component analog video full amplitude and possible down a single transmission loss of color saturation. channel such as coax, fiber, or a satellite channel. Usu- ally involves digital processes to achieve the time compression.

page 46 NTSC (National Television phase error – A picture resolution – The number of Systems Committee) – Orga- defect caused by the incor- bits (four, eight, ten, etc.) nization that formulated rect relative timing of a sig- determines the resolution of standards for the NTSC tele- nal in relation to another the digital signal. vision system. Now describes signal. 4-bits = a resolution of 1 in the American system of color phase shift – The movement 16 telecasting which is used in relative timing of a signal 8-bits = a resolution of 1 in mainly in North America, in relation to another signal. 256 Japan, and parts of South pixel – The smallest distin- America. 10-bits = a resolution of 1 guishable and resolvable area in 1024 Nyquist sampling theorem – in a video image. A single Eight bits is the minimum Intervals between successive point on the screen. In digital acceptable for broadcast TV. samples must be equal to or video, a single sample of the less than one-half the period picture. Derived from the RP 125 – See SMPTE 125M. of highest frequency. words picture element. routing switcher – An elec- orthogonal sampling – Sam- PRBS – Pseudo random tronic device that routes a pling of a line of repetitive binary sequence. user-supplied signal (audio, video signal in such a way video, etc.) from any input to production switcher (vision that samples in each line are any user-selected output(s). mixer) – A device that allows in the same horizontal transitions between different sampling – Process where position. video pictures. Also allows analog signals are measured, PAL (Phase Alternate Line) – keying and matting often millions of times per The name of the color televi- (compositing). second for video. sion system in which the V propagation delay (path sampling frequency – The component of burst is length) – The time it takes for number of discrete sample inverted in phase from one a signal to travel through a measurements made in a giv- line to the next in order to circuit, piece of equipment, en period of time. Often ex- minimize hue errors that or a length of cable. pressed in megahertz for may occur in color video. transmission. quantization – The process of converting a continuous SAV – Start of active video in parallel cable – A multi-con- analog input into a set of dis- component digital systems ductor cable carrying simul- crete output levels. taneous transmission of data scope – Short for oscillo- bits. Analogous to the rows quantizing noise – The noise scope (waveform monitor) or of a marching band passing a (deviation of a signal from its vectorscope, devices used to review point. original or correct value) measure the television which results from the quan- signal. patch panel – A manual tization process. In serial dig- method of routing signals scrambling – 1. To transpose ital, a granular type of noise using a panel of receptacles or invert digital data accord- only present in the presence for sources and destinations ing to a prearranged scheme of a signal. and wire jumpers to inter- in order to break up the low- connect them. rate conversion – 1. Techni- frequency patterns associated cally, the process of convert- with serial digital signals. peak to peak – The ampli- ing from one sample rate to 2. The digital signal is shuf- tude (voltage) difference another. The digital sample fled to produce a better spec- between the most positive rate for the component for- tral distribution. and the most negative excur- mat is 13.5 MHz; for the sions (peaks) of an electrical serial digital – Digital infor- composite format it’s either signal. mation that is transmitted in 14.3 MHz for NTSC or serial form. Often used infor- phase distortion – A picture 17.7 MHz for PAL. 2. Often mally to refer to serial digital defect caused by unequal used incorrectly to indicate television signals. delay (phase shifting) of dif- both resampling of digital serializer – A device that ferent frequency components rates and encoding/decoding. within the signal as they pass converts parallel digital Rec. 601 – See CCIR-601. through different impedance information to serial digital. elements – filters, amplifiers, reclocking – The process of SMPTE (Society of Motion ionospheric variations, etc. clocking the data with a Picture and Television Engi- The defect in the picture is regenerated clock. neers) – A professional orga- “fringing,” like diffraction nization that recommends rings, at edges where the standards for the television contrast changes abruptly. and film industries.

page 47 SMPTE 125M (was RP 125) – sync word – A synchronizing time-multiplex – In the case The SMPTE recommended bit pattern, differentiated of CCIR-601, a technique for practice for bit-parallel digi- from the normal data bit pat- transmitting three signals at tal interface for component terns, used to identify refer- the same time on a group of video signals. SMPTE 125M ence points in the television parallel wires (parallel defines the parameters re- signal; also to facilitate word cable). quired to generate and dis- framing in a serial receiver. TRS – Timing reference sig- tribute component video sig- telecine – A device for cap- nals in composite digital sys- nals on a parallel interface. turing movie film as a video tems (four words long). SMPTE 244M – The SMPTE signal. TRS-ID (timing reference sig- recommended practice for temporal aliasing – A visual nal identification) – A refer- bit-parallel digital interface defect that occurs when the ence signal used to maintain for composite video signals. image being sampled moves timing in composite digital SMPTE 244M defines the too fast for the sampling rate. systems. It’s four words long. parameters required to gener- A common example is wagon truncation – Deletion of ate and distribute composite wheels that appear to rotate lower significant bits on a video signals on a parallel in- backwards. digital system. Usually terface. time base corrector – Device results in digital noise. SMPTE 259M – The SMPTE used to correct for time base VTR (video tape recorder) – recommended practice for errors and stabilize the tim- A device which permits 525-line serial digital compo- ing of the video output from audio and video signals to be nent and composite interfaces. a tape machine. recorded on magnetic tape. still store – Device for storage TDM (time division multi- waveform – The shape of an of specific frames of video. plex) – The management of electromagnetic wave. A synchronous – A transmis- multiple signals on one graphical representation of sion procedure by which the channel by alternately send- the relationship between bit and character stream are ing portions of each signal voltage or current and time. slaved to accurately synchro- and assigning each portion to word – See byte. nized clocks, both at the particular blocks of time. receiving and sending end.

page 48

For further information, contact Tektronix: World Wide Web: http://www.tek.com; ASEAN Countries (65) 356-3900; Australia & New Zealand 61 (2) 888-7066; Austria, Eastern Europe, & Middle East 43 (1) 701 77-261; Belgium 32 (2) 725-96-10; Brazil and South America 55 (11) 3741 8360; Canada 1 (800) 661-5625; Denmark 45 (44) 850700; Finland 358 (9) 4783 400; France & North Africa 33 (1) 69 86 81 08; Germany 49 (221) 94 77-400; Hong Kong (852) 2585-6688; India 91 (80) 2275577; Italy 39 (2) 250861; Japan (Sony/Tektronix Corporation) 81 (3) 3448-4611; Mexico, Central America, & Caribbean 52 (5) 666-6333; The Netherlands 31 23 56 95555; Norway 47 (22) 070700; People’s Republic of China (86) 10-62351230; Republic of Korea 82 (2) 528-5299; Spain & Portugal 34 (1) 372 6000; Sweden 46 (8) 629 6500; Switzerland 41 (41) 7119192; Taiwan 886 (2) 722-9622; United Kingdom & Eire 44 (1628) 403300; USA 1 (800) 426-2200 From other areas, contact: Tektronix, Inc. Export Sales, P.O. Box 500, M/S 50-255, Beaverton, Oregon 97077-0001, USA (503) 627-1916

Copyright © 1997, Tektronix, Inc. All rights reserved. Tektronix products are covered by U.S. and foreign patents, issued and pending. Information in this publication supersedes that in all previously published material. Specification and price change privileges reserved. TEKTRONIX and TEK are registered trademarks of Tektronix, Inc. All other trade names referenced are the service marks, trademarks, or registered trademarks of their respective companies. 9/97 TD/XBS 25W–7203–3