Does differential signalling really need complimentary signal <<Back
Does differential signalling really need complimentary signal
Q: Hello Experts,
I have confusion over differential signalling so please help to clear my clouded mind.
I was reading a book by Howard Johnson where "differential signaling is defined as transmission of *two complementary signals* over two-wire signaling and if both wires/traces are identical & matched *so that both wires/traces have equal coupling to the reference system*, any AC currents induced in the reference system by one wire are counteracted by equal and opposite signals induced by the complementary wire"
That means complimentary signal in differential signalling helps to mitigate the effect of parasitic coupling between differential trace and reference system and I think that's the biggest motivation of using differential signal at high frequency.
Complementary signals if I understood correctly is sending opposite signal *simultaneously *to both traces that means +V to one trace and -V to another trace but in many designs I don't see complementary signals are used for differential signalling like in USB 2.0 I see + 200 mv (dp) and 0 V (dn) for active high and 0 V (dp) and + 200 mv (dn) for active low ( If it was complimentary signal it would have been +100 volt (dp) and -100 volt (dn) for active high and -100 volt (dp) and +100 volt (dn) for active low )
My question is how can we mitigate the effect of parasitic coupling between differential trace and reference system when signal is not complimentary or is there something am I missing ??
Digital differential signals are two complementary CODED signals that may have little line to line coupling: loosely coupled, or perhaps as much as 15% line to line coupling: tightly coupled. Differential signals have several advantages, one of which is better noise rejection
relative to the power common at the receiver and a much smaller common mode current component throughout the channel than either of the constituent signals by themselves. Differential signals do not have to swing the voltage rails, and most do not. Neither do they have to swing above and below Vss. Differentially coded signals need to swing in opposite directions on each transition. USB uses differential signaling.
it is not complementary signal per se that is important but the signal that goes opposite directions on each transition as Steve pointed out.
Remember, It's AC component of common mode current that creates all EMI/noise and when signal in two wire line changes direction on each transition, it effectively cancels this AC common mode current flowing between line and reference.
Hope that helps !
You have supposed that the reference level must be zero, but it can be any DC level (e.g. gate threshold). Only the AC component of the signal matters for differential signaling.
Far from a differential pair, it does not radiate a magnetic field since the two conductors have opposite currents. If perfect, there is no capacitive coupling to a far away reference plane since when one conductor swings upward the other one swings downward.
It is also true that a fluctuating field impinging on a differential pair is likely to cause mostly a common mode interference - ie, both conductors are affected the same way - and within limits common mode is ignored by the receiving chip.
If you take a single line over a gap in a ground plane there is a significant impedance bump because the single line was coupled to the reference plane, that was its return path. If you take a differential pair over a gap in a ground plane there is much less of an effect, because the two lines are tightly coupled, one is the return path for the other,
and they are not as much affected by the reference plane.
If you REALLY don't want to have differential pairs of lines, then you COULD reduce cross talk etc :
1) make sure you single line is closer to its reference path than to any other single line, by a substantial factor.
2) Separate the thing that radiates from your single line with shielding.
I don't think either of those are good options. That's why PCI Express (8Gb/s), Serial ATA (SATA) 6Gb/s, and Serial Attached SCSI (SAS) 12Gb/s, and USB3 5Gb/s soon to be 10Gb/s, all use differential signaling. I could go on into display interfaces (HDMI, Display Port) too.
I know I'm not giving details here. This is not mathematically rigorous. I'm just suggesting why differential lines are almost always good thing, if you can do that. The alternatives to get the same performance are not as good, at least that's my opinion.
---- Joe S.
"it is not complementary signal per se that is important but the signalthat goes opposite directions on each transition "Rohit, are not two signals that go in "opposite directions on each transition" the very definition of a complementary signal? How is that different from a complementary signal?
To answer this question, yes it does need a complimentary signal! That is the
very definition of differential (difference of two signals).
When I said complementary signal, I meant antipodal signal that's one signal lie on 1 and other on -1. Differential signaling doesn't necessarily need -ve voltage to cancel ac component of common mode current.
The quote you provide says, "any AC currents induced... are counteracted...," but it doesn't say anything about DC currents, which are not cancelled. Fortunately, we don't have much trouble in high-speed digital systems with DC effects, so the lack of complementary behavior at DC causes no difficulties.
If I understood correctly, Ac component of current generates only when signal is transitioning from high to low or low to high but I am not very clear that how dc common mode current is flowing between trace and reference plane. Once capacitor between trace and reference plane is charged/discharged after transition of signal, as per capacitor theory no current should flow through this charged/discharged capacitor.
In short, current flows through capacitor only during transition and we call this current ac common mode current but once line is at high/low level no current flows i.e. no dc current flow ( assuming that there is no leakage current between trace and reference plane i.e G = 0 and differential termination is used between dp and dn )
Am I missing something ??
You are right in your assumption that no dc common mode current will flow if you use differential termination.
Dc current always follows the least resistive path so until there is some resistive path between trace and reference (as in common mode termination), no dc current will flow.