Phase expalined....

[Robert D]

Robert,



In addition to Glen's good point about complex waves, your example applies only to one tiny and highly restricted situation. Sort of like all dogs are animals but not all animals are dogs.

--Ethan

Ok, if you take two identical dogs howling at.........

I get the point about complex waveforms and their complex phase relationships. My point is that no matter the complexity of the waveform, inverting it's polarity has the effect of inverting the phase of any and every component of that waveform, relative to the same point in the non-inverted wave. So I don't subscribe to the argument that phase reversal is a mis-nomer, or that phase can only be changed by time delay. When you push the phase reverse button on a mic or a channel strip, you are indeed inverting the polarity of the signal, but in doing so you are inverting the phase relationship to the source, regardless of frequency and harmonic content, or waveform complexity. Polarity reversal is the method, phase inversion is the result.

 

[SouthSIDE Glen]

When you push the phase reverse button on a mic or a channel strip, you are indeed inverting the polarity of the signal, but in doing so you are inverting the phase relationship to the source, regardless of frequency and harmonic content, or waveform complexity. Polarity reversal is the method, phase inversion is the result.

I understand what you're saying, Rob, and it does make some good sense with the "givens" used. However, there are problems with some of the "givens". The first is the use of the phrases "phase reverse" and "phase inversion". Yes, I know that many desk manufacturers ID their polarity inversion buttons as "phase invert" or "phase reverse", but those really are somewhat misnomers in and of themselves. (I know, I know, I'm going to get replies getting on my case for who the ____ am I for questioning or disputing the fine folks at SSL or Soundcraft? I'll just have to live with that and hope that those repliers read on first .)

What is "inverse phase"? By definition, phase is defined by an angle somewhere between 0° and 360° relative to a reference signal. It is a continuum of possibilities. There is no "inversion" in phase, there is only rotation of some percentage through the continuum of 360 possible degrees. There is nothing "special" about a 180° phase rotation; the only reason it appears so special and the only reason that it's (erroneously) called an "inversion" is because in very special circumstances, a 180° rotation happens to resemble the result one gets when they invert polarity.

Those special circumstances include not only a simple periodic signal, but *also* one in which there is perfect symmetry around 0V. This is the second erroneous "given".

Think of this hypothetical (or perform the real experiment with an oscilliscope): You have a perfect sine wave of 1V amplitude, let's say. But it's range is not from -0.5V to +0.5v, but rather it's from 0V to +1V. Now rotate it's phase 180°. It's crests and troughs will flip flop, but it's range will remain 0V to +1V. Take that same original sine and reverse it's polarity and the results will be quite different. The crests and troughs will still flip flop, but so will the voltage range. The new wave will be a true inversion, with the mirror at 0V. The phase rotation, however, will not be an inversion, but rather just a time shifting.

Now, many reply at this point, "So what? I remove any DC offset anyway, so what you describe is itself a special circumstance that bears no relation to real life audio engineering." Well, sure it does. First of all it does illustrate what the actual difference between rotation and inversion actually are, and that they are indeed different animals (to inadvertantly and apologetically mangle Ethan's excellent analogy ).

But second, and more to the point, is that symmetry around DC matters and makes a real difference. Anybody who uses a waveform editor knows that there are many instruments - most notably drums, but also other instruments - for which, even if there is no DC offset, there is still asymmetry in the waveform. The skins just don't rebound as far as they are initially compressed, and the waveforms tend to be either top-heavy or bottom-heavy (depending upon mic placement and polarity). On such DC non-symmetrical waveforms, the result of a polarity inversion and a phase rotation (or "phase inversion", if you will) will NOT be identical, even if they were simple and periodic waveforms.

G.

 

[cusebassman]

Think of this hypothetical (or perform the real experiment with an oscilliscope): You have a perfect sine wave of 1V amplitude, let's say. But it's range is not from -0.5V to +0.5v, but rather it's from 0V to +1V. Now rotate it's phase 180°. It's crests and troughs will flip flop, but it's range will remain 0V to +1V. Take that same original sine and reverse it's polarity and the results will be quite different. The crests and troughs will still flip flop, but so will the voltages. The new wave will be a true inversion, with the mirror at 0V. The phase rotation, however, will not be an inversion, but rather just a time shifting.

To think - I just learned there's a difference between inverse polarity and phase rotation... and you didn't try to sell me a thing.

Thanks for the info

 

[Ethan Winer]

There is nothing "special" about a 180° phase rotation; the only reason it appears so special and the only reason that it's (erroneously) called an "inversion" is because in very special circumstances, a 180° rotation happens to resemble the result one gets when they invert polarity.

Exactly.

--Ethan

 

[Robert D]

Phase rotation is more appropriate for discussions of polyphase things like alternators and stepper motors. And there is something special about 180 degrees, it's the only "phase shift" that can be accomplished outside of the time domain, by simple inversion. The mirror problem is no problem at all, with a variety of solutions. I can do it from a single ended source with 1 transistor and 4 resistors, or a simple diferential amp. Phase inversion has no effect on the time-domain nature of the signal...... really. I think the guys at SSL and Soundcraft get this.

 

[flatfinger]

here is a nice little web tool that's an eye opener

(probably not for ethan; you already know this stuff like the back of your hand)

Note that with audio frequencies of 20Hz to 20,000Hz the range of physical sizes of wavelengths is substantial (1000:1), and this causes no end of problems with loudspeaker system design.

or room tuning
or mic placement
or.................




http://www.mcsquared.com/wavelength.htm

 

[PeteHalo]

There's one important factor that hasn't been discussed here yet. The severity of the phase problem depends on the relative strengths of the combined signals which makes it easier in real life to cope with these situations as you can ignore signals that differ greatly in strength from each other. Human ear cannot tell apart two signals from same source unless theres more than about 15 to 30 ms. time difference between them when the later sound is heard as echo and if the difference is smaller only the frequency content of that sound is altered. I read it from somewhere that if the signals mixed together differ more than 10 db in strength the resulting dip in the frequency response cause by the comb filter effect is less than 1dB that most people can't even hear and from this fact comes the legendary 1:3 rule of thumb used in mic placement.

Especially while recording drums this is important knowledge because kick, snare and tom toms bleeding to OH track will cause trouble later in the mix if you don't realize what's going on. The sound of snare and kick arrive to the close up mics about 5ms. before they hit the OH's and when you bring up the snare track fader to boost the snare in the mix there will be horrible phase problems when those two snare signals reach equal strength. What you'd expect is the snare to become 3dB louder but there will also be deep dips at every 200Hz in frequency content because of the comb filtering making it sound weird and unnatural.

Before the days of modern DAWs you had to come up with all kinds of tricks and rules of thumb - like cutting the bass frequencies from OH track so that bringing up the bass track would not cause audible comb filtering - to avoid these phase problems or to hide the consequences of phase difference but fortunately now you can easily phase align tracks by moving them by a few ms. to make snare hit on snare track match the snare hit on OH track.

 

[SouthSIDE Glen]

Phase rotation is more appropriate for discussions of polyphase things like alternators and stepper motors.

I don't see the distinction, other than an arbitrary divide between waveform properties and hardware or circuit properties. Phase is a derived property of any cyclic waveform, and phase rotation the method for adjusting the value of that property. It is equally as valid for an audio signal as it is for any other such use.

And there is something special about 180 degrees, it's the only "phase shift" that can be accomplished outside of the time domain, by simple inversion. The mirror problem is no problem at all, with a variety of solutions. I can do it from a single ended source with 1 transistor and 4 resistors, or a simple diferential amp. Phase inversion has no effect on the time-domain nature of the signal...... really. I think the guys at SSL and Soundcraft get this.

The engineers surely do . Whoever is in charge of the labeling the controls don't...or rather, probably more correctly, if they do, they are simply falling back on a common misnomer of a title because of it's ensconced popular use. It's kind of like calling the flatbread creation you pull out of the oven a "pizza pie" even though it's not really a pie by common definition, and to call it such can lead to misunderstandings by those new to the culinary topic .

You yourself said that the button accomplishes what it does by inverting the polarity, not by shifting phase, and I'm just trying to illustrate that there are common real live cases where the real results between the two processes are indeed different. A phase shift alone will not create a true inversion (and vice versa), the special exception being if the waveform is symmetrical around 0V and it is a simple periodic.

I'm not saying that there is a "mirror problem", I'm saying that the experiment demonstrates that a 180° phase rotation - or phase shift, if you will - is not the same thing as a polarity inversion. Yes you can correct for it by manipulating the signal using processes outside of and unrelated to the actual phase rotation itself. Just like you can adjust the calibration on the 'scope to make the shifted signal indeed appear the same as the inverted signal did. But those are add-on processes that are tacked on to the actual phase shift process and not a result of the phase shift itself. Yes you can make 2*4 equal the same thing as 3*3 if you add a 1 to the end of it, but that doesn't make 2*4 the same thing as 3*3.

And yes, you're right that phase can be rotated without having to introduce a delay into the recipe; i.e. it can be time-independent. That still doesn't address the point that for offset or asymmetrical components of the wave, the results between flipping polarity and shifting phase will still be different.

Or am I missing something in what you're trying to say?

G.

 

[PeteHalo]

Why is this phase thing so difficult to grasp? After all it all comes down to just a couple of simple principles. With today's DAW you can deal with phase issues by taking a look at the waveforms and if a sound source appears on two tracks you move one of the tracks so that the source is in phase on both tracks. This is not to say that by carefully positioning the mics in multi miced situation you couldn't avoid some of the phase issues and as I wrote earlier if the source appears much more quiet (the result of proper mic positioning) on one track than the other you can ignore that track. Unless of course you alter the dynamics or volume of the source by using compression or bringing up the channel fader or boosting or cutting certain frequencies with EQ.

The only use for phase button on mixer channel I can think of is when you capture a source from opposite directions or when there's some other reason signal polarity is flipped in signal chain. Trying to remedy phase problems cause by time difference using the polarity switch is not going to make the problem disappear no matter how hard or frequently you push it. This much should be obvious to anybody who's working with audio at any level.

To sum it all up when you deal with phase issues first thing to do is to decide if the issue is caused by time difference or reversed phase. If you can't figure it out you better leave it at that before you make things even worse by doing something you can't undo later when you finally realize what the real problem is.

 

[SouthSIDE Glen]

With today's DAW you can deal with phase issues by taking a look at the waveforms and if a sound source appears on two tracks you move one of the tracks so that the source is in phase on both tracks.

True, but only when assuming that there is only a single sound source on both tracks and that there is not a phase problem between the two tracks at the source of each track. This is what causes much of the misunderstanding, I think.

If there is more than one sound source - as when dealing with excessive cross-bleed, for example - it's not always possible to get both sources on both tracks in phase simultaneously by just sliding a tracks back and forth in time. Perhaps if the distance from the mic to the primary source on each track were identical. But if there is a signifigant difference in the miking technique for each primary source, there will not be symmetry in the time delays for each source, and one not be able to line both up exactly by a single time adjustment as sliding tracks will give you.

And even if it's a single source, but different tracks (i.e. not just a copy of one source track, but rather one source from 2 mics), then you can potentially have the comb filtering problem in which sliding one track back and forth will only be able to phase align certain frequencies at any given adjustment.

This is another example of where technology simply cannot replace technique. This is why old-school rules like the 3:1 rule (and it's many related guidelines) remain most solid keystones of recording technique. Because good miking technique to reduce or eliminate potential problems at the source - phase is only one of them - is a far superior solution than letting the problems into the tracking and trying to fix them in the mix later, no matter how much magic our magic boxes have in them.

G.

 

[Robert D]

It is equally as valid for an audio signal as it is for any other such use.G.

Yes, it is valid, it's just not usually used in that context (in my experience and in most of the published literature I've encountered) because we visualize audio using a straight x axis for time.

The engineers surely do . Whoever is in charge of the labeling the controls don't...or rather, probably more correctly, if they do, they are simply falling back on a common misnomer of a title because of it's ensconced popular use.

It's not just SSL and Soundcraft, it is the overwhelming majority of audio equipment manufacturers. I find it difficult to believe that in all of these companies, the engineers, when reviewing the labeling, had their well grounded objections overruled by marketing due to popular use. Hell, the Portico has a Phase reversal switch. Are you telling me someone in marketing overruled Rupert Neve?

You yourself said that the button accomplishes what it does by inverting the polarity, not by shifting phase......

What I said, and this is the crux of the matter, is that "Polarity reversal is the method, phase inversion is the result." The button is labeled for what it accomplishes, not how it accomplishes it. That's what function labels are supposed to do.

A phase shift alone will not create a true inversion (and vice versa), the special exception being if the waveform is symmetrical around 0V and it is a simple periodic.

Glen, get your head out of the time domain. I'm not talking about a phase shift, I'm talking about phase inversion, which occurs entirely independant of the time domain, and is, to qoute James B. Calvert, Professor Emeritus, University of Denver, "frequency-independent, both in phase shift and amplitude (within the bandwidth of the circuit), so that they act the same for all the frequency components of a complex signal. Phase inversion has no effect on the time-domain nature of the signal, and neither delays, advances nor distorts it."

I'm not saying that there is a "mirror problem", I'm saying that the experiment demonstrates that a 180° phase rotation - or phase shift, if you will - is not the same thing as a polarity inversion.

Agreed, except that I'm not talking about 180 degree phase rotation or phase shift, nor was the video that started this discussion. If it had, I'd be agreeing with Ethan, and with you.

Yes you can correct for it by manipulating the signal using processes outside of and unrelated to the actual phase rotation itself. Just like you can adjust the calibration on the 'scope to make the shifted signal indeed appear the same as the inverted signal did. But those are add-on processes that are tacked on to the actual phase shift process and not a result of the phase shift itself. Yes you can make 2*4 equal the same thing as 3*3 if you add a 1 to the end of it, but that doesn't make 2*4 the same thing as 3*3.

There is nothing to correct for. A phase reversal switch doesn't cause a phase shift, and has no need to correct for it.

And yes, you're right that phase can be rotated without having to introduce a delay into the recipe; i.e. it can be time-independent. That still doesn't address the point that for offset or asymmetrical components of the wave, the results between flipping polarity and shifting phase will still be different.
G.

Again, neither myself nor the auther of the video are talking about shifting phase, we're talking about inverting or reversing phase.

Or am I missing something in what you're trying to say? G.

Yes, that a phase shift or rotation of 180 degrees is very different from a phase inversion or reversal. The first is accomplished by phase lead or lag, and has all the problems you correctly defined. The second is accomplished by signal inversion, is time, frequency, and waveform complexity independant, and is well recognized both scientifically and by marketing as "phase reversal". It is not a misnomer, it is an accurate description of the result. I don't need to fire up the scope, I did this experiment in 1977 in my first year of EE, and nothing has changed since.

I have a great deal of respect for you and Ethan's vast knowledge regarding audio, and still do. But I've never met anyone who was right all the time, and on this one subject, you guys are missing the point.

 

[SouthSIDE Glen]

I have a great deal of respect for you and Ethan's vast knowledge regarding audio, and still do. But I've never met anyone who was right all the time, and on this one subject, you guys are missing the point

And likewise, Rob, I hope you do know the respect I have for you. I agree with what you have to say far more often than not, and you are one of the members here for which I have the most respect.

Also, I hope you understand that I find this to be an interesting discussion , not an argument. I have the feeling that it is semantics, and not real concept that is seperating us here; that we both understand pretty much the same thing that the engineers at the Big Companies do, and are in far more agreement than disagreement. Let's see if that's the case...

Yes, it is valid, it's just not usually used in that context (in my experience and in most of the published literature I've encountered) because we visualize audio using a straight x axis for time.

I'm not even talking about a time shift. As both of us have said, phase relationships can be time independant. Phase is the measure of the difference in wave cycle rotation. Often this will be in a given point in time, in which one wave form is offset in the X/time axis. It is when there is a time function involved that you have what can truely be called a "phase shift". It's the shifting in the time axis that's causing a concurrent shift in phase.

But when we are talking in a time-independant domain - as it appears both of us are here - there is no phase "shift" per se, because there is no shift in a timeline. There is only a change in the phase angle, as measured in degrees.

Now, where we are reaching the point of confusion and disagreement, I think, is twofold: the first has to do with the definition of "inversion" itself. The second has to do with the difference between "phase inversion" and "polarity inversion". A lot of semantics and nomencalture there.

As to the first, phase angle is a linear measurement, from 0 to 360 degrees. There are no "inversions". While 180° may appear in some special cases to be a mirror image of 0°, the same can be said of 30° and 210°, 90° and 270°, and so on. There is nothing special about 0 and 180 in that regard. Even more to the point, there mathematically and physically is no actual "inversion" going on. It's just a change of phase angle that happens to be a 180° change. And it just so happens that if one assumes the source waveform as being 0° relative phase angle, and that if the wave is DC symmetric, that a change of 180° happens to macth the results we get if we inverted polarity.

But that's just a change of phase of 180°. It's no more special than a change in the phase angle of 37° or 222°. The fact that 180 happens to be the halfway point on a linear scale of 0 to 360 does not make 180 the inverse of 0 or 360, any more than 5 is the inverse of 0 or 10 on a linear scale of 0-10. It simply looks like what our brains like to think of as "special" because of the DC symmetry (synmmetry around the X axis). In other words, a symmetry around 0DC makes a 180° phase change identical to a polarity inversion, which while the result is the same in that one special case, is an entirely different mechanism from phase rotation.

BTW, I'm not sure which literature you're referring to, but I grew up on plenty of physics and electronics texts, almost all of which describes phase by using a diagram that related points in a sine wave to the degree of rotation on a circle. They usually keep it simple and more easily understandable by showing the "start" of the wave, or 0°, at the center line of the chart; i.e. it has symmetry around the X axis.

The problem is that in real life audio, true symmetry around 0DC is as much the exception as it is the norm. Asymmetry in the sound waves from various natural physical sources, combined with introduction of random electrical noise that does not know from positive or negative, and even possible DC offset, large or minute, means there is in real life a whole lot of breaking of symmetry around the 0DC centerline. It doesn't matter at that point whether the wave is simple or complex. It will, of course by definition be complex, but that's beside the point at this point in the discussion. What's important is that a major component of that complexity is loss of symmetry around the X axis.

Once that symmetry is lost, a polatity inversion and a "phase inversion" show their true character as being two different things, because the results of each one of those manupilations will be different. The polarity inverted waveform will flip everything around the 0DC axis. The "phase inverted" waveform, however, will simply be the mirror of itself; it will be "X axis independant", so to speak.The simple fact is that for most real-world waveforms, that if those switches are performing polarity inversions, they are NOT necessarily performing actual phase inversions as labeled on the switch.

I know and understand that everybody and their uncle calls those switches "phase inverters". They also call what they blow their noses into "Kleenex", regardless of whether they are usng Scott tissue or a Bounty paper towel. And I doubt that Rupert got into an argument with anybody; I'll bet you he calls it "phase inversion" the same way he called his photocopies made on his Canon photocopier copier "Xeroxes".

While "phase inversion" is not a brand name, it's misuse is just like the misuse of the words "Kleenex" Or "Xerox" (and you thought I could never type a sentence with three X's in it! .) It is similar in that it is a misuse adopted by virtually everyone, even those in the industries related to the term, and it is similar in that it's use is only actually true or correct in special circumstances (It is really only correct to say "Kleenex" when one is actually using a Kleenex brand tissue.)

What's the big deal, so we say "Xerox" when we do it on an IBM copier. Same difference, right? Yep, not a big deal. And I never go around correcting anybody about that. But when discussing actual phase, the difference between phase inversion and polarity inversion are real, relevant, and rather important. Rupert (Neve, not Murdoch ) no doubt undertands the difference and takes the common misnomer in stride because he does understand. But for those who may not understand, like those budding engineers who hang out in end-user forums such as this one, it cam be important t to make the real life distinction.

EDIT: View the attached image for a schematic example of the difference between a 180° "phase inversion" and a true polarity inversion. This is a stylized example illustrating the effect that a simple DC offset has, because it's easy to draw and to see . But any time there is an asymmetrical component in the waveform, a polarity inversion and a "phase inversion" can have similar types of differences. Polarity inversion means a reflection around the 0DC center line, including guaranteed sign change. "Phase inversion" by itself, OTOH, performs a phase angle change in the waveform, independant of any given centerline.

G.

 

[Robert D]

Been away for a few days, thanks for the return compliment Glen. Of course we were discussing, not arguing. I think we'll have to agree to disagree on this. To me it's like saying the brightness button on an LCD monitor is misnamed because it's really an LED backlight bias control. The phase reverse button does different things, depending on if it's on a microphone's balanced output, in a single ended circuit in a channel strip, or a piece of code in the software realm. In all cases it doesn't do what you showed as a polarity inversion. But what it most certainly is doing is creating a phase relationship to the before the switch signal that is 180 deg inverted on the Y axis, while staying perfectly in place on the X axis. That's a phase inversion or reversal, not a phase shift or rotation, and it's a very usefull thing, as was demonstrated in the video. Call it what you like, I think it quacks like a duck, which is what practically everyone else is calling it.

 

[flatfinger]

Gentlemen,
The correct term is Phase AVERSION!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Your Welcome

 

[SouthSIDE Glen]

In all cases it doesn't do what you showed as a polarity inversion. But what it most certainly is doing is creating a phase relationship to the before the switch signal that is 180 deg inverted on the Y axis, while staying perfectly in place on the X axis.

That is exactly correct. Whether you want to call it a phase inversion or a phase doohickey, that definition is entirely correct for a time independant change in phase angle measurement. We are in exact agreement there. That's fig B. in the graphic.

That, however, is just simply not the same thing as a polarity inversion, which is a roatation around the X axis. Or put another way, it's simply a change in sign of the voltage values. As shown in Fig C.

Those two descriptions only have the same result when the waveform just happens to be symmetric around the X axis. Any asymmetry, and the results of those two functions are different.

If a phase switch on a mixer truely performs a change in phase angle only, then OK, it's a phase switch. If, however, that circuit is actually performing a polarity inversion, then the "phase switch" label is technically a misnomer, because it's not doing the same thing as a phase change alone does with asymmetrical waveforms.

I'm not sure just how else to explain it, so I sould probably just quit now .

G.