# How is a Twisted Pair's +-2.5v used to send a 1 or a 0?

This is a multi-part question that I'm asking as a follow on to this answer from my other question.

I know 100BASE-TX transmits through a pair of wires, that are twisted around each other so that the effective of ambient EMI affects them both identically (for the most part). I know that the signal is sent through one wire (TX+), and the exact opposite signal is sent through the other wire (TX-).

1. If the 100BASE-TX runs +-2.5V, does that mean the TX+ wire uses 0 through +2.5V, and the TX- wire uses -2.5V through 0?

I also know that "sending a signal" is effectively "applying a voltage" to the wire. And that a certain level of voltage correlates to a 1 or a 0.

1. What is the voltage range that equates to 1, and what is the voltage range that equates to 0?
2. Who and how is this range set? Is it a standard, or dynamically learned based on percentages of what either end of the wire is able to push through the wire (aka, different each physical connection)?

EMI is unavoidable, but due to the twisting of the pairs, the effect on the TX+ and TX- wires are identical, and as a result the noise added by EMI can be extracted on the receiving end.

1. How is the noise removed?

I am assuming if the wires send +1 and -1 respectively, and the EMI noise (for simplicity) adds +3 to both, the receiving side would receive +4 and +2. If you subtract those values from each other, you get a difference of 2, and if you put the 2 on either side of a 0 on a number graph, you end up with +1 and -1 again. Is that about it for the process?

Lastly, if noise is nothing more then adding or subtracting from the voltage on the wire, to what degree can noise affect the voltage.

1. If 100BASE-TX uses +-2.5V, in a typical office environment, about how much mV of noise can you expect?

I'm mostly curious to see how much voltage is added/subtracted by noise, proportionate to the sending signal. AKA, if typical noise might apply +-50mV, that wouldn't be all that much compared to a sending signal of 1000mV (1V). Not looking for a technical deep dive on this, just an idea so I have a basic understanding of the numbers / number ranges.

• Due to the magnetics isolating the lines, there really is no absolute voltage that specifying made sense – PlasmaHH Nov 24 '15 at 18:05

If the 100BASE-TX runs +-2.5V, does that mean the TX+ wire uses 0 through +2.5V, and the TX- wire uses -2.5V through 0?

Not usually.

What is the voltage range that equates to 1, and what is the voltage range that equates to 0?

Usually, you'd transmit a 0 with, for example TX+ = -2.5 V and TX- = +2.5 V. And you'd transmit a 1 with TX+ = +2.5V and TX- = -2.5V.

While, as Andy says, in theory it isn't important if the two signal lines have different bias points, in practice it's easiest to make a receiver that compares the two lines to each other. So that any situation where RX+ > RX- is interpreted as a one and any situation where RX+ < RX- is received as a zero.

Who and how is this range set?

As I said above, any situation where RX+ > RX- is interpreted as a one and any situation where RX+ < RX- is received as a zero.

The specifications for the allowed common mode voltage and signal amplitude are usually chosen based on the available technology (semiconductor process and known circuit designs) so that all users can design their circuits to be interoperable. But basically this is an arbitrary choice made for different standards (LVDS, Ethernet, etc.) based on the technology expected to be used to implement them.

How is the noise removed?

Noise is removed because the receiver is designed to be mostly sensitive to only the difference between the input voltages, and have as little as possible response to the common mode voltage of the inputs.

If 100BASE-TX uses +-2.5V, in a typical office environment, about how much mV of noise can you expect?

You can see how much noise and interference the system designers allowed for by looking at the minimum signal amplitude allowed at the receiver. To give very low error rates, the maximum noise allowed is probably 1/15 or less of the signal amplitude.

PlasmaHH's comment has a critical point in it: Ethernet in particular is not directly connected to the transceiver but has a small transformer at each end. So voltages aren't referenced to the system ground of either end (which may be different!) but relative to each other.

But when you have a really detailed question, it's always best to go to the source if possible, and the IEEE802.3 specs are available for free (registration required): http://standards.ieee.org/getieee802/download/802.3-2012_section1.pdf

Turn to "14.3.1.2.1 Differential output voltage":

The peak differential voltage on the TD circuit when terminated with a 100 Ω resistive load shall be between 2.2 V and 2.8 V for all data sequences for a type 10BASE-T MAU that is not a type 10BASE-Te MAU. For a type 10BASE-Te MAU, the peak differential voltage on the TD circuit when terminated with a 100 Ω resistive load shall be between 1.54 V and 1.96 V for all data sequences.

Later section 14.3.1.3.1 describes a receiver specification with a threshold of 585mV (and a particular pulse shape). It also says to reject

All signals that when measured at the output of the following filter would produce a peak magnitude less than 300 mV. The filter is a 3-pole low-pass Butterworth with a 3 dB cutoff at 15 MHz (refer to B.4.2).

Even later 14.4.4.1 cable specification "Impulse Noise":

The average rate of occurrence of impulses greater than 264 mV shall be less than or equal to 0.2/s as measured at the output of the following specified filter. Following the start of any particular impulse that is counted, any additional impulse shall be ignored for a period of 1 μs. The simplex link segment shall be terminated at the far end in 100 Ω. The filter is a 3-pole Butterworth low-pass with a 3 dB cutoff at 15 MHz (refer to B.4.2).

Further documents are at http://standards.ieee.org/about/get/802/802.3.html , presumably including those for 100BASE-T.

You need to do some googling - there are tons of info on different differential signalling schemes from RS485, thru LVDS and PECL etc.. There is no one standard and I feel you should acquaint yourself with what is out there to get a better feel.

You can have a signalling system that uses +/-1V differentially but if both signals are superimposed on a dc level of 40 volts it don't make a difference EMC wise. Neither does it make a difference if one is centred at 100V whilst the other is at 200V. Data on these types of systems is regarded as AC and the DC levels are unimportant for decent transmission and reception provided both AC signals are of quite similar amplitudes.

This is an example Ethernet port circuit: simulate this circuit – Schematic created using CircuitLab

If the 100BASE-TX runs +-2.5V, does that mean the TX+ wire uses 0 through +2.5V, and the TX- wire uses -2.5V through 0?

Potential level of a wire in a pair is not important. The important thing is the voltage difference (potential difference) between the two wires.
For example if the potential of +line is 600V, the potential on the - line must be smaller than 600V (i.e.; 597.5V) in order to make a logic-0 signal. For a logic-1 signal, for example, you can send 350V on the - line and 352.5V on the + line.
The DC component of the signal will be blocked in the transformer part; the only 2.5V pulse is transferred. If you can guaranty that there will be no DC component in the signal, and the voltage on the each wire stay in the allowed limits in the Ethernet Phy datasheet, you can remove the transformer.

What is the voltage range that equates to 1, and what is the voltage range that equates to 0?

if the voltage difference is positive and above some value $$\\left(V_+ - V_- > V_{th}\right)\$$, the signal on the line is interpreted as logic-1. And if the voltage difference is negative and below some value, it is interpreted as logic-0.

Who and how is this range set?

For value of $$\V_{th}\$$, see the datasheet of the Phy chip you are using. For example, KSZ9031RNX is a popular Phy chip. Onthe page 56 of its datasheet, typical value of Squelch Threshold is given as 400mV.

How is the noise removed?

As far as I know, the noise is removed by using a twisted pair cable, 50$$\\Omega\$$ differential pair traces on the PCB and selecting a transformer with a common mode choke (e.g.; HX5008).

If 100BASE-TX uses +-2.5V, in a typical office environment, about how much mV of noise can you expect?

Maximum allowed noise can be calculated by your design parameters; those are:
1) The voltage level of the signal. (Ex: $$\\pm2.5V\$$)
2) The threshold value $$\\left(V_{th}\right)\$$ given in the Phy datasheet.