Skip to main content

some questions about CSMA/CD,hubs and full/half duplex

More
18 years 9 months ago #9366 by x1ng
i dont know if that is that is the loop back that i wanna know or not..

it came in a section in cisco press book under "how 10baseT HUBS Works"

1- the NIC sends a frame

2- the NIC loops the sent frame onto its recive pair internaly on the card

3- the hub recives the frame

4- the hub's internal wiring propagates the signal to all other ports,but not back to the port that the signal was recived upon

5- the hub repeats the signal to each recive pair to all devices

so the NIC always Loop back the packet to its recive pair ..is that to prevent the hub from sending him back the packet that he sent to save throhput or efficency?!!!!!!!
i realy dont know why the NIC loops back.
More
18 years 9 months ago #9367 by nske
ah sorry, I was talking about a different thing then. I am as puzzled as you about that, but I trust someone else here can enlighten us :idea:
More
18 years 9 months ago #9434 by TheBishop
Replied by TheBishop on topic Loopback
Phew! Had to dig out my really old Ethernet notes for this one... I think this is the answer:-
The hardware loopback within the NIC itself is essential for CSMA/CD to function. What happens is that when the NIC begins to transmit, it also has to listen to the wire at the same time in case a collision happens during the transmission. So the explanation that the transmitted frame is looped back to the reciever is a bit misleading if I understand it properly. What is really going on becomes easier to understand if you think of the NIC as containing a transmitter and a receiver connected to the ethernet wire along with some logic that controls them. Then you can see that the thing can, indeed must, be able to listen to the wire while it is transmitting.
The collission detection works rather cleverly. The actual signal on the ethernet cable is an a.c. waveform (the data) superimposed on a d.c. offset (the carrier). So all the card needs to do is to watch the d.c. offset voltage on the wire. If it sits around the normal level, all is well. If two stations transmit at the same time (collision) the offset goes up to around twice the normal value which is easily detected
More
18 years 9 months ago #9440 by Chris
Bishop's just returned from his holidays and is nailing questions on the forums once again :)

I was aware that network cards also listen while transmitting data on the network - this is as you mentioned a CSMA/CD function, but I didn't know this function is made possible with the use of the hardware (nic) loopback!

Are you sure about this ?

Chris Partsenidis.
Founder & Editor-in-Chief
www.Firewall.cx
More
18 years 9 months ago #9446 by TheBishop
Replied by TheBishop on topic Loopback
Yeah, I'm getting all this out of my old BICC Ethernet Theory and Implementation notes from 1989!
Actually there are two things in danger of getting mixed up here. Firstly, the CSMA/CD collision detection does work by the NIC transmitting and recieving at the same time exactly as I've described. However whether you choose to call that mechanism a 'hardware loopback' is a matter of preference. Personally I'd prefer not to as I think it confuses the issue.
There is also something else I spotted when I read through the notes again more deeply. And it makes more sense when you remember that back in the old days you used to have a separate ethernet transciever (the RF bit) and an MAU (media access unit, the logic bit) connected vie a fat AUI (attachment unit interface) cable. Nowadays we just have a NIC and that's it. Imagine the 'NIC' sending a frame. The frame is assembled by the MAC layer and passed for transmission, but the MAC layer has no way of knowing whether or not the frame got recieved by the other station. But what it does need to know is whether it was transmitted successfully or not (so it can report back to the higher layers that invoked it). To do this, the outgoing frame is copied back into the reciever so the MAC layer can see that it went out to the physical transciever okay. And as added confidence the MAC layer has the fact that the transciever did not raise the collision signal so the frame made it onto the wire without interruption. Now i'm a little confused by the notes as to exactly where this internal loopback takes place - the notes seem to suggest it happens within the MAC layer itself. Suffice it to say, though, that there is one, and that's why
Time to create page: 0.147 seconds