Ethernet Signal Specifications

[PixieDust]

Hi, I'm trying to find information on Ethernet signals. Mostly interested in rise time to figure out what bandwidth would be ideal to see the signals on an oscilloscope. I'm researching 10BASE-T as a starting case, but I would like to figure things out for any bandwidth. Does such information exist? Every search I've plugged into good 'ol Google has come up empty on this. Surely there must be a specification somewhere? I'm looking at twisted pair.

[Berni]

10BASE-T is so slow that any modern digital scope should see it fine.

Things get trickier when it gets faster Ethernet, but still not too crazy. 100BASE-T is specified for working over 100Mhz bandwidth CAT5 cables so technically a 100MHz scope should see it fine. When it comes to 1Gbit it needs 500MHz of bandwidth. Special coding used by both of those to keep the bandwidth requirement lower.

But if you want to certify it as being compliant you need a scope that can see much more bandwidth than it operates at so that you can see weird oddities out of the spec, so that takes the big expensive above gigahertz boat anchors.

If you just want to snoop data on a ethernet link you don't actually want to look at the raw signal. The encoding makes it look like a complete mess at 100Mbit and 1Gbit. The better way to do it is to get a Ethernet PHY chip and connecting that to the line. That chip will do clock recovery, synchronization, decoding etc. then spit out the raw data out of the other end on a serial or parallel bus that then can be fed directly into a logic analyzer to see the raw bits going over the wire.

[jonpaul]

https://en.wikipedia.org/wiki/Ethernet

https://en.wikipedia.org/wiki/Ethernet_physical_layer

http://www.cablek.com/technical-reference/cat-5---5e--6--6a---7--8-standards

RS-485 is related but different,

https://blog.robotiq.com/what-is-rs485-communication-protocol

Jon

[PixieDust]

10BASE-T is so slow that any modern digital scope should see it fine.

Yes, I have seen what it looks like on say a 100MHz scope. But I can't help but wonder if 100MHz is enough for a good representation of the signal. Looks too sinusoidal.

Things get trickier when it gets faster Ethernet, but still not too crazy. 100BASE-T is specified for working over 100Mhz bandwidth CAT5 cables so technically a 100MHz scope should see it fine. When it comes to 1Gbit it needs 500MHz of bandwidth. Special coding used by both of those to keep the bandwidth requirement lower.

But if you want to certify it as being compliant you need a scope that can see much more bandwidth than it operates at so that you can see weird oddities out of the spec, so that takes the big expensive above gigahertz boat anchors.

If you just want to snoop data on a ethernet link you don't actually want to look at the raw signal. The encoding makes it look like a complete mess at 100Mbit and 1Gbit. The better way to do it is to get a Ethernet PHY chip and connecting that to the line. That chip will do clock recovery, synchronization, decoding etc. then spit out the raw data out of the other end on a serial or parallel bus that then can be fed directly into a logic analyzer to see the raw bits going over the wire.

That's what I'm trying to figure out, how do I calculate the required bandwidth if I'm looking for signal integrity? At the moment I'm just curious, but even for 10BASE-T I might need more precise measurements in the not too distant future.

https://en.wikipedia.org/wiki/Ethernet

https://en.wikipedia.org/wiki/Ethernet_physical_layer

Jon

Yep I've seen those, wiki doesn't address my query. I searched for all sorts of terms - standards, electrical properties, specifications, rise time etc etc. Found the original Xerox specification: https://ethernethistory.typepad.com/papers/ethernetspec.pdf Nothing that talks about signal properties.

[Berni]

Yes, I have seen what it looks like on say a 100MHz scope. But I can't help but wonder if 100MHz is enough for a good representation of the signal. Looks too sinusoidal.

BASE10 uses manshester encoding. This already tells you that the minimum occupied bandwidth of the signal is 2x the bitrate. So 10Mbit will need 20MHz.

You do want signals on long cables to be rounded rather than sharp because this reduces the problem of signal reflections on short cables and long cables will eventually round off a signal anyway. As long as you can clearly determine a 1 and 0 it is fine enough. If you wanted a razor sharp square wave signal then a 10Mbit digital signal would need a bandwidth of >10GHz, something that long cable runs are not capable of.

That's what I'm trying to figure out, how do I calculate the required bandwidth if I'm looking for signal integrity? At the moment I'm just curious, but even for 10BASE-T I might need more precise measurements in the not too distant future.

You can do a FFT to see the spectrum of the signal.

The whole thing gets much more complicated at 100Mbit and above because simple Manchester encoding is no longer used. The signal is not even digital anymore, it has 3 levels instead of 2 binary ones. All this trickery is the reason that 100Mbit Ethernet can fit inside of 100MHz of bandwidth (It actually uses only 1/3 of the bandwidth, rest is overhead). Gigabit uses similar tricks, except doubles the number of pairs hence why it can fit in 500MHz of bandwidth. Gigabit on top of that does extra tricks like preemphesis, so it will pre-distort the signal in a special way so it comes out looking nicer out the other end.

Once you go to 10Gbit they still fit it into the 500MHz bandwidth, but uses even more sophisticated encoding like PAM16 to squeeze more information into the signal at the expense of making it less resilient to noise. Trying to look at that with a scope will just look like random junk all over the screen.

All modern RF links like Wifi, DSL internet, DVB TV...etc uses various fancy modulation and encoding techniques to squeeze more data into the available bandwidth. The QAM modulation is popular there.

[PixieDust]

Thanks Berni, very useful info! Clears up a few points. And some new leads for me to chase up!