Gigabit Transmission
What’s the Limit?
Fanny Mlinarsky
www.scope.com
Updated November 06
Page 1
What’s the Limit?
• Speed
→ Faster→ higher frequency → higher
attenuation → less headroom
• Distance
→ Longer → higher attenuation →
more jitter → less headroom
• Headroom
→ How close to collapse?
Headroom
Updated November 06
Page 2
Noise
Signal
www.scope.com
SNR
Required SNR
Noise+Jitter Headroom
Voltage
1
0
1
0
0
1
1
1
Time
www.scope.com
Updated November 06
Page 3
Effect of Low Headroom -Slow Network Response
Help!
Help!
www.scope.com
Updated November 06
Page 4
• A single bit error can cause
re-transmission of the
entire message
• Re-transmission cuts
throughput to less than half
due to ...
• … overhead of detecting
corrupt data and arranging
the re-transmission
Headroom Guarantees
Expected Throughput
• Bit errors can significantly slow down throughput
• Headroom is needed to withstand
– Ambient noise
– Build-up of jitter over distance
• On twisted-pair the major cause of errors is inadequate
SNR
• On fiber the major cause of errors is excessive jitter
www.scope.com
Updated November 06
Page 5
Noise Environment is
Application-Dependent
RX
TX
100 Base-TX
NEXT
x4
TX
RX
1000 Base-T
ECHO
www.scope.com
Updated November 06
Page 6
FEXT
Gigabit Ethernet SNR
Environment
Data Signal
Noise
Power
Signal
power
Headroom
Updated November 06
Page 7
Noise
www.scope.com
Signal
1000Base-T
Transceiver
Model
SNR
SNR
Headroom
Required SNR
Required SNR
Gigabit Ethernet
Headroom vs. Distance
Headroom
0
0.3 m
11.4 dB
Headroom
5
20 m
9.8 dB
Required
SNR
Headroom
10
50 m
7.0 dBSNR
Required
Headroom
4.0 dB
15
20
Cat 5 channel limit
25
100 m
Required
SNR
Headroom
2.0 dB
Required SNR
IEEE worst case model
dB
Required SNR
30
40
MHz
www.scope.com
Updated November 06
Page 8
50
60
70
80
90 100
Noise
20
Signal
0 10
Gigabit Ethernet
Headroom vs. Distance
12
Based on cabling with
typical attenuation
SNR Margin (dB)
10
8
6
4
2
0
www.scope.com
Updated November 06
Page 9
0
20
40
60
80
Meters
100
120
140
Cabling Requirements of
1000Base-T
• 1000Base-T is expected to work on category 5
installations that meet draft TSB95
– Attenuation, NEXT, ELFEXT, return loss, delay and skew
• Will existing installations meet the requirements of
TSB95?
– Some parameters, such as return loss and ELFEXT, might not
meet TSB95 limits, but ...
– … 1000Base-T might still have sufficient headroom
• Category 5E will have sufficient headroom by design
– > 3 dB
• Category 6 will offer more headroom
www.scope.com
Updated November 06
Page 10
Fiber Optic Transmission
Transmitted
pulses
LED
MMF
www.scope.com
Updated November 06
Page 11
* Micron = 10-6 meter
Cladding
Received
pulses
Core
50 or 62.5 microns*
Jitter
Jitter Tolerance vs. Data
Rate
1
0
Clock period = 800 ps for gigabit Ethernet
1
0
www.scope.com
Updated November 06
Page 12
Fiber Optic Transmission
Transmitted
pulses
LED
MMF
Laser
SMF
Laser
MMF
www.scope.com
Updated November 06
Page 13
* Micron = 10-6 meter
Cladding
Received
pulses
Core
50 or 62.5 microns*
Jitter
Dispersion
<10 microns*
Differential
Mode Delay
Differential Mode Delay
(DMD)
Laser
launched
into center
Multi Mode Fiber
Cladding
Received
pulses
Core
jitter
Transmitted
pulses
Multi Mode Fiber
Cladding
Laser
launched
off-center
www.scope.com
Updated November 06
Page 14
Core
Received
pulses
Offset Patch-cord
MMF
MMF
RX
TX
SMF
To
Equipment
Offset
MMF
To MMF
cabling
• Offset patch-cord can be used with 1000Base-LX (1310
nm) gigabit Ethernet devices to address the DMD
problem
www.scope.com
Updated November 06
Page 15
IEEE 802.3z Gigabit Ethernet
Limits
Gigabit
Ethernet
Specification
Type of
Fiber
Wavelength
(nm)
Fiber Core
Size
(microns)
Modal Bandwidth
·(MHz * km)
Maximum
Distance
(m)
Attenuation
1000Base-SX
MMF
850
50
400
500
3.37
500
550
3.56
160
220
2.38
200
275
2.60
50
400, 500
550
2.35
62.5
500
550
2.35
5,000
4.57
62.5
1000Base-LX
MMF
SMF
1310
1310
10
Attenuation budget is based on length restrictions.
Link power budget, based on transmitter launch power
and receiver sensitivity, is 7.5 dB for multi-mode fiber.
www.scope.com
Updated November 06
Page 16
(dB)
1000Base-X Loss Budget -LED vs. Laser
LED Source Laser Source
Approximate measured loss
of one connection
0.5 dB
0.2 dB
• When LED source is used for field testing…
– ... a link with 4 connections can easily fail the tight attenuation
limits for 1000Base-X
• When a laser source is used …
– … the same link might pass
www.scope.com
Updated November 06
Page 17
IEEE 802.3z Fiber Field
Testing Issue
• Issue
– IEEE 802.3z channel loss limits cannot be met by some installations
when measured with an LED source but may be within limits when
measured with a laser source
• 802.3z references for channel insertion loss
measurement (sec 38.10)
– ANSI/TIA/EIA-526-14A/method B (MMF)
– ANSI/TIA/EIA-526-7/method A-1 (SMF)
– IEC1280, IEC 1280-4-1 and IEC 1280-4-2
» Based on above TIA standards
• TIA-526-14A specifies the use of an LED source
• No guidance from the IEEE 802.3z standard as to the
required source for channel loss measurement
www.scope.com
Updated November 06
Page 18
Liaison Letter Regarding
Fiber Loss Measurements
Liaison letter of May 13, 1998 from TIA TR41.8.1states:
"We are concerned that users may reject
acceptable systems because of attenuation
based on LED test sets as established in TIA526-14-A"
www.scope.com
Updated November 06
Page 19
Cabling Requirements for
802.3z
• Several different length and loss limits depending on
type of fiber and type of equipment (1000Base-LX,
1000Base-SX)
– Not defined by any generic cabling standard
• Issues with the designated TIA loss measurement
method
– Possible false failures due to the use of LED sources
• Clearly, 802.3z calls for an application-specific cable
test
• Fiber length measurement is very important because
jitter -- the major disturber in the optical systems -builds-up as a function of length
www.scope.com
Updated November 06
Page 20
Standards Chain Reaction
New LAN
Equipment
Standards
• IEEE 802.3 Gigabit Ethernet
Standards
– 1000 Base-SX, 1000Base-LX
(Optical Fiber)
– 1000 Base-T (Category 5,
Category 5E)
• TIA Cabling Standards
Structured
Cabling
Standards
– TSB95
– TIA-568-A Addendum 4
• ISO/IEC Cabling Standards
– Proposed Draft Amendment 3
to ISO/IEC 11801
www.scope.com
Updated November 06
Page 21
Conclusion
• Characterization of application-specific headroom is
not just a good idea
• It is practically a “must” for some applications due to
the lack of generic cabling standards
– For example, 802.3z gigabit Ethernet over fiber
• Noise and jitter environments are applicationdependent
• Application headroom carries information about
network robustness and its ability to deliver the
expected throughput
www.scope.com
Updated November 06
Page 22
Announcement
For more discussion on related topics …
You are cordially invited to join Mohawk/CDT
for a lunch and a presentation of “Next Generation
of High Performance Fiber Optic and Copper Cables”
in the Grand Ballroom of The Mirage Convention
Center…
… starting right now
www.scope.com
Updated November 06
Page 23