Question about Phase shifting

[Olivier]

Hi,
do we have something like http://www.voxengo.com/product/pha979/ in reason ? I'm especially interested in the phase shifting part of it.

The only thing i can find that does phase shifting is Echobode, or are there more ?

[Benedict]

There are option on the SSL or in Selig Gain to reverse Phase but I don't believe there is an option to move the phase of a signal by n degrees.

You can look into Echobode which is worth owning regardless. The phasers effects like T2 seem tempting but I would assume not a good idea as they are based on comb filters so you aren't moving the whole signal.

You could also look into any delay line that lets you move incredibly small nits of time. I am guessing that would be only the JP Bucket Brigade stuff, then you need to work out how time relates to degree of phase. Too scary for me.

Do you mind letting us know why you want this function?

[Exowildebeest]

You could also look into any delay line that lets you move incredibly small nits of time. I am guessing that would be only the JP Bucket Brigade stuff, then you need to work out how time relates to degree of phase.

The VMG offers the smallest steps possible, single samples of delay

[Olivier]

I'm very familiar with normen hansens sample delay https://www.youtube.com/watch?v=cNWT8X4tgNY

I'm looking into this for an assignment for audio classes i'm in. The teacher was quite adamant that phase shifting a signal was a better option then using a sample delay when for example: layering kicks.
I'm not so sure, so i wanted to test it for myself
I think he used a very exeggarated example to show its usefullness whereas a simple sample delay would achieve the same thing.

He didn't like the idea of using a sample level delay. Because well.. its a delay.. transients would start to dissalign etc etc.. I think the timeframe we're talking about, especially by stuff thats higher pitched then a kick, is way to small to run into that kind of problem...

Anyhow.. thats why i wondered if we have something like that in reason

[Benedict]

Sure, Teach is right in that time shifting will introduce Flanging or double-hit whereas a pure Phase shift will (in theory) simply have all your phases exactly the same for a greater punch.

Personal choice as to whether that is a good thing. Imgine telling your Drummer and Bassist to play in phase LOL

[selig]

I'm very familiar with normen hansens sample delay https://www.youtube.com/watch?v=cNWT8X4tgNY

I'm looking into this for an assignment for audio classes i'm in. The teacher was quite adamant that phase shifting a signal was a better option then using a sample delay when for example: layering kicks.
I'm not so sure, so i wanted to test it for myself
I think he used a very exeggarated example to show its usefullness whereas a simple sample delay would achieve the same thing.

He didn't like the idea of using a sample level delay. Because well.. its a delay.. transients would start to dissalign etc etc.. I think the timeframe we're talking about, especially by stuff thats higher pitched then a kick, is way to small to run into that kind of problem...

Anyhow.. thats why i wondered if we have something like that in reason

Delays do not in any way cause transients to "dis-align". Not sure where that idea is coming from.

The problem is that what you are calling a "phase shifter" is a modulation effect, not a static way to shift phase. Maybe you/he meant to say an All Pass filter?

Here's how I understand this subject: Phase shift IS delay, but as it is related to frequency rather than time. An all pass filter can be used to shift the phase at certain frequencies, but this will not produce any change in sound on it's own (unless you combine the output with the original signal to produce a comb filter). The problem with this approach is that only one frequency will be at the desired phase, and the rest will ALSO be shifted but by differing amounts usually negating any advantage.

When combining two kicks it IS important where each one starts (delay), and also possibly the polarity of the kicks to make sure you're not loosing any low end energy by phase cancellation. But using an all pass filter would not be on my list of options when layering any drums as IMO it's overkill and has far too many side affects to be useful in my experience.

But I'd love to hear what this is all about, maybe something new I'm not familiar with?

[Olivier]

I didn't mean to say that a delay would dis-align a transient WITHIN a sound wave. I'm aware that that doesn't happen.
I also don't really understand how i'm talking about a modulation effect. I also don't find the term "all pass filter" on the http://www.voxengo.com/product/pha979/ description. But maybe i'm reading it wrong

Let me try to paint a bigger picture of what happened. I'm just trying to make sense of my teachers story



so It was day one of the course, he was doing sound 101 to quickly bring everyone to the same level, explaining frequency/amplitude and phase. He wanted to stress the importance of being aware of the concept of phase.

To illustrate how important it is he had prepaired an example with 2 kicks. You could see the attacks (or transients he said) were lined up nicely. The rest of the waveform was out of phase.
He then slapped http://www.voxengo.com/product/pha979/ onto one of the kicks and started to jog the dials. You could hear that at a certain setting the samples blended together better, no hollow sound.. He called this "changing the phase".

So thats what he told us.

I then asked, how is the process different from introducing a small delay to one of the kicks ?

He answered by dragging the kick sample a tiny bit backward in the sequencer and said: Now the phases line up.. there are no more phase issues, but do you see how the attacks don't line up anymore ? "You see" he said.. the transient would not line up if you do this by adding slight delays.
This plugin doesn't delay the sound, it changes phase.

At that point i'm confused as to how that plugin would even be able to do that. He was adamant that changing the phase of the entire kick was something different then delaying it. That in itself sounds valid. But how can you change the phase of a "complex" waveform as a kick ?

I didn't ask that question. I didn't want to confuse the rest of the audience..

So on the car ride home i came up with this.. i dunno if this would even work in the real world.
The only way i can think of that you can phase shifting an entire complex sound wave and end up with something that sounds the same is when you would deconstruct it using fourier analysis to get all the partials with their volume and phases. Then change the phase of every partial (whilst preserving the phase relations between the partials) and blend it together again. But i have a nagging feeling that this may not work. I'm going to have to set up some experiments to test that.

The thing is, now that i'm home and read what the plug is doing, i get the impression its just a delay :S They call it phase shifting though.
Voxengo says one of the applications of this VST is a "Track phase fixer "

Anyway.. teacher said chaning phase is not like delaying. So i think to myself thats fine but i want him to explain the details or show me what its doing because to me now it doesn't make sense
Teacher promised me to render one of the "phase changed" kicks to show me that the waveform actually changed. We'll see I'm still pretty confident the VMG-01 is all i really need to change phasing issues, and its a sample level delay.

[Raveshaper]

I could be completely wrong here, but I believe what the plug is doing is focusing in on certain periods of the waveform and then acting on them. This would mean that you could think of what's happening in terms of a spring. The spring represents the entire waveform. The phase shifting would be pinching or stretching only a specific section of that spring. In short, only certain sections of the waveform are either being shortened or lengthened. Possibly amplitude is being adjusted as well (or in place of the shrink/expand possibility).

[Olivier]

I'm gonna dig into what this all pass filter actually is. Its time i'm gonna really understand all this tech

[ScuzzyEye]

I'm gonna dig into what this all pass filter actually is. Its time i'm gonna really understand all this tech

Filters work by phase shifting, and then combining the phase shifted signal with the original. The more out of phase the parts become the more they cancel.

When someone says "linear phase", they don't mean all frequencies are shifted by the same number of degrees, but that the amount of shift doesn't have curves. Or said another way, the amount of shift from frequency x to frequency y goes (basically) straight from 0 to 180.

An all-pass filter is the phase shifting part, but it doesn't recombine with the original signal, so there's no canceling.

A device that can shift all frequencies the same number of degrees is a mythical beast. Or at least a rare one. It does require a Fourier transform for anything but 0 and 180 degrees. Well, there's a special case Hilbert transform that works for 90° (and 270 by inverting that).

[Olivier]

Cheers, that certainly helps!
Now i'm curious how circuitry is able to make the phase shifted signal i know what has to be achieved... Just not how :p

[ScuzzyEye]

Let me back up. I simplified things too much, and probably muddied the waters. That's one part of a filter, especially one that has a variable Q factor.

What's really happening in a filter is there's a part that's taking in the signal and limiting how quickly it can change. This in itself introduces a phase shift. So when combined with the original signal it affects the frequency response of the output.

This is difficult to explain without calculus (as calculus is the math that deals with rates of change). Along with trigonometry (sine and cosine, and logarithms), and complex numbers (voltage and current).

A trivial circuit of a resistor and a capacitor is a low-pass filter even without combining it with the original signal. A capacitor won't pass DC, and the closer a signal gets to DC the less is passed. Even that becomes a low-pass filter, but a charge will build up in the capacitor, so it's a really junk filter. A resistor is added to constantly pull the voltage back to 0.

The more the rate of change is limited, the more the phase will lag. At the highest frequencies the capacitor passes the signal unaltered in amplitude and phase.

My initial description of an all-pass is flawed. I thought about it for a minute. It still uses a low-pass filter, and compares it with the original signal to generate negative feedback. So the more of the signal that's removed, the more gain is applied to the original to restore unity gain (op amps are good for this).

If you can get through the math, the Wikipedia pages on RC, and All-pass filters are excellent.
https://en.wikipedia.org/wiki/RC_circuit
https://en.wikipedia.org/wiki/All-pass_filter

Unfortunately there's not an easy way to explain instantaneous rates of change. As I said, that's exactly what calculus was invented to do.

[Exowildebeest]

ScuzzyEye perfectly illustrates the brick wall I always hit when trying to really understand this stuff. My forgotten highschool level mathematical knowledge, which certainly didn't include calculus, just doesn't cut it. I think that goes for a lot of people. At some point, you've got to accept that there's a limit to comprehension without deep specialization in the subject. Practical knowledge, luckily, is mostly accessible to everyone and doesn't require a deep understanding of how something works - it's a matter of knowing that something works. I guess quantum physicists tell themselves that on a daily basis

[ScuzzyEye]

ScuzzyEye perfectly illustrates the brick wall

Now, a brick-wall filter is something different.

[Raveshaper]

Someone famous, I believe Einstein, once said the following:

"If you can't say something simply, you don't understand it."

His opinion has merit, but what I'm getting at is whether or not this could be described easily in visual terms. Can this be simplified into pictures or even animations that illustrate what's happening to the waveform? Has someone already done this?

[ScuzzyEye]

OK, one more time, approaching from a different direction.

The idea of instantaneous rate of change in calculus, is that you can divide a curve into an infinite number of slices, and go and look at any one of those, and find if the curve was going up or down at that instant, and how quickly. This rate of change of an instant, when related to how far along the curve it is will give the frequency at that point. That's in the analog world.

In the digital world, you don't get an infinite number of slices of a curve, and they are at fixed intervals. The second part of that makes figuring out the frequency based on rate of change even easier. Also while calculus is needed to figure out the initial relationship between time and rate of change, you only need simple math to work with the values.

Back to the abstract form. If rate of change at a particular instant is directly related to the frequency, if you can limit that rate of change from being over a certain amount, you can also block frequencies over an amount. Put more simply, if a frequency of 10kHz needs a 50% change, but you limit the change to be 25% you've blocked frequencies over 5kHz. (These aren't exact numbers, there is where calculus is absolutely required, to relate frequency to slope).

So rate of change equals frequency. We've established that. Thus, slowing rate of change limits frequency. How does phase come into this? You'll find people talking about how much the phase of an altered signal lags behind. Think about what's happening when you limit the rate of change of a signal. If it's value wanted to to from 0 to 0.5, but you limited its rate of change to 25% it only reached 0.25. That would seem to simply decrease the amplitude. What I haven't covered yet is that concept of an instant in time. What happens to the next instant? Say it would have normally been at 0.75, which would have only been a 25% change from 0.5, but now that you limited the change from the previous instant the next now exceeds the limit. How is this dealt with? Again, handwavey calculus. But you can think of it as the cumulative effect of limiting the change ripples through each instant to the next. Allowing some instants to change faster than the would have, while slowing others, but no instant is exceeding the limit. The original signal is still trying to express itself through this limit. Maybe think of the energy of the signal being squeezed through the limit, building up in places, and then releasing in others. This dragging, or slowing of the release of energy is what causes the phase to lag. The original signal wants to show itself, but the filter is preventing that from happening all at once, so it takes its time. How long it gets delayed from being expressed is related to how much of a limit is being placed on the change at that instant so different frequencies lag by different amounts.

The astute amount you may be wondering how long it will take for the original signal to release all of its energy through since each previous instant has an effect on the following. The answer is forever. That's why the type of filter I've been describing is called a Infinite Impulse Response. In theory IIR filters work with an input that started at the beginning of time, and continue on until the end of the universe. But in practice once the values get small enough you can treat them like zero.

Also, going back to my "aren't exact numbers", this is also assuming a brick wall filter. One that can absolutely limit the rate of change. These don't really exist. What really happens is the percent is more like a percent of a percent (of a percent of a percent, ad infinitum). This is why filters have slopes listed in dB per octave. Large changes also with high amplitudes can only be partly limited.

[Exowildebeest]

Thanks Scuzzy, I think that might be a very helpful explanation