Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 7, 2021 16:51:49 GMT -6
Sorry, I will read through this properly later but have you got any technical or engineering documents to support this? Especially around phase shift. There's a lot of articles that seem to disagree with your phase / time statement alone.. audiouniversityonline.com/polarity-vs-phase/
|
|
|
Post by Guitar on May 7, 2021 16:52:08 GMT -6
If I understand correctly, those little phase deviations in say, the high pass filter of this thread, the ones on bode plots, mean your signal is out of phase with itself. The bass end will come out of the speakers at a slightly different time than the treble end of the same signal. It's still a function of time. No, sir. The periodic waveform will look as if there is a delay because the phase changes, but an the signal is not slowed down. This should be fairly simple to test. Take a balloon, set up a mic and a y-cable to two identical mic pres. Engage the HPF on one, don't engage on the other. Record the balloon pop. Look at the waveforms. Zoom in as far as you like. Put the HPF as high as you like. They won't be time-shifted. Thank you, sir! For the fascinating discussion. Now this is my kind of test. I'm not sure how I would measure this without a fourier transform? Is that right? I wouldn't know how to run this test is all I'm saying. All audio signals are composed of periodic frequencies, that is the nature of sound traveling through air. The constant swing from positive to negative polarity, in cycles. Maybe my definition is not so nuanced, don't know. Also I was wrong about Crave EQ. Minimum Phase (digital) is the only one that is zero latency. All the other modes introduce latency, which makes it not a fair test in light of matt's post.
|
|
|
Post by matt@IAA on May 7, 2021 17:05:16 GMT -6
No need for a Fourier, just look at the waveform. If the waveform is slid left or right, then the time component has changed. If the only difference is magnitude, some information has been lost, but time isn't involved.
All audio signals are composed of periodic frequencies, yes, clearly. But they're dynamic periodic signals, not-steady state sine waves. A balloon is also composed of periodic frequencies, but its basically all frequencies at once, just like a square wave can be described as an infinite Fourier series.
No idea, that's a black box to me. As far as I know, our ears are not sensitive to phase by itself - we are sensitive to, for example, speakers with asymmetrical response at different polarities. And obviously the amplitude component is audible, as is the destructive or constructive interference to amplitude caused by phasing (which can include things made by EQ!).
|
|
|
Post by Guitar on May 7, 2021 17:09:57 GMT -6
No need for a Fourier, just look at the waveform. If the waveform is slid left or right, then the time component has changed. If the only difference is magnitude, some information has been lost, but time isn't involved. All audio signals are composed of periodic frequencies, yes, clearly. But they're dynamic periodic signals, not-steady state sine waves. A balloon is also composed of periodic frequencies, but its basically all frequencies at once, just like a square wave can be described as an infinite Fourier series. No idea, that's a black box to me. As far as I know, our ears are not sensitive to phase by itself - we are sensitive to, for example, speakers with asymmetrical response at different polarities. And obviously the amplitude component is audible, as is the destructive or constructive interference to amplitude caused by phasing (which can include things made by EQ!). Sorry matt, I was editing while you were responding. Now I'm wondering if I can hear phase without relation to some other signal. If not, half of this thread is irrelevant. I can hear polarity, of a mono bass guitar signal, I know that much. That's the extreme example. I can certainly hear an all pass filter. I am filled with great wonderment.
|
|
|
Post by matt@IAA on May 7, 2021 17:24:29 GMT -6
Sorry, I will read through this properly later but have you got any technical or engineering documents to support this? Especially around phase shift. There's a lot of articles that seem to disagree with your phase / time statement alone.. audiouniversityonline.com/polarity-vs-phase/Well, I'm a mechanical engineer, so I'm using mechanical analogues. Putting on my very-old, very-rusty EE hat I could be completely and unequivocally wrong on the audio time phase statement, simply because there are time terms in the equations of inductance and capacitance with respect to current and voltage respectively. I would have to work the math out. At any rate the articles you're citing start with the premise of periodic signals. But, again, a true impulse is not a periodic signal. You cannot express the frequency of an non-periodic signal with period (which is one thing that the word phase refers to - some portion of time relative to how often the signal repeats). I'd even go so far as to say that other than pure functions with only one periodic component (like, a simple and perfect sine wave) all signals are only more or less defining frequency content on a per-sample basis which is limited by signal bandwidth. Hence a square wave being represented by an infinite Fourier series, which exists in theory only, yeah? For example, the article you linked summarizes this - "Polarity is a function of positive and negative wiring, while phase is a function of time." Clearly polarity is a function of positive and negative, this is quite true. But we need to be more specific when we handle the second part of that. Phase being a function of time is carried over from the period of the function that it is describing. The phase defines a portion of the period, it is defined by the period. Phase angle is expressed in radians, and you cannot convert that angle from one signal to another without first converting it back as time, based on that signal's period. So a pi phase shift on a signal with one frequency is a different amount of time than the same phase shift on a signal of another frequency. In the case of that article, they're introducing a single time delay, but that time delay results in a unique phase angle shift for each frequency under question. That's why I keep saying it is contentious because people throw the same word around to mean different things.
|
|
|
Post by Guitar on May 7, 2021 17:43:49 GMT -6
I've reached my absolute limit in this thread just now. I'm sitting here reading phase formulas, and listening to white noise through a zero latency all pass filter, which I can absolutely hear, reliably. It's free if you want to laugh with me, phasenudge by Airwindows. Not going to try to say anything intelligent, I'm just amused where this has led.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 7, 2021 17:52:44 GMT -6
I've reached my absolute limit in this thread just now. I'm sitting here reading phase formulas, and listening to white noise through a zero latency all pass filter, which I can absolutely hear, reliably. It's free if you want to laugh with me, phasenudge by Airwindows. Not going to try to say anything intelligent, I'm just amused where this has led. Yeah, I think if we're going to get any deeper into this we'd need a conference hall .. Matt@iaa, I get where you're coming from (roughly) although to steer the talk away from techie stuff for a moment most AE's aren't that deeply technical and neither do they need do they be. I understand what they're saying even if it is contentious, I'm not going to point each of them to get a masters in electrical engineering . Although I have enjoyed this discussion, I rarely get to do so thanks for chiming in.
|
|
|
Post by jmoose on May 7, 2021 18:25:43 GMT -6
Ok let's try and put this in simple guitar player terms & try to look past the fact that Ethan Winer is a world class putz...
If we plug into a wah pedal or mxr phase 90 those are filters that effect phase and amplitude via frequency. There's no time component involved.
Rock the wah pedal back & forth as much as you want and there's no delay.
That's what happens with a single band of EQ, more or less.
Now imagine that 3 bands of EQ plus high & low pass is like having 5 wah pedals on. Probably wouldn't be much left of the original guitar signal yeah? But there'd still be absolutely no trace of time delay.
|
|
|
Post by Guitar on May 7, 2021 18:27:19 GMT -6
One last hail mary before bed, this graphical representation of the balloon test matt proposed. I ran a 440 Hz wave, just a single cycle, through the 50 Hz steep zero latency HPF shown, no where near the test signal. the first wave is the test wave, the reference. The second one has gone through the filter. The distance between the peaks still appears to be about 440 cycles per second (.00227 seconds), if I measured correctly (I used a ruler.) It still sounds like the note "A" through my speakers. Matt is right, jmoose as well, in this digital example, there is no time delay. Both signals start at zero and seems to end at the same moment, although I'm not sure what that trailing line is where the second one stops its rotation above zero, that could be a graphical thing in Cubase, or it could be a sound, not sure, some leftover voltage. This one reminds me of the "water in a bathtub" metaphor I've seen engineers using. There's the same amount of signal in the same amount of space, but it's higher in some spots and lower in others, it's been sloshed around. Phase shift seems not really to be a time change more of a position change. Whereas polarity is an instantaneous difference of 180 degrees, phase is an instantaneous difference of any number of degrees from zero to 360. Measured at a specific frequency, at a specific time always.
|
|
|
Post by Guitar on May 7, 2021 18:36:39 GMT -6
Ok let's try and put this in simple guitar player terms & try to look past the fact that Ethan Winer is a world class putz... If we plug into a wah pedal or mxr phase 90 those are filters that effect phase and amplitude via frequency. There's no time component involved. Rock the wah pedal back & forth as much as you want and there's no delay. That's what happens with a single band of EQ, more or less. Now imagine that 3 bands of EQ plus high & low pass is like having 5 wah pedals on. Probably wouldn't be much left of the original guitar signal yeah? But there'd still be absolutely no trace of time delay. Here's a fun link. www.electrosmash.com/mxr-phase90I think I get it now, after all these years. A chorus and a flanger use time delays to create the sweeps but apparently a phaser pedal only uses phase rotation.
|
|
|
Post by theshea on May 8, 2021 0:07:38 GMT -6
so which plugin eq‘s are good/ok/better for hpf? linear phase ones?
|
|
|
Post by Guitar on May 8, 2021 3:23:10 GMT -6
so which plugin eq‘s are good/ok/better for hpf? linear phase ones? I think generally speaking, shallower (lower order) slopes tend to have less phase shift, with zero latency EQ's, that's the most obvious rule of thumb, I'm sure someone already said this. Linear phase has perfect phase response, but adds the artifact of pre-ringing, which sounds like a tiny reverse reverb if you really listen for it. More noticeable at lower frequencies due to longer wave lengths. More noticeable on transients. Linear phase is sort of complex, I don't fully understand it, but here's a really good article that spells it out in easy to read form: cravedsp.com/blog/linear-phase-eq-explainedFor whatever it's worth, I've seen a lot of mastering engineers saying they don't use linear phase EQ, sticking to traditional EQ styles. But it's worth a listening test and forming your own conclusions. That's where all this theory lives and dies, the ear test. I'm curious if @pueblo has some insight into what makes a good high pass filter. I myself will use any of them, like a barbarian, as long as they sound good. That's my own answer. If you really want to listen to filter shapes, some different topologies (Butterworth, Bessel, Chebyshev (Type I & II), Elliptical, and Resonant Butterworth) you might have a fun listen with Plugin Alliance Cleansweep Pro. The Butterworth type is one of the most common styles you will see around. For one example, they Cheybshev will have a slight bump around the resonant frequency, and a steeper slope than a butterworth, which may be why the old Harrison filters are famous for sounding good.
|
|
|
Post by gravesnumber9 on May 8, 2021 7:40:18 GMT -6
Ok let's try and put this in simple guitar player terms & try to look past the fact that Ethan Winer is a world class putz... If we plug into a wah pedal or mxr phase 90 those are filters that effect phase and amplitude via frequency. There's no time component involved. Rock the wah pedal back & forth as much as you want and there's no delay. That's what happens with a single band of EQ, more or less. Now imagine that 3 bands of EQ plus high & low pass is like having 5 wah pedals on. Probably wouldn't be much left of the original guitar signal yeah? But there'd still be absolutely no trace of time delay. Hmm… can we put it in terms that even a drummer would understand? I kid! This is a great way to visualize the concept. Now I have an excused to set up five wah pedals. This would be fun to try.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 8, 2021 8:09:57 GMT -6
Ok let's try and put this in simple guitar player terms & try to look past the fact that Ethan Winer is a world class putz... If we plug into a wah pedal or mxr phase 90 those are filters that effect phase and amplitude via frequency. There's no time component involved. Rock the wah pedal back & forth as much as you want and there's no delay. That's what happens with a single band of EQ, more or less. Now imagine that 3 bands of EQ plus high & low pass is like having 5 wah pedals on. Probably wouldn't be much left of the original guitar signal yeah? But there'd still be absolutely no trace of time delay. The issue is it's not just Ethan saying it and that's not how EQ's work. There are so many doc's / articles on this, plus a wave is in a continuously cycling stage anyway. I think the point most are missing here is with a decent EQ it happens so quickly nobody should be able to tell.. Ironically some EQ's are actually preferred due to their audible phase artefacts. Before I'd agree I would need to see at least a few articles backing your opinion especially as this is contrary to how technical engineers say they work. Where's Paul Frindle when you need him? . Read Svart's post as well on the previous page, that is 100% correct and AFAIK he's an EE. soundfirst.com/EQ_Phase.htmlintegraudio.com/linear-phase-vs-minimum-phase-eq/en.wikipedia.org/wiki/Equalization_(audio)
|
|
|
Post by matt@IAA on May 8, 2021 9:45:12 GMT -6
Ok, so I did some mulling. I was wrong on the phase-time thing, because for any periodic signal or anything that can be described by a periodic object, any phase representation corresponds to a time based on the period of that portion of the signal.
I will say that while I think it is a correct understanding of this particular case, the word delay still really bothers me because of the implication it has in a mechanical system which can be described by the exact same mathematics. As I explained in mechanical system sometimes that phase change can be positive, which would result in a negative time delay. I think limiting it to being described as a phase change keeps things clear, and also prevents the confusion from talking about actual delays which are also a thing in audio.
|
|
|
Post by Guitar on May 8, 2021 10:49:01 GMT -6
I ran the "balloon test" again but this time really zoomed in on the start of the signal, if you look closely where the cursor is, you can see that there is a several sample delay when running through the HPF. Or call it a phase shift. These are two channels receiving the same exact signal, except one has an HPF and the other has nothing.
|
|
|
Post by matt@IAA on May 8, 2021 10:59:01 GMT -6
But you’re not doing the balloon test, you’re doing a pure sine wave test.
Anyway the result is the same, the difference is the particular frequency would be masked by the other non-shifted frequencies.
Also I suspect if you showed a waveform in 3D on the complex and imaginary planes as a frequency sweep through a phase-shifting circuit it would make what’s going on clearer - like the video svart posted earlier, except with a changing period.
|
|
|
Post by Guitar on May 8, 2021 11:13:09 GMT -6
Thanks for humoring me, I'm not nearly on the level of many people who have posted, but I have learned a lot from this thread, so I appreciate it.
|
|
|
Post by jmoose on May 8, 2021 14:12:57 GMT -6
Hmm… can we put it in terms that even a drummer would understand? I kid! This is a great way to visualize the concept. Now I have an excused to set up five wah pedals. This would be fun to try. Maybe start with two... make a pair of wah boots. (no time to read & catch up right now)
|
|
|
Post by gravesnumber9 on May 8, 2021 14:34:11 GMT -6
Hmm… can we put it in terms that even a drummer would understand? I kid! This is a great way to visualize the concept. Now I have an excused to set up five wah pedals. This would be fun to try. Maybe start with two... make a pair of wah boots. (no time to read & catch up right now) Speechless. Pure genius.
|
|
ericn
Temp
Balance Engineer
Posts: 14,941
Member is Online
|
Post by ericn on May 9, 2021 19:37:30 GMT -6
No need for a Fourier, just look at the waveform. If the waveform is slid left or right, then the time component has changed. If the only difference is magnitude, some information has been lost, but time isn't involved. All audio signals are composed of periodic frequencies, yes, clearly. But they're dynamic periodic signals, not-steady state sine waves. A balloon is also composed of periodic frequencies, but its basically all frequencies at once, just like a square wave can be described as an infinite Fourier series. No idea, that's a black box to me. As far as I know, our ears are not sensitive to phase by itself - we are sensitive to, for example, speakers with asymmetrical response at different polarities. And obviously the amplitude component is audible, as is the destructive or constructive interference to amplitude caused by phasing (which can include things made by EQ!). Hey Matt I think the problem your having with the phase and time issue is that while phase is time, it’s not linear time in the minutes, seconds way. It’s a time based on frequency. It gets confusing as hell.
|
|
|
Post by matt@IAA on May 9, 2021 21:16:06 GMT -6
The problem I have is the terminology.
Phase has more than one meaning - it can be either a portion of a waveform such as so many radians, or describe the translation of the beginning magnitude of a waveform like Asin(wt+ phi) where phi is phase.
And that’s the thing - it is time. But it’s time relative to a period. A pi phase is half a period, but that means that same phase is different for a 1k and 10k frequency component. So a ten degree phase shift or whatever is meaningless without the period of the element you’re describing. Which again is why something that isn’t periodic (like an impulse) is meaningless to try to assign time to phase. All you can do is sample and describe the system response as if it were composed of a portion of an infinite Fourier series. Which is useful and practical but not technically true - you impose an artificial bandwidth limit where there is none.
The other thing is that a phase shift is direction agnostic. There’s no mathematic reason why it can’t be positive or negative. And sometimes in mechanical systems it can drift, the phase angle between two components can lead or lag, or lead and then lag, based on the frequency of the system. In a control system you can happily introduce a positive or negative phi and the controller will not care. Because for a periodic system it doesn’t matter.
So when people say “delay” that’s a very confusing thing. Delay has a definite direction to it, and it also can be described as affecting the whole waveform, as well as distinctly linked to the distance a mic is from the object.
|
|
|
Post by ragan on May 9, 2021 23:28:54 GMT -6
Here's an animation that's pretty cool: Whoa. I wish I had seen this a couple of years ago in the second or third quarter of physics. What a beautifully simple way of looking at the complex plane.
|
|
|
Post by ragan on May 10, 2021 0:06:10 GMT -6
The problem I have is the terminology. Phase has more than one meaning - it can be either a portion of a waveform such as so many radians, or describe the translation of the beginning magnitude of a waveform like Asin(wt+ phi) where phi is phase. And that’s the thing - it is time. But it’s time relative to a period. A pi phase is half a period, but that means that same phase is different for a 1k and 10k frequency component. So a ten degree phase shift or whatever is meaningless without the period of the element you’re describing. Which again is why something that isn’t periodic (like an impulse) is meaningless to try to assign time to phase. All you can do is sample and describe the system response as if it were composed of a portion of an infinite Fourier series. Which is useful and practical but not technically true - you impose an artificial bandwidth limit where there is none. The other thing is that a phase shift is direction agnostic. There’s no mathematic reason why it can’t be positive or negative. And sometimes in mechanical systems it can drift, the phase angle between two components can lead or lag, or lead and then lag, based on the frequency of the system. In a control system you can happily introduce a positive or negative phi and the controller will not care. Because for a periodic system it doesn’t matter. So when people say “delay” that’s a very confusing thing. Delay has a definite direction to it, and it also can be described as affecting the whole waveform, as well as distinctly linked to the distance a mic is from the object. I'm with you, I think. Unless we're purposefully doing a transient analysis or hitting the system with an impulse and looking at the response, there's no inherent meaning to "before" or "after" and so the term "delay" is wonky. There are of course states involved, with storage elements like capacitors and inductors, so that the initial conditions need to be known or prescribed, but once the signals exist and the system has been excited, describing things as "ahead" or "behind" becomes weird. If we pick a reference, like current, we can say that a capacitance will cause the voltage to lag by pi/2 or that an inductance will cause the voltage to lead by pi/2, but that's arbitrarily picking something to reference so that we can talk about anything at all. And with cyclical things, which cycle are we talking about? It can be "ahead" or "behind" depending on where we choose to take a look at it.
|
|
|
Post by Guitar on May 10, 2021 4:08:46 GMT -6
Eventide calls it "time alignment" for their positive or negative offset plugin, maybe that's a cool term for it that doesn't imply ahead or behind. I guess the term negative delay sounds funny but I kind of like this idea. Negative time is a fun concept.
I'm sure you educated folks already know this one but I'll toss another term into this discussion, since it seems relevant, "group delay," being very simply as I understand it, a time shift, phase shift whatever you want to call it, that is consistent within a certain bandwidth. Could be 90 degrees from 40 hz to 400 hz, whatever, just an example. Just an example of an accepted, phase-related, technical term using the word "delay." And yes there's been a lot of casual non technical language in this thread, speaking for myself especially, as a journeyman informal electronics enthusiast. As far as my basic understanding goes a typical filter can only have a positive group delay, so that agrees with what you guys are saying. Maybe a predictive filter with latency could have negative group delay. It seems as though, as ragan was talking about, that just about all passive components, and transistors as well, can add group delay to a circuit. svart is the tiny-time master if you're reading svart I wonder if this could be explained in technical language.
In a lot of my educational type searching I see group delay mentioned at the same times as phase, regarding filters. So I wanted to add this EE talk to the filter thread.
|
|