F. D. KELL, Ldr. - Recording Mechanics and J. D. RITTENHOUSE, Ldr. - Recording Systems and Development - Applied Research DEP, Camden, N. J.

F. D. KELL received the BSME degree from Drexel Institute of Technology in 1957. As an undergraduate, he worked with RCA on cooperative assignments. Now, at the University of Pennsylvania he is a candidate for the MSME. In 1957, he became a permanent member of DEP Applied Research. He has since made major contributions to the development of a family of helical-scan video recorders and was responsible for the introduction of air lubrication to the headwheel mechanism of commercial video recorders. In addition, Mr. Kell also headed the development of a precise satellite tape transport which featured extremely low power consumption and inherent angular momenta cancellation. His contributions to this and other programs have resulted in 14 patent applications. In 1962 he was promoted to Leader, Recording Mechanics in DEP Applied Research.

J. D. RITTENHOUSE received the BSEE from Drexel Institute of Technology in 1958. Upon joining RCA in 1958, he became a member of the RCA graduate study program, on which he received his MSEE from the University of Pennsylvania in 1960. He then joined DEP Applied Research, where he is now an Engineering Leader. He has concentrated on recording and data-processing systems. Recently he has been concerned with the systems concepts of transverse and helical-scan equipments for defense applications. His group is responsible for the development work on helical scan in RCA. On recent projects he has been concerned with various encoding techniques for recording information on tape and has studied narrowband FM modulation techniques, synthesizing the record-playback processes on a computer. Mr. Rittenhouse is a member of Eta Kappa Nu, Tau Beta Phi. and Phi Kappa Phi.

Reviewed are recent developments in recording electronics and tape transports for wideband recorders - longitudinal, transverse, and helical-scan transports; solutions to switching transients and time-base stability problems; and data-reduction techniques. Reductions in switching transients and advances in time-base stability, combined with multichannel capabilities of helical-scan equipment, answer many demands of radar data reduction by eliminating need for data multiplexing. A further advantage in the instrumentation field is the ability to combine real-time data storage and data readout into a single magnetic recording system.

Problems encountered in recording wideband data on magnetic tape derive from the implications of the simple, basic equation:

Formel

Where: lambda = wavelength of the recorded signal, v = relative velocity between the record-reproduce head and the tape, f = frequency of the signal being recorded.

For a given frequency, the shorter the recorded wavelength the lower the head-to-tape velocity required. However, recorded wavelength (or, more exactly, the resolvable wavelength) is fixed by the state of the art in head design and manufacture. Therefore, the shortest wavelength is, in general, fixed by the best head structure available. Then, for a specified frequency, the head-to-tape velocity required is calculated. With the available wavelength resolution fixed, the method of obtaining the appropriate scanning velocity becomes the primary consideration in wideband recording.

Since this velocity is obtained by maintaining a relative motion between the tape and heads, design of the tape transport is of paramount importance.

To appreciate problems of designing tape transports, consider some typical parameters.

For example, if a recording head is capable of resolving a wavelength of 150 microinches (0.000150 inch) and if a frequency response of 7 Mc (MHz) is desired, the head-to-tape speed required is in the order of 1,000 ips. The "fm" bandwidth resulting from using the 7-Mc direct-record bandwidth is of the order of 3.5 to 4.0 Mc. The 1,000-ips head-to-tape speed poses the mechanical problem of obtaining such head-to-tape speed while realizing a practical and economical transport.

There are two basic approaches to solving this problem.

(1) The longitudinal-scan method, used several years ago in the RCA Broadband Recorder and recently in the recorders for Tradex1 and the Precision Tape System, moves the tape at the required speed over a stationary head.

(2) The second approach holds the tape essentially motionless while the head moves past it; this technique, featuring rotary scanning, has been realized at RCA in two forms.

(2a) The more familiar rotary scan is the transverse-scan technique employed for television recording; this technique has been extended in many cases into the instrumentation field.

(2b) The other form of rotary scan is the helical-scan technique.

Several laboratory models, one field-type 3-channel radar recorder, and six 2-channel radar recorders have been constructed using this technique, which offers several advantages over the longitudinal scan and the transverse scan.

Each type of machine (the longitudinal scan, the transverse scan, and the helical scan) has some advantages over the other two. The selection of type of scan for a particular program must be based on an evaluation of these advantages and disadvantages.

The main advantages of the longitudinal scan are: the availability of many parallel channels (up to 15 channels on a 1 inch tape), the absence of switching transients caused by commutation of heads (there is no commutation of heads), and the realization of a speed reduction of data in a relatively simple manner.

The main disadvantage of the longitudinal scan is that at the head-to-tape speed involved, massive tape reels are required for a practical recording period. For instance, the Tradex machine (5-minute recording period) requires 36-inch reels holding 7 miles of tape.

Contrary to common belief, each of the three types of scan exhibits tape interchangeability problems, even the longitudinal type. The specific problems of the longitudinal scan are manifested in the strict requirements for channel alignment across the multichannel head, for precision mounting of the head, and for the tape-tension servo.

The main advantages of the transverse-scan technique are its high-level performance and its acceptance as the standard for the television industry.

Approximately 2,000 television broadcast units are in use throughout the world. A typical transverse-scan assembly is shown in Fig. 1. The most important elements of this unit include the head-wheel (i) in which the video heads (2) are mounted, the headwheel motor (3), and the vacuum guide {4).

The vacuum guide forms the tape around the head-wheel and controls the depth the heads indent the tape. The scan assembly shown is a two-channel unit which requires eight heads to continuously record two channels; each head records and reproduces during only slightly more than 90° of every rotation. In operation, the tape moves through the vacuum guide parallel to the axis of rotation of the headwheel, thus producing the pattern of the transverse scan.

The helical scan combines the advantages of the transverse scan with the simplicity and multichannel capability of the longitudinal scan. However, it is not well known to most people in the recording industry, whereas both the transverse and longitudinal scans are.

The helical scan also has a problem of head commutation, but the rate of commutation is usually slower than that employed on a transverse-scan recorder. However, new switching techniques, described later in this paper, permit practical negation of the switching transients which result from commutation of the heads.

The helical-scan recorder generates a diagonal scan pattern on the tape by moving the tape helically around a scanning wheel in an assembly which is illustrated in Figs. 2 and 3.

The most significant elements of this helix assembly are the two helical guides (1) and (2), the headwheel (3) and the video heads (4).

The drives for this scan assembly are similar to the transverse scan, with separate drives for headwheel and tape. The wrap of the tape about the headwheel is formed naturally by the helical tape path, while the head pressure against the tape is determined by the balance between tape tension and the pressure of the air supplied to lubricate the tape as it moves around the helix.

In Fig. 3, the helix is shown pivoted into the position for headwheel maintenance. The particular headwheel shown is a two-channel assembly which contains three video heads spaced 120° apart. During every rotation of the head-wheel each of these heads records and reproduces during 250°, which through appropriate switching, is sufficient for recording two continuous channels.

In general, the number of heads in a helical-scan system is greater by one than the number of information channels, whereas four heads are used for every information channel in transverse scan because of the limited tape wrap angle.

Although transverse-scan and helical-scan equipments are best for obtaining very large bandwidths and reasonable storage times, they suffer from two imperfections: time-base error (a problem common to all types of recorders) and commutation transients (a problem peculiar to rotary-head recorders). Solutions to both of these problems are presented below. Although developed specifically for a transverse-scan recorder, the techniques described are applicable also to a helical-scan recorder.

Frequency modulation is the mode of encoding information for most instrumentation applications; this fact is the key to complete negation of switching transients through a technique known as "fade switching".

Assume the rotary headwheel in Fig. 4 to be rotating in a clockwise direction. During this rotation, the tape, as shown, may be considered to be advancing out of the page.

During the recording process, the four heads can be energized in parallel. Therefore, we can be certain that the one or two heads contacting the tape will be recording. Because of the multiplicity of heads, recording presents no particular problem. There is no transient generated in the record process.

In the reproduce mode, however, a different situation prevails. Since only the head, or heads, in contact with the tape can reproduce a signal, the output from any given head is discontinuous, and a continuous output can be achieved only by a proper combination of the outputs of all four heads.

The envelopes of the outputs from the four heads are shown in Fig. 4. Note that overlap of the envelopes provides signal redundancy.

In recently constructed recording systems (the Dual-Channel, Dual-Speed equipments) where the headwheel speed is 480 rps, the overlap time is approximately 70usec.

During these 70 usec, the "switcher," which combines the waveforms of Fig. 4 into a single continuous waveform, must "switch" its output from the head leaving the tape to the head engaging the tape. It is during this time interval, when the switcher moves its output from one head to another, that the "switching transient" is generated.

Fundamentally, the switching transient results from the difficulty of maintaining an exact time relationship between the signals from the head leaving the tape and the signals from the head engaging the tape. Angular errors in the positioning of the heads, head wear, and changes in tape dimensions combine to defeat efforts to eliminate the transient mechanically.

It is not always sufficient to provide an adjustment so that an operator can tune out these errors. Tape jitter and vibration, plus inability to maintain absolutely constant headwheel speed, cause errors of a highly dynamic nature, changing the time base of the recorder not only during the switching interval, but also during the passing of a single head across the tape.

The reduction of these dynamic errors to about 100 nsec is the best attainable at present with electromechanical systems, an accuracy not approached in conventional longitudinal tape recorders.

Switching transients and tenth-microsecond timing errors can be tolerated in many applications. For example, in broadcast television the transients are hidden in the horizontal retrace of a picture line. Also, for certain instrumentation applications, the generation of a switching transient is not particularly bothersome since the switching transient may be synchronized with the recorded information, or it may be identified and dealt with accordingly.

With good switching diodes, the switching discontinuity can be reduced to a width of 100 nsec with an amplitude no greater than the peak noise level at the output of the recorder. This technique, known as rapid "video" switching, has been applied successfully in several types of instrumentation recorders.

However, the ideal rotary-head recorder should have no switching transients or information dropout due to head commutation; techniques for the elimination of these objectionable parameters are described below.

The 100-nsec dynamic error remaining after final electro-mechanical servo correction may be composed of slow-rate headwheel errors, switching errors due to inaccurate positioning of the vacuum shoe, and "quadrature" effects due to interchangeability problems.

In order to explain the effects of fade switching we will, for the moment, neglect the quadrature and headwheel servo effects, and will concentrate on only those due to the vacuum guide (or tape shoe).

The magnitude of a typical shoe error might be 30 nsec; this error occurs as an instantaneous jump in recorder time base at the commutation rate of the heads. The transient elimination scheme is initiated by combining the four outputs of the heads in Fig. 4 into two channels of information, as shown in Fig. 5.

From this figure we can draw two important conclusions:

1) redundant data exists during overlap time,
2) the maximum instantaneous phase error during overlap is 30 nsec.

Neglecting for the moment that the 30-nsec time-base error may be reduced further, we may conjecture as to transient levels as we commutate these two channels of information into one final serial channel.

It was previously mentioned that frequency modulation is the key to the transient-free switch. If the "fm" signal is switched instantaneously, a 30-nsec time-base error will occur and will appear as a 30-nsec phase discontinuity in the carrier. Such a discontinuity will provide an impulse transient at the output of the fm discriminator.

If, however, these two channels fade into one another during the overlap time, a gradually changing phase error exists on the "fm" carrier. The rate of change of phase is small, producing a negligible transient at the output of the discriminator. Assuming an "fm" carrier of 10 Mc, an interval of 30 nsec corresponds to (30/100) x 360 = 108° of phase shift which occurs through a fade period of 50usec.

The nonlinearity of the "fm" system requires that the exact level of this transient be obtained on a computer. Normally, a total of 2 Mc will be deviated so that the switching transient becomes (2 x 106) (6 x 103) = 333 units below our peak output signal. It should also be noted that fade switching is effective only up to 180° of phase differential between the switched waveforms. At this point the 180° relationship between the added carriers results in complete carrier cancellation, which produces a transient from the fm discriminator. Fig. 6 shows typical fade switch performance.

Based on these results, it can be said that for all practical purposes, there is no switching transient.

Another major problem, time-base stability, has been solved in response to the demands for more sophisticated radar systems. The time-base system described below is the elite of a group of systems that employ electronically variable delay lines for time-base correction.

It has been noted that 100 nsec of error is the minimum obtainable from the electromechanical system. Consequently, if further reductions in time-base error are made, they must be achieved through devices other than the basic servos. Analysis of the error rates involved reveals that the sampling rate and the bandwidth of the error-nullifying device must be higher than those of the basic servos if rapid rate errors are to be significantly reduced.

The familiar "once-around" indicator, or "tonewheel," no longer possesses enough information to describe the time base of the video signal as the head scans the tape. For this reason, other signals which describe time-base errors must be utilized. Television tape recorders have such a signal in the form of horizontal sync pulses. The general-purpose radar or sine-wave recorder, however, has no preknowledge of the characteristics of its input signal; therefore, another means of generating an accurate time-base signal is necessary.

Happily, the unique vestigial sideband fm system employed in the record-reproduce process permits a means of establishing such a time-base signal. A portion of available tape bandwidth is not occupied by any significant sideband spectra. It is, therefore, used to provide a record of time base during recording by adding in an accurate small-amplitude sinusoid from a crystal source (see Fig. 7).

During playback this sinusoid, when compared with the same crystal, provides a means of measuring time-base error. It is important to realize that this time standard is recorded on the same track with the information signal, thus providing an accurate measure of the time-base error experienced by that signal during recording and reproduction.

After a means of error detection has been established, the means for nullifying that error must be accomplished. As is often the case, techniques developed by the Broadcast Division's Electronic Recording Group were borrowed and changed to provide the desired error-correction capabilities. Specifically, electronically variable delay lines are employed to nullify the time-base errors.

The closed-loop block diagram shown in Fig. 8 is employed on the Dual-Channel, Dual-Speed recording equipment. A reference sinusoid frequency of 1 Mc is used. The closed-loop approach permits the realization of several additional
benefits. Notice that the lines are used in the fm domain, permitting time-base correction before final fade switching, thus reducing the phase error prior to the fade switch.

Furthermore, time-base correction in the fm domain, if accomplished quickly enough, can dynamically compensate for quadrature effects, thus solving many interchangeability problems. This closed-loop system, under test at this writing, is achieving a time-base accuracy of ±10 nsec. The rms value of that error is less than five nanoseconds, equivalent to less than 5 feet of range error, a new standard of excellence for radar recording accuracy.

So far we have described recent improvements in the general performance capabilities of scan-type recorders used as data-storage devices. In many instances, however, tape equipment may also facilitate data reduction or data presentation by providing time compression, time expansion or repetitive readout.

An equipment developed recently provides a 200:1 expansion of real time, which permits on-line computer reduction of data. This equipment incorporates two transverse-scan tape machines. One records and reproduces two channels of wideband data in real time; the second accepts tapes prepared by the first and reproduces them at l/200th the real time rates.

The tape and head-wheel drives developed for this equipment provide extremely accurate, yet slow, speeds; both use hysteresis synchronous motors as prime movers. The speeds of the output members are reduced through spliceless Mylar belts similar to those employed in Tiros recorders. The speed of both assemblies is precisely controlled through servos which modulate the frequency of the motor power.
Fig. 9 shows the head-wheel assembly for this time-expansion equipment.
.

One potential capability of the helical-scan equipment which has been explored for several data-processing systems is the ability to repetitively read out in real time a desired segment of data. Basically, this is accomplished simply by stopping the tape with the desired information track around the scanning wheel.

However, since the diagonal track was recorded with the tape moving, it does not align perfectly with the scanning path of the headwheel when the tape is stationary. Alignment, however, can be effected by reading on a helix assembly which is slightly larger in diameter than the record helix assembly. The slight difference in diameter causes a slight difference in the scan angle between head and tape and permits a complete track to be continuously scanned by the wheel. The length of the time segment which can be read out repetitively in such an equipment is limited only by the system bandwidth and the diameter of the helix assembly. The system bandwidth determines the head-to-tape speed, and the diameter of the helix assembly determines the length of the track.

The combination of time compression and repetitive readout has been proposed for data which is to be presented on visual displays. In a typical application the data is received at data and frame rates which would cause flicker upon display. A typical solution to this problem is a tape equipment which incorporates two helical-scan assemblies.

One of these assemblies records incoming data with a slow-moving recording head; the second assembly reproduces the data at a higher rate. If the recorded tracks are made sufficiently wide, the readout head will reproduce the same data repetitively. This repetitive readout will increase the output frame rate. Continuous processing by this technique requires that the ratio of frame rates be equal to the ratio of data rates.

The advances attained in time-base accuracy and switching-transient reduction have significantly improved the usefulness of scan-type recorders for instrumentation applications. The combination of these techniques with the multichannel capabilities of the helical-scan equipment will answer many demands of radar data-reduction systems by eliminating the need for data multiplexing.

A further enhancement of the role of tape equipment in the instrumentation field in many instances is yielded by the ability to combine real-time data-storage requirements and data-readout requirements into a single magnetic-recording system.

In conclusion, the tailoring of high-performance tape equipment to the needs of a specific system has become an advanced art.

1. J. M. Uritis and B. A. Cola, "A High-Speed Precision Tape Recorder," RCA
2. A. C. Luther, "Modem Color Television Tape Recorders Using Automatic Timing Correction," RCA

pictures
Fig. 1 - Typical transverse-scan assembly.
Fig. 2 - Helical-scan headwheel assembly (closed).
Fig. 3 - Helical-scan head-wheel assembly (open).
Fig. 4 - Waveforms before switch
Fig. 5 - Waveforms of FM after first (4x2) FM switch.
Fig. 6 - Theoretical fade switch performance.
Fig. 7 - FM record spectrum showing injected pilot tone for time base reference.
Fig. 8 - Closed-loop electronically variable delay line system.
Fig. 9 - Headwheel assembly for time-expansion equipment.
.