Black Box 4

home *** CD-ROM | disk | FTP | other *** search

/ Black Box 4 / BlackBox.cdr / textinfo / cartphas.arj / CARTPHAS.BEI

Wrap

Text File | 1988-10-21 | 22.7 KB | 492 lines

********************************************************************* * NOTE: The following ASCII text file (without graphics) * * is contained in a printed technical paper available * * from Broadcast Electronics Inc. Unfortunately, it * * was not possible to reproduce the graphics portions * * of this paper within this text file. If you find the * * information in this file of interest, you may request * * a complimentary, printed, copy including figures and * * graphics from: BROADCAST ELECTRONICS INC. * * P.O. BOX 3606 * * 4100 N. 24TH STREET * * QUINCY, IL. 62305-3606 * * ATTN: SALES DEPARTMENT * * PH 217-224-9600 * * FAX 217-224-9607 * * * * The contents of this technical paper are * * Copyrighted (c) 1986, by Broadcast Electronics Inc. * * All rights reserved. * ********************************************************************* Originally from AV-Sync BBS (404) 320-6202 AUTOMATIC PHASE CORRECTION FOR TAPE CARTRIDGE MACHINES BY: James R. (Rick) Carpenter Manager of Audio Engineering Broadcast Electronics, Inc. Quincy, Illinois Published: March 1986 Time delay (phasing) errors between the left and right channels in stereo tape cartridge machines causes erratic high frequency loss and other compatibility problems when the channels are "summed" to monaural either before transmission or within a monaural receiver tuned to a stereo broadcast. These problems are created by azimuth errors between the playback head gap and the material recorded on the tape. The common causes of azimuth errors are head misalignment, gap scatter and the time dependent variations of the tape cartridge and its related tape guidance system. Presently used cartridge machine phase correction systems use a one time alignment during recording to correct for an average phase error value. This average correction only partially addresses the problem since the amount of correction needed changes as the tape cartridge mech- anism revolves. Real time correction is required to completely correct phase errors. Stand alone systems are available to correct for phase errors, but they are expensive and require encoding of a reference signal on each tape. This paper describes a technology that corrects phase errors in real time during tape motion without the encoding of a reference signal. LIMITS OF PERFORMANCE In order to set the design limits of the correction circuits, the performance limits of the tape cartridge medium need to be defined. The parameters that we are most interested in are azimuth error and its effects, signal-to-noise, frequency response, separation and distortion. Azimuth Azimuth refers to the head gap orientation with respect to the di- rection of tape travel. Absolute azimuth denotes that the head gap is perfectly perpendicular to the direction of tape travel as shown in Figure 1. FIGURE 1. ABSOLUTE AZIMUTH FIGURE 2. RELATIVE AZIMUTH ERROR Relative azimuth error depends on the difference between the azimuth of the material recorded on the tape and the azimuth of the playback head as shown in Figure 2. Relative azimuth errors cause two problems in tape cartridge machines. The first is a non-recoverable loss of high fre- quency information during playback. Figure 3 shows the amount of high frequency loss versus relative azimuth error according to the formula: L=20 log sin (180 T)/3.14159*T L= loss in dB T= tan (A)*F*W/V A= relative azimuth error (degrees) F= the frequency of interest (Hz) W= track width (inches) V= tape speed (inches/second) Azimuth errors also cause time delay (phase shift) errors between the two audio channels. This interchannel time delay can cause image shift in stereo systems and cancellation of some signal components in summed monophonic systems. The cancellation effect is most pronounced at high frequencies because the differential time delay results in a phase error which increases with frequency according to the formula: Pe=360*Td*Fs Fs is the frequency of interest. Td is the interchannel time delay. Pe is the interchannel phase error. From Figure 3 it can be seen that an azimuth error of 0.5 degrees causes a 10.5 dB loss at 15 kHz. This is an interchannel phase error of 40 degrees at 1 kHz, which is equivalent to an interchannel time delay of 115 microseconds. FIGURE 3. LOSS AS A FUNCTION OF FREQUENCY AND AZIMUTH This 115 microseconds of delay causes complete cancellation of 4.3 kHz and 13 kHz components in a mono sum signal as shown in Figure 4. The maximum tolerable system delay error would seem to be less than 16 micro- seconds. This is equivalent to 90 degree interchannel phase error at 15 kHz (relative azimuth error of 0.07 degrees) which gives a maximum stereo frequency response loss of 0.15 dB at 15 kHz. The maximum mono sum loss would then be 3 dB at 15 kHz. FIGURE 4. SUMMED MONOPHONIC FREQUENCY RESPONSE VS. DELAY It is obvious from these figures that the onset of monophonic signal degradation occurs with much smaller relative azimuth errors than those which affect the stereo channels only. This information on the amount of error necessary to degrade the monophonic versus the stereo program ma- terial sets the necessary time delay correction range. Along with the typical cartridge machine performance data discussed next, it sets the boundaries for the performance of the phase correction system. Signal-To-Noise The signal-to-noise of a tape machine is determined by the tape type and operating level. Without some form of noise reduction, an unweighted S/N of 60 dB below 250 nWb/m pulling tape is achievable at this time. Frequency Response The frequency response of a tape cartridge machine is limited by the amplitude of the low frequency contour effects of the head. Newer head designs provide a frequency response of 1 dB from 30 Hz to 16.5 kHz. Distortion The distortion performance of a tape cartridge machine is limited by the tape type and the recorded level. Newer tape formulations allow distortion figures of under 1% at record levels of under 250 nWb/m. Separation The separation performance of any tape machine is dominated by the tape heads. While separation performance of 60 dB is possible at fre- quencies less than 1 kHz, inductive coupling between the coils of each channel limit the separation performance to the neighborhood of 35 dB at 16 kHz. In order for the phase correction circuitry to be transparent to the cart machine performance, it needs to exceed the performance specifica- tions given below in Table 1. PHASE CORRECTION PERFORMANCE MINIMUMS Distortion 0.5% Signal-To-Noise 70 dB Frequency Response 0.25 dB 40 Hz-16 kHz Separation 60 dB 30 Hz-16 kHz Correction Range 115 microseconds (620 degrees at 15 kHz) Correction Error 16 microseconds 90 degrees at 15 kHz) (30 degrees at 5 kHz) Mono Compatibility As was detailed above, the onset of monophonic signal degradation occurs with much smaller relative azimuth errors than those which affect the stereo channels only. Monophonic listeners still form the largest group of listeners for all broadcasters. According to industry standards groups, over 90% of AM and TV listeners and 50% of FM listeners regularly listen in mono. Compared to the smooth frequency response loss of the stereo channels with increasing azimuth error, the degradation of the summed monophonic signal takes on a particularly offensive quality. For example, an inter- channel time delay error of 50 microseconds totally cancels any 10 kHz response in the summed signal. The response loss at 5 kHz is only -3 dB and the signal amplitude also rises again above 10 kHz as shown in Figure 5. FIGURE 5. SUMMED MONOPHONIC RESPONSE VS. FREQUENCY (50 uS) The objectionable nature of this degradation is due to the fact that this mid-frequency notch is not normally present in naturally occurring sound sources. The time dependent skew of the typical tape cartridge magnifies the effect of this notch by shifting the notch back and forth in frequency at a slow rate dependent on the speed of the tape, the cartridge machine positioning system and the cartridge construction. The notch then audibly "swishes" up and down in frequency as the tape is played. These consid- erations require a real time phase correction system to remove all annoy- ing artifacts from the monophonic signal. METHODS OF PHASE CORRECTION There are several available methods to correct phase in cartridge systems. They tend to fall into two broad categories: encoding systems and non-encoding systems. Encoding Systems Encoding phase correction systems usually inject some type of control signal onto the tape when it is recorded to define the correct phase rela- tionship of the right and left channels. The advantage of the encoding systems is that the correction circuitry can be optimized for a known reference signal. For example, one system uses a modulated 19 kHz pilot signal recorded on both audio channels. Another system records left channel audio on the cue track for a reference. While not usually thought of in these terms, a system which uses sum and difference matrixing techniques is also an encoded system. This process uses one channel for sum (L+R) information and the other channel for difference (L-R) information. Each playback machine must have a decoder (dematrix) and each recorder must have an encoder (matrix). When they are actually used, encoded systems usually give the best phase correction performance. The disadvantage of encoded systems is the need to encode each cartridge in the system to take advantage of the cor- rection capability. In order to actually use the correction performance of an encoding system, the entire cartridge library, all agency spots and any other cart not recorded with the encoding system must be re-recorded. Encoding systems also require each playback cartridge machine to have a decoder assigned to it, which imposes large cost, maintenance and com- plexity penalties on the entire audio system. A subset of the encoding phase correction system is a cart machine system that mechanically adjusts the relative azimuth of the record head to that of the playback head during a setup procedure. This allows the machine to correct for the average phase error of that one tape machine/ tape cartridge system. This system ignores the questions of machine to machine interchangeability, cartridges that were not recorded on that system (eg. agency spots) and changes in tape cartridge phase performance due to wear and minor damage. It cannot correct for real time changes in phase error due to the rotation of the cartridge mechanism, which creates time dependent variations in the amount of tape skew. Matrix encoding systems do not solve any phase problem, they just introduce the problem in another form. With phase problems in the dis- crete audio channels, there will be phase problems in the matrixed audio. In matrix form the result is usually poor separation and a poor stereo image for the stereo listener. To insure good separation, the amplitude and phase characteristics of the L+R and L-R channels must be tightly controlled. Most tape head amplitude balance specifications are on the order of 3 dB and phase dispersion is rarely specified. Figure 6 gives the resulting separation if amplitude and phase errors between the L+R channel and the L-R channels are known. Since the channel signal-to-noise ratio for a tape machine is fixed, it is also likely that the final discrete signal-to-noise ratio will de- grade. Most stereo signals have much more L+R than L-R information. In the worst case (L or R only), the L+R channel will be 6 dB lower than a discrete channel. This is because when L=R the amplitude is twice the single channel value (6 dB). When the decoded noise contribution of the L-R channel is added during dematrix, there can be as much as a 9 dB de- gradation of signal-to-noise. FIGURE 6. SEPARATION VS. DIFFERENTIAL AMPLITUDE AND PHASE Non-Encoding Systems Non-encoding phase correction systems are based on the fact that stereo program material has a considerable amount of monophonic content and that this monophonic content can be used to guide the correction process. As was shown in the section on azimuth effects, degradation of the monophonic content of the stereo program occurs well before any de- gradation of the actual stereo information. In order to use the monophonic content of a stereo program to cor- rect for time delay (phase) errors, it is necessary to find a way to ex- tract the time delay information from the audio signals on the tape. Signal theory points out a way by using a valuable property of signals called the auto-correlation function. The auto-correlation function is a time function of the signal. It indicates the degree to which the signal is related to values of itself in time. This function is obtained by multiplying the current signal by a delayed sample of itself and averaging over the sample time. At zero delay, the signal is multiplied by itself and the value is the signal power. As the amount of system delay is increased the value of the auto-correlation will decrease. The auto-correlation function will always have its largest amplitude at zero delay, except for sine waves which have additional, equal value peaks at multiples of the period. If a signal is multiplied by a delayed sample of another signal, the result is the cross-correlation function, which represents the amount of common information in the signals. Using the two channels of a stereo system, the cross-correlation will then represent the auto-correlation of the monophonic components. This is the information needed to extract the time delay information from the stereo signals. Using a servo system to maximize the value of the correlation function by linearly delaying the leading channel, will allow real time correction for interchannel time delays. Unfortunately, finding the peak of the correlation function requires the evaluation of the function at several points and a search to find the largest peak. This process does not lend itself to cost effective real- time implementation in a servo system. However, knowledge of the per- formance limitations of the system to be corrected allows the design of real-time signal processing techniques to simplify the evaluation of the correlation function. Because each channel is processed before the cor- relation, the system design can be limited to detecting the zero point of the function, the central concept of any servo design. This design allows a low cost, high performance real time phase cor- rection to be designed for almost any stereo system. The design of the signal processing circuitry will be dependent upon the performance char- acteristics of the stereo source device. Since the optimum signal pro- cessing circuitry for each source device type is different, there is no easy way to use this technique to develop a "universal" phase corrector. Each type of stereo source device would require different signal proces- sing circuits. Using one signal processing device would require re- adjustment for each source. If a compromise setting is attempted, degra- dation of both monophonic and stereo performance would result. SYSTEM EXPLANATION AND PERFORMANCE System Explanation A block diagram of the system is shown in Figure 7. The right and left audio channels are low pass filtered, then input to linear audio delay lines. The audio delay lines were picked as appro- priate technology for a cartridge machine corrector system because they have a linear time delay versus frequency characteristic. It is very difficult to build analog phase delays that have a linear delay versus time characteristic. The delay line audio performance, while not up to compact disc audio quality, is more than adequate for tape machine use. The harmonic distortion is less than 0.25% and the signal-to-noise is greater than 75 dB below normal signal level. The left delay is fixed at one millisecond, the right delay is variable from 0.8 milliseconds to 1.2 milliseconds. This gives a maximum delay range of 200 microseconds ( 1080 degrees at 15 kHz). FIGURE 7. TAPE CARTRIDGE PHASE CORRECTION SYSTEM BLOCK DIAGRAM After the delays, the audio is low-pass filtered to remove sampling artifacts and is passed on to the noise reduction circuitry. The audio is also sampled at this point for the phase detector signal processor. After signal processing the signals are correlated, rectified and fil- tered. This filtered, level shifted signal is used to run the right de- lay clock, completing the phase tracking servo loop. System Performance The Lissajous figure shows before and after correction results for a single test tone of 7.5 Hz in Figure 8. The error is slightly greater than 90 degrees (oval trace) and the correction is better than 3 degrees (45 degree line). Figure 9 shows a before-and-after correction results for a pink noise test tape. These pictures were created by manually mis- aligning the playback head of the machine and then enabling the phase correction. FIGURE 8. TONE-PHASE ERROR BEFORE FIGURE 9. PINK NOISE PHASE ERROR AND AFTER CORRECTION BEFORE AND AFTER CORRECTION The trace in Figure 10 is a five minute sample of relative uncor- rected cartridge machine playback phase error (degrees) versus time (minutes). A new 6 minute cartridge was used. It was recorded with a 7.5 kHz tone on the same cartridge machine used for playback. The large excursion in-phase at about 4 minutes is the tape splice. The trace in Figure 11 is a five minute sample of the same machine and tape as Figure 10, but with the phase correction in circuit. Without phase correction in circuit, the maximum peak phase error (excluding the splice) is 17 degrees, with short term variances of as much as 15 degrees. With the phase correction in circuit, the maximum peak phase error (excluding splice) is 1 degree, with short term variations of 2 degrees. FIGURE 10. UNCORRECTED PHASE ERROR VERSUS TIME FIGURE 11. CORRECTED PHASE ERROR VERSUS TIME This is an obviously an optimum system. It has a brand new cart- ridge and freshly tweaked record and playback alignment. Figure 12 shows a non-encoded five minute sample of the same machine, but with a different tape that was recorded on a different machine. The phase offset is 80 degrees. The maximum peak phase error is 85 degrees, with short term variations of 10 degrees. With the phase correction in circuit (Figure 13), the phase offset is eliminated. The short term error is reduced to less than 3 degrees. FIGURE 12. OLD CART-UNCORRECTED PHASE ERROR VERSUS TIME FIGURE 13. OLD CART-CORRECTED PHASE ERROR VERSUS TIME CONCLUSION A built-in, non-encoding interchannel phase correction system has been profiled in this paper. This cost effective, operator transparent system eliminates the monophonic compatibility problem for any tape cart- ridge machines playing any tape, without the inconvenience and expense involved with encoded systems. ACKNOWLEDGEMENTS The author would like to thank T. Whiston for the testing, J. Houghton for the drawings and C. Steffen and L. Foster for putting this paper into readable form. THE AUTHOR James R. "Rick" Carpenter earned his BSEE, and is pursuing an MSEE, from West Virginia University in Morgantown, West Virginia. Mr. Carpenter has designed instrumentation for the U.S. Bureau of Mines. He was project engineer for the Harris MX-15 FM exciter develop- ment, the Broadcast Electronics TZ-30 TV MTS generator, and the Broadcast Electronics PT-90 cartridge machine. The author has extensive experience in solid-state RF design and analog tape equipment design. The author is presently Manager of Audio Engineering for Broadcast Electronics Inc. in Quincy, Illinois. Mr. Carpenter has authored numerous technical papers, including co- authorship of the NAB Handbook chapter on "Analog Magnetic Recording" and is a member of AES. REFERENCES Burstein, Herman, "How Important is Tape Azimuth", Audio VOL.68, No.9, pp. 40-746, (c) 1984. Moris, A.H., Mullen, J.T., "Phase Error In Tape Cartridges for Radio Broadcast Service", Journal Audio Engineering Society, VOL.31, No.1/2, (c) 1983. Heinrich, Mer, "Delay-Lock Tracking of Stochastic Signals", IEEE Transactions on Communications, VOL. COM 24, No.3, (c) 1976. Cabot, R.C., Pavlok, R., "A High Accuracy Analog Cross Correlator", AES Preprint 1362, 60th Convention, (c) 1978.