LeCroy DDA-3000A, Руководство оператора

Страница: 428 / 437

426
WM-OM-E Rev I
PLL and AGC
The sample, phase-steering (PLL) and gain adjust (AGC, level computation) work together, one
sample at a time. They are used to compute sampled data for input to the Viterbi detector.
ML Detector
Samples on the output of the ADC ideally have a small number of levels: {1,0,+1} for the PR4
system, for example. A threshold detector could be used to classify a current sample value
comparing it to an amplitude threshold. For example, if sample > 0.5, sample = 1; if sample < 0.5,
sample = 1; if sample 0.5, sample = 0.
For the sequence of samples 0.8 0.3 -0.7 -0.2 0.6 0.9 1.1 0.2, the threshold detector output would
be: 1 0 -1 0 1 1 1 0.
The difficulty here is that the sequence of three ones in a row ("111") is impossible: the pulse is too
wide. The only possible combinations are 011, 110, -111, and so on.
A threshold detector, such as the peak detector on a Peak- Detect drive, does not use the previous
and subsequent samples. But the maximum likelihood (ML) detector "knows" that "111" is a
forbidden sequence of samples and tries to determine the most probable data pattern for this
sequence of samples (21 samples used for PR4).
Proposing several close allowable sequences --- {1 0 -1 0 1 1 0 0} or {1 0 -1 0 0 1 1 0} or {1 0 -1 0 0
0 1 1} --- is easy. But which is the most probable?
The decision is made based on a sequence of samples, instead of only a single, current sample,
and the sequence with the minimum distance (maximum likelihood) is selected as the detection
result.
Viterbi Detector & Trellis
The Viterbi detector is a state machine consisting of two distinct parts: states and transitions. While
state is the current magnetization of the disk and some history (memorization of the latest states),
transition relates the current state to the next state. For the detector, only two possibilities exist:
either the state (medium magnetization) is the same between the current and the next bit periods or
it is not.
The detector’s trellis works according to this dichotomy: "0" or "1" is followed by either "0" or "1",
and so on. The trellis is a mechanism that keeps track of a sequence of magnetization states. When
the ML detector makes decisions, it keeps the states of several consecutive time instants and
estimates the likelihood of possible "histories" (higher probability). The higher the order of the
PRML system, the larger and more complex the trellis. However, some trellises do not allow certain
transitions (d=1 constraint), thereby limiting the extent to which the pulses can overlap.
SAM
Sequenced Amplitude Margin (SAM) measures the error margin of every sample taken by a PRML
channel chip. Determining that a written bit is either a "0" or a "1" is the disk drive’s most basic
decision. SAM measures the margin by which the Viterbi detector has made this decision, the
margin or distance being a function of the path metrics and current sample taken together.
SAM can provide a prediction of the error rate, and can be used both for characterization and for