128GFC preliminary link budget studies

128GFC preliminary link budget studies
April 2014
T11/14-100v1
Preliminary study of link budget sensitivities
‹
In T11/14-021v0 we identified the following parameters to
be of special interest for 128GFC link budget:
ƒ
ƒ
ƒ
ƒ
‹
Receiver sensitivity
Tx min and max OMA
Q (forward error correction signal-to-noise, uncorrected)
VCSEL spectral width ∆λ
Start with 32GFC spreadsheet T11/12-376v0
ƒ “M5E 850S4500 32GFC 2xCDR” worksheet
ƒ Note: T11 Docs doesn’t have a copy of this; need to ask me if
you don’t have a copy from 2012 Boise meeting
ƒ Vary each parameter of interest, leaving all the others at nominal
value for 32GFC
ƒ Plot impact on link reach
‹
Repeat for 100GBASE-SR4 spreadsheet
2
Link reach vs. receiver sensitivity
3
Rx sensitivity comments
‹
0 dB on the previous plot means the 128GFC
receiver sensitivity is identical to the 32GFC
sensitivity (-10.2 dBm; 10-6 BER; 28.05 Gbd) or to the
100GBASE-SR4 sensitivity (-11.2 dBm; 5x10-5 BER;
25.78 Gbd).
‹
My initial expectation was to see a degradation in
sensitivity for a 4-channel transceiver (whether
QSFP+ or CFP4) because of increased crosstalk
from neighboring lanes.
‹
As we’ll see in later slides, this is not the most critical
parameter for a 4-lane solution.
4
Link reach vs. Tx min OMA
‹
The previous plot was given from the perspective of
impact of receiver sensitivity on link reach.
‹
However, this same analysis also applies to impact of
transmitter min OMA on link reach.
‹
It might be conceivable to compensate for degraded
receiver sensitivity by increasing Tx min OMA on a
128GFC link.
‹
However, ”Case 2” for breakout to individual 32GFC
SFP+ transceivers won’t be compensated. There
would be a link budget discrepancy between “Case 1”
and “Case 2”.
5
Link reach vs. uncorrected signal-to-noise Q
6
Warning: Rx sensitivity not adjusted
‹
This analysis only adjusted the Q parameter box in
the spreadsheet (see above)
‹
Rx sensitivity is dependent upon uncorrected BER
allowed; we will calculate that later in this
presentation
‹
The following analysis did not adjust Rx sensitivity.
‹
With adjustment of Rx sensitivity, the dependence
shown would be even more pronounced.
7
Link reach vs. VCSEL spectral width
8
Comparison of link budgets
Parameter
100GBASE-SR4
FC-PI-6
Tx min OMA
-3.0
-3.2
Rx sensitivity
-11.2
-10.2
Power budget
8.2
7.0
Connector loss
1.50
1.50
Fiber attenuation
0.36
0.36
ISI
3.16
3.14
DJ_center
1.76
0.93
MPN
0.11
0.15
RIN
0.74
0.61
MN
0.13
0.14
Cross
0.44
0.19
8.20
7.02
Total
9
Comparison of transmitter specifications
Parameter
100GBASE-SR4
FC-PI-6
Min Tx wavelength (nm)
840
840
rms spectral width (nm)
0.60
0.57
Min Tx OMA (dBm)
-3.0
-3.2
Tx risetime (20-80, ps)
21.0
20.8
RIN(OMA) (dB/Hz)
-128
-129
DJ & TP4 eye (ps)
21.4
14.15
DCD_DJ (ps)
1.94
0.00
10
Comparison of receiver specifications
Parameter
100GBASE-SR4
FC-PI-6
Nominal sensitivity (dBm)
-11.20
(5x10‐5 BER; 25.78 Gbd)
-10.20
(10‐6 BER; 28.05 Gbd)
Receiver bandwidth (GHz)
18.047
20.0
11
Scaling receiver sensitivity
‹
100GBASE-SR4 sensitivity is better than FC-PI-6
sensitivity mostly because of
ƒ Lower signaling rate
ƒ 50 times higher uncorrected bit error tolerance
‹
Q for 100GBASE-SR4 is 3.89, Q for FC-PI-6 is 4.75
10 log10 (4.75 / 3.89) = 0.867 dB
‹
Signaling rate for 100GBASE-SR4 is 25.781 Gbd,
signaling rate for FC-PI-6 is 28.050 Gbd
10 log10 (28.050 / 25.781) = 0.366 dB
12
Initial search for 128GFC link budget
‹
Option #1: Start with 100GBASE-SR4
ƒ Bump up signaling rate
ƒ Was 25.781 Gbd
ƒ Should be 28.050 Gbd
‹
Option #2: Start with FC-PI-6
ƒ Bump up link DJ + CDR allocation
ƒ Was 14.15 ps & 0.0 ps
ƒ Should be 21.4 ps & 1.94 ps (?)
‹
Results summarized on next page
13
Initial efforts at 128GFC
Parameter
Tx min OMA
Rx sensitivity
Power budget
Connector loss
Fiber
attenuation
ISI
DJ_center
MPN
RIN
MN
Cross
Total
Total - budget
100GBASESR4
FC-PI-6
100GBASESR4
scaled
FC-PI-6
+ more DJ
-3.0
-11.2
8.2
1.50
0.36
-3.2
-10.2
7.0
1.50
0.36
8.2
1.50
0.36
7.0
1.50
0.36
3.16
1.76
0.11
0.74
0.13
0.44
8.20
3.14
0.93
0.15
0.61
0.14
0.19
7.02
3.96
2.30
0.14
1.64
0.13
1.79
11.82
3.65
2.23
0.18
1.82
0.14
0.99
10.83
0.00
0.00
3.62
3.83
14
Summary and next steps
‹ The
single biggest difference for a 4-lane
PHY compared with a single channel PHY is
“DJ” contribution to link penalties.
‹A
link budget for 100GBASE-SR4 has been
presented.
‹ We
are still struggling to find ways to close
the budget for 128GFC.
15