I don't understand phase distortion

[Farview]

Actually if you consider that the RED active wire is doing all the pushing and pulling of the signal, having the wires back to front will invert the sound. Even worse, having them right on one speaker and wrong on the other will give you out of phase playback... We test masters with such configurations for the benefit of the unknowing consumer......

Only if a balanced cable is wired backwards on one connector.

I was talking about unbalanced instrument and speaker cables. It really doesn't matter which end you plug into the guitar as long as the other is connected to the amp. It just doesn't.

 

[RDMSstudio]

Only if a balanced cable is wired backwards on one connector.

I was talking about unbalanced instrument and speaker cables. It really doesn't matter which end you plug into the guitar as long as the other is connected to the amp. It just doesn't.

http://www.klhaudio.com/hookupoverview.htm
http://www.rahul.net/thinker/audionotes.html
http://www.stevehoffman.tv/forums/archive/index.php/t-41802.html

 

[DM1]

Phase distortion occurs when some frequencies get through a system at a different time than other frequencies, modifying the phase relationship between the frequencies.

When all you are doing is changing the phase relationship of different parts of the waveform.

OK, that helps a lot. Thanks very much for the responses.

Here's a mini-article I wrote about EQ and phase shift, and maybe you'll find it useful:
www.ethanwiner.com/EQPhase.html

Thanks Ethan, I actually came across that article as I was researching the issue for myself originally. It helped me understand how phase shift is used to EQ a sound. And paired with apl and farview's responses, I can see the potential for "distortion." But I'm still not sure I understand how it sounds.

So if I wanted to contrive an experiment to exaggerate the effect, to make it unmistakably audible, what would be my best approach? Is there a particular EQ operation that's more sensitive to this type of phase distortion?

 

[Farview]

http://www.klhaudio.com/hookupoverview.htm
http://www.rahul.net/thinker/audionotes.html
http://www.stevehoffman.tv/forums/archive/index.php/t-41802.html

You aren't getting what I am saying. (or what those links are telling you, one or the other)

All those links are telling you is to be consistant by hooking + to + and - to - to make sure that both speakers are in phase with each other. There is no difference between the red wire and the black wire, they are just marked so you don't accidently swap them.

What I am talking about is (for example) with a guitar cable, they are telling you that it makes a difference which 1/4 connector gets plugged in to the guitar and which one gets plugged into the amp. Not swapping + and -, but swapping the ends of the length of cable. It's so stupid, that it is hard to explain it to people because they always think you are trying to say something else.

 

[Ethan Winer]

DM,

> Is there a particular EQ operation that's more sensitive to this type of phase distortion? <

You may have missed the point of my mini-article. If someone boosts the treble using an EQ and then hears a "phasey" and hollow sound that wasn't present before, it's not that the EQ added that. That hollowness was always present and the EQ just brought it out while doing it's job of boosting highs with respect to lows.

--Ethan

 

[DM1]

You may have missed the point of my mini-article. If someone boosts the treble using an EQ and then hears a "phasey" and hollow sound that wasn't present before, it's not that the EQ added that.

Yep, I definitely missed the point!! Heh.

So I guess I'm still confused. It reads like you're saying "the only effect introduced by an equalizer is that of boosted or cut frequencies. Anything 'negative' brought out by the EQ was actually in the captured sound to begin with."

Is that right?

And if that's the case, why are there so many articles by professionals admonishing us amateurs for overusing EQ? (I mean, aside from the obvious "capture it right at the source, and you won't need EQ" lessons.)

 

[Ethan Winer]

> It reads like you're saying "the only effect introduced by an equalizer is that of boosted or cut frequencies. Anything 'negative' brought out by the EQ was actually in the captured sound to begin with." <

Yes, but with a caveat. There's always a caveat!

If you limit yourself to high and low shelving EQ, or very low Q (wide bandwidth) boosts, the EQ is not likely to introduce new artifacts of its own. But if you use a narrow bandwidth / high Q, the EQ can add ringing when you add a lot of boost.

> why are there so many articles by professionals admonishing us amateurs for overusing EQ? <

EQ often is overused! Sometimes it's because a "home recordist" is inexperienced and doesn't know when enough is enough. Just as often it's due to having a poor monitoring environment, especially when EQ is misused on bass instruments in an untreated room.

That said, I've often seen "professionals admonish" people to avoid EQ because of the phase problems EQ causes. But you know what? They are wrong! I don't care how good their ears are, or how many hit records they have mixed. Just because someone has the talent to turn knobs and make music sound good does not mean they understand the science behind those knobs. This is how so many audio myths get started. In fact, I'd say that the majority of articles in popular magazines are wrong on most aspects of audio science.

When I wrote my first Audio Myths article, I submitted it to a popular magazine I had written for previously. A few weeks later the magazine editor called me to say he and his staff had a meeting, and they disagreed with every single one of my ten myths! So clearly, as often as not, conventional wisdom is wrong.

--Ethan

 

[DM1]

Ethan, I can't thank you enough for all your help. You're a Godsend for newbs like me. (And that "myths" article you wrote is great. It should be required reading if they ever make an official audio-engineer certification.)


My confusion's mostly cleared up, but there are two things I'm still not sure of. Maybe someone has some thoughts:

- If the phase shift in (well made) equalizers doesn't distort the sound, then why do manufacturers sell expensive "linear phase" EQs? Are they just capitalizing on consumer igornance? I've never tried (for example) Waves LinEQ, but the extra price they charge for it certainly makes it seem like there's something special about the plugin.

- I've read in a few places that in an equalizer, when the phase shifted signal is mixed back with the original signal, the fact that it's shifted in time introduces a small, imperceptible echo. Simplified math: to boost a signal at 1000Hz, one would mix a signal delayed by 500 microseconds (i.e. half a wavelength) to achieve the desired EQ effect. Now, 500 microseconds is obviously beneath the levels of human perception. But it's an echo nonetheless. And the contention in the articles I've read is that repeated EQing (e.g. if there's EQ on the way in, then EQ on the track, then EQ on the master bus,) will compound this "echo" effect which can lead to smearing of the sound.

Further, at lower frequencies, the amount by which the signal must be delayed approaches human perceptibility. (To EQ at 50 Hz the delay time is 5ms.) So shouldn't this echo effect actually start to become obvious when EQing lower frequencies?

Or has my simplified math overlooked some complexity of an equalizer that makes this moot?

 

[Ethan Winer]

> why do manufacturers sell expensive "linear phase" EQs? Are they just capitalizing on consumer igornance? <

You now understand fully the deepest mystery, and the origins of audio mythology. And consumerism in general.

Next up - alternative medicine.

> Now, 500 microseconds is obviously beneath the levels of human perception. But it's an echo nonetheless. <

Yes, but all frequencies are not delayed the same 500 microseconds. That would create a comb filter having a series of many peaks and nulls, which is the basis for phaser and flanger effects. With an EQ based on resistors and capacitors, or a digital equivalent, it's really phase shift that creates the "delay" as opposed to a true broadband delay.

> And the contention in the articles I've read is that repeated EQing (e.g. if there's EQ on the way in, then EQ on the track, then EQ on the master bus,) will compound this "echo" effect which can lead to smearing of the sound. <

There's no such audio parameter as "smearing." So when you see someone use that term you know right away they don't understand this stuff at a technical level. Read: 95 percent of audio magazine writers. (And 99 percent of audio forum posters.)

Echoes can give an effect you might call "smearing," but that's for echoes in the milliseconds range (and more than one echo). For example, "smearing" might be an appropriate term when you have loudspeakers near reflective walls and no acoustic treatment at the first reflection points. Since sound travels at about one foot per second, distances in the "several feet" range give delays in the "several milliseconds" range. It's not possible to perceive a single echo shorter than 20 milliseconds or so, so what is really heard is the effect of comb filtering.

In the case of loudspeaker first reflections, this "smearing" is not only one echo, but three - one each from the left, right, and ceiling. If the floor is reflective make that four echoes, all with slightly different delay times.

> (To EQ at 50 Hz the delay time is 5ms.) So shouldn't this echo effect actually start to become obvious when EQing lower frequencies? <

Again, this would be the case if the EQ were done using a broadband delay, where mids and highs were also delayed, but that's not how they do it.

--Ethan

 

[tarnationsauce2]

sorry for the OT here, but ...
Holy crap! Ethan!
I'm kind of star struck, I saw your Real Traps videos a while ago. haha! they really taught me a lot about treating rooms and the importance of bass traps.
I love the part when you're like "jut eyeball it" and your co-hort broke out the laser level LOL!

I'm forever in debt to you.
Small world.

 

[SouthSIDE Glen]

> why do manufacturers sell expensive "linear phase" EQs? Are they just capitalizing on consumer igornance? <

You now understand fully the deepest mystery, and the origins of audio mythology. And consumerism in general.

Next up - alternative medicine.

> And the contention in the articles I've read is that repeated EQing (e.g. if there's EQ on the way in, then EQ on the track, then EQ on the master bus,) will compound this "echo" effect which can lead to smearing of the sound. <

There's no such audio parameter as "smearing." So when you see someone use that term you know right away they don't understand this stuff at a technical level. Read: 95 percent of audio magazine writers. (And 99 percent of audio forum posters.)

I'd like to ask how these two responses jive with what is said in parts one and two of of the article written by Dr. David Berners (Universal Audio Director of Algorithm Development) and Dr. Jonathan Abel (UA's CTO). Now I'll grant you that these guys are representing a product and I am always suspicious of the bias of those with products to sell, but I'd frankly be hard-pressed to find anybody who knows more about this subject than these guys.

In part 1 of this article they do say that phase distortion is not extremely audible, but they do discuss where and why it does become audible, along with sample WAV files demonstrating their explanations. They claim that phase problems are most audible when associated with transients, Then they go further to say:

"The audible artifacts from the phase warping are perceived by most people as time-domain problems. The warped-signal has a more diffused, almost reverberant sound, while the dry track sounds much tighter. As a rule of thumb, if the phase-warping smears the signal by less than 10mS, the artifacts will not be offensive. Since the amount of time-smearing is defined as the phase of the response at a given frequency divided by that frequency measured in radians/second, for a given phase shift, artifacts may be more noticeable at lower frequencies."(Boldface added)

"Smearing" may not be a technical term for a technical cause any more than "sharpness", "warmth", "harshness" or "muddiness" are, but it is a valid adjective for describing sonic effects that result from a perceived de-focusing of the sound in either the spacial soundstage dimension or (as described above) in the time-domain dimension.

The second part of this articles talks about linear-phase vs. minimum-phase EQ. Using sibilance as an example, they highlight what they see as one advantage of linear phase EQs:

"If a filterbank is made to be linear phase, it can be completely transparent as long as no modifications are made to the signal while it is split into separate bands. An application for this may be a de-esser which is being used to limit occasional sibilance. If the de-esser is made with a linear-phase filterbank, it can be made to be absolutely transparent except during those instants when the de-esser is active, reducing sibilance. On the other hand, if the filterbank is not linear-phase, then anything sent through the processor will incur phase distortion all the time, even when no de-essing is taking place. So for transparency in a filterbank, the linear phase property is valuable."

I'm not picking sides here at all, just curious as to which experts selling which products are correct.

I suspect it probably the usual answer: Both and neither.

G.

 

[tarnationsauce2]

I think the "smear" term means that the EQ'd frequency is happening at a different time than all the rest of the un-EQ'd sound. Especially during a transient, you're spreading the power (sperad / smear) over more time making it have less impact than if all frequencies happened at once.
Yeah 500us is hardly any time... and I don't think someone could hearsomething delayed for that amount of time, but maybe at that quick of time we don't percieve it as a delay, but rather a "smear" so to say.
Also different EQ's have more or less phase distortion. I imagine real cheap stomp boxes, or RadioShack EQ's have a lot more delay, since the transistor circuits/ opamps they use probably have a horrid slew rate and delay time.
Also, the "flange" effect really needs some sort of modulation to make it apparrent (or sound like a flanger). Otherwise without any modulation, it just has a kind of strange "hollow" sound to it.

 

[DM1]

In part 1 of this article they do say that phase distortion is not extremely audible, but they do discuss where and why it does become audible, along with sample WAV files demonstrating their explanations.

I think the first few examples they give are somewhat misleading. In order to hear the phase shift, you need to A/B the WAV files against eachother. That's illustrative, I suppose, but certainly not representative of what happens in an equalizer, which only outputs a single audio stream.

The warped-signal has a more diffused, almost reverberant sound, while the dry track sounds much tighter.

The drum sample they're talking about was generated with an allpass filter, rather then an EQ. Is that a fair comparison? (I'm genuinely curious. An allpass filter is explicitly designed to induce phase shift.)



All that said:

Especially during a transient, you're spreading the power (sperad / smear) over more time making it have less impact than if all frequencies happened at once.

That does seem like an inescapable conclusion, doesn't it?

In 500μs, sound travels 17cm, about the distance between our ears. Since that distance is so important to sound localization, obviously we're sensitive to at some level to such small changes in the time domain.

 

[SouthSIDE Glen]

The drum sample they're talking about was generated with an allpass filter, rather then an EQ. Is that a fair comparison? (I'm genuinely curious. An allpass filter is explicitly designed to induce phase shift.)

The point there is not so much, I think, what device caused the phasing; the point was to get phase shift so that one cound indeed hear the audible difference introduced by it.

Those technical issues aside, after 24 hours of reflection I think my main greif with the direction that this thread was starting to take was that I have a problem with introducing any kind of distortion to my signal unless I have to. The idea that was being implied that because phase distortion is debatably inaudible and admittedly not very audible at it's best, that it's perfectly OK to inject into the signal and not worry about just goes against nature.

Except in very special circumstances where distortions cancel, *all* distortion is additive. It is possible to add together two different distortion components that are both individually inaudible, but summed together produce audible distortion. It is also possible for an otherwise inaudible distortion component to have an audible effect on normal signal processing (e.g. transient phase distortion thrown into fast attack compression.)

G.

 

[Ethan Winer]

tarnationsauce2,

Thanks man!

> I love the part when you're like "just eyeball it" and your co-hort broke out the laser level LOL! <



--Ethan

 

[Ethan Winer]

Glen,

> In part 1 of this article they do say that phase distortion is not extremely audible <

When I see stuff like this in an article meant for non-technical types like most recording engineers:

a(f ) sin(2pf ·[t – t (f )]) = G{sin(2p f t)}

My first thought is they're BS'ing and trying to snow us with techno-babble. Reading further confirms this. I'm not saying they're wrong! I'm saying the article is irrelevant and contradictory, and probably intentionally confusing. For example:

the human ear is not completely linear, so some harmonic content is actually 'manufactured' in the ear.

Here they're admitting that the audible difference between those square wave files is due entirely to playing them so loudly that your ear distorts.

On the other hand, if the filterbank is not linear-phase, then anything sent through the processor will incur phase distortion all the time, even when no de-essing is taking place.

This is simply not true. At least not if the equalizer is designed properly. For an example of a very capable parametric EQ circuit that could be the basis for pretty much any type of EQ needed, see Figure 3 in this article:

www.ethanwiner.com/spectrum.html

When the boost/cut knob is in the center, the EQ is effectively out of the circuit and will not have any effect on frequency or phase of signals passing through it.

--Ethan

 

[SouthSIDE Glen]

When the boost/cut knob is in the center, the EQ is effectively out of the circuit and will not have any effect on frequency or phase of signals passing through it.

From your wounderfully-written article:

"One feature of this circuit is that the entire filter is effectively out of the picture when the control is in the center position, thus eliminating the need for a separate in/out switch for each band."

Wouldn't that then be the schematic for one band of what would turn out to basically be a linear phase EQ design?

Or am I missing something there?

G.

 

[Ethan Winer]

Glen,

> Wouldn't that then be the schematic for one band of what would turn out to basically be a linear phase EQ design? <

No, you're missing something here.

The EQ is effectively out of the circuit when the boost/cut knob is in the center. But when it's not in the center the EQ is doing it's job via phase shift. I'm not an expert with the implementation of linear phase EQs, but I recall reading somewhere that they do shift phase. Just that they shift less than a conventional EQ, or in a slightly different way.

The important point is that regular EQ works just fine, thank you very much. And if a regular EQ shifts phase as it works, this is not damaging and there's no reason to worry about it. This is the part that kills me every time. People obsess over minutiae like phase shift and jitter and dithering - none of which matter one iota - all the while ignoring huge problems in their rooms that are very damaging.

--Ethan