Can all kind of noise be decomposed into differential-mode and common-mode noise?

Dong-gyu Jang

Joined Jun 26, 2015
115
Hello.

I've studied the electrical noise and it looks there are only two class of noise; differential-mode and common-mode noise.

Can all kind of noise be decomposed into differential-mode and common-mode noise? Is there any other kind of noise which can't be handled by typical counter-measurement for these two?

And right now, I'm confusing concept of common-mode noise when the noise source is external radiation shining whole electronic system. I guess the typical Common mode noise picture is the attached picture. This may represent the fact that noise source is 'external' to circuit. I've learn that common-mode noise has equal magnitude and polarity (phase) on both lines (Ex: live and neutral). But If chaotic radiation is the noise source, Is it really true that the noise are the same on both lines? I hardly imagine this.

Attachments

• 96.8 KB Views: 13
Last edited:

crutschow

Joined Mar 14, 2008
28,533
Common-mode and differential aren't different kinds of noise, they are noise that occurs for two different types of conditions.
Common-mode noise occurs in both the signal and the return lines equally. A differential receiver can mitigate that type of noise.
Differential-mode noise occurs between the signal line and the return line. You use a filter to mitigate that type of noise.
Those are the only two classes of signal noise conditions there are.

DickCappels

Joined Aug 21, 2008
8,158
If you only discuss how noise passes from a noise source to a susceptible circuit via two or more conductors, then, all of that noise can be described as the combination of common mode and differential mode signals as crutschow explained.

Noise can also be induced, by electromagnetic fields.

Dong-gyu Jang

Joined Jun 26, 2015
115
If you only discuss how noise passes from a noise source to a susceptible circuit via two or more conductors, then, all of that noise can be described as the combination of common mode and differential mode signals as crutschow explained.

Noise can also be induced, by electromagnetic fields.
Oh. Yes EM wave is actually what makes me confused. Is EM wave incident on the entire circuit common-mode noise? As fas as I know the common-mode noise has its noise source external to the circuit. In this sense, EM wave should be common-mode. But I hardly imagine this noise is equal in both lines if the wave is chaotic.

DickCappels

Joined Aug 21, 2008
8,158
You could probably call it either differential or common mode depending up on the mode of coupling.

For example, if magnetically noise were induced voltage in a trace on a printed circuit board (as it would a turn in a coil) the circuit could probably be referred as being affected by a common mode source.

In the case like an unshielded microphone connection picking up 50 Hz power, it could be said that the circuit is affected by a differential noise source -such as between the capacitvely coupled noise source and the earthed microphone amplifier.

MrAl

Joined Jun 17, 2014
8,609
Oh. Yes EM wave is actually what makes me confused. Is EM wave incident on the entire circuit common-mode noise? As fas as I know the common-mode noise has its noise source external to the circuit. In this sense, EM wave should be common-mode. But I hardly imagine this noise is equal in both lines if the wave is chaotic.
Hi,

It's not really dependent on whether the noise is random or not, it more depends on the distance and orientation of the conductors. We can get into all sorts of cases here, so i'll quote three simple cases.

First case, the two conductors are at an equal distance from the noise source. In this case both conductors should (ideally) receive the same signal because there is no difference in phase or amplitude of the received signal.
Second case, one of the conductors is farther away from the noise source than the other. The one farther away gets a delayed signal and therefore there is a phase difference, and this will mean one wire gets a different level signal than the other and that will result in a differential signal. The difference will depend on the separation distance as well as the noise component frequency (or frequencies) and this can be viewed as a comparison of the wave length to the separation distance.
One more case, where part of one wire is closer than the other for a distance along the wires and the other part of the wire is closer for some other distance along the wires such that they are both closer for the same wire length and both farther for the same wire length just not in the same exact place. If the signal source was some distance away opposite the midpoint of the two equal wire distances, one wire would receive a higher amplitude at one spot and lower at the other, while the other receives a lower amplitude at the same spot as the first received the higher and higher amplitude where the first received the lower, then the signal levels could average out back to common mode.

Dong-gyu Jang

Joined Jun 26, 2015
115
Hi,

It's not really dependent on whether the noise is random or not, it more depends on the distance and orientation of the conductors. We can get into all sorts of cases here, so i'll quote three simple cases.

First case, the two conductors are at an equal distance from the noise source. In this case both conductors should (ideally) receive the same signal because there is no difference in phase or amplitude of the received signal.
Second case, one of the conductors is farther away from the noise source than the other. The one farther away gets a delayed signal and therefore there is a phase difference, and this will mean one wire gets a different level signal than the other and that will result in a differential signal. The difference will depend on the separation distance as well as the noise component frequency (or frequencies) and this can be viewed as a comparison of the wave length to the separation distance.
One more case, where part of one wire is closer than the other for a distance along the wires and the other part of the wire is closer for some other distance along the wires such that they are both closer for the same wire length and both farther for the same wire length just not in the same exact place. If the signal source was some distance away opposite the midpoint of the two equal wire distances, one wire would receive a higher amplitude at one spot and lower at the other, while the other receives a lower amplitude at the same spot as the first received the higher and higher amplitude where the first received the lower, then the signal levels could average out back to common mode.
Hello.

I guess I have to accept that every (even the radiation-induced) noise should be able to decomposed to differential mode and common mode although I absolutely prefer mathematical proof of this.

Can radiation-induced noise make current flow? If it is right then I have to conclude that noise counter-measure for conductive noise can also work for the radiation-induced noise as there is essentially no difference in view of the filter.

nsaspook

Joined Aug 27, 2009
9,075
If by radiation-induced you mean particles then the main defense is mass shielding to reduce noise from sources (or the elimination of the source) that generate things like alpha particles or come from cosmic rays (increasing due to low sunspot activity reducing the earths shielding).
http://www.kurzweilai.net/how-to-turn-your-android-phone-into-a-cosmic-ray-detector
Some types of soft errors in digital systems (like DRAM chips) are caused by these 'noise' sources.

An interesting story, this was 'predicted' in a science fiction story long before it happened in real devices.
http://www.epubsbook.com/ScienceFiction/Asimov42/27421.html

MrAl

Joined Jun 17, 2014
8,609
Hello again,

Proof? How does an antenna work? It picks up lots of noise, both intended signals and unintended signals. If the noise is random then the average will be zero, but the time 'between' averages will not be zero.
Electromagnetic radiation isnt new. Even a changing mostly magnetic field can generate currents in remote wiring. If the field is constantly increasing it can cause a current in the wire that is constant, at least for the time the field is increasing, which can be any length of time we want it to be in theory (other than infinite).
How about a spark gap? That transmits radiation usually unintended and radios can easily pick this up.

One of the things that separates this kind of discussion from the more usual electronics discussion is *distance*, ie the variable 'x' comes into play whereas in most of electronics we are only concerned with voltage and current as they relate to time, not how they relate to distance. Once distance comes into play all things change drastically. The distance between conductors, the length of the conductors, the orientation of the conductors, the type of conductors, all this comes into play and the new variable means we have a more complicated expression which means partial derivatives rather than regular derivatives for example because we have to know what is happening along x as well as in time t. So instead of V(t) and I(t) we might have V(t,x) and I(t,x). And that is the simplest of cases as in three dimensions we would have V(t,x,y,z), which would mean the voltage changes not only with time but possibly between 3d points in space as well. Obviously if we freeze time we have V(x,y,z) which means we could have a different voltage along the wire 'x' and also between wires 'y'. Because a wave travels along a distance and it's amplitude varies along that distance, when it encounters a wire it will immediately react with that wire, but a wire father away will experience a different amplitude at that same point in time t. If the distance is small as compared to the wavelength, then the difference in amplitude will be small, but if larger, then it depends highly on the phase difference which is caused by the distance of separation.

Last edited:

Dong-gyu Jang

Joined Jun 26, 2015
115
Hello again,

Proof? How does an antenna work? It picks up lots of noise, both intended signals and unintended signals. If the noise is random then the average will be zero, but the time 'between' averages will not be zero.
Electromagnetic radiation isnt new. Even a changing mostly magnetic field can generate currents in remote wiring. If the field is constantly increasing it can cause a current in the wire that is constant, at least for the time the field is increasing, which can be any length of time we want it to be in theory (other than infinite).
How about a spark gap? That transmits radiation usually unintended and radios can easily pick this up.

One of the things that separates this kind of discussion from the more usual electronics discussion is *distance*, ie the variable 'x' comes into play whereas in most of electronics we are only concerned with voltage and current as they relate to time, not how they relate to distance. Once distance comes into play all things change drastically. The distance between conductors, the length of the conductors, the orientation of the conductors, the type of conductors, all this comes into play and the new variable means we have a more complicated expression which means partial derivatives rather than regular derivatives for example because we have to know what is happening along x as well as in time t. So instead of V(t) and I(t) we might have V(t,x) and I(t,x). And that is the simplest of cases as in three dimensions we would have V(t,x,y,z), which would mean the voltage changes not only with time but possibly between 3d points in space as well. Obviously if we freeze time we have V(x,y,z) which means we could have a different voltage along the wire 'x' and also between wires 'y'. Because a wave travels along a distance and it's amplitude varies along that distance, when it encounters a wire it will immediately react with that wire, but a wire father away will experience a different amplitude at that same point in time t. If the distance is small as compared to the wavelength, then the difference in amplitude will be small, but if larger, then it depends highly on the phase difference which is caused by the distance of separation.
Hello.

The proof what I meant was a kind of theorem in which every noise can be decomposed into common-mode and differential-mode noise, not antenna. Maybe we can think this way; we can draw noise 'function' over the x. negative x side represents forward path (live) noise while positive side is for return path (neutral). As we know, every function of any shape can be represented as summation of even function and odd function. If we corresponds even function to common-mode and odd function to differential-mode, it is concluded that any noise is summation of common-mode and differential-mode noise.

And your idea has important point that EM wave can generates current flows at least for instantaneous time due to EM wave has E-field. Maybe I should accept that my trial to distinguish radiation-induced noise from conductive-noise is fail. They're all classified into two category although EM-wave noise looks more complicated and its fundamental counter-measure is external shielding, not using filter as NSASPOOK said. (I've tried to conclude that conductive filter can be also good for EM-wave noise but It looks not good idea right now.)

Last edited:

nsaspook

Joined Aug 27, 2009
9,075
For EM field noise the primary countermeasures are shielding (primarily electric but also magnetic at low frequencies), shunting with filters out of band noise and the generation of an equal but opposite polarity (like twisted pairs or differential signal systems) to cancel the noise energy. These are all mingled in their operation with reflection, diversion, cancellation or conversion of noise energy being the common operators as energy can not be destroyed.

Common-mode and differential-mode are circuit terms that are come from the physical geometry of matter and how that energy is directed or modified by that matter. Noise as just field 'energy' is the same and obeys the same rules as any other type of electrical field energy in a circuit.

Last edited:

Tesla23

Joined May 10, 2009
495
Hello.

I've studied the electrical noise and it looks there are only two class of noise; differential-mode and common-mode noise.

Can all kind of noise be decomposed into differential-mode and common-mode noise? Is there any other kind of noise which can't be handled by typical counter-measurement for these two?

And right now, I'm confusing concept of common-mode noise when the noise source is external radiation shining whole electronic system. I guess the typical Common mode noise picture is the attached picture. This may represent the fact that noise source is 'external' to circuit. I've learn that common-mode noise has equal magnitude and polarity (phase) on both lines (Ex: live and neutral). But If chaotic radiation is the noise source, Is it really true that the noise are the same on both lines? I hardly imagine this.
This thread is making this much more complicated than it is.

For an interface with two wires, call the currents at the interface I1 and I2. We can simply change variables to talk about
• the common mode current Icm = (I1 + I2)/2
• the differential mode current Idm = (I1 - I2)/2
All we have done is change variables. I1 = Icm + Idm, I2 = Icm - Idm.

We can call the voltages on the wires V1 and V2. Again we can change variables to:
• the common mode voltage Vcm = (V1 + V2)/2
• the differential mode voltage Vdm = (V1 - V2)/2
Again, all we have done is change the variables, V1 = Vcm + Vdm, V2 = Vcm - Vdm.

This is general, it can be done for any signals on any two wires, it doesn't have to be noise. You can talk about the common mode impedance (Vcm/Icm), the common mode power etc... One place where you may have met this is on op-amp datasheets where the common mode and differential mode input impedances are often specified, common mode rejection ratio specified etc..

This analysis is often useful when dealing with noise at interfaces, but each case is specific. For example, a twisted pair will tend to pick up the same noise on each conductor, so induced noise is often common mode. As the signal is usually differential mode, a common mode filter will filter out the noise and leave the signal. If you don't twist the wires uniformly, or have loops in one conductor for example, you may couple more noise in to the differential mode and your common mode filter won't reduce it. To answer your question:

But If chaotic radiation is the noise source, Is it really true that the noise are the same on both lines? I hardly imagine this.
The coupling to both lines will be very similar if the twists and the gaps between the wires are a small fraction of the wavelength of the interference. You may have to wait until you learn more electromagnetics, but for a noise source to have significantly different fields at points a distance 'd' apart requires frequencies of the order of c/(10*d) where c = 3 x 10^8 m/s. At really high frequencies we use coax cable.

Glenn Holland

Joined Dec 26, 2014
705
Here's a practical example of differential VS common mode noise (or an unwanted signal).

If the input to a device (like an op amp) is through two unshielded (and closely spaced) wires, and they are both subject to a varying magnetic field, an EMF will be induced in both wires. This EMF will have the same the magnitude and direction in both wires (commonality) and it is referred to as a "Common Mode" signal.

However, where's there a difference in the magnitude or direction of the EMF in both wires, it's called a differential mode signal. An example of a differential EMF could be signal from a moving coil transducer where one wire is + and the other is -.

Most op amps have a differential (a "symmetric pair" of common emitter amps) at the input stage so a common mode signal will turn on both transistors by the same amount, but the gain will be close to zero.

Last edited:

Dong-gyu Jang

Joined Jun 26, 2015
115
This thread is making this much more complicated than it is.

For an interface with two wires, call the currents at the interface I1 and I2. We can simply change variables to talk about
• the common mode current Icm = (I1 + I2)/2
• the differential mode current Idm = (I1 - I2)/2
All we have done is change variables. I1 = Icm + Idm, I2 = Icm - Idm.

We can call the voltages on the wires V1 and V2. Again we can change variables to:
• the common mode voltage Vcm = (V1 + V2)/2
• the differential mode voltage Vdm = (V1 - V2)/2
Again, all we have done is change the variables, V1 = Vcm + Vdm, V2 = Vcm - Vdm.

This is general, it can be done for any signals on any two wires, it doesn't have to be noise. You can talk about the common mode impedance (Vcm/Icm), the common mode power etc... One place where you may have met this is on op-amp datasheets where the common mode and differential mode input impedances are often specified, common mode rejection ratio specified etc..

This analysis is often useful when dealing with noise at interfaces, but each case is specific. For example, a twisted pair will tend to pick up the same noise on each conductor, so induced noise is often common mode. As the signal is usually differential mode, a common mode filter will filter out the noise and leave the signal. If you don't twist the wires uniformly, or have loops in one conductor for example, you may couple more noise in to the differential mode and your common mode filter won't reduce it. To answer your question:

The coupling to both lines will be very similar if the twists and the gaps between the wires are a small fraction of the wavelength of the interference. You may have to wait until you learn more electromagnetics, but for a noise source to have significantly different fields at points a distance 'd' apart requires frequencies of the order of c/(10*d) where c = 3 x 10^8 m/s. At really high frequencies we use coax cable.
Yes! this is exactly what I originally wanted!

There are two wires (forward and return path) and signal for wire 1 or 2 is called S(1) or S(2).

Common-mode signal (even signal over the lines) is Scm(n) = [S(n) + S(n')]/2 where n and n' are either 1 and 2 and should be not the same. ( Scm(1) = [S(1) + S(2)]/2, Scm(2) = [S(2) + S(1)]/2) The way of this notation rule is based on the even and odd function decomposition in math. (See: https://en.wikipedia.org/wiki/Even_and_odd_functions)

Differential-mode signal (odd signal over the line) is Sdm(n) = [S(n) - S(n')]/2 and the rule of n and n' is kept.

And as you know, S(1) = Scm(1) + Sdm(1) and S(2) = Scm(2) + Sdm(2).

In this way of analysis, we can easily see that flipping line (setting n = 1 or 2) clearly shows Scm (Sdm) is truly common-mode signal (differential-mode signal).

I really appreciate your clear help again!

Last edited:

Dong-gyu Jang

Joined Jun 26, 2015
115
Here's a practical example of differential VS common mode noise (or an unwanted signal).

If the input to a device (like an op amp) is through two unshielded (and closely spaced) wires, and they are both subject to a varying magnetic field, an EMF will be induced in both wires. This EMF will have the same the magnitude and direction in both wires (commonality) and it is referred to as a "Common Mode" signal.

However, where's there a difference in the magnitude or direction of the EMF in both wires, it's called a differential mode signal. An example of a differential EMF could be signal from a moving coil transducer where one wire is + and the other is -.

Most op amps have a differential (a "symmetric pair" of common emitter amps) at the input stage so a common mode signal will turn on both transistors by the same amount, but the gain will be close to zero.
Yeah, What made me confused at the first time to upload this question is that some people said common-mode and differential mode is only determined by relative signal travel direction with respect to each other (common-mode: same direction, differential-mode: opposite direction.) rather than taking condition that magnitude of signals for both lines are the same.

I finally got to know that every signal is just summation of common-mode and differential mode so question is now clear.

WBahn

Joined Mar 31, 2012
26,398