In order to guarantee collision detection, signal attenuation must not exceed 11.5db. Amplifiers/Repeaters (hubs) restore signal levels but add delays. Repeater delays are specified as no more than 9.6mSec for 10Mbps operation and to 960 nanoseconds at 100Mbps. Distance delays plus repeater delays must not exceed the smallest packet duration -- 512bytes for 10/100 Mbps Operation. An interpacket time gap of the same duration is required. 8.4us distance delays (1.2km roughly) are allocated at 10Mbps, 840ns (120 meters) at 100Mbps. Both allow four repeaters with 100ns delays at 10Mbps and 10ns delays at 100Mbps. This has been codified as follows:
1. Thick Net Ethernet: Five 500 meter segments. Four Repeaters. Only four segments can have devices tapped off*. Minimum tap separation 2.5 meters.*
2. Thin Net Ethernet: Five 185 meter segments. Four Repeaters. Only four segments can have devices tapped off*. Minimum tap separation 0.5 meters.* Maximum tap separation 100 meters.* Maximum taps per segment 30.
3. 10 Mbps Unshielded Twisted Pair: Five 100 meter segments. Four repeaters.
4. 100Mbps Unshielded Twisted Pair
With short segments, additional repeaters may be possible.
Unlike hubs, Ethernet switches ingest entire packets and retransmit them. While delays will be longer than those in amplifiers, collision and retransmission can be handled in the switch. Therefore, when switches are used, segment length limits still apply, but repeater limits apply only with subnets connected to the switches.
*No, I don't know why this limit exists. Neither can I find any documentation explaining the limit.
Return To Index Copyright 1994-2002 by Donald Kenney.