Jam signal and collision detection

i figure that a jam signal would cause the receiving NIC to detect a bad CRC and reject the frame.

but i'm wondering about the purpose of the jam signal in relation to the transmitting NICs. Given 2 transmitting NICS A and B - does A's jam signal trigger the backoff in B's NIC. Or is B's backoff only triggered by its(B's) own collision detection?

one site says that a jam signal is required because although the collision will propagate, the collision loses energy(current/voltage?) and may not reach the transmitting NIC that is further from the collision. So the jam signal ensures that the transmitting NIC that is further away knows that a collision has occurred. and i imagine that implies that A's jam signal received by B will trigger backoff in B

Reply to
Loading thread data ...

If B detects the collsion himself then no he sends his own jam and goes into backoff. If B doesn't detect the collision then the jam sent by A will corrupt any frames and they will be discarded by B because of CRC error. B will not go into backoff unless he detects a collsion himself.

Yes, backoff is only initiated when a NIC detcts a collsion.

Reply to

If i'm right, a jam is used in half-duplex environments to notify the transmiter to holddown the output queue and retransmit the current packet a few miliseconds later on.

It is mostly found when a dual-speed hub is used with a 10Mbit host on one side and a 100Mbit on the other side that "pushes" its data to fast to handle.

======================================================= NEW on the Internet: cisco irc chat groep irc://irc.intersecting.net/cisco.certification =======================================================

Reply to

Yes, in a full duplex environment collision-detection is turned off. Without collisions there are no jam signals.

Jam signals were around long before 100M ethernet was developed.

Reply to

The station that was transmitting goes into backoff.

Reply to

its sort of a moot point - once the jam signal is on the wire, any device sending at that point should backoff and retry (and there is no limit - 3 or more devices could be involved in a collision).

No - the distance limits on a segment are chosen to make sure signals get transferred reliably - jam is just another kind of signal (although badly formed, so might decay a bit more quickly).

this is the correct bit - but the reason is more about timing skew.

imagine a max length fibre segment @ 10 Mbps, no repeaters (4 Km for 1/2 duplex) - signals take a finite time to move across that glass.

if a collision happens at 1 end of the signal, the jam is needed to extend the time for the event - as the jam is propagating back down the cable towards the far end. The jam is long enough that both ends "see" a collision, event though the overlap in packets is different due to propagation delay

So the jam signal ensures that the transmitting NIC that is


Reply to

The jam is sent to ensure that all devices on an Ethernet segment know that a collision has occurred and that errors detected by other devices receiving the signal do not mistake this as noise. When a collision occurs both transmitters will detect the collision. Receivers will only detect the collision by receiving the jam signal. Transmitters detect the collision by comparing what is being sent, by what is being received (this is why it is called a "transceiver" because it transmits and receives at the same time). If the received signal differs from the transmit, a collision has occurred, and it sends a jam to notify the receivers to drop the frame and mark it as a collision. The collision is what triggers the backoff timer on the transmitter, not the jam signal. Both transmitters will always detect the collision.


Reply to

Well, no. "transceiver" is a device that can receive and transmit, but not necessarily both at the same time.

formatting link

Reply to
Walter Roberson

Good definition Walter.

Scott, when ethernet is in half-duplex mode then it can only transmit OR receive not both at the same time.

Reply to

but then howcome it says * in the 802.3 spec

the spec seems to say that the jam signal elongates the duration of the collision for it to be detected by transmitting stations!

yet you and many others, seem to me to say that the collsiion is detected by the transmitters, without the jam. The jam is so the receivers reject the frame.

either way, a prob I noticed with all this, is that if transmitters immediately transmit a jam signal after detecting a collision, then wouldn't jam signals collide?

infact, according to one site, the jam signal 'unambiguously destroys the colliding transmissions'. So the stations better detect the collsion before that happens!

btw, do the standard CCNA Cisco books from ciscopress iron out these details?

  • (quoted from 802.3 spec - available free from IEEE) ""Transmit Media Access Management enforces the collision by transmitting a bit sequence called jam. ....

--This ensures that the duration of the collision is suf=EF=AC=81cient to be noticed by the other transmitting station(s) involved in the collision--. After the jam is sent, Transmit Media Access Management terminates the transmission and schedules another transmission attempt after a randomly selected time interval." "

Reply to

Kind of. What happens is if you have two stations spaced out to the max distance, first A transmitts, because of propagation delay it takes a certain amount of time to travel down the pipe. Now say station B who is at the max distance from A transmits a few microseconds after A., A collision will occur closer to B than A, since both A and B were transmitting, they both start to transmit the Jam signal. Now the signal has to travel from B back to A, After B sends out it's jam signal, after a few microseconds where A notices the collision, it as well will transmit the jam signal. B finishes its 48 bit jam signal, but A still hears this jam signal from B because of the propagation delay. When the B jam signal at A finally ends, A is still in the process of transmitting it's own jam signal.

All receivers here the jam signals from A and B, but only A and B go into backoff before attempting to retransmit.

This is why frame length is important. If frame length was shorter, more collisions would occur, as the stations would be transmitting before a signal propagates through the medium between the two stations separated the most.

By the transmitting stations, not the actually transmiter in the NIC. Remember this is a transceiver, it can transmit and receive, simultaneously in full duplex only.

Yes, and this can be done by either looking at the current, as if only one station is transmitting, the current level would be X, (I think I remember this as being in the neighborhood of 16 to 20 miliamps.) now if another station transmits, this current being seen by a receiver now increases to X+.(Which would probably be over 22 miliamps). Another way is for the transmitter to compare what it received to what it transmitted, during the transmission phase. If the CRC's match, probably no collision.

Well they will reject the frame as it would be garbled, the jam signal is notification that the collision happened.

At some point on the transmission media yes they would. To see how this works look at two stations separated the max distance of

2800 meters. Lets have A transmit first, now B transmits before the signal from A reaches B, somewhere closer to B the signals collide, since the collision occured closer to B, B detects the collision first, and sends it's jam signal. After B sends it's jam signal, A also dected the collision and it also sends it's Jam signal. After B sends its jam signal and stops transmitting, the entire jam signal has still not made it to A, so while A is transmitting it's jam signal, B's jam signal is still on the line, After B's jam signal is off the media, A is still transmitting it's 38 bit jam signal.

This is where frame length and max distance between stations come into play. When station A or B transmitts, there is a certain amount of time before either B or A receives this transmission.

CCNA does not go into this much detail, so the answer is no. Some older CCNA books may have this. As look at the CCNA book now compared to one that came out in the mid 90's, quite a bit of material has been dropped off the test.

Reply to

There is lots of stuff on this in comp.dcom.lans.ethernet (newsgroup recommended by charles - the author of the 'definitive ethenet guide').

A big authority on this is richard seifart on that newsgroup mentinoed before.

The transmitting NIC has to stop transmitting - it doesn't want the receiving station to read the frame. But if it stops transmitting, then the frame may be so so small (< 96 bits) that - on 500m coax, the collisino frame will degrade and be unreadable if the other transmtiting station is at the other end. (as seifart says regarding the jam "it was designed in the coax days.") So, the jam signal is

32-bits long. Station B that detected the collision first, will complete its preamble and SFD(if it hasn't already), then will transmit the jam signal as a continuation, so it transmtis a frame with a bad CRC that is long enough not to degrade. (this is a different kettle of fish to the much larger minimum frame size of 512 bits). That collision frame from B will reach station A, (it will reach A while A is still transmitting, because the 512 bit minimum frame size ensures that). A will then stop transmitting and emit the jam signal which will append onto its transmission. Suppose this reaches B !!! If B is still sending its jam, i.e. its transmitting, then it won't pay an attention to any more collisions while its sending its jam (to quote an earlier post fro 1998 "a station sends the fixed length jam once and ignores what is happening while it is transmitting it. ") . Suppose it reaches B and B isn't transmitting - so B like any other receiving frame isn't detecting collisions, and will read the bad CRC and reject the frame.

why a min frame size ? so that a transmitting station will always hear collision while its transmitting. Why is that important? because then it can resend the frame. If however, the transmitting frame were to complete tarnsmitting and then etect a collisionl it won't be storng the farme anymore, it doesn't archive old frames. It clears its memory once its transmitted - it assumes that tarnsmission is successful. once it has transmtitd it isn't listening for a collision.

The 96 bit thing is to last out twice the cable length while still emitting. its purpose is just to be long enough to reach the other end

- it can stop being emitted long before .

as mentioned. jam signals don't trigger backoff. collisions do, and jam signals are bits within a collision anyway. the jam signal just ensures that the collision is long enough (96 bits) to reach the other end of a

500m coax cable without degrading. They also give/maintain a bad CRC. (i say 'maintain' because i'm sure that a collided frame would have a bad CRC anyway). (i say 'give' because if the collisino is detected during the preamble and SFD then the premable and SFD is finished, and the jam signal would give a bad CRC). since calculation of the CRC begins immediately after the preamble and SFD. .

i'm dizzy now. i hope that's more-a-less correct!! do check other posts - tere are many on this topic in that newsgorup mentinoe earlier, specially check richard seifart's posts.

Reply to

its not about signal degrading in this co-ax - its about sending a long enough signal that it still looks like a "jam" locally, and on other segments after passing through up to 4 repeaters.

That collision

No (or not in the general case) - the signal has to propagate a lot further. at 10 Mbps the collision domain spans multiple repeaters, and several segments in parallel. or on a point to point fibre without repeaters - up to

4 Km.

The minimum length bit string is to make sure signals propagate end to end across the entire collision domain.

Reply to

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.