Network Uptime: Fast Ethernet Collisions

January 15, 2001

Fast Ethernet Collisions

Although many networks are implementing switched networks and full-duplex links, collisions are still micro-managed by many network administrators. Although we don't have enough room in this issue to help folks understand why ONE collision isn't the end of the world, we can shed some light on what a collision might look like on a protocol analyzer.

In the 10 megabit Ethernet world, the jam sent after a collision looks like an alternating pattern of 01010101 or 10101010, depending on where the collision might hit the frame alignment. The hexadecimal decode of this would be a pattern of 5's or A's, respectively. When looking at the hex decode, this pattern really jumps out of the screen!

ADDR  HEX                                               ASCII
0000: 08 00 20 06 cc a4 02 60 8c 2a 4d 0d 08 00 45 00 | .. .I.`O*M...E.
0010: 00 9c c7 0a 00 00 0f 11 d2 d1 c0 09 c8 96 c0 09 | .o?.....O?A.E-A.
0020: c8 cb 00 7f 08 01 00 88 00 00 20 05 09 cf 00 00 | EE...^.. ..I..
0030: 00 00 00 00 00 02 00 01 86 53 55 55 55 55 55 55 | ........+SUUUUUU
0040: 55 55 55 55 55                                  | UUUUU

It's interesting to note that the 802.3 specification doesn't actually specify the signal that should be sent for a jam. However, every manufacturer has used the alternating pattern of 0's and 1's, which makes it easy to find the jam in a hex decode.

When 100 megabit Ethernet networks arrived, the familiar AA and 55 collision information disappeared, and a new pattern of D0 and 43 appeared.

ADDR  HEX                                               ASCII
0000: 08 00 09 d4 8d 24 08 00 20 71 39 94 08 00 45 00 | ...O?$.. q9"..E.
0010: 05 dc cc d0 40 00 ff 06 1c 41 b1 b1 14 d0 b1 b1 | .?ID@.?..A??.D??
0020: 14 d7 80 28 00 14 73 f8 56 5e 16 c1 32 01 50 18 | .x?(..soV^.A2.P.
0030: 60 f4 28 c5 00 00 00 00 00 00 00 00 00 00 00 00 | `?(?............
0040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 | ................
0050: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
0060: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
0070: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
0080: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
0090: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
00a0: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
00b0: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
00c0: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
00d0: 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 43 | CCCCCCCCCCCCCCCC
00e0: 43 43 43 43 43 43 43 43 43 43 43 43 43          | CCCCCCCCCCCCC

Although it was a bit alarming to see 'new' collision jam, we soon realized that we were looking at the same jam in a new Ethernet signal! 10Base-T uses Manchester encoding to send data across the wire. The signaling type of 100Base-T uses 4B5B, which changes the way information is sent over Fast Ethernet. Here's a chart that compares Manchester to 4B5B:

Decimal   4-bit data group   5-bit symbol
-------   ----------------   ------------
0         0000               11110
1         0001               01001
2         0010               10100
3         0011               10101
4         0100               01010
5         0101               01011
6         0110               01110

If the 4-bit equivalent of 43 43 is put together, the 5-bit result is a string of 0101010101010101's!

So, sit back and relax. Those new Ethernet jam codes are the same as the old Ethernet jam codes. Who needs full duplex, anyway?

Posted by james_messer at January 15, 2001 11:00 PM

Comments