[802.3ae_Serial] Sine interference in stressed eye
All
-
Since the rush of
the meeting, I have done some more analysis on the amount of sine interference
required to generate the stressed eye we have prescribed. In summary, I was able
during the meeting to determine that 1/4 closure with the sine signal was not
sufficient to create the level of required jitter, but I was not able in time to
determine if 1/3 was enough. (For those that did not attend, D4.3 prescribes
that at least 2/3 of the vertical eye closure should be accomplished with a
filter - see page 480, line 26. D4.2 prescribed at least
1/2.).
My latest analysis
shows that approx 1.1 dB of sine interference is required, meaning that in
some cases, 1/3 is also not sufficient (for example, -L where the total required
closure is 2.2 dB).
Background:
- Jitter due to sine
interference and sine jitter will be high-probability in nature. I fear that
all other unintentional sources of jitter will have tails with low probability
outside of the 0.3 UI 99% window. Such tails will lead to higher stress than
we desire.
- Because of
this, we have gone out of our way to specify minimal data-dependent
jitter, minimal noise, etc. If one
strictly follows these recommendations,
sine interference and sine jitter should
be the only sources of jitter and
tails will be minimized.
- We have
limited sine jitter to 0.15 UI pk-pk. Given the above, this requires that sine
interference should be able to generate 0.15 UI of jitter to sum to the 0.3 UI
total.
Model:
- For a simple
model, I have assumed that the inner eye closed by the filter will be
essentially sinusoidal in shape. Graphical analysis with Bessel filters shows
this to be VERY close.
- To achieve
the 0.15 UI pk-pk, the sine model shows the peak amplitude of sinusoidal
interference must 0.15*pi/2 (-1.1 dB). This is accurate for small amounts
of interference, which I believe is satisfied by 0.15 UI.
- The 1.1 dB value is
relative to the inner eye created by the filter. For example, if the filter
closes the eye down to 80%, sine interference needs to further close it down
to 61%.
- In the case of -L,
this means that 50% of the required 2.2 dB must be done with sinusoidal
interference.
- I have confirmed
this through my waveform simulation tool (the one that created the stressed
eye in D4.3).
Other:
- Stretch observed in
the lab that low probability DDJ increased as the filter was cranked down to
get a higher percentage of closure. This is not a surprise, but again, it
caused more stress than desired.
- I understand the
concern expressed by Piers about having too much sine interference. It is
basically causing pulse shrinkage jitter (more than 5 psec) which is harsh on
bandlimited receivers. It should be minimized closely to the 1.1 dB
value.
- Do we want to
increase the amount of sine jitter, thereby reducing the required amount of
sine interference?
- Minor, but my final
point is that reducing the range of sine jitter (fine knob) tightens the
bandwidth range of filters (coarse knob).
Comments?
Tom
Stratos
NW
(425)
672-8035 x105
BEGIN:VCARD
VERSION:2.1
N:Lindsay;Tom
FN:Tom Lindsay
ORG:Stratos Lightwave;Opto NW
TITLE:Prin. Engr.
NOTE;ENCODING=QUOTED-PRINTABLE:Home: lindsayta@xxxxxxx=0D=0A=0D=0Aom
TEL;WORK;VOICE:(425) 672-8035 x105
TEL;HOME;VOICE:(425) 775-7013
TEL;CELL;VOICE:(206) 790-3240
TEL;WORK;FAX:(425) 672-8014
ADR;WORK;ENCODING=QUOTED-PRINTABLE:;;6100 219th Street SW=0D=0ASuite 520;Mountlake Terrace;WA;98043
LABEL;WORK;ENCODING=QUOTED-PRINTABLE:6100 219th Street SW=0D=0ASuite 520=0D=0AMountlake Terrace, WA 98043
EMAIL;PREF;INTERNET:tlindsay@xxxxxxxxxxxxxxxxxxxx
REV:20020219T195001Z
END:VCARD