IBM System/370 Manual page 63

Page of GC20-1730-0
Added 11/20/70
By TNL GN20-2277
Phase-encoded recording. The phase-encoded (PE) recording technique
offers superior error detection and reliability as compared with the
non-return--to-zero (NRZI) technique. In both cases, magnetic recording
of one and zero bits is accomplished by means of flux reversals or
changes in polarity. In NRZI recording, only one bits are recorded
as magnetized spots, and a flux reversal occurs only for one bits.
In PE recording both zero and one bits are recorded (the zero bit and
one bit being opposite in polarity), and a flux reversal is required
in every bit position. Thus, the PE dual flux recording technique
differentiates between no recording and the presence of a zero bit,
and the absence of any signal is detected as an error.
The positive recording of all zero and one bits in PE eliminates
the need for horizontal parity bits (longitudinal redundancy check
used in NRZI recording), and vertical parity bits are used to correct
single-bit read errors in-flight. During reading, if a single track
fails to respond with a suitable pulse in any bit position, reading
of the rest of that track is immediately disabled for the remainder
of the data block, and the remaining bits for that track are
automatically generated by use of the vertical parity bits. In-flight
single-track error correction eliminates the time normally lost in
backspacing and rereading NRZI tape for correction of single-track
dropouts or defects.
Phase encoding offers other advantages. If a string of zeros is
recorded on tape, successful reading in NRZI requires close
synchronization to ·count· the correct number of zeros. With PE, this
synchronization is provided by the flux reversal in every bit position;
hence, PE recording (and reading) is self-clocking. In addition, each
block written on a PE tape is preceded and followed by a coded burst
of bits in all tracks to set up the individual track-clocking rates.
The read circuitry is designed to recognize these bursts and thereby
minimize the effect of noise in the gap.
The critical nature of vertical skew (alignment of bits within a
byte) that is imposed by NRZI recording is minimized
by
this individual
track clocking scheme (one clock per track versus one clock for the
entire tape subsystem), and by the use of one-byte (nine-bit) capacity
skew buffers that can be in the process of collecting up to four data
bytes at the same time, as the tape passes the read head. Because
of the positive recordi~g of all bits, once a skew buffer contains
nine bits, one from each horizontal data track, it is an indication
that a byte has been read. Thus, the 3420 can handle the situation
in which the tape is not exactly aligned, and bits from up to four
adjacent bytes can be read concurrently.
Like 2400-series tape units, the 3420 utilizes a two-gap read/write
head that performs readback checking during write operations. The
3420 also has a separate erase head that erases the entire width of
the tape during any write operation before waiting occurs. Full-width
erasure reduces the likelihood of leaving extraneous bits in interblock
gaps or skip areas and minimizes the interchangeability problems that
can occur when tape is written on one tape unit and read on another.
Advanced engineering design. The tape path in the 3420 tape unit
is designed for ·soft handling· of tape volumes to minimize tape wear
and thus improve tape reliability. other features, such as the single-
drive capstan and optical tachometers, result in faster data access
and rewind times than those of the 2401 and the 2420, for models with
comparable data transfer rates.
51.2