Super Motion System

By L. J. Thorpe, T. Nakamura, and K. Ninomiya

A radically new television field-acquisition system designed to enhance The quality of a given picture the capture of motion is described. The Super Motion system is com- frame. The spatial resolution of that prised of a new genre of camera and a new VTR system. frame, which is obviously adversely The motion judder and image blur associated with conventional TV impacted by the fact that the subject is in motion for the brief period during pictures when played on a standard slow-motion recorder have been which the frame is being exposed or overcome. This is accomplished by shooting the scene with a camera scanned, and complicated further by whose frame rate is three times as high as that of a normal 525/30 the finite integration time of today’s camera. New technologies incorporated in the system - mixed-field photoconductive pickup tubes. pickup t ubes,f iber-optic transmission, digital processing and standards These two factors contribute di- conversion, and high-speed C-MOS digital LSI chips to encode RGB to rectly to the subjectively disturbing NTSC-are described. phenomena that mar conventional Type-C slow-motion playback. The specific symptoms are: ince the introduction of the Type- Development of the Super Motion Staccato motion. A direct result S C 1-in. videotape recorder, slow- System” at the 126th SMPTE Tech- of inadequate temporal resolution motion replay has gained wide nical Conference in November 1984. and the repeat field mechanism em- popularity by permitting close exami- This paper dealt in some detail with ployed in Type-C slow-motion play- nation of movement in sports, science, the theory of the special VTR em- back. and drama. Increasingly, it is even ployed. The emphasis of this article Picture blur. Associated with the used as a special effect in commer- will be on the special Super Motion frame rate of the taking camera and cials and drama. Dynamic tracking camera. the finite integration time of the (DT) technology (essential to this fa- pickup devices used. cility) has become an indispensable The Slow-Motion Problem It has long been recognized that the element of this format. This slow-mo- poor reproduction of movement (typi- tion ability is eagerly sought within While adding immeasurably to the fied by the erratic or staccato nature smaller-format systems, such as 3/4-in. production capabilities of today’s of this motion), particularly when the and the new l/z-in. VTR, the technical quality of conven- playback is substantially slowed In the film domain, the problem is tional 1- in. Type-C slow-motion play- down, is subjectively a much more tackled by increasing the shutter back is fundamentally flawed. TV disturbing aspect. The associated speed to multiples of the normal 24 cameras shoot at a fixed field rate of problem of picture blur actually alle- frames/sec (camera frame rate). 60 Hz, and thus deliver to the VTR a viates this problem by masking the When the film is subsequently pro- finite number of video “snapshots.” jerkiness via the integration or blur- jected at a fixed speed of 24 frames/ This is a relatively slow taking rate, ring phenomenon. In playback of sec, the motion is reproduced at a rate particularly when fast motion is in- many sports activities, the split-sec- proportionally slower (by the same volved, and much motion resolution is ond relative motion (as in racing) or multiple as the shutter was increased) inherently lost. the beauty of form (as in gymnastics) but with the same quality as would be The Type-C VTR can be controlled is primarily impaired by the poor time seen in the projection of a normal- to play back the recorded tape at any resolution of the system. speed picture. The viewer is looking at linear speed between still-frame and The Super Motion system was de- all of the original high-speed frames, normal speed of 24.4 cm/sec. To re- veloped specifically to improve the over a longer period of time, and thus play in slow motion, the appropriate first of these problems. That is, it is sees slow motion with higher tempo- intermediate speed is set, and for the designed to substantially enhance the ral resolution. VTR to produce a standard 525/60 fluidity of the slow motion (also inci- The Super Motion system was de- NTSC output, the video head must dentally reducing the motion blur), veloped to solve a specific problem trace the same track multiple times rather than to improve the clarity of a associated with conventional slow- (this multiple varying according to stop-action freeze-frame picture. motion playback of television record- speed of slow-motion replay). The ings. Sony Broadcast Products pre- limited number of frames recorded sented a paper entitled “ T h e initially is thus further reduced, caus- Fundamentals of the Super ing increased motion impairment. Motion System There are two important aspects re- The Super Motion system emulates Presented at the 19th Annual SMPTE Television Conference in San Francisco (paper No. 19-20) on lating to the impaired quality of the high-speed photography in the video February 16,1985. L. J. Thorpe (who read the paper) reproduced slow-motion image: domain. This is accomplished by is affiliated with Sony Broadcast Products Corp., Tea- 0 The resolution the motion it- shooting the scene with a camera op- neck. N.J. T. Nakamura and K. Ninomiya are with of Sony Communications Products Group, Atsugi. Ja- self. The clarity and smoothness of erating at a frame rate that is three pan. This article was received February 6, 1985, and the movement, which is purely a func- times as high as that of a normal TV also appears in Components ofthe Future. published scan rate (30 frames/sec). The Super SMPTE, 1985. Copyright 0 1985 by the Society of tion of the system’s temporal resolu- Motion Picture and Television Engineers, Inc. tion. Motion camera runs at 90 frames/

896 SMPTE Journal, September 1985

sec. In '/3o sec, it therefore photo- and it fits nicely with the generally ments in the spatial resolution, for graphs three sequential video frames accepted 4.2 MHz for the NTSC sig- example, to double the horizontal and in place of the one single frame cap- nal. vertical spatial resolution as in the tured by a conventional camera. If an In the Super Motion camera, all 1125/60 system. That same band- object is in motion, three snapshots of parameters are assumed to remain width, however, can just as well be the progress of that motion are cap- the same, except fF which now be- applied to the third important dimen- tured, rather than just one, in '/30 sec. comes 90, and this produces: sion of TV, the temporal or time di- The BVP-3000 Super Motion cam- mension. This is the premise of the era scans 525 at a 90- fmax = 112 X 0.7 X 1.33 X 525 X 90 Super Motion camera. The spatial Hz frame rate, or 180 fields inter- X I resolution is essentially that of today's = 12.6 MHz laced 2:l. To record subsequently this NTSC signal, and all of the band- unusual signal and play it back as a The significant fact is that there is a width is employed to sustain the de- standard 525/30-Hz NTSC signal, so proportional relationship between the mands of the higher frame rate (Eq. that it will deliver an appropriate sig- frame rate of a camera and the band- 1). Thus, in its deployment of band- nal into the existing TV signal, re- width of which it must be capable. In width, the Super Motion system is a quires the process of standards con- conventional TV color cameras, the variation on the theme of high defini- version. The 90-Hz frame rate must originating RGB set is actually wider tion - high temporal definition. be converted down to the 30-Hz stan- than the 3-dB minimum of 4.2 MHz In discussing the special processing dard. This conversion is accomplished for the NTSC system, being more required, we refer to a set of three in the camera control unit (CCU) pri- typically 6 or 8 MHz to allow proper high-speed fields occupying the same or to recording. aperture-correction schemes to be time duration as a TV field in a con- The elevation of the camera frame employed and to produce full band- ventional 525/60 camera. Each of the rate, while keeping the number of width for picture processing, etc., high-speed pictures is expanded in scanning lines constant at 525, has within a studio. The Super Motion time in the standards conversion pro- important repercussions on two key camera has been designed with a cess, and is then separated into three facets of camera design: pickup de- bandwidth closer to 18 to 20 MHz. individual pictures. Each of these re- vice technology and bandwidth. The In one sense, the Super Motion constructed fields is recorded in real bandwidth consideration is important camera is a high-definition camera. time on a specially designed high- because it forces us into the realm of All of the design techniques - the speed Super Motion VTR. The three high-definition television. wideband pickup requirements, high- original high-speed fields from the The minimum bandwidth required er scanning rates, and wideband camera totalled '/60 sec in time. Their to reproduce all of the information preamplifier and video processing - recorded time, on tape, now totals 3/60 within a specific TV standard is given are those required, for example, in an sec. Therefore, playback of the tape at approximately by the following ex- 1 125/60 HD camera. The use of that standard Type-C speed (24.4 cm/sec) pression: bandwidth, however, is quite differ- will reproduce the original picture ent. with one-third the speed of motion. In an HDTV camera, as we nor- This picture has all the resolution of a mally speak of it, the wide bandwidth normal 525/60 camera/recorder is employed to gain radical improve- combination, running a t normal where K is the Kell factor, generally assumed to have a value of approxi- mately 0.7; A is the aspect ratio; N is the total number of scanning lines in a single interlaced frame; fF is the frame repetition rate; Kh is the hori- zontal retrace ratio; and K, is the ver- tical retrace ratio. As a basis for examination, it is useful to consider first the 525/30 NTSC system where the relevant pa- rameter values are as follows:

A = 413 = 1.33 N = 525 f~= 30 Kh = 63.5110.9 = 5.8 K,, = 262.5121 = 12.5.

which produces for fmax a value of finax = 1 1 2 X 0.7 X I .33 X 525 X 30 X 1.08 = 4.17 M H z This equation defines the minimum bandwidth to cater to a given line number, frame rate, and aspect ratio, 7gure 1. Super Motion system basic architecture.

SMPTE Journal, September 1985 897

speed, but with the important addi- capable of playback on any standard the CCU can simultaneously deliver a tion that the VTR plays back three Type-C machine, regardless of manu- standard NTSC output and the re- times the number of frames. facturer. quired signals for the special high- The camera system will deliver a speed recorder. Figure 1 shows a Operational System fully standard 525/30 NTSC video block diagram of the total Super Mo- Requirements output, in real time, capable of feed- tion system. The camera and the VTR have cer- ing directly to air. In addition to these prime require- tain operational system requirements 0 The system will be genlockable ments, there are other key operational which form the foundation for much so that it can be incorporated into a requirements, as follows: of the system architecture. These re- total TV system. The camera head will also be ca- quirements were pivotal in making The camera head will be porta- pable of operation with large field key design decisions in the camera ble to allow for hand-holding or shoul- lenses and a large studio-type view- processing and VTR recording meth- der mounting, thus permitting mobil- finder to permit a more traditional ods. Briefly stated, they are as fol- ity in a crowded sports environment. type of sports coverage with the cam- lows: While the recording process is spe- era mounted on a tripod. The final output from the system cialized, the playback is standard. The CCU and VTR will be con- will be a 1- in. recording, fully meeting The camera is also specialized (in that nected via a short multi-core cable, so Type-C SMPTE specifications and it runs at higher scanning rates), but that these units can be located togeth- er in a mobile van for operational con- trol convenience. Existing test equipment such as waveform monitors and picture moni- tors will be used for operational ad- justment and maintenance. The camera head will connect to the CCU via a single cable capable of being extended to 1 km length, again for operational flexibility in the field. The CCU operation will be as conventional as possible to minimize operator training requirements. A separate remote video opera- tor panel will allow flexibility in cam- era control. 0 The Super Motion VTR is capa- ble of playing back any standard Type-C tape. That is, the machine can be used as a normal playback machine when not required for Super Motion effects. The Super Motion VTR uses standard time code and is fully com- patible with the RS-422 remote-con- trol system employed by standard Figure 2. BVP-3000 camera system and BVH-2700 VTR. Sony 1-in. VTRs.

898 SMPTE Journal, September 1985

Figure 2 is a photograph of the sys- tem. Super Motion Camera System The Super Motion camera head is relatively simple. Figure 3 shows the basic details of one of the RGB video channels. It is an RGB camera em- ploying three2/3-in. mixed-field Plum- bicon pickup tubes. Following pream- plification, the wideband RGB signals first undergo basic linear processing: black-and-white level normalization and balancing, black-and-white shad- ing corrections, and linear matrixing. This is followed by nonlinear process- ing such as black-and-white clipper and gamma correction. These three wideband 20-MHz analog signals are then fed to the camera cable for deliv- ery to the CCU. Within the CCU, the final process- ing is performed on the RGB signals. They are then standards-converted, Figure 4. Short mixed-field compared with standard s i n . tube. and the three high-speed frames are individually separated into three par- allel 525/60 video signals. Because the CCU is almost entirely digital, it subsequently directed to produce a In the Super Motion camera, the is called the digital processing unit smaller 2/,-in. mixed-field tube (Fig. total charge removed is one-third that (designated DPU-3000). 4). This tube was developed to sup- of a conventional camera, assuming port future portable HDTV camera constant lumens of light on the face- Camera Head development, and with appropriate plate in either case. The total scan- refinements, a high-performing ver- ning period for one field is also re- Because the camera is still required sion was produced to form the prima- duced by a factor of three, so the ratio to scan 525 lines within every high- ry pickup element for a new range of of AQlAt remains constant. Hence, speed frame, the horizontal scanning 525/60 standard broadcast cameras the signal current level from a pickup must also be increased in speed. That currently being introduced by Sony tube is essentially independent of is, to scan 525 lines (two fields) in ‘/go (BVP-360, BVP-3A, BVP-30, and scanning speed. sec instead of ‘/30 sec, the horizontal BVP-150). This same 2/3-in. tube is The total generated in the scan rate must be increased to 47 kHz employed in the Super Motion BVP- front-end preamplifier is represented (15.734 X 3 kHz). Horizontal scan- 3000. by an effective root-mean-square ning is, even today, extremely diffi- (RMS) noise current according to: cult to accomplish satisfactorily at the Camera Sensitivity standard rate of 15.734 kHz using The BVP-3000 employs a standard conventional magnetic deflection high-performance, high-index, techniques. It would be virtually im- prism-splitting block. It interfaces possible at three times this rate. with all lenses currently available for where the incremental noise current Sony’s recently introduced mixed- the standard BVP-300 and BVP-3A/ squared is: field tube, however, employs totally 30 series cameras. Optical efficiency electrostatic deflection while retain- is therefore identical to that of Sony’s S: o( ~ 2 f [2 c , + ciSsR]2, (4a) ing a dc magnetic focusing system. standard three-tube broadcast cam- where is the output capacitance of This new electron optical system was eras. The efficiency of transformation C,, developed specifically to permit preci- the tube assembly; Cis, is the input of the light lumens on the individual capacitance of the field-effect transis- sion deflection for high-definition tube faceplate to signal current out- tor (FET); Re is the equivalent input tubes operating at, for example, an put requires some explanation. 1125/60 scanning rate. A new diode resistance of the FET; and B is the The output signal current from a bandwidth. gun and a new target layer also con- pickup tube is given by: tribute to the very small beam capa- This RMS noise current thus has bility. High corner resolution is a ma- I, = dQ/dt (2) the familiar triangular noise spec- jor benefit of this new tube, in that trum represented in Fig. 5. In the much better orthogonal landing of the or approximately by: downconversion employed in the camera DPU, the high-frequency beam all over the raster can be main- I,?= AQ/At (3) tained with this technology. noise is also converted down. Thus, This pickup tube research, which where A Q is the total charge removed the effective triangular equivalent initially centered on a 25-mm tube for in one field; and Af is the period of one noise current becomes as shown. The an 1125/60 H D camera, was also field. total equivalent noise current is in-

SMPTE Journal, September 1985 899

I

Figure 5. Process of noise downconversion in the Super Motion system.

creased approximately by a factor of Sony designed and manufactured separate field memories (Fig. 7). This three, thus degrading the camera sig- large-scale integrated (LSI) circuits memory bank converts the single 180- nal-to-noise ratio (SNR) by approxi- designated CX-20116, which have a field/sec video into three separate and mately 9 dB. This deterioration is al- capability of 100-MHz maximum distinct 60-field/sec signals. Each of leviated, however, by use of a new sampling rate but are operated here at the R, G, and B 8-bit parallel digital FET developed by Sony, which has a a 43-MHz rate (or 12 X 3.58-MHz signals enters its memory bank as a lower Cis,a nd a lower R,. This same subcarrier). It is an 8-bit system. serial sequence of high-speed fields. FET employed in a normal camera The digitized video is then passed Each of these component video sig- would produce an SNR in excess of 60 through a digital aperture-correction nals (which go into memory as a 525/ dB for a 4.2-MHz bandwidth. The system, both horizontal and vertical, 180 scanning format) is subsequently Super Motion BVP-3000 system has and is then delivered to a field mem- clocked out at the slower 60-field rate. an output SNR of about 51 dB for a ory system. This is a large digital The input high-speed fields, which 4.2-MHz bandwidth with a green sig- memory organized into a bank of nine are serial in time, are read out in par- nal current of 200 nA. The capacitive lag (for constant light level and output signal current) is increased by a factor of three as a result of the high-speed scanning. This effect is reduced somewhat in the BVP-3000 by the use of a totally new mixed-field Plumbicon tube that fea- tures a diode gun and a low-capaci- tance photoconductive layer. Bias light is also operated at a higher level than in a normal 2/3-in. camera. Camera Processing Unit The wideband 20-MHz RGB ana- log signals are received from the cam- era cable into the DPU-3000 camera processing unit (Fig. 6) and are passed through a variable equalizer that can be switched to compensate for the particular cable length em- ployed. Where fiber-optic link is used, 1 I this equalization is automatic. The three video signals are then fed to three A/D converters. These are Figure 6. Block schematic of the DPU-3000 camera system. 900 SMPTE Journal, September 1985

allel according to the scheme shown in Fig. 7. That is, high-speed field 1, for the red video, is clocked out of mem- ory block 1A as a 525160 scanning format. Field 2 is clocked out '/go sec later from memory block 2A, also as a 525160 signal. Similarly, field 3 comes from memory block 3A, de- layed by a further '/go sec. Field 4 comes from memory block 1B and so on. These signals are clocked out of memory at 14.3 MHz (4 X 3.58-MHz subcarrier). Thus nine separate digi- tal video signals are produced at a converted scanning rate of 525160 - three red, three green, and three blue. Each original high-speed 180-Hz ser- ial set of three fields becomes three ~~ ~ parallel individual standard 60-Hz Figure 7. DPU-3000 digital memory organization. fields, displaced in time relative to each other (Fig. 8). The next step is to formulate three encoded NTSC-type video signals from these three sets of RGB. Each set of time-coincident R, G, and B is ma- trixed, filtered, and encoded, all in the digital domain. A number of custom- built (by Sony) low-power digital C- MOS LSI chips perform this encod- ing (Fig. 9). Each of the three encoders is formulated from combi- nations of four basic LSI chips: CX-23059 Y,I.Qm atrix chip CX-23056 1.Q filter and Y delay chip CX-23057 I,Q modulator and sync mixer blanking chip CX-23058 delta delay block Three of the labeled chips are self- explanatory. The delta delay block LSI is required to match data timing Figure 8. Separation and expansion of high-speed TV fields.

Figure 9. Digital encoding using LSI building blocks.

SMPTE Journal, September 1985 001

between the other LSIs. The low- dinal recording speed, allows three Employment of current video re- power C-MOS devices employ pipe- times the amount of information (of a cording technology line techniques to overcome speed normal Type-C recorder) at standard Basing the BVH-2700 on a prov- limitations. scanner speed. Three video tracks are en tape transport system The DPU-3000 arranges the three recorded per revolution of the scan- Enhanced flexibility of the sys- channel signals so that the correct col- ner. Recording of the three fields tem (can play back all Type-C record- or field sequence is recorded. This en- takes place simultaneously because of ings). coding is performed in the digital do- the spatial 120° displacement of the The substantial increase in longitu- main to ensure absolute equality of three heads and the corresponding dinal tape speed, by a factor of three, the three luminance and 1 20° relative time displacement of actually increases the writing speed signals. the three incoming video signals. Fig- by about 2% above that of Type-C. ure 1 1 indicates the recording meth- The precise Type-C footprint on tape Super Motion BVH-2700 VTR odology. The adherence to essential is carefully ensured by an increase in Type-C writing speed allows: The Super Motion VTR has three the helix angle, also by about 2%. recording heads, each positioned on the scanner at 12 0° relative spatial separation. During recording, the drum rotates at 60 Hz,a nd each of the three heads is separately driven by its previously described respective indi- vidual video signal from the DPU- 3000 (Fig. 10). The BVH-2700 has a parallel three-channel recording sys- tem in contrast to the single video recording channel emplayed in a stan- dard Type-C recorder. The recording tape linear speed is 73.2 cm/sec, or three times that of a conventional Type-C recorder (3 X 24.4 cm/sec). The use of three record heads, at three times normal longitu-

Figure 11. BVH-2700 recording method.

902 SMPTE Journal, September 1985

Figure 12. Operational configurations for the BVH-2700: (a) Super Motion system; (b) VTR playback only.

This, coupled with an effective reduc- speed, in which instance the VTR will machine. It cannot be used to record a tion in head speed, guarantees the play back every third video track and standard NTSC signal. correct Type-C track angle on tape. display a picture identical to that To reduce the head speed appropri- from conventional Type-C recording Conclusion ately, one of two approaches is possi- techniques. If variable slow motion is The Super Motion system was spe- ble: a reduction in the drum diameter required, this utilizes field-repeat cifically developed to alleviate the or a small amount of time compres- methods similar to standard Type-C. most pressing limitation of conven- sion of the incoming record video. The Because we are starting with three tional slow-motion playback in televi- former expedient could cause difficul- times the number of fields, the clarity sion, the poor portrayal of movement. ties in recovering sufficient signal for of motion remains substantially im- The system was not designed to ad- stable vertical sync detection. The proved all the way down to still- dress exclusively the clarity of a still- time-compression technique, using frame. Even at very slow speeds and frame picture. HDTV technology the digital techniques of the Betacam still-frame, the picture clarity is con- permitted the design and practical system, was therefore employed. This siderably improved because of the implementation of a high-frame-rate solved the potential problem of still- shorter exposure time of the frame. camera. The high frame rate of 90 frame playback and allowed the gen- Image blur, due to prolonged subject Hz, compared to the standard 30 Hz, eration of the proper head-speed vec- movement within a given frame, is dramatically elevates the capture of tor. considerably reduced because of the fast motion and preserves, in slow- There are two operational configu- 5-msec (approximately) field time. motion VTR playback, a superb fluid- rations for the BVH-2700 VTR sys- Most of the residual image blurring is ity and smoothness of movement. tem. When it is employed in the Super a consequence of the inherent integra- This has been verified in a consider- Motion configuration (Fig. 12a), the tion characteristics of the pickup de- able number of sports events covered recorder operates in the high-speed vice over that time period. by the Super Motion system during mode during recording, as described The BVT-2700 time-base corrector the past year. above. In playback, one head is used (TBC) is similar to that employed A novel method of separating and to separately trace one track at a time. with a standard Sony Type-C ma- recording all of the captured high- The requirements are essentially chine. It has a wider correction win- speed video frames, while maintain- those of a Type-C VTR except in the dow to meet the extended DT require- ing absolute adherence to Type-C re- area of still-mode. In playback, with ments of slow-motion playback on the cording specifications, was developed. the VTR running at one times stan- BVH-2700. This imbued the system with an im- dard playback speed of 24.4 cm/sec, Figure 12b shows the second portant operational flexibility. High- the result is one-third slow-motion configuration, where the BVH-2700 definition technology, digital video playback, with full NTSC bandwidth VTR and the BVT-2700 TBC can be processing, and Betacam technology, pictures and greatly enhanced motion used to play back any Type-C tape, along with further refinement of DT portrayal. independent of the DPU-3000 proces- technology, all coalesced to produce For real-time playback, the tape sor. Thus, the BVH-2700 system can the Super Motion recording system. must travel at three times normal be used as a normal Type-C playback @

SMPTE Journal, September 1985 903