Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Ethernet expert required

Status
Not open for further replies.

zape

Full Member level 2
Joined
Jul 10, 2003
Messages
120
Helped
14
Reputation
28
Reaction score
5
Trophy points
1,298
Location
Spain
Activity points
1,044
I have an old design implementing an AUI interface. Recently we connected it to a 10BaseT ethernet Hub (without connecting the collision pair) and a PC to the hub. This was done without thinking to much and it worked. My question is why did it work? The transformers and levels are different and perhaps the AUI Tx to 10BaseT Rx may work, but in the other direction I do not find a valid explanation. Any idea?
 

Hi zape,

although I'm not an ethernet expert, I have a comment about this. The first ethernet interface was (I believe) the AUI, then Coax in at last the Twisted pair interface. The AUI and the coax are at the "signaling or protocol perspective" the same interfaces. However, the twisted pair has differences, like "link alive" pulses (I think every few hundred msec), full duplex capability, ... which the AUI does not have, but generally they are very alike. Both have transformers for isolation, the levels are (by my knowledge) the same, btw AUI doesnt have preemphasis, TP doesnt need the collision signaling.

I did an AUI interface circuit (using CPLDs) for wireless communication experiments and was thinking about upgrading it to the TP interface, since AUI is for quite a while obsolete, but 10Mbps TP still exists. The main diferences were the link alive pulses and the TPs preemphasis of the signals, which should be neglected for short cabling...
 

    zape

    Points: 2
    Helpful Answer Positive Rating
Hi rfmw,

first thanks for your answer. I have analyzed the physical levels for the following transceiver implementing AUI and TP (pls see attached file). If you consider the effect of the TP tx transfomer (1:1.414) the AUI receiver levels do not match (without considering the different load).

I am quite interested in the info about link alive pulses and preemphasis. Could you please provide me your info source?
 

Hi zape,

sorry for the late reply. Regarding the TP pre-emphasis and "link alive" pulses (100ns pulse every 16ms) look at datasheet of DP83902A at the National Semiconductor Web page under search or any other PHY for 10Mbps Twisted Pair ethernet or probably the best way - look at the IEEE 802.3 standard, which I believe is free to download at the IEEE standards page. There should be everything you'll need solving your problem.

Hope this helps,
rfmw

PS: regarding the AUI or TP TX levels that you've mentioned, I'm no expert, so perhaps someone else should explain the problem and solution.

EDIT: your document also explains the pre-emphasis and link pulses ;)
 

Hi rfmw,

I am now working in the 802.3. My question about the link alive, preemphasis was in the way of implementing them was mandatory or optional.
My current main doubt regards the use of the AUI collission lines in case of communication through a hub in a 3 actors network where the communication is full duplex.

Thks for your support
 

Hi zape,

I believe that link alive pulses are mandatory in the TP network (that is, there should be at least one pulse 100ns or packet every 16ms or less). Pre-emphasis is not, because it's funtion is to compensate for the higher frequency components - spectrum of the ethernet signal in the case of the long twisted pair cable, where higher spectrum attenuates more due to the TP cable loss. I guess that you don't need to implement the pre-emphasis and must the link alive pulses circuitry.

If you have two TP NICs (Network Interface Cards) and one AUI, connected to the hub without the collision pair, there should be a problem with the AUI device, because it wouldn't know when the collision happened on the network. Thats the difference of the AUI towards the TP - the AUI has to have an additional circuitry to tell it there is a collision on the network, where the TP interface does it by itself - it knows if the collision happened, if there appears the RX signal while it transmitts on the TX pair already.

But perhaps the AUI would back-off in the case of incoming signal on the RX pair, while it transmitts already on the TX pair (you should try this somehow, but according to the standard, the AUI is not that smart, so it needs a Collision pair).

Full duplex would not work with the AUI, since AUI could step over and not knowing that the collision happend, just as described above. I guess with the low traffic, where there are not many collisions, perhaps it could work nevertheless...


What is that you could do, is to isolate the AUI device to its own network, that is replace the hub with the switch, where everyone has its own communication link. The hub re-transmitts everything that it hears to everywhere, so the AUI has a higher chance creating the collision...



zape said:
Hi rfmw,

I am now working in the 802.3. My question about the link alive, preemphasis was in the way of implementing them was mandatory or optional.
My current main doubt regards the use of the AUI collission lines in case of communication through a hub in a 3 actors network where the communication is full duplex.

Thks for your support
 

    zape

    Points: 2
    Helpful Answer Positive Rating
Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top