Proposed improvements to jitter specification text
Piers Dawe, IPtronics
For comment 399
Change 92.8.3.6 as follows:
92.8.3.6 Transmitter output jitter
Start with the material that’s common to several jitter metrics. May need a sentence of introduction, e.g.
“Four measured of jitter are defined as follows.”
Need to put the jitters in some memorable order, e.g. alphabetical or the order that they appear in Table 92–
5.
92.8.3.6.1 Even-odd jitter
Even-odd jitter is measured from the two symbols in the middle of a sequence of no fewer than 8 symbols of
alternating polarity. If PRBS9 is the test pattern, a suitable sequence may be found starting at either bit 161
or bit 383 where bits 1 to 9 are the run of 9 ones. A correct measurement of even-odd jitter requires that the
period of the test pattern is an even number of symbols. When the base pattern period is an odd number of
symbols, such as PRBS9, the test pattern for the purpose of even-odd jitter measurement must be two periods
of the base pattern.
Even-odd jitter is defined to be half of the magnitude of the difference between the mean width of the
positive pulse and the mean width of the negative pulse. The reference voltage for pulse width measurements
is the mid-point between the positive pulse amplitude and the negative pulse amplitude. The pulse amplitude
is defined to be the mean amplitude of the pulse within the central 20% of the nominal unit interval.
Even-odd jitter shall be less than or equal to 0.035 UI regardless of the transmit equalization setting.
NOTE—Even-odd jitter has been referred to as duty cycle distortion by other Physical Layer specifications for operation
over electrical backplane or twinaxial copper cable assemblies (see 72.7.1.9). The term even-odd jitter is introduced to
distinguish it from the duty cycle distortion referred to by Physical Layer specifications for operation over fiber optic
cabling.
92.8.3.6.2 Total jitter (TJ)
The total jitter (TJ) of a signal is defined as the range (the difference between the lowest and highest values)
of sampling times around the signal transitions for which the BER at these sampling times is greater than or
equal to 10
–12
.
TJ may be measured directly, or estimated by fitting the measured jitter distribution to the
dual-Dirac mathematical model used for effective random jitter (RJ).
92.8.3.6.3 Data dependent jitter (DDJ)
Data dependent jitter (DDJ) is
characterized using
defined by
the procedure
defined
given
in 85.8.3.8.
Total jitter excluding data dependent jitter is the difference between TJ and DDJ and shall be less than or
equal to 0.28 UI regardless of the transmit equalization setting.
Editor’s note (to be removed prior to final publication):
Clause 85 was not clear on what was meant by “excluding” DDJ from TJ. In this draft, it has been interpreted to
be the value of TJ as defined above minus the value of DDJ as defined in 85.8.3.8.
92.8.3.6.4 Effective random jitter (RJ)
The effective random jitter (RJ) of a signal is defined to be the difference between the TJ and effective
deterministic jitter (DJ). Effective DJ is derived from a fit of the measured jitter distribution to a dual-Dirac
mathematical model. The fit is computed as follows.
a)
Measure the jitter
J
n
which is defined to be the range of sampling times around the signal transitions
for which the BER at these sampling times is BER
n
. Measure two values J
0
and J
1
where BER
0
is
10
–9
and 10
–5
.
Use headings to make
the document easier to
use.
Comment [PD2]:
Should not use 3 names
for (almost) 1 thing.
Symbol, bit or unit
interval? Note Clause
91’s use of “symbol” for
10 bits on the line, and
there is no definition of
“symbol”here.
Comment [PD3]:
For a waveform like this,
when I see “amplitude”
I think of the swing.
“Voltage” or maybe
“level” might be better.
Comment [PD4]:
Want the abbreviation
to appear in the
contents and
bookmarks.
b)
For each BER
n
, determine the associated Q
n
from the inverse normal cumulative
probability
distribution
function
adjusted for an assumed transition density of 0.5. For example, Q
n
is 5.7
68
7
if
BER
n
is 10
9
and Q
n
is 3.94
4
if BER
n
is 10
–5
.
c)
Calculate the effective DJ as (
Q
0
J
1
– Q
1
J
0
) ⁄ (Q
0
– Q
1
).
d)
If TJ is found by the fitting method, the effective RJ is estimated by Q
min
(J0 – J1) / (Q0 – Q1),
where Q
min
is 6.839 for a specification BER of 10
–12
.
The effective RJ shall be less than or equal to 0.15 UI regardless of the transmit equalization setting.
Editor’s note (to be removed prior to final publication):
Clause 72 specifies that transmit equalization is off (preset) for the measurement of jitter. Clause 85 does not
specify the conditions for jitter measurement and this is interpreted to mean that it applies for all transmit
equalization settings. There has been some discussion as to which approach is appropriate.
The effect of a single-pole high-pass filter with a 3 dB frequency of 10 MHz is applied to the jitter. The test
pattern for TJ and RJ measurements is either PRBS31 (see 83.5.10) or scrambled idle (see 82.2.10). The
voltage threshold for the measurement of BER or crossing times is the mid-point (0 V) of the AC-coupled
differential signal.
Replace 93.8.7.1 with the following:
93.8.1.8 Transmitter output jitter
Even-odd jitter is defined in 92.8.3.8. Even-odd jitter shall be less than or equal to 0.035 UI regardless of the
transmit equalization setting.
Total jitter is
characterized using the procedure
defined in 92.8.3.8. Data dependent jitter is
characterized
using the procedure
defined in 85.8.3.8. The total jitter, excluding data dependent jitter, shall be less than or
equal to 0.28 UI regardless of the transmit equalization setting.
The e
E
ffective random jitter is
characterized using the procedure
defined in 92.8.3.8. The effective random
jitter shall be less than or equal to 0.15 UI regardless of the transmit equalization setting.
Comment [PD5]:
This paragraph should
go in the common
section at the
beginning.