Understanding Phase

[amirm]

I have been working on a tutorial on what phase is. This is one of those mysterious topics that is foreign to many. As I got into it, the text became longer and longer and as such, I have not been able to finish it.

In the interim I thought I do a search to see if there are any good video tutorials and ran into this two part series that covers it in the context of loudspeakers from Bob McCarthy as an adjunct to his book, Sound Systems: Design and Optimization (his link shows this to be out of print).

First part does a very nice job explaining the odd displays of phase you see in programs like REW:

Part 2 starts somewhat repetitive but then gets interesting half way through as he analyzes the response of a real speaker (Meyer Sound?):

They are not the easiest videos to follow but are short and worthwhile to watch.

 

[RayDunzl]

In my room, measuring at the listening position, I have a 20~25 dB dip at about 48Hz.

I didn't know it was there until I measured and found it.

Measuring left or right speakers alone (red and green traces), the dip is small, maybe 5dB, vs speakers playing (black).

The dip appears with correlated pink noise, is almost absent with uncorrelated.

For some reason, my room provides 165 degrees relative phase difference between the left and right speaker/subs around 50Hz. The room is wide open on the left rear corner, but symmetrical around the speakers. At the present time the subs are simply "stereo" and sit next to the mains, and the sub output overlaps the full range mains.

What will I do about it?

Right now, nothing.

Correlated pink noise:

Uncorrelated pink noise:

 

[Cosmik]

I have been working on a tutorial on what phase is. This is one of those mysterious topics that is foreign to many.

I'm not sure that phase itself is all that difficult to understand. I think that what is generally missing is context:

• Why are we talking about sine waves in the first place?
• Why are we concentrating on regular periodic waveforms when music waveforms are rarely that?
• Why do people interchange phase and delay as if they are equivalent? ("I am seeing 20 degrees of phase shift which is 2 milliseconds of delay". No it's not.)
• What are the mechanisms that cause phase shift? What are the mechanisms that cause delay?
• Why live with phase distortion if we don't have to?

 

[fas42]

And also, how important is phase accuracy to getting competent sound? For myself, I have never, ever worried about phase - it can do what it likes in a system, - and such has never helped or hindered me getting the quality I was after ...

 

[dallasjustice]

Why do people interchange phase and delay as if they are equivalent? ("I am seeing 20 degrees of phase shift which is 2 milliseconds of delay". No it's not.)

The answer is that phase does express delay with respect to a specific frequency. Here's a handy conversion tool.
http://www.sengpielaudio.com/calculator-timedelayphase.htm


Phase is very useful only when referenced to a particular frequency. It can help identify non-minimum phase frequency changes.

I think the most common confusion is when folks confuse electrical polarity, acoustic polarity and phase.

 

[RayDunzl]

Why are we talking about sine waves in the first place?

Why are we concentrating on regular periodic waveforms when music waveforms are rarely that?

They're a tool with which to simplify or probe more complex situations

Why do people interchange phase and delay as if they are equivalent? ("I am seeing 20 degrees of phase shift which is 2 milliseconds of delay".

How would you characterize the idea that something comes later or earlier than expected after passing through some system?

No it's not.)

It is, at a specified frequency. If it is not, then what is it?

What are the mechanisms that cause phase shift? What are the mechanisms that cause delay?

Reactance, reactance, and in my case DSP that can vary phase by frequency by moving that frequency ahead or behind in time relative to the rest of the signal.

Why live with phase distortion if we don't have to?

You know how to eliminate it?

 

[RayDunzl]

And also, how important is phase accuracy to getting competent sound? For myself, I have never, ever worried about phase - it can do what it likes in a system, - and such has never helped or hindered me getting the quality I was after ...

If you take note of phase, it will give you a whole new area in which to be conjuring up solutions.

 

[Keith_W]

Why do people interchange phase and delay as if they are equivalent? ("I am seeing 20 degrees of phase shift which is 2 milliseconds of delay". No it's not.)

I have to admit that I don't understand phase, either. But my understanding of phase is - that it is only delayed if it is with reference to another wave. i.e. if there is 20deg of phase shift of xxx frequency compared to yyy frequency, then it would be fair to say that xxx frequency is delayed by 2ms with respect to yyy frequency. After all, this is how all pass filters work?

If you have a better explanation, I would love to learn.

 

[RayDunzl]

My experimentation with phase shows it has some importance in the reproduction of transients. The whack and the thump parts sound better if they aren't smeared in time by phase anomalies.

Otherwise, it is not particularly noticeable.

Unless the phase is changing during the rest of a sound's envelope. Ever hear a flanger/phaser, or a Leslie speed up and slow down on a B3?

 

[fas42]

If you take note of phase, it will give you a whole new area in which to be conjuring up solutions.

I agree that if someone was aiming for absolute accuracy in reproduction, in the classic sense, that it would be important to deal with - however, in terms of achieving clarity and a sense of "realness" in the playback it's not relevant. At one stage the system I was using had a subwoofer covering up to about 180Hz with a simple switch to easily reverse polarity while running - yes, it altered the perspective, some elements of what I was hearing. But then I seemed to quickly adjust to the new presentation, it just "sounded the same".

 

[RayDunzl]

I agree that if someone was aiming for absolute accuracy in reproduction, in the classic sense, that it would be important to deal with - however, in terms of achieving clarity and a sense of "realness" in the playback it's not relevant.

That's what I try to conjure - accuracy of the direct wave in the room as compared to the signal contained in the recording. The second part of your statement, I don't know if I particularly agree.

At one stage the system I was using had a subwoofer covering up to about 180Hz with a simple switch to easily reverse polarity while running - yes, it altered the perspective, some elements of what I was hearing.

Well, you didn't necessarily make the phase "right", just differently "wrong".

But then I seemed to quickly adjust to the new presentation, it just "sounded the same".

We're adaptable.

That's why I like to have some objective tools around to supplement my subjectivity.

 

[fas42]

That's what I try to conjure - accuracy of the direct wave in the room as compared to the signal contained in the recording. The second part of your statement, I don't know if I particularly agree.

Well, you didn't necessarily make the phase "right", just differently "wrong".

We're adaptable.

That's why I like to have some objective tools around to supplement my subjectivity.

Where we differ on what we're trying to achieve - in very simple terms - is that you're aiming for the wave envelope at a zoomed out perspective being spot on; what I focus on is that the notorious "micro-detail", that audiophiles obsess about, is well reproduced - a zoomed in, close examination of the wave envelope viewpoint.

Yes, I accept that the phase could have been "wrong" in both instances. However, that adaptability I would assert is key, it's what allows us to listen to live music, in breathtakingly different POV as far as the wavefront reaching our ears is concerned, yet still register, subjectively, that nothing significant has altered as far as the music-making is concerned. I believe in exploiting that human behaviour, because it results in the greatest satisfaction in listening to a wide variety of music recordings.

 

[RayDunzl]

Where we differ on what we're trying to achieve - in very simple terms - is that you're aiming for the wave envelope at a zoomed out perspective being spot on; what I focus on is that the notorious "micro-detail", that audiophiles obsess about, is well reproduced - a zoomed in, close examination of the wave envelope viewpoint.

Really?

Source, corrected in-room, uncorrected in-room:

Show me your micro-nannette.

 

[DonH56]

Be sure to include the impact of crossovers. Probably a good idea to define group delay and relate it to time response as well.

 

[RayDunzl]

Another micro-detail example, using lower frequencies than the example above:

When the amplitude is off, I'll put that down to imperfect frequency response.

When the direction is wrong - that's a phase trouble. Movement from another time is relocated to the wrong moment in time.

Again, top - source, middle - in room with correction, bottom - in room without correction.

 

[fas42]

Really?

Source, corrected in-room, uncorrected in-room:

View attachment 2765

Show me your micro-nannette.

When I say micro-detail, I really mean milli-detail, as in 1 thousandth, 60dB down from the envelope peaks . This is where the differences seem to be important, for me at least, and of course in the normal waveform display that area of behaviour is invisible, lucky to be able to see 1% variations.

 

[Cosmik]

How would you characterize the idea that something comes later or earlier than expected after passing through some system?

What is that something? A peak? A zero crossing? These things can be caused by a 'distortion', not a delay. If something comes "sooner" than expected, are we seeing into the future? No, we are just seeing a type of distortion. If I differentiate a sine wave I get a cosine. If I high pass filter a sine wave I get some other phase shift. If I low pass filter it I get a phase shift in the opposite direction. I have 'distorted' the waveform in a specific way, not delayed it, nor seen into the future. A phase shift can be undone in real time perfectly using a network with the opposite 'distortion'. If the phase shift was a delay, how could we correct it 'now' without seeing into the future?

On a continuous, repeating waveform, a genuine delay (e.g. an acoustic delay) will give me a phase difference, but I would not call this a phase shift. For anything other than continuous waveforms, phase differences caused by a delay cannot be corrected by phase shifting networks, so a system that appears 'perfect' for continuous waveforms may not be so for transients. This is why time alignment between drivers must be achieved using acoustic delays, or the DSP equivalent. This is also why over-simplifying audio into the concept of purely continuous repeating waveforms results in systems that do not work properly for real music.

You know how to eliminate it?

Simple: appropriate application of phase shifting networks and delays to oppose the original phase shifts and delays. In short, DSP - with the caveat that with real physical drivers the correction will not be 100%, nor perfect at all locations. (I do not throw my hands up and say "What about the effect of the room on phase? Why bother correcting it?". I am one of those heretics who believes that humans hear the direct sound from the speakers and filter out the room reflections, so for me the phase of the direct sound matters at some level.)

 

[fas42]

I am one of those heretics who believes that humans hear the direct sound from the speakers and filter out the room reflections, so for me the phase of the direct sound matters at some level.)

Perhaps at some level, but not a terribly important one - that's been my experience.

 

[fas42]

I am one of those heretics who believes that humans hear the direct sound from the speakers and filter out the room reflections, so for me the phase of the direct sound matters at some level.)

I agree with this position - but the interesting thing is that if the direct sound is of high enough quality then purely reflected energy has sufficient integrity to allow the musical content to remain intact, subjectively. The current system I'm playing with, tweaking, faces completely away from me when sitting down at the laptop - the only direct sound are the reflex ports at the back of the speakers; yet, the bounce from the walls, and glass that see the direct sound passes on the message with full coherency ... it's fascinating how the brain can reconstruct the sound event if enough untainted information is available.

 

[tomelex]

I would describe defective phase in audio as thus: the wiggles move out of alignment with one another when time is the reference.

1)Pretend you are looking at the moving element in a microphone from the side moving back and forth due to the air vibrations. Now that music signal is sent through the whole recording and mixing chain and playback back out to another same size speaker element placed right below the mic one. In theory, if the whole amplification process and mix and master process and storage process and playback process did nothing to the signal, those two elements would be moving in perfect synchrony.

Those of us in the real world, realize that is never going to happen, but we still enjoy the music none the less. The best we can hope for if accuracy is the goal is that the wiggles we pull off of the CD or whatever storage medium also wiggle in unison with what we pull off the CD after they are launched from our speakers, and those of us in the real world know that just does not happen either.

As others have said, luckily, we don't seem to be too concerned about phase, however, certainly, the relationship between these wiggles as they are affected by phase shifts can cause sound to be fat or thin sounding, whether the upper frequencies get out earlier or later than the lower ones.

If in 1) above, the output signal moved the same as the input signal but was shifted or delayed a few milliseconds or even a second, to our ears we would not know the difference, that is what perfect group delay is all about, all the signals maintain time phase with each other but the whole lot of them is delayed, as if going through a perfect delay line (the recording and playback and speaker output chain).