CA2101860A1 - High speed data transfer over twisted pair cabling - Google Patents

Abstract

Abstract A method provides for a first network node in a plurality of network nodes to transmit a data packet to a hub. The hub and the network nodes are interconnected within a local network system. Control signals are exchanged between the first network node and the hub. The exchange of control signals is done in a first signal frequency range. A data packet is sent from the first network node to the hub. The data packet is sent using data signals within a second signal frequency range. The first signal frequency range and the second signal frequency range do not overlap.

Description

:

i8~ ~

HIGH ~3PEED D~TA T~ 3FElR O~ER TWISI~13D P~R C~BLI~G
~,Dd The present invention concerns a high speed data transfer over twisted pair cabling.
Generally, for high speed data transfer in a local area network (LAN) where data transfer rates are in excess of 25 MHz, data has been transferred using a fiber optic media. For example, the fiber distributed data interface . (FDDI) protocol is a common network protocol which operates using a fiber `,. optic medium.
' 1~ The use of fiber optic media for local area networking presents various problems. Particularly, most e~cisting buildings do not have an ; installed base of f;ber optic cable. Therefore, to utilize a fiber optic network it .i~ is generally required to specially install fiber optic cabling. This can be cost prohibitive.
There has been some work done to increase the rate over which data can l~e transferred over installed twisted pair cabling. Twisted pair cabling ~A, iS used for voice grade telephone transmissions. See for example Patent Number 5,119,402 issued to Simon A. Gin~burg et al. for a Method ~nd . 20 Unshiçlded~wiste~airs. However, in the prior art there has been no work Ii which has sufficiently increased the speed of data transm~ssion so that - transmission over a four pair voice grade twisted wire pair network can . .
rival the speed of data transmissions over fiber optic cabling. I
, . .

In accordance with the preferred embodiment of the present .
-i invention, a method is presented which provides for a first network node in .,. .` , .
:i a plurality of network nodes to transmit a data packet to a hub. The hub ,. . .
i . .
, ~ Hewlett-Packard Company HP10926~-1 ~, ;
", . :. . , ~ . ~ . : . : : . .

2 21~
and the network nodes are interconnected within a local network system.
Control signals are exchanged between the first network node and the hub.
The exchange of corltrol signals is done in a first signal frequency range. A
data packet is sent from the first network node to the hub. The data packet is i ~ sent using data signals within a second signal firequency range. The first signal frequency range and the second sigrlal frequency range do not `` overlap.
In the preferTed embodiment, the first network node is connected to - the hub using a plurality of twisted wire pairs. YVhen the control signals are exchanged between the first network node and the hub, a fir~t set of the first plurality of twisted wire pairs is used to send control signals from the first network node to the hub. A second set of the first plurality of twisted wire pairs is used to send control signals from the hub to the first network node.
When the data packet is sent ~om the first network node to the hub, all twisted wire pairs in the first plurality of twisted wire pairs are used to senddata packet from the first network node to the hub. For example, the first set -of the first plurality of twisted wire pairs includes two twisted wire pairs.
i~
`3! The second set of the first plurality of twisted wire pairs includes two twisted ; ~ wire pairs.
21) Additionally, in a preferred embodiment, control signals are exchanged between the hub and each of the plurality of network nodes excluding the first network node. This e~change of control signals is done .
` in the first signal frequency range for the purpose of providing arbitration to ~, determine which network node will transfer a next data package to the hub.
.:
, 25 Further, this exchange of control signals is done simultaneously to the data packet being sent from the first network node to the hub.
In one embodiment of the present invention, while receiving the data packet from the first network node, the hub checks a destination address for ~' Hewlett-Packard Company ~IP 1092665-1 .~ .

2 ~ $~
the data packet. When the destination address is for a second network node ;; in the plurality of network nodes, the hub sends the data packet to the second network node as the hub receives the data packet from the first network node. When the destination address is for a first subset of at least two of the plurality of network nodes, the hub stores the data packet until the hub completely receives the data packet from the first network node. Then the ` hub sends the data packet to the first subset of network nodes.
., ...
~s,~
.;~
Figure 1 shows a simplified block diagram of the interconnection of ~- various networks.
Figure 2 shows a simplified block diagram of a network in accordance ` ~ with a preferred embodiment of the present invention.
Figure 3 shows a simplified block diagram of a network device of the ~ 15 network shown in Figure 2 in accordance with a preferred embodiment of ;~q the present invention.
Figure 3A is a block diagram which shows logical flow of information within the network device shown in Figure 3 in accordance with a preferred embodiment of the present invention.
2û Figure 4 shows a simplified block diagram of the hub of the network ' ~ shown in Figure 2 in accordance with a preferred embodiment of the ` present invention.
Figure ~ shows a simplified block diagram of a t~ansceiver within the .;; hub shown in Figure 4 in accordance with the preferred embodiment of the 25 present invention.
i`-?, Figure 6 shows a simplified block diagram of a repeater within the ~ hub shown in Eigure 4 in accordance with the preferred embodiment of the : i present invention.
...
. -,,'.,.
Hewlett-Packurd Company HP109266 ..
.
~ .

. , . ~, ~ . .. . . .
: . , : .: .: .

.:

- 4 ',~ 3 Figure 7 is a state diagram for a repeater state machine within the :: ~
repeater shown in Figure 6 in accordance with the preferred embodiment of the present invention.
Figure 8 is a state diagram for a training sta$e maehine within the 5 repeater shown in Figure 6 in accordance with the preferred embodiment of the present invention.
Figure 9 is a state diagram for a client state machine within the network device shown in Figure 3A in accordance with ths preferred embodiment ofthe presentinvention.
Figure 10 is a state diagram for a client training state machine within , . .
the network device shown in Figure 3A in accordance with the preferred embodiment ofthe presentinvention.
Figure 11 is an example of a filter design which may be used the hub shown in Figure 4 and the network device shown in Figure 3 in accordance with the preferred embodiment of the present invention.
Figure 12 shows logic blocks within a network interface which prepare data to be forwarded to a hub in accordance with the preferred embodiment of the present invention.
i` Figure 13 shows a diagram which explains data flow within the logic .~, blocks sho~wn in Figure 12 in accordance with the preferred embodiment of .:
- ~ the present invention `i`l Figure 14 is a tirning diagram showing the ti~ing of signals through a group of four twisted pairs of copper wires in accordance with a preferred embodiment of the present invention.
....
Figure 15 is a block diagram of a circuit which facilitates collision detection in a network in accordance ~unth a preferred embodiment of the :~ present invention.
~. .
~,,, ~`''''` .
: ~ Hewlett-Packard Company HP1092666-1 .,! ~
'-.~ ' .

.
Figure 16 and Figure 17 show potential frequency spectrums for signals sent across a twisted pair shown in Figure 15 in accordance with a preferred embodiment of the present invention.
~i Figure 18 shows a block diagram of a circuit which proYides common-;~ 5 mode collision signaling in a network in accordance with a pref0rred ;i embodiment of the present invention.
Figure 19 shows a block diagram of a circuit which provides in-band collision signaling in a network in accordance with a preferred embodiment ofthe presentinvention.
Figure 20 is a block diagram which shows how the implementation . ~i ,j circuit in Figure 15 or the circuit shown in Figure 18 may be used in a network where a network device sends data to a network device over four twisted wire pairs in accordance with a preferred embodiment of the present invention.
: ' ~ 15 Figure 21 shows a hub connected to network nodes through four ,, twisted wire pairs in accordance with a preferred embodiment of the present nventlon.
, '7, Figure 22 shows an example of signal timing packages w~thin a - network in accordance with a prefe~Ted embodiment of the present .,i. :
invention.
Fi~ure 23 shows an example of signal timing packages within a ~,j ni network in accordance with an alternate preferred embodiment of the r.~,^
present inYention. I
rr Figure 24 shows a hub connected to network nodes in a system where r75 there is a collision window before each packet transmission in accordance with a preferred embodiment of the present invention.

Hewlett-Packard Company NP109266 . . : -:

6 ~ 60 Fi~ure 25 shows a circuit within a receiving network device which can be used to adjust squelch for information rece*ed from a tw~sted pair in accordance with a preferred embodiment of the present invention.
Figure 26 is a ti~ung diagram which illustrates operation of the 5 circuit shown in Figure 25 in accordance with a prefe~red embodiment of the presentinvention.

, .
~'7 Figure 1 shows a simplified block diagram of the interconnection of ` 10 various networks. A local area network 11, a local area network 12 and a , . .
local area network 13 are connected, for example, through a bridge to network 10. Network 10 operates, for example, using the fiber distributed data interface (FDDI) protocol. Local area network 11 and local area network 13, may operate, in accordance with any number of protocols. For example, if connected through a router, these local area networks could ,,, operate in acçordance with the IEEE 802.3 protocol, with the Token Ring protocol, with ISDN protocol or with WAN protocol.
Various network devices may be connected to the local area networks.
For example, a network device 14 and a network device 15 are shown ` ~0 connected to local area network 11. A network de~nce 16, a network device 17 and a network device 18 are shown connected to local area network 12. A
network device 19, a network device 20 and a network device 21 are shown `~ connected to local area network 13. Network devices 14 through 21 may be, .~ , for example, a work station, a personal computer, a network server, or some .~.` 25 other device.
;~ Figure 2 shows a block diagram of local area network 12. Local area '~' network 12 includes a hub 30. Hub 30 is connected to network device 16, ~
.~( through four twisted pairs of copper cable 31. Hub 30 is connected to :'."
.
... Hewlett-Packard Company HP10926~

7 ',~ . 8 ~ ~
network device 17 through four twisted pairs or copper cable 32. Hub 30 is connected to network device 18 through four twisted pairs of copper cable 33.
Figure 3 shows a simplified block diagram of a network interface 41, which is used by each of network devices 16, 17 and 18 to interface with hub 30. A backplane interface 42 provides an interface between computer system RAM and the network device. A random access memory (RAM) 43 is used to temporarily store data packets received from or to be transferred out on the network. A media access controller (MAC) 44 is used to control !
i the flow of data within network interface 41. A transceiver 45 is used to send and receive through the network. A transformer and filter 46 is used to adjust voltage and provide noise filt~ring for signals transferred between transceiver 45 and a connector 47. A connector 47 is connected to the bundle ~: of four tw~sted pairs of copper wire from hub 30.
Figure 3A is a block diagram which shows logical flow of information within network interface (client) 17. A client training state machine 501 is used in initializing connection between network device 17 and hub 30. A
~, client state machine 502 is used to control dat,a transactions between network device 17 and hub 30. A DMA controller 503 is used to control DMA
transfers between RAM 43 and a data buffer 504. Twisted pair transmit '.t' æ logic forwards data to transceiver 45. Twi6tedl pair receive logic receives data from transc~iver 45.
~; Control signals flow from DMA controller 503 to RAM 43 through aninformation channel 507. Control signals flow from DMA controller 503 to I
data buffer 504 through an information channel 510. Data flow between DMA controller 503 and RAM 43 through an information channel 508. ;
~,~!, DMA controller 503 signals client state machine 502 through information channel 509 when there is a packet to transmit. Data buffer 504 sends data ;`,, to twisted pair transmit logic 505 through information channel 511. Data ~;

Hewlett-Packard Company HP1~926~

:

:`

2 ~ 6 ~
buffer 504 receives data frorn twisted pa~r receive logic 506 through information channel 512. Transceiver 45 receives data ~rom twisted pair transmit logic 505 through information channel 513. Transceiver 46 sends I
. .
data to twisted pair receive logic 506 through information channel 514.
5 Twisted pair receive 506 signals client state machine 502 through an .~ information channel 515 at the start of a packet and the end of a packet (RXDONE). Client state machine 502 signals twisted pair receive 506 through an information channel 516 when a packet is to be received.
.; Twisted pair transmit 505 signals client state machine 502 through an 10 information chiannel 517 when a transmit is omplete. Client state machine 502 signals twisted pair transmit 505 through an in~ormation channel 618 . when a packet is to be transmitted.
. Figure 4 shows a block diagram of hub 30. A backbone physical ,,,.:
interface 61 provides a physical interface of hub 30 to network 10. A
backbone media access controller (MAC) 52 controls data flow between hub 30 and network 10. A bridge buffer RAM 53 provides temporary storage for .;, data flowing between hub 30 and network 10. A repeater 67 directs data flow on locial area network 12. A content addressable memory (CAM) 54 is addres~able with a network address and outputs an associated port. A
~; 20 broadcast SRAM 56 is used for temporary storage of multi-port messages " i,, which are to be broadcast across LAN 12.
;., A network management system 58 provides network management.
~;c Network management system ~8 includes a processor 60, an EPROM 62, a -.; RAM 61 and a memory access controller (MAC) 59. EPROM 62 stores 25 program information used by processor 60. RAM 61 stores programs used i. by processor 60. MAC 59 provides a means for processor 60 to commun~cation with other nodes on the network.

.; j~ ; .
.~ Hewlett-Packard Company HP109265~
~,, . .

.~

A transceiver 63 is used to send and receive data to and from a network device connected to a connector 67. A transceiver 64 is used to send and receive data to and from a network device connected to a connector 68.
A transceiver 65 is used to send and receive data to and from a network . 5 device connected to a connector 69. A transceiver 66 is used to send and receive data to and from a network device connected to a connector 70. While only transceivers 63 through 66 and connectors 67 through 70 are shown, , many more transceivers and connectors can be added. For e~ample, in the preferred embodiment of the pre~ent invention, the hub has 24 ports. A
transformer/filter 73 connects transceiver 63 to connector 67. A
transformer/filter 74 connects transceiver 64 to connector 68. A
transformer/filter 75 connects transceiver 65 to connector 69. A
transforrner/filter 76 connects transceiver 66 to connector 70.
:.';!

3 Figure 5 shows a block diagram of transceiver 63. Transceiver 63 is 15 connected to connector 67 through four twisteld wire pairs. The first twisted ~, wire pair includes a connector line 81 and a connector line 82. The second ~i twisted wire pair includes a connector line 83 and a connector line 84. The third twisted wire pair includes a connector line 8~ and a connector line 86.
The fourth twisted wire pair includes a connector line 87 and a connector ~0 line 88.
. ", ~
~,~ When transceiver 63 receives data over the four twisted pairs, the data is received by an equalization circuit 91, an equalization circuit 92, an equalization circuit 93 and an equalizatiorl circuit 94. Each of equalization circuits 91 through 94, is used to provide a clean and amplified signal. In : ~ , ''''','! ~6 addition, equalization circuit 91 also provides a carrier detector signal on a ..
carrier detector line 180.
A phase locked loop (PLL) clock and data recovery circuit 101 receives a clean and amplified signal on line 181 and 182. PLL clock and data ~ ..
~ Hewlett-Packard Company NP1092655-1 lo 2 ~
recovery provides a data signal on a line 171, a clock signal on a line 175 and a clock valid signal on a line 279. A phase locked loop (PLL) clock and data recovery circuit 102 receives a clean and amplified signal on line 183 and 184. PLL clock and data recovery provides a data signal on a line 172, a clock signal on a line 176 and a clock valid signal on a line 280. A phase locked loop (PLL) clock and data recovery circuit 103 receives a clean and amplified ;~ signal on line 185 and 186. PLL clock an~1 data recovery provides a data signal on a line 173, a clock signal on a line 177 and a clock valid signal on aline 281. A phase locked loop (PLL) clock and data recovery circuit 104 ,:
:. 10 receives a clean and amplified signal on line 187 and 188. PLL clock and data recovery provides a data signal on a line 174, a clock signal on a line 178 , and a clock valid signal on a line 282.
An elasticity buf~er 111, an elasticity buffer 112, an elasticity buf~er 113 . and an elasticity buf~er 114 synchronize the data signals from PLL clock and ., .
data recovery circuits 101 throuigh 104 to a single clock. Elasticity buf~er 111receives the data signal and clock signal ~om PLL clock and data recovery circuit 101 and produces a synchronized data signal on line 191. Elasticity . buffer 112 receives the data signal and clock signal from PLL clock and data recovery circuit 102 and produces a synchronized data signal on line 192.
ao Elasticity buf~er 113 receives the data signal and clock sigllal from PLL clock and data recovery circuit 103 and produces a synchronized data signal on , line 1~3. Elasticity buf~er 114 receives the data signal and clock signal from - ~:
PLL clock and data recovery circuit 104 and produces a synchronized data sig~al on line 194.
. 25 A logic OR gate 170 receives the clock signal on line 175, the clock valid signal on line 279, the clock valid signal on line 280, the clock valid i~ signal on line 281 and the clock valid signal OIl line 282. Logic OR gate 170 `; produces a clo~ signal (Clk 0) on a line 190. The Clk 0 signal passes ::

.... .
. .~, Hewlett-Packard Company HP10926~
,~

11 '2~8~
through Lo~ic OR gate 170 when the four clock valid signals are asserted low. Driver buffer 106 forwards data to repeater 57 on a line 126, a line 127, aline 128 and a line 129. Driiver 106 also provides a clock signal on a line 129.Receive line state logic 105 is used to receive and forward transfer set-up requests over the first and the second twisted wire pairs. Receive line state logic 105 receives the carrier detector signal on carrier detector line 180, the data signal on line 171, the data signal on line 172 and the clock : .
signal on line 175. Receive line state logic 105 produces a priority (PRI) `'~! request signal on a line 121, a receive line state signal (RLS0) on a line 122, , 10 a receive line state signal (RLS1) on a line 123, an receive line state signal . .
(RLS2) on a line 124 for forwarding to repeater 57. Repeater enables receive ~q line state logic 105 by placing a receive line state enable signal on a line 125.
~:. A receiver enable signal (RXEN) is generated by repeater 57 to select receive line state logiic 105 or driver buffer 106 to forward information to repeater ~7.
When transceiver 63 is used by repeater 57 to transmiit datai, repeater ; 1 57 places a first datai signal (TDATA0) on transmit data line 137, a second data signal (TDATA1) on transmit data line 138, a third data signal (TDATA2) on transmit data line 139 and a fourth data signial (TDATA3) on .'~ transmit data line 140. Whien transceiver 63 is used by repeater 57 to " ~
`.;~ 20 transmit control signals, repeater 57 places a first transmit line signal .
. ~ (TLS0) on a line 132, a second transmit line signali (TLS1) on a line 133, a ~:
third transmiit line signal (TLS2) on a line 134~ and a transmiit line clock ~:. (TLSCK)onaline135. TLSCKisusedtostoretheTLSvalues. Transmit ~1 :
i .. .
line state logic 115 generates tones and drives the TLS values to be forwarded 25 to a multiplexor/transmitter 116.
Multiplexor/transmitter 116, in response to a transmit enable signal ~ i (TXEN) on aline 136, selects eitherdatasignals onlines 137,138,139 and i.. ~ 140 to be forwarded to the four twisted wire pairs 81 through 88, or the tones ".1 ~ .

`~ Hewlett-Packard Company HP1092655-1 ~ ,. . .

.!. ~ i ~

8 ~ ~
and driver enables from transmit line state logic to be forwarded to the third and fourth tw~sted wire pairs 85 through 88. A transmitter clock (TXCLK) is provided to transmit line state logic 115 and multiplexor transmitter 116 on a ; line 141.
Figure 6 is a block diagram showing data flow within repeater 57.
Repeater 57 essentially functions to channel transferred data. Twisted pair receive logic 212 receives data and control signals from the transceivers, e.g., transceivers 63, 64, 6~ and 66. For e~ample, twisted pair receive logic 212 is connected to lines 121 through 131 of transceiver 63. Broadcast RAM
. 10 read-back logic 211 receives data from broadcast SRAM ~6. Backbone receive logic receives data from bridge buffer RAM 53.
~ Twisted pair transmit logic 220 sends data and control signals to the transceivers, e.g., transceivers 63, 64, 65 and 66. For example, twisted pair transmit logic 220 is connected to lines 132 through 141 of transceiver 63.
15 Broadcast write logic 219 sends data to broadcast SRAM 56. Backbone transmit logic sends data to bridge buffer RAM 53.
The data rece*ed by tw~sted pair receive logic 21a, broadcast RAM
read-back logic 211 and backbone receive logic 210 is channeled through a . first-in-first-out (FIF(:)) b~er 215 to either backbone transmit logic 218, .. .
broadcast write logic 219 or twis~ed pair transmit logic 220. FIFO buffier 215 ~3 also provides arbitration information to a receiver port arbiter 214.
~ i :
Receiver port arbiter 214 selects firom which port to receive a data transmission. In general, a simple arbitration scheme is used. For ... . .
exampleJ a round robin arbitration scheme may be used in which the last 25 port from which a data transmission is received is given lowest priority. A
, .
~;j transmit arbiter 217 determines to which port of backbone transmit logic 218, broadcast logic write logic 21~ or twisted pair transit logic 220 data is to '~" be transmitted. Transmit arbiter 217 determines where to send a message ;~
' Hewlett-Packard Company HP1092655-1 ~, ~;~

by forwarding a network address of the message to CAM 54. CAM 64 retuIns a port number to transmit arbiter 217. Repeater 57 also includes a repeater state machine 216 and a training state machine 213.
Figure 7 shows a state diagram for repeater state machine 216. After start-up and training of all ports (as explained below) has taken place, repeater state machine is an idle state 131. Upon receiving a request for .: transfer from receiver port arbiter 214, repeater state machine enters an acknowled~e port state. When repeater state machine 216 is in the acknowledge port state, repeater 57 sends an acknowledge signal to the port ,- 10 which was selected by receiver port arbiter 214. If repeater 57 times out ~ii before repeater 57 begins to receive a data packet from the selected port, ` repeater state machine enters a set retrain port state 239. In set retrain port state 239, repeater state machine 216 signals training state machine 213, to . 1 retrain the port. Repeater state machine 216 then returns to idle state 231. :
',~,! 15 From acknowledge port state 232, upon repeater 57 beginning to ~ ~ .
.~' receive a network data packet, repeater state machine 216 enters a ,, .
determine destination state 233. While repeater state machine 216 is in : ~ :
determine destination state 233, transmit arbiter 217 determines where to ;~i send a message by forwarding the network adldress in the network data` ~0 packet to CAM 54. CAM 54 returns a port nw[nber to transmit arbiter 217. ~ :
i If transmit arbiter 217 determines the destination is to a port within . .
,~
~ local network 12, repeater state machine 216 enters a transmit to port state ~ :
- 235. In transmit to port state 235, data as it is received from the port selected by receive port arbiter 214 is forwarded immediately to ~e port selected by ` ~ 25 transmit arbiter 217. Upon repeater 57 receiving the complete network data 5 '" packet and completion of the forwarding of the data, repeater state machine .i.~., ::
;.! 216 returns to idle state 231. If complete data packet is not received within a . specilSed time, repeater state machine 216 enters set ret~ain port state 239.
, :
`, ':; ~
.. ` Hewlel;t-Packard Cornpany HP10926~
.,~."

1~
In determine destination state 233, if transmit arbiter 217 determines the destination is to multiple ports within local network 12, repeater state machine 216 enters a buffer to local RAM state 237. In buffer to local RAM
state 237, data as it is received from the port selected by receive port arbiter214 is forwarded to broadcast SRAM 56. If a complete data packet is not received vrithin a specified time, repeater state machine 216 enters set retrain port state 239. Upon repeater 57 receivilng the complete networl~
data packet and completion of the forwarding of the data to broadcast SRAM
56, repeater state macliiine 216 enters a transmit to all ports state 238. In transmiit to all ports state 238, repeater 57 reads the broad cast message in broadcast SRAM 56 and forwards the message to each of the ports specified.
Upon completion of the data transmissions, repeater state machine 216 returns to idle state 231. :~
. .ji .
In determine destination state 233, if transmit arbiter 217 determines .~. 15 the destination is to the backbone of local network 12, repeater state machine ~i` 216 enters a buffer to bridge state 234. In buffer to bridge state 234, data as it is received from the port selected by receive port arbiter 214 is forwarded to :.~ bridge buf~er RAM 53. If a complete data packet is not received within a ,,s specified time, repeater state machine 216 enters set ret~ain port state 239.
ao IJpon repeater 57 receiving the complete network data packet and ~:
completion of the forwarding of the data to buffer RAM 53, repeater state machine 216 returns to idle state 231. If buffer RAM 53 runs out of available memory locations before completion of the transfer of the network data packet, repeater state machine 216 enters a set busy signal state 236. In set 26 busy signal state 236, repeater 57 sends a busy signal to the transmitting :~
data port and throws away the network data packet~ Upon completion of the :~ transfer of the network data packet to repeater 57, repeater state machine ~ . 216 returns to idle state 231.
:.:
... - .
,.....
~. .
. Hewlett-Packard Company HP109266~-1 . ,.
,. `

15 ~ 8 ~ ~
Figure 8 is a state diagram for training state machine 213. A~ter a ,~ reset or whenever it is necessary to train a port, training state machine 213 proceeds through the training states. Before training a port, training state machine 213 is in an idle state 241. When training state machine 213 5 receives a training idle up signal from a port which requests training, ; training state machine 213 enters a drive transmit idle down state 242. In ;.
training idle down state 242, repeater 57 sends a training idle down signal to ~ the port requesting training.
;~ Upon receiving a request to transmit signal from the port to be 10 trained, training stat0 machine 213 enters a request to repeater state ";
machine state 243. In request to repeater state machine state 243, training ; state machine 213 waits for repeater state machine 216 to acknowledge the port to be trained. Upon repeater state machine 216 providing the .~ acknowledgment, training state machine 213 enters an acknowledge client 15 state 244. In acknowledge client state 244, training state machine 213 waits for the port to start sending a training packet.
i Upon the port starting to send a packet, training state machine 213 .. , ~ .
enters a receive training packet state 245. In receive training packet state ~ ~
,.
`~ 245, training state machine 213 waits for completion of the sending of the -~
aD training packet. When the training packet has been received, training state machine enters a training completion state 246. In training completion state 246, a check is done to see whether receive training is complete. For example, in the pre~erred embodiment, training is complete if 25 l~
consecutive training packets have been received without errors. If there are 25 errors in reception, the equalization and clock frequencies are adjusted in the transceiver for the port. If receive training is not complete, train~ng ~`1 state machine 213 returns to drive training idle down state 242.

, ~ :
~ Hewlett-Packard Company HP1092656-1 : . ~

When receive training is complete, training state machine 213 enters - a request to repeater arbiter state 247. In request to repeater arbiter state 247, training state machine 213 requests transmit arbiter 217 to initiate the transmission of a training packet to the port being trained. Upon receiving 5 an acknowledgment from repeater state machine 216, training state machine 213 enters a transmit training packet state 248. Upon completion of the transmission of the transmit training packet state, training state machine 213 enters a training complete state 249. If transmit training is not -, complete, training state machine 213 returns to request to repeater arbiter ` 10 state 247.
,. When transmit training is complete, training state machine 213 , enters a port is ;n line state idle state 250. Training state machine 213 remains in port is in line state idle state 250 during normal operation of the ''~s port. When the port requests retraining, training state machine 213 returns ~ ~;
15 to drive training idle down state 242.
Figure 9 shows a state diagram for client state Imachine 502. Client state machine 502 is initially in an idle state 251. When hub 30 signals that a packet will be incoming to the client, client state machine 502 enters a wait ~or packet state 252. In wait for packet state 252, if the client sees the receive 20 line state transition to a state other than incoming, such as idle, client state machine 502 returns to idle state 251. Upon the client beginning to receive a packet, client state machine 502 enters a receive packet state 253.
In receive packet state 263, client state machine 502 waits for the end ~i of the packet. When the end of the packet is received and client state 25 machine 502 is not waiting on a busy signal, client state machine 502 returns to idle state 251. When The end of the packet is received and client .. `:~!
;i'~ state machine 502 is waiting on a busy signal, client state machine 502 .:'!
~'' enters a wait to re-transmit state 258.
```!
~ , ,, . - Hewlett-Packard Company HP1092656-1 .:

. ,. ~ .: ., ~ . .

17 b~
From idle state 251, when the client desires to transmit a packet, the client state machine 502 enters a send request state 254. In send request state 254, sends a request to transmit signal to hub 30. The client then waits . for an acknowledgment from hub 30. While waiting, if hub 30 signals that a 5 packet will be incoming to the client, client state machine 502 enters wait for packet state 252. In send request state 254, when the client receives an acknowledgment from hub 30, client state machine 502 enters a transmit ;t packet state 256. In transmit packet state 256, the client sends a data packet ~ -:~1 i-, to hub 30. Upon the end of packet being sent, client state machine 502 enters ',,t 10 a wait ~or idlelbusy state 257. In wait for idle/busy state 257, if the client ~:
receives an idle signal from hub 30, trans~ission of the packet was : i successful and client state machine 502 returns to idle state 251. If the clientreceives a busy signal from hub 30, client state machine 502 enters wait to re-transmit state 258. :
;,.
In wait to re-transmit state 258, client state machine 502 waits for hub :
30 to stop sending a busy signal. If hub 30 signals that a packet will be incoming to the client, client state machine 502 enters wait ~or packet state 252. VVhen in wait to re-transmit state 258, client state machine 502 detects ;.g; the busy signal from hub 30 being de-asserted, client state machine ~02 returns to send request state 254. When in wait to re-transmit state 258, client state machine 502 times out waiting for hub 30 to de-assert the busy ~:
.; .
signal, client state machine 502 enters a discard packet state 259. ~-:
~^ When client state machine 502 is in discard packet state 259, the client :~i discards the network packet. Then, client state machine 502 returns to idle `1 2~ state 251.
Figure 10 shows a state diagram for client tra~ning state machine 501.
~- Upon receipt of a reset, client training state machine 601 enters a training ` idle up state 261. In training idle up state 261, the client forwards to hub 30 a ' ' ~

- Hewlett-Packard Company HP109265~
. .:

18 2 ~
training idle up signal. Upon hub 30 signaling the client that the training idle up signal has been rece3ived, client training state machine 501 enters request state 262. When client training state machine 501 is in request state 262, the client signals to hub 30 a request to transmit a training packet.
: 5 Upon an acknowledgment from hub 30, client training state machine 501 3i enters a transmit training packet state 263. When client training state ;~ machine 501 is in transmit training packet state 263, the client transmits the training packet to hub 30. When the transmission is complete, client '- training state machine 501 enters a wait for response state 264.
VVhen client training state machine 501 is in wait for response state 264 and the client receives a training idled signal from hub 30, client .;~ training state machine 501 returns to request state 262. When client ;
training state machine 501 is in wait for response state 264 and the client receives from hub 30 an incoming packet signal, client training state .
machine 501 enters areceive trainingpacketstate 26 In receive training packet state 265, when the client has received the ~ ;
entire training packet, client training state machine 501 enters a training complete state 268. If training is not complete, client training state machine ~~ 501 enters a wait for incoming state 267. When client training state! ., machine 501 is in wait for incoming state 267 and the client receives from ., ~ hub 30 an incoming packet signal, client training state machine 501 enters ,:, receive training packet state 265.
;:~ When client training state machine 501 is in training complete state i .3 s~ 268 and training is complete, client training state machine 501 enters 25 training idle state 269. When client training state machine 501 is in training idle state 269, the client is in a normal operating state. Upon a ..
~` transmission error occurring, the client receives a transmit idled signal . .~
:-, `'~
~: Hewlett-Packard Company HP1092665-1 .'~

19 '2 ~
from hub 30. Upon receipt of the transmit idled signal from hub 30, client training state machine 501 returns to transmit idle up state 261.
Figure 11 shows an example of a filter design which may be used to implement filters 73 through 76 within hub 30, and may also be used to implement the filter portion of transformer and filter 36 within network interface 41. For example, filter has a return loss of less than or equal to -20dB for signals from 100 KHz to 15 MHz. The 3 dB cutDff frequency is between 19 - 21 MHz. Stopband attenuation is greater than or equal to 13.5 dB at 30 ~ MHz. The filter includes, for example, a resistor 281, a resistor 289, a ;i 1~ capacitor 282, a capacitor 283, a capacitor 284, a capacitor 290, an inductor 285, an inductor 286, an inductor 287 and an inductor 288, connected as shown. For example, resistor 281 is 50 ohms, resistor 289 is 50 ohms, capacitor 282 is 33 picofarads, capacitor 283 is 110 picofarads, capacitor 284 ~l is 160 picofarads, capacitor 290 is 33 picofarads, inductor 285 is 330 nano- ;
1 15 henries, inductor 286 is 680 nano-henries, inductor 287 is 330 nano-henries, `~ inductor 288 is 680 nano-henries.
In the preferred embodiment of the present invention, the physical , . .
`' layer implementation of the connection betwleen hub 30 and network . interface 41 is intended to provide a high speed communications link over 20 low cost wiring. The below described specific application provides a 100 ~-~ megabit communication channel over voice grade telephone wire. This is ~` done by multiplexing 4 adjacent channels at 25 megabits each.
~; The media type for the twisted pairs is, for example, Category III
UTP, Category IV UTP or Category V UTP. The media distance is for example 100 meters when using Category III UTP, 120 meters when using Category IV UTP, or 150 meters when using Category V UTP. The media configuration is a 4 pair, 25 pair bundles, (10-BASE-T compatible wiring systems).
,:~.- , .~
Hewlett-Packard Company ~IP1092666~

2~8~
In implementing the physical layer, a method of transmitting 25 megabits of information v.~ithin a similar bandwidth to 10 BASE T encoding is desired to provide comparable attenuation and crosstalk characteristics.
Cornparable SNR and DC balance is also desired. For this purpose non-5 return to zero (NRZ) encoding using a 5B/6B block code is utilized to providemaximum balance. This is done by taking all balanced 6B symbols and associating them with particular 5B symbols. Then, the remaining 5B
:~ symbols are associated v~ith 6B symbols tha$ are unbalanced by a single bit.
:~ The same 5B symbols are also associated with the inverse of that 6B symbol.
Duriing transmission, a status bit determines whether the last unbalanced ~-,: symbol sent was positive or negative. If the status indicates the last unbalanced symbol was positive, the encoder then inverts the next .~ unbalanced symbol and toggles the status bit. This way, DC balance is ~1 maintained in the data stream. Care is taken to ensure that the unbalanced 15 symbols have no more than three consecutive bits on the symbol boundary.
This way the run-length is limited to no greater than six bit times. The following block code listed in Table 1 below meets the above c~iteria.
;~
,...
., .
.,, ::!
,iij, ~,;
.~.
:`~
,:~
,~., I
i ::, , ~ ~

;''~
:;~
,.,~

~ Hewlett-Packard Company HP1~92666-1 .. ~

Table 1 ..
#$~3/6B # Qf 1s _ # 5!316B # of 1sALT5E3!6B
O000111 3 aD011011 4 100100 . 5 1001011 3 21011101 4 100010 2001101 3 ~20111iO 4 100001 -3 3001110 3 23100111 4 011~0û
4010011 3 ~4101011 4 010100 6011001 3 æ lolllo 4 010001 ~ 9100011 3 ~9110110 4 001001 ;, 18010101 3 ,'. ' .' ; 25 Table 2 below lists the frame format for the transfer of data at the s physical level.
,~ Table 2 ,,.
Preamble - 8 symbols (sextets) of alternating Os and 1s.
Start Delimiter - 1 symbol (sextet) of a specific one-zero patteIn.
Destination Address - 48 bits which are split up among the four pairB.
:j ` Source Address - 48 bits which are split up among the four pairs.
i Type/Length Field - 8 bits which are split up among the four pairs.
- Data block - 64-1500 bytes split up among the four pairs..~ 35 Cyclic Redundancy check - 32 bits used to ensure frame integrity.
End Delimiter - 2 symbols (sextets) of continuous ones.
~, Abort symbol - 2 symbols (se~tets) of continuous zeroes.
., .
., .
The frames are distributed among the four channels by breaking up I
40 the address, data, Type/Length and Cyclic Redundancy Check (CRC~
.. ~,.
~ segments and multiplexing this information. The first ~ bits are coded into i `~ a 6 bit symbol and transm~tted onto channel 0. The second ~ bits are coded . into a 6 bit symbol and transmitted onto channel 1, and so on. CRC is Hewlett-Packard Company HP109265 . ..
....

8 ~ ~
generated based upon the dat~ ~rame's bit sequence, and compared on the receive end after de-multiplexing the frame.
Because data is typically bounded on octet boundaries, and the . symbols are gathered on quintet boundaries, it is likely that a ~ew extra bits ;~ 5 will be stuf~ed into the final symbol to ensure encoding on proper -, boundaries. VVhen the data is returned to octet boundaries, those bits will be pushed into obliv~on, thus making the recovery of the data complete.
:, Table 3 below gives control symbols used in the frame format .~; (described below) used at the physical layer in the prefen ed embodiment.
~j 10 Table i,.J End Delirniter (ED) 111111 6 111111 Preamble (PlREAMBLE) 010101 3 010101 StartDclimiter (SD) 100101 3 100101 1~ Abort Symbol (ABORT) 000000 0 000000 -.
The code according to the preferred embodiment provides for a 15 megahertz tone on PREAMBLE which will allow the minimum time for ~! clock synchronization. Further, the code uses the same symbols in the data 20 stream, but makes the decision that PREAME~LE/SD is only being looked for immediately after energy is detected on the link.
, ... .
:i The transition from PREAMBLE into SD1 is designated by a " 11 " or "00" occurrence, and the SD symbol is balanced. The probability of misdetection can be reduced by requiri~g that the transceiver will not pass any received bits through until the clock has been secured, and by requiring six valid preamble bits to occur prior to accepting a valid SD. Inversion of the data stream can be determined by the polarity of the PREAMBLE-SD
symbol boundary.
; ` The End Delimiter is composed of all ones, and would be two such 30 symbols back to back. This provides a sequence of twelve consecutive ones , `::
.....
,, .
Hewlett-Packard Company HP 1092666-1 - which cannot be generated by any valid data pattern. If a bit error occurs in the ED, it would appear as an invalid symbol.
The ABORT symbol is pro~Tided to allow HUB-to-HUB data transfers to be dropped with minimal effort. If a valid ED has no$ occurred, and two 5 consecutive ABORT symbols appear, the receiving node considers the packet dropped.
The physical layer according to the preferred embodiment also includes scrambling. Scrambling is necessary to provide for clock recovery.
3~ In order to provide a system that operates within minimal e~cessive .,.~ .
10 bandwidth (approximately 35%), a very low bandwidth PLL is required.
~; This means that the distribution of spectral components must be random in ~-order to prevent clock drift. ~ ~
Scramblingis also necessaryto provide for crosstalkreduction. By ~ -spreading the energy in the transmitted signal, it has been found that 15 crosstalk is reduced by a few dB. This improves the signal to noise ratio . ~
(SNR) of the system. Scrambling is also necessary to provide for emissions ; reduction.
.1 For the preferred embodiment of the present invention, a stream cypher of 11 bits provides the spectral dispersion necessary to ensure the 20 above characteristics are met. Unlike a synchronous scrambler, the stream !cypher does not propagate eITors, nor does it exhibit the potential to "lock-.,up". The primary issue with stream cyphers has to do with -. .
sync}~onization. Since the data out of the cypher is a function of the incoming data and a pseudo random bit sequence (PRBS) of a time~
25 dependent value, it is necessary on the receive end to know e~actiy what ;l; point in the sequ0nce the data is associated with. This can be done by using :., ,j ~` a cypher on the data and presetting the cypher before performing an XOR
function on its contents with the data. The polynomial factors are S[n] = 1 +
, .

~. ~
`- Hewle~t-Packard Comp~ny HP109266~-1 P~
.
..

;

. S[n-9] ~ S[n-11]. The four channels each have a different cypher which is initialized to a different quadrant of the PRBS to avoid the likelihood of ~ common patterns on each wire.
`.` Figure 12 shows logic blocks within network interface which prepare ~ 5 data to be forwarded to hub 30. A scrambler/descrambler 293 scrambles ~;-.
message, residing in a memory 291, which are to be forwarded to hub 30.
;~ Scrambler/descrambler 293 descrambles message, residing in memory 291, ~ ~ which have been received from hub 30. Ser~ali~ation and blosk coding logic .-. 292 blr,ck codes and serializes scrambled data which is then forwarded to :
hub 30 via a data path 295. Deserialization and block decoding logic 293 ,.
deserializes ancl block decodes scrambled data which is received from hub 30 .~
; via a data path 296.
Figure 13 shows a diagram which explams data flow within the logic blocks shown in Fi,gure 12. A row of twenty bits 303 are shown in groups of 15 five bits. Byte boundaries 301 show where byte boundaries for the twenty bits ;, would exist in memory 291. Scrambling bits 303 yields a row of twenty bits 304. Bits 304 are serialized and block coded to produce four serial data streams 305 of six bits each. Each data stream is packetiæed and put onto a separate twisted pair. After being sent across local area network 12, a 20 network de~rice receives and depacketizes four serial data streams 307 which are identical to data streams 305. Streams 307 are deserialized and , .. .
`~ decoded to produce a row of twenty data bits 308. Data bits 308 are then ~ .
descrainbled to produce a row of twenty data bits 309. Data bits 309 are ` identical to data bits 303.
.. 25 In order to maximize data flow in network 12 and avoid crosstalk, , .
using four twisted wire pairs, for data channels, half-duplex data channel ; ` is used. However, full duplex is used for control/status channels. This . allows for noise immunity comparable to IEEE 10 BASE-T standards. Using .;', ,~
'!,',~'.1 Hewlett-Packard Company ~P109265~-1 ,. ..
,, , 25 2 ~ 8 $ ~
four twisted-pairs in a 10 BASE-T cable and half-duplex transm~ssion, 25 megabyte throughput is required through each twisted wire pair.
In order to maintain adequate noise immunity, the channel bandwidth must not be significantly increased. Through empirical measurements, it has been determined that crosstalk is acceptable when system bandwidth is kept below 21 Megahertz. In addition, a simple binary ~- (two level) code provides lower cost implementation.
Operating with a two level NRZ block code of reasonable efflciency, the , .
bandwidth of the system can be constrained to less than 21 megahertz. This keeps noise down, and the two level code provides robust noise-immunity.
The block code must be balanced and effioiency must be above 80%.
Therefore, as discussed above, a 5B/6B block code is used. `
This enabling scheme utilized by the present invention allows various ;~, other protocols to operate by either doing a 25 megabit full duple~ channel on two-pairs (e.g., as in 25 megabit 10 BASE-T), a 50 Megabit full duple~
communication channel on four-pairs (e.g., as in 50 megabit 10- BASE-T, or . 45 megabit ATM), or dual-100 Megabit channels on separate four-pair cables (e.g., as in FDDI, ATM).
ControVstatus information is full duplex in order to keep latency .,.
down. Therefore, it is possible to use two pairs for upstream -comnnunication, and two pairs for downstream communications. The ., :
`. transition rate of these channels is kept very low in order to minimize , .~ . . .
crosstalk effects on adjacent wires. By using tones of 0.9375 megahertz - 3.75 !
:~ megahertz crosstalk in bundles is minimized. Three tones per wire (plus a ~ 2S lack of tones) can allow up to ten different control status signals.
:~;i, ~n the preferred embodiment, eight line states are provided by the i~ transceiver state machine. For the purposes of the description below, hub .:

. Hewlett-Packard Company HPlOg2665~
:t ~ :
:

~ 30 is the master and the network devices are the slaves. Table 4 below .~ summarizes the extant line signals.
.`j Table 4 CODE ~I,AVE MASTE~_TX PR 1 IX PR 0 RX PR 1 ~ ~ p ~:~ 000 SILENCE SILENCE 0 0 0 0 :~ 001 IDLE IDLE 16 16 1~17 1~17 . 010 REQ 0 N/A 16 8 15-17 79 011 REQ 1 SYNCH 8 16 7-9 1~17 ; 101 RSVD Incoming 16 4 15-17 3-5 ~ 110 RSVD RSVD 4 16 3-5 15-17 '.i 111 RSVD RSVD 8 4 7-9 ~5 The listed transmits numbers (under TX PR 1 and IX PR 0) are the ;`~ 15 number of clock cycles each pulse contains. The listed receive mlmbers :~ (under RX PR 1 and RX PR 0) account for sampling error.
Line state (Code 000) provides for the transmitter to be turned off completely. As seen by Table 4, the first transmitter wire pair (17~ PR 0), the second transmitter wire pair (IX PR 1) the first receiver wire pair (RX PR 0) and the second rece*er v~rire pair (TX PR 2) are all at 0 (i.e., silent). In theevent that the MASTER detects silence on it's receiver for an extended pe~od of time~ it will transmit silence to prevent transmitting onto an . .~
unterminated line. The SLAV13 and the MASTER indicate SILENCE in the event they are about to begin reception of data. The SILENCE state allows :.
~;~i 25 for the twisted pair media to settle before data is inserted onto the wires.
`~ Line state (001) indicates that the SLAVE and MASTER are connected, and the link is inactive. The state is entered upon the end of a l~data transmission in one of two ways. In the event of a proper transmission, the ETD/ABORT sequence would create the first IDLE symbol which would tell the receiver on the opposite end of the link to disable it's data receptioncircuits. In the event of an aborted frame, the ETD would not appear, and ~:b,~, the ABORT symbol would provide the first component of the IDLE tone.

~i /
~1 Hewlett-Packard Company HP1092666~
". ~ : ~

21~8~

;~ Line state (010) is used by the SLAVE node to i~dicate a low priority request. Line state (011) is used by the ~3LAVE node to indicate a high priority request. This line state is used by the MASTER node to provide a synchronization pulse to end nodes.
6 Line state (100) is used to initiate a link connection sequence by theSLAVE node. Upon detection of this tone, the MASTER will indicate T IDLE
which will indicate to the SLAVE that a connection exists. Then, the connection arbitration cycle (training) is then executed. Line states (101,110,111) are not implemented in the preferred embodiment.
In a preferred embodiment of the present invention, the network ' device (client) transmits on pairs 0,1 and the hub transmits on pairs 2 and 3.
Three frequencies, .975 MHz, 1.8~ MHz and 3.7~ MHz are used. Table 4 gives assigned control signals for the preferred embodiment.
The transceivers within hub 30 and network interface 41 generate and measure frequency of the tones. The acknowledge from the hub to a network device is not a tone frequency pair. Rather it ;s the event of the transition '1 .:: ~
from the hub driving a tone to the hub driving no signal. ;
` Figure 14 is a simplified timing diagram which illustrates a :.j transaction on four pairs. For the example transaction, two frequencies, 20 e.g. one megahertz and two megahertz, are used. The network device ; ~ (client) transmits on pairs 0,1 and the hub transmits on pairs 2 and 3.
. Table 5 below gi~es the assigned control signals for a preferred embodiment.
;~.; , Table 5 '~'' ' 25 :~:
~; Frequency of Tone Oscillation Signaled Control Signal ~ ~ First Pair (Qor 2) Second Pair (1 or ~ Client Hub : ~ 1 MHz 1 MHz Req 0 Busy .~ 1 MHz 2 MHz T-ldle T-ldle 2 MHz 1 MHz Idle Incoming ` ~ 2 MHz 2 MHz Req O Idle .... .~
Hewlett-Packard Company HP109265~
~. .

8~

The transceivers within hub 30 and network interface 41 generate and measure frequency of the tones. The acknowledge from the hub to a network - device in not a tone frequency pair. Rather it is the event of the transition from the hub driving an IDLE to the hub driving no signal.
In Figure 14, signal waveform 310 represents a signal on a first ''''! twisted wire pair between network interface 41 and hub 30. A signal waveform 311 represents a signal on a second twisted wire pair between network interface 41 and hub 30. A signal waveform 312 represents a signal on a third twisted w~re pair between network interface 41 and hub 30. A
signal waveform 313 represents a signal on a fourth twisted wire pair between network interface 41 and hub 30.
In a time period 315, network interface 41 is driving an idle signal on :
the first and second twisted wire pairs. Likewise, hub 30 is driving an idle ~i signal on the third and fourth twisted vvire pairs.
; 15 In a time period 316, network interface 41 is driving a request signal ~ !
on the first and second twisted wire pairs. Hub 30 continues driving an idle signal on the third and fourth twisted wire pairs.
In a time period 317, hub 30 acknowledges t}le requested by allowing .~ signals on the third and fourth twisted wire pairs to float to a middle voltage.
~, 20 In a time period 318, network interface 41 transmits a data packet on "~3 all four twisted wire pairs.
In a time period 319, the end of packet has been reached. Network interface 41 stops driving the third and fourth twisted wire pairs. Network 41 starts driving an idle signal on the first and second twisted wire pairs.
In a time period 320, network interface 41 continues driving an idle signal on the first and second twisted wire pairs. Hub 30 begins driving an ~ ~
idle signal on the third and fourth twisted wire pairs. ~ ~ ;

- s ~ Hewlett-Packard Company HP1092~-1 :
.. :.: .
.

While this example has been for transmission of a single packet, multiple packets may also be transmitted after a single arbitration.
Key constraints of the above described system include the use in a network of four twisted wire pairs to attach each network node to the hub.
During data transmission or reception, the direction of data flow on all four twisted wire pairs is in a single direction. It is thus not possible to reliablydetermine whether two nodes are transmitting simultaneously because there is no way to signal a transmitting node since it is not receiving at that , time. In the hardware configllration, twisted wire pairs from several ~, clients can be combined into one cable b~mdle. Due to near end crosstalk, during data reception, the hub is not allowed to transmit the data packet to more than one port. However, the hub may send data packets to multiple :, ~ ports while the hub is not receiving data packets. Additionally, when the - ~ hub is receiving data from one of the network nodes, the hub exchanges 15 control signals with other network nodes. The control signals are tones Y which are at firequencies well below the data rate.
` l In order to facilitate the operation of the system over a broad range of , possible cables, a period of characterizing of the cable is performed before '~ transrnission of user data. This is the training periods described above.
, '"J 20 In the preferred embodiment, no existing network protocol is used to ~` arbitrate network usage for this topology. Rather, as descIibed above, a port ;~ of the hub is in one of three states at any particular point in time. The first --~ state is where the port is transmitting a packet (four twisted wire pairs t driven by client). The second state is where the port is receiving a packet 25 (four twisted wire pairs driven by hub. The third state is during arbitration ;: ~
::s~ for a link (two tw~sted wire pairs driven by client, 2 twisted wire pairs driven ~i , ~`t' the hub). At any one point in time, different ports of the hub can be in ~ ., ,, ~ ',,~1 Hewlett-Packard Company HP10926~5-1 ~: ' - 3" 21~18~Q
different modes, e.g., one port transmitting, one port receiving and the rest arbitrating for the next cycle.
.~ During arbitration, pairs of low frequency tones are sent by $he hub and client. These allow the hub and client to determine who gets to transmit 5 next. In addition, other control in~ormation may be sent.
During the training sequence the client notifies the hub of its network address. Also, network protocol errors retrigger the training sequence.
To support applications which require low latency and guaranteed ."
. network bandwidth availability, two priority levels of client data are 10 supported. These two priority levels are preserved through the bridge to the ": backbone network.
To ~void packet loss through the bridge, a busy signal which indicates the buffer memory is full is sent to the client that has transmitted a packet which could not be stored. This signal is held until space is available in the :.
:~1 15 bridge buffer. The advantage is that the packet can be retransmitted by the ;'iJ
client hardware without depending on a software protocol timeout to !''.'~ retransmit.
~ To address the limitation of transmitting to only one client, the , -following method is used during reception of a packet. During reception of a ~0 packet, the repeater identifies the destination client before transmitting the packet. The data is transmitted to that port only. This has the added benefit of providing protection against an eavesdropping node. The hub does not - fully receive the packet before retransmitting it. In the event the packet is intended for multiple destinations, the packet is buf~ered in the repeater and 25 then retransmitted once it has been fully received.
During network operation, the hub checks all ports for requests. The ~`i, following priority is used. Highest priority is granted high pr~ority `` messages from the backbone. The next highest priority is granted high '.' '' ~
Hewlett-Packard Corrlpany HP1092666-1 ~ . . : , , :' priority local messages. Then priority is granted to data priority messages from the backbone. Lowest p~iority is granted to data priority messages from the local network. When there are multiple clients requesting at the : same priority lev~l, they are satisfied in a round robin order.
The advantages of the above described embodiment of the present invention include good support for bridging, multiple priority levels and a predictable arbitration method under heavy loads.
Various preferred enobodiments of the present invention can be adapted for use with various protocols. For example, preferred - 10 embodiments may be adapted to run similar to the IEEE 802.3 protocol. In ~..
one such embodiment, at the start of a packet, a client transmits on pairs 1 .. and 2. The hub repeats the data onto pairs 3 and 4 to the other clients. The transmitting client monitors pairs 3 and 4 for activity and the hub monitors pairs 1 and 2 ~or activity. Once the 802.3 arbitration has completed without a -collision, i.e. the slot time has passed, the client is able to transmit on all 4 :
pairs. The 802.3 arbitration state machine can be used in a form to the version described above.
;;:, In an alternate embodiment, the clierlt7 at the start of a packet transmits on a first set of twisted wire pairs 192,3 and the hub repeats on a ,........................................................................... .
;~ ao second set of twisted wire pairs 2,3,4. The client monitors pair 4 for activity ; ;i and the hub monitors pair 1 for activity. A~er the arbitration is complete, ~. the client can transmit data on all four twisted wire pairs.
~ , ~
- Alternately, after the arbitration is complete the client can continue to transmit on only three pairs; however, in order to maintain a 100 megabit transmit throughput over the network, the transmission rate through each twisted wire pair would need to be correspondingly increased. For example, in order to transmit at 100 megabits on three lines with a 5B/6B two level .

;:
Hewlett-Packard Company ~IP 1092666-1 ."

';
` 32 æ~s~
code would require 40 megabauds per twisted pair, i.e., a maximum bandwidth per tw~sted pair of approximately 25 -30 MHz.
In networks where bundles of twisted w~re pairs are used, during .
arbitration a low frequency preamble is sent because data frequencies would generate too much crosstalk. If bundles are not used, the packet is transmitted during the arbitral~on at half or three quarters the final data rate, depending on whether 2 or 3 pairs are available.
: In order to implement training in such an embodiment, a methodsimilar to that described in the above described training state machines is used. In such an embodiment, for e~ample, idle signals are sent on the cable to indicate whether a port has been trained or not. Until the training .,;
:.~ is complete, the port is not allowed to send regular packets.
The above-described embodiment of the invention has utilized a ;i protocol in which for control signals tones are transmitted in full duplex at low frequency relative to data signals which are transmitted in half duplex.
~j However, alternate embodiments of the pres~nt invention allow for ~`~ additional adaptations to existing protocols. For example, forprotocols which require collision detection, such as IEEE 802.3 protocol, various alternate embodiments may be implemented in accordance w~th the present invention. ;~ -For example out of band signaling may be used. Collision ~6~ ~ information is propagated back to the end node with a frequency that can be filtered out from the data stream. A low or high frequency could be used.
Even a DC signal could be sent back on one pair to indicate collision. This would allow the data packet t be sent on all 4 pairs immediately, thus . .
- increasing network effilciency.
` Figure 15 gives one block diagram of an implementation which allows collision detection by sending collision information over a different i:
.. -~ , .

` Hewlett-Packard Company HP10926~-1 . ....................................................................... .

~",,,.",~"",,~

33 '~ 8 ~ 6~
frequency than data. A data signal generator 331 operating at a first frequency range is connected in parallel to a collision signal generator 332 - and a resistance 333. Collision signal generator 332 generates a collision signal at a frequency different than the first ~equency range used for data.
5 Data signals and collision signals are transmitted through transformer 334, over a twisted wire pair 335 and through a transformer 337. In parallel with a resistance 337 a filter 338 ~or data frequencies forwards the data signals ' using an amplifier 340. Likewise, a filter 339 for the collision ~equency 339 .. forwards the collision signal using an amplifier 341.
Figure 16 and Figure 17 show potential frequency spectrums for signals sent across twisted pair 33i5. Data BigIlalS are in a riange between a ; first frequency f 1 and a second frequency f~. Collision signals are sent a frequency f3. Figure 16, illustrates the case where the spectrum for data frequencies 346 are at lower frequencies thian the spectrum for collision 15 frequency 347. Figure 17, illustrates the case where the spectrum for data frequencies 346 are at higher frequencies than the spectrum for collision i~; frequency 347. ~ ;
Alternately, common-Mode signaling may be used. In this case, the twisted wire pairs carry the data stream with differential mode signaling. -20 Again, all 4 twisted wire pairs send data immediately, and collision -signaling is sent back at a common mode signal on one pair. Alternately, ;., .
`~ radio frequency interference (RFI) is minimized by sending a common mode :,~
AC signal on two twisted wire pairs.simultaneously. The AC signal is 180 ~ degrees out of phase on each pair to cancel the electromagnetic fields i` 25 created by a single transmitter.
. Figure 18 shows a block diagram of an implementation which , ~.
, provides common-mode collision signaling. A data signal generator 360 ."; transmits data signals through transformer 361, over a twisted wire pair , ,.~.
, ., ~`. Hewlett-Packard Company ~IP1092656-1 '.-~

34 21~18~0 ~- 363 and through a transfiormer 365. A fil$er 366 forwards the data signals using an amplifier 367. Likewise, a data signal generator 370 transmits data signals through transformer 371, over a tw~sted wire pair 373 and through a transformer 375. A filter 376 forwards the data signals using an amplifier 377. A collision generator 382 is connected to a transmission resistance 364 and a transmission resistance 374. In response to an enable ~ signal on a line 383, collision generator 382 generates a differential signal : ! through twisted wire pair 363 and twisted wire pair 373. A collision detector consisting of an amplifier 381 and a resistance 380 is coupled between a ; 10 reception resistance 362 and a reception resistance 372. The collision detector detects and forwards a collision signal generated by collision generator 382.
In an alternate embodiment, in band signaling can be used. In this embodiment, collision information is driven to a transmitting node with an 15 in ~and frequency signal. The receiving node has a hybrid transformer that allows echo cancellation of the outgoing data stream. The network node is thus able to verify a received collision signal in addition to the data being ~i sent. Also, active circuits which provide echo cancellation may be used ~, which allows half duplex signaling on a four twisted wire pairs.
Figure 19 shows a block diagram of an implementation which provides for in-barld collision signaling. A transmit amplifier 390 transmits - data signals through a transformer 391 over a twisted wire pair 393 and , . .
through a transformer 395 to an amplifier 397. Likewise, a transmit ~; amplifier 398 transmits data signals through transformer 35~ over twisted .
26 wire pair 393 and through transformer 391 to an amplifier 399. A hybrid transformer 92 and a hybrid transformer 394 serve to cancel energy from ` ` being received by a nodes own transmitter. However, hybrid tr~nsformer ~`1; 392 and hybrid transformer 394 will not block the incoming reception from ,: ..
,i .
Hewlett-Packard Company HP1092656-1 ., ,,;,, :,: . , , . , , ... ;; . . ~ .

~5 21~18~0 another node. If in either case, a node is transmitting and receiving data at the same time, this indicates a collision has occurred.
; Figure 20 shows how the implementation shown in Figure 15, or the implementation shown in Figure 18 could be used in a network where a 5 network device 419 sends data to a network device 420 over four twisted wire .. . .
pairs 411, 412, 413 and 414. Within network device 419, a trans~ttmg amplifier 401 sends data over twisted wire pair 411, a transm~tting amplifier 403 sends data over twisted wire pair 412, a transmitting amplifier 406 sends data over twisted wire pair 413 and a transmitting amplifier 407 sends data 10 over tw~sted wire pair 414. Within network device 420, a receiving amplifier 402 receives data over twisted wire pair 411, a receiving amplifier 404 receives data over twisted wire pair 412, a receiving amplifier 406 ~' receives data over twisted wire pair 414 and a receiving amplifier 408 receives data over twisted wire pair 415.
Collision detection is accomplished, for exarnple, using a separate :~
~requency as in the implementation shown in Figure 15 or Figure 18.
Collision detection circuitry 415 within network device 419 detects collisions by listening for a collision signal sent on twisted wire pair 413 and/or twisted "
-~ wire pair 414. Collision detection circuitry 4:l6 within network device 420 ~0 detects collisions by listening for a collision signal sent on twisted wire pair -'~ 411 and/or twisted wire pair 412.
. . I .
In this embodiment, when network device 419 desires control of the .. network, network device 419 begins transmission on all of twisted wire pairs ., .
';i 411, 412, 413 and 414. In addition, network device 19 sends a collision signal 25 on one or both of twisted wire pairs 411 and 412. Collision detection circuitry 415 then listens for a collision signal on twisted wire pair 413 and 414. When ~, a transmit enable signal 417 and a collision detection signal from collision ,.
, j .~r :: Hewlett-Packard Company HP1092656-1 ', ~
, ~ 6 detection circuitry 415 are both activated, a logical AND gate 418 signals a collision.
In another alternate embodiment, time multiplexing is used. In this embodiment the network nodes transmit on all four twisted wire pairs.
;"
5 After making an initial transmission, transmission ceases or a low frequency tone is sent during a collision window period. The collision . window is used by the repeater to signal that a collision on the network has occurred. The original transmitting node would continue with packet ., transmission if no collision signal is returned during the collision window.
10 Otherwise, the network node will back of~, for example, in accordance with the IEEE 802.3 backoff algorithm. The use of a low frequency tone (or single ~., .
tone) allows the collision signal to be sent back as a different tone. This allows a simple frequency detection circuit to be used to detect the collision tone.
Figure 21 shows a hub 430 connected to a node 431 through four twisted wire pairs 433. Hub 430 is connected to a node 432 through four twisted wire pairs 434. In a time multiplexed hub based collision detection ... .
scheme, each node desiring to send information sends first tones during a ., ~, collision interval. Hub 30 listens for the tones. If tones from more than one ~; ~ 20 node is heard, hub 30 sends to all nodes a second tone indicating a collision ,~,i has been detected.
Figure 22 shows an example of signal timing packages used in a time-multiplexing scheme. Signal line 425 represents potential signals sent ;, ~ .
- by node 431. Signal line 426 represents potential signals sent by hub 30. In a ;; 25 time period 427, node 431 finishes a sending a last data packet over four . .~
twisted wire pairs 433. In a collision detection period 428, node 431 and any other nodes which desire to send data send the first tone to hub 430. If hub .
''''' ;
Hewlett-Packard Company HP1092655-1 ~ .;
., ~.

2~ sga 30 detects a collision, hub 30 sends the second tone. Othe~wise in a time period 429, node 431 can begin transmission of a new network packet.
In another alternate embodiment of the present invention, a collision signal is sent aflGer the data packet. A modification to the IEEE 802 3 protocolin accordance with this embodiment allows half-duplex operation on all 4 twisted wire pairs immediately for each data transmission. When transmitting a packe$, a network node transmits a complete packet using all four twisted pairs. At the end of the packet, a collision window is opened by all nodes, allowing the repeater (Hub) to send a collision signal back to the original transmitting nodes. A network with low collision counts can ha~re a significant increase in throughput efficiency by allowing all four pairs to transmit at the beginning of the data packet, as allowed by this embodiment.
Figure 23 shows signal timing packages for this altelnate scheme.
Signal line 435 represents potential signals sent by node 431. Signal line 436 represents potential signals sent by hub 30. When there is a collision, a collision indicator (e.g., a tone) is sent to the nodes which collided. All colliding nodes would be informed after the packets were complete. Each node would then back o ff per the algorithm of the network program, (e.g., the 802.3 protocol).
In the preferred embodiment, a tone preamble occurs during a collision slot time. In this embodiment, a single tone is sent throughout the collision window as the data packet preamble. The single tone allows a collision signal to be sent back at an in band tone of another frequency, allowing ease of collision detection.
. --26 Figure 24 illustrates the case where there is a collision window before ~,..
- each packet transmission. During the collision window, network nodes ' which desire to transmit data inform a hub 440 by sending first tones to hub 440. For example, a network node 441 sends to hub 440 the first tone over a .,.
, : .
~, Hewlett-Packard Company HP1092655-1 i:,..

2 1 ~

first set of twisted wire pairs 444. A network node 442 sends to hub 440 the first tone over a first set of twisted wire pairs 446. As soon as hub 440 receives the first tone from any network node will begin sending an incoming signal to all nodes. For example, hub 440 will send an incoming signal to node 441 over a second set of twisted wire pairs 443. Hub 440 will ~; s0nd an incoming signal to node 442 over a second set of twisted wire pairs 445. All nodes will then be allowed to make a request to send data for a time .li duration set by the protocol. Each node measures the time duration from the time the node receives the incoming signal from hub 440. Hub 440 will wait until all possible requests to transmit have been heard. Then, if there has been more than one request to transmit data, hub 440 will send the ,;, collision tone in place of the incoming tone. Otherwise, hub 440 will cease transmissions allowing the one node requesting data transmission to proceed with the transmission.
.,, Because of losses when transmitting over twisted wire pairs, low frequency tones have a higher voltage amplitude when received than do i'l higher frequency data signals. In order to take advantage of this, in one .
embodiment of the present inven1ion, the noise threshold (squelch) i~ set s higher when receiving control tones, and is set lower when reeeiving data æ signals.
` j For example Figure 25 shows a circuit within a receiving network device which can be used to adjust squelch for in~ormation received from a twisted pair. An incoming signal is placed on lines 451 and 452. A
", `~: comparator 454 compares voltage on lines 451 and 452 and is used supply j~ 25 data on a line 459. Additionally output from comparator 454 is received by start and end of data pattern search logic 457. Logic 457 also receives a receive enable signal (RXEN) on a line 461. A comparator 453 compares the signal on line 451 with a threshold voltage on a line 460. A comparator 45~ ;

~ t Hewlett-Packard Company HP1092655~
, ,:
'`''' ":: ~' 3 6~ ~

compares the signal on line 452 with the threshold voltage on a line 461.
Output from comparator 453 and comparator 455 are received by tone period squelch logic 456.
- When the receiving network device receives from another device a 5 tone which indicates data is to soon to be sent to the receiving network device, RXEN is asserted on line 461. At this point the receiving network -device listens for the signal voltage transitions detected by comparator 464.
i~, This allows for accurate detection of the lower voltage data signals. Whenlogic 457 detects an end of data pattern, the network listens for tone signals based on a squelch with the threshold voltage placed on lines 460 and 461.
A logic OR gate 458 generates on line 462 a link/data OK. signal which indicates when data or tones is being reliably received by the receiving ~A~ network device. Logic OR gate 458 receives an activity energy signal on a line 463 when tones with signal voltages greater than the threshold voltage are detected on lines 451 and 452. Logic OR gate 458 receives an ETD signal from Logic 457 when both RXEN on line 451 is asserted and data ~,' transmission is detected by comparator 454. Alternately, logic OR gate 458'~"! may be replaced with a multiplexor controllecl by RXEN on line 461.
`:f Figure 26 is a timing diagram which illustrates operation of the ,,, ~i 20 circuit shown in Figure 25. A waveform 473 represents voltage variation of -~- signals on line 4~1 and 4~2. Waveform 473 is shown in comparison with a zero or midpoint voltage 470, a positive threshold voltage 471 and a neg~tive `i threshold voltage 47~.
`~; Du~ng a period 475, a tone signal is received, as is represented by ;, 25 waveform 473. Because the tone signal has a voltage amplitude exceeds the positive and negative voltage threshold, the activity signal on line 463 is asserted, as represented by a waveform 478. Also during period 475, the link OK signal on line 462 is asserted, as represented by a wave~orm 481. Also at ! -.:.`
Hewlett-Packard Company HP1092655 -~:

i*?~ `'r ",'""' ' ;' ~, ' some time within period 475, RXEN is asserted, as illustrated by a waveform . 480.
During a period 476, a data signal is received, as is represented by waveform 473. When logic 457 detects the data signal, the ETD signal on 5 line 464 is asserted, as represented by a wavefolm 480.
During a period 477, the da$a transmission has ceased and RXEN is , de-asserted, using information encoded in the received data stream. THe higher threshold is therefore re-enabled. Noise on the line, as represen$ed -~ by w~veform 473, is ignored because it does not have a voltage amplitude $hat exceeds the positive and negative voltage threshold. Therefore, the link OK
signal on line 462 is de-asser~ed, as represented by waveform 481.
The foregoing discussion discloses and describes merely exemplary methods and ernbodiments of the present invention. As will be understood by those familiar with the art, the invention may be embodied in other .j . .
specific forms without departing from the spirit or essential characteristics thereof.
For example, the method described above can be modified in order to reduce the possibility of undetectable errors occurring owing to noise bursts ` affecting all four channels simultaneously a]nd thereby corrupting several ,.
aD successi~e 5B/6B symbols propagating in parallel through $he channels. In i~ this modification the 6B symbols on two channels are o~set in time by half ~.j the time for transmission of a symbol, relat*e to the symbols on the .; remaining two channels. As a result a noise burst af~ecting the channels for the duration of transmission of up to four bits can COITUpt at most six ~sl 25 consecutive 5B symbols (thirty consecutive bits). Such corruption can always be detected using a 32-bit CRC code as described herein.
.. . :

. ,.; .
. . .
. :

~ Hewlett-Packard Company HP10926~6-1 " , ~larious different 5B/6B block codes may be used in place of the block code given in Table 1 above. One possible alternate 5B/6B code is shown in Table 6 below.
- Table 6 ,i ~
~ 5-bit data blo~k 6-bit co~le value ternate 6-~c~de '.'i O 00000 000110 11100 '.' 1 00001 001110 . 2 00010 110010 . 10 3 0~011 000111L

6 00110 1~0001 011110 : 7 00111 011000 1û0111 ` 15 8 01000 110100 i' ~ 01001 010110 . 10 01010 000101 111010 .' 12 01100 11~001 13 01101 001~01 110110 . 14 01110 011010 `.:,', 1~ 01111 010101 i. 16 10000 01û100 101011 . 17 10001 100100 011011 . j 25 18 10010 100101 :,'.' 19 10011 101~10 ao 10100 001011 i, 21 10101 101~01 :'............. 22 10110 101~00 010111 .1 30 23 10111 001~)10 110101 ~ ~. 24 1100() 011~)01 .~"'. 25 11001 101100 ~',`, 35 2~ 11100 100010 101110 ~'ii 29 11101 0011~0 110011 i.'`~, 30 11110 001101 :~ I
.~ Accordingly, the disclosure of the present invention is intended to be ;~. 40 illustrat*e, but not lim~ting, of the scope of the invention, which is set forth ~ in the following claims.
, .
.-.,-,~ .~
.",.*j - :

~ Hewlett-Packard Company EP1092656-1 ~:
-, ?

Claims (29)

1. In a local network system in which a hub is connected to each of a plurality of network nodes, a method which provides for a first network node to transmit a data packet to the hub, the method comprising the steps of:
(a) exchanging control signals between the first network node and the hub, the exchange of control signals being done in a first signal frequency range; and, (b) sending a data packet from the first network node to the hub, the data packet being sent using data signals within a second signal frequency range, wherein the first signal frequency range and the second signal frequency range are different.

2. A method as in claim 1 wherein:
the first network node is connected to the hub using a plurality of twisted wire pairs;
in step (a) a first set of the first plurality of twisted wire pairs is used to send control signals from the first network node to the hub and a second set of the first plurality of twisted wire pairs is used to send control signalsfrom the hub to the first network node; and, in step (b) all twisted wire pairs in the first plurality of twisted wire pairs are used to send data packet from the first network node to the hub.

3. A method as in claim 2 wherein the first set of the first plurality of twisted wire pairs includes two twisted wire pairs and the second set of the first plurality of twisted wire pairs includes two twisted wire pairs.

4. A method as in claim 1 additionally comprising the step of:

(c) exchanging control signals between the hub and each of the plurality of network nodes excluding the first network node, the exchange of control signals being done in the first signal frequency range for the purpose of providing arbitration to determine which network node will transfer a next data package to the hub, wherein step (c) and step (b) are performed simultaneously.

5. A method as in claim 4 additionally comprising the step of:
(c) checking by the hub, a destination address for the data packet sent in step (b), wherein time of performance for step (b) and step (c) overlap;
and, (d) when the destination address is for a second network node in the plurality of network nodes, performing the following substep:
(d.1) sending the data packet from the hub to the second network node, the data packet being sent using data signals within the second signal frequency range, wherein time of performance for step (b) and step (d) overlap.

6. A method as in claim 5 additionally comprising the step of:
(e) when the destination address is for a first subset of at least two of the plurality of network nodes, performing the following substeps:
(e.1) storing the data packet until completion of step (b), and (e.2) upon completion of step (b) sending the data packet from the hub to the first subset of network nodes.

7. A method as in claim 5 wherein:
the first network node is connected to the hub using a plurality of twisted wire pairs;

in step (a) a first set of the first plurality of twisted wire pairs is used to send control signals from the first network node to the hub and a second set of the first plurality of twisted wire pairs is used to send control signalsfrom the hub to the first network node; and, in step (b) all twisted wire pairs in the first plurality of twisted wire pairs are used to send data packet from the first network node to the hub.

8. A method as in claim 7 wherein the first set of the first plurality of twisted wire pairs includes two twisted wire pairs and the second set of the first plurality of twisted wire pairs includes two twisted wire pairs.

9. A method as is in claim 1 wherein the first signal frequency range is lower than the second frequency range;
step (a) includes filtering by the hub of control signals received from the first network node using a first squelch level;
step (b) includes filtering by the hub of data signals received from the first network node using a second squelch level;
the first squelch level is higher than the second squelch level;
the squelch level is changed from the first squelch level to the second squelch level based on receipt of a particular control signal; and, the squelch level is changed from the second squelch level to the first squelch level based on receipt of particular control information within the data signals.

10. In a local network system in which a hub is connected to each of a plurality of network nodes, each network node being connected to the hub using a separate plurality of twisted wire pairs, each plurality of twisted wire pairs having a first set of twisted wire pairs and a second set of twisted wire pairs, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps:
(a.1) beginning transmission, by the first network node durlng an arbitration time slot, of a first data packet to the hub on the first set of the plurality of twisted wire pairs connecting the first network node to the hub, (a.2) listening, by the first network node during the arbitration time slot for activity over the second set of the plurality of twisted wire pairs connecting the first network node to the hub, and (a.3) upon the first network node detecting activity over the second set of the plurality of twisted wire pairs connecting the first network node to the hub during the arbitration time slot, ceasing transmission by the first network node; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) upon detecting more than one network node beginning to send a data packet to the hub, sending a collision signal to each of the network nodes on each second set of twisted wire pairs.

11. A method as in claim 10 wherein step (a) additionally includes the substep of:
(a.4) upon the first network node not detecting activity over the second set of the plurality of twisted wire pairs connecting the first network node to the hub during the arbitration time slot, upon completion of the arbitration time slot, continuing transmission, by the first network node of the first data packet on both the first set and the second set of the plurality of twisted wire pairs connecting the first network node to the hub.

12. A method as in claim 11 wherein for each plurality of twisted wire pairs, the first set of twisted wire pairs consists of two twisted wire pairs and the second set of twisted wire pairs consists of two twisted wire pairs.

13. A method as in claim 11 wherein for each plurality of twisted wire pairs, a total number of twisted wire pairs in the plurality of twisted wire pairs is four, the first set of twisted wire pairs consists of three twisted wire pairs and the second set of twisted wire pairs consists of three twisted wire pairs.

14. In a local network system in which a hub is connected to each of a plurality of network nodes, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps (a.1) beginning transmission, by the first network node during an arbitration time slot, of a first data packet to the hub, the transmission being in a first data frequency range, (a.2) listening, by the first network node during the arbitration time slot, for a collision signal sent from the hub to the network node, the collision signal, when sent, being sent in a second data frequency range wherein the first data frequency range and the second data frequency range do not overlay, and (a.3) upon the first network node receiving the collision signal from the hub during the arbitration time slot, ceasing transmission by the first network node; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) upon detecting more than one network node beginning to send a data packet to the hub, sending the collision signal to each of the network nodes.

15. A method as in claim 14 wherein each network node is connected to the hub using four twisted wire pairs.

16. In a local network system in which a hub is connected to each of a plurality of network nodes, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps (a.1) beginning transmission, by the first network node during an arbitration time slot, of a first data packet to the hub, (a.2) listening, by the first network node during the arbitration time slot, for a collision signal sent from the hub to the network node, the collision signal, when sent, being sent as a common mode signal, and (a.3) upon the first network node receiving the collision signal from the hub during the arbitration time slot, ceasing transmission by the first network node; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) upon detecting more than one network node beginning to send a data packet to the hub, sending the collision signal to each of the network nodes using common mode signaling.

17. A method as in claim 16 wherein each network node is connected to the hub using a plurality of twisted wire pairs and in substep (b.1) to each network node the hub sends common mode AC signal on two twisted wire pairs simultaneously and 180 degrees out of phase.

18. In a local network system in which a hub is connected to each of a plurality of network nodes, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps (a.1) beginning transmission, by the first network node during an arbitration time slot, of a first data packet to the hub, the transmission being in a first data frequency range, (a.2) listening, by the first network node during the arbitration time slot, for a collision signal sent from the hub to the network node, the collision signal, when sent, being sent in the first data frequency range, the listening including the substep of (a.2.1) echo cancellation of the transmission of data being performed in step (a.1), and (a.3) upon the first network node receiving the collision signal from the hub during the arbitration time slot, ceasing transmission by the first network node; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) upon detecting more than one network node beginning to send a data packet to the hub, sending the collision signal to each of the network nodes using the first data frequency range.

19. A method as in claim 18 wherein each network node is connected to the hub using four twisted wire pairs.

20. In a local network system in which a hub is connected to each of a plurality of network nodes, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps (a.1) beginning transmission, by the first network node during a first data transmission time slot, of a first data packet to the hub, (a.2) listening, by the first network node during a arbitration time slot following the first data transmission time slot, for a collision signal sent from the hub to the network node, and (a.3) upon the first network node receiving the collision signal from the hub during the arbitration time slot, ceasing transmission by the first network node; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) upon detecting more than one network node transmitting a data packet during the first data transmission time slot, sending the collision signal to each of the network nodes during the arbitration time slot.

21. A method as in claim 20 wherein step (a) additionally includes the substep of:
(a.4) upon the first network node not receiving the collision signal from the hub during the arbitration time slot, continuing transmission by the first network node during a second data transmission time slot.

22. A method as in claim 20 wherein the collision signal is a tone at a lower frequency than a frequency used for transmission of the first data packet.

23. In a local network system in which a hub is connected to each of a plurality of network nodes, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps (a.1) transmitting, by the first network node during a first data time slot, a first data packet to the hub, (a.2) listening, by the first network node during a arbitration time slot following the transmission of the first data packet, for a collision signal sent from the hub to the network node, and (a.3) upon the first network node receiving the collision signal from the hub during the arbitration time slot, retransmitting by the first network node of the first data packet at a later time; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) upon detecting more than one network node transmitting a data packet during the first data transmission time slot, sending the collision signal to each of the network nodes during the arbitration time slot.

24. A method as in claim 23 wherein the collision signal is a tone at a lower frequency than a frequency used for transmission of the first data packet.

25. In a local network system in which a hub is connected to each of a plurality of network nodes, a method comprising the steps of:
(a) when a first network node desires to send a data packet, performing by the first network node the following substeps (a.1) transmitting, by the first network node during a arbitration time slot, a first request tone within a first frequency range, (a.2) listening, by the first network node during the arbitration time slot, for a collision signal sent from the hub to the network node, (a.3) upon the first network node not receiving the collision signal from the hub during the arbitration time slot, transmitting a first data packet to the hub, wherein the first data packet is transferred by the first network node, and (a.4) upon the first network node receiving the collision signal from the hub during the arbitration time slot, backing off, by the first network node; and, (b) during the arbitration time slot, performing the following substep by the hub (b.1) when the hub detects more than one network node transmitting the first request tone, sending the collision signal to each of thenetwork nodes during the arbitration time slot.

26. A method for transmitting first data across a network, the method comprising the steps of:
(a) scrambling the first data to produce scrambled data;
(b) serializing and block coding the scrambled data to produce four serial data streams;
(c) transmitting the four serial data streams across the network;
(d) deserializing and block decoding the four serial data streams to recover the scrambled data; and, (e) descrambling the scrambled data to recover the first data.

27. A method as in claim 27 wherein in step (a) the scrambling is based on the following polynomial factors:
S[n] = 1 + S[n-9] + S[n-11].

28. A method as in claim 26 wherein in step (b) the block coding is performed using a 5B/6B code.

29. A method as in claim 28 wherein step (c) is performed using binary NRZ modulation