LeCroy DDA 7 Zi series, Руководство оператора

Страница: 490 / 544

WavePro 7Zi
489
WP700Zi-OM-E-RevA
the NLTS calculations are adjusted by dividing their value by the correlation coefficient value at
an integral number of pattern-length delays.
The following table gives the standard deviation of the nlts parameter for varying amounts of
auto-correlation signal-to-noise, and numbers of repetitions of the pseudorandom sequence in
the input waveform. The sampling rate used was four samples/bit cell, and the input waveform
had 20% NLTS.
Parameter
Settings
When you select nlts from the PRML dialog's With parameter field (Drive Analysis Measure
Parameter Set PRML ) an nlts Pattern Length field and pattern Delay field appear also.
You can adjust the mantissa, exponent or number of mantissa digits using the pop-up numeric
keypad. The pattern length should be set to the pattern period.
Although the DDA searches for the correct pattern length, the value provided needs to be
sufficiently close to the actual pattern length for
nlts to perform the search. A 1 μs pattern may,
for example, accept a range of 1 μs ± 40 ns. Within this range, a value for nlts will be provided.
Otherwise "---" appears on the screen, indicating that no measurement can be made.
The pattern Delay setting is a percentage of the pattern length. The DDA will internally scale the
delay value entered by the ratio of the pattern length calculated internally to the pattern entered
by you. Several disk drive waveform attributes can be measured by using different delay values.
The following table provides delay values to enter for the commonly used 127-bit pseudo-
random sequence (x
7
+ x
3
+ 1 polynomial) when measuring various waveform attributes:
Notes
The pattern Delay tells the DDA where in the repeating pattern to measure NLTS. NLTS is
calculated from the correlation coefficient at that time. Correlation is calculated at 3 sample
times: the nearest sample time and one on each side. A curve fit is performed to calculate a
better estimate of the true peak height.
As described above, the nlts parameter only requires one waveform. Each acquisition must
contain only the pseudorandom repeating sequence (PRS), not a servo wedge or a preamble. It
does not matter where in the PRS the acquisition begins; its echo properties are independent of
starting point. There must be at least two repetitions of the pattern in each acquisition. A
reasonable number would be about 25 repetitions (32 repetitions of a 127-bit sequence should
fit in a sector).
NLTS requires at least 5 samples per PW50 for acceptable accuracy. More is better.
The PRS data must correspond to the transitions on the media. This means it must be written in
direct write mode; it must not be scrambled or encoded.
NLTS calculated by correlation techniques and by the fifth harmonic elimination techniques do
correlate, but they are not identical. Fifth harmonic elimination uses a pattern including only
dibits and widely spaced transitions. Of course, a dibit is the worst case for NLTS. The
pseudorandom pattern has some dibits, some tribits, some transitions 1 apart, etc. The reading
that any method based on a PRS comes up with is affected by all of the pattern. This tends to
make the correlation method come up with a somewhat lower value than fifth harmonic. That is
the better value to look at when determining write precomp, since the PRS is a lot like real data.
The fifth harmonic method is not sensitive to amplitude asymmetry. Its drawbacks are
x
If the fifth harmonic is small, noise will tend to inflate the value read: for 20 dB SNR it
may not be possible to read below 10% nlts
x
Amplitude loss due to partial erasure will contribute to the fifth harmonic about as much
as actual NLTS.
The correlation method is less sensitive to amplitude loss due to partial erasure: one reference
cites experiments showing that 25% nlts and 25% amplitude loss only inflates the nlts reading to
30%.
However, it is more sensitive to PW50/T ratio (it is most accurate at high ratio, approaching 3.0).