Phase The Music
The ins and outs of phasing explained.
A brief resume of phasing and flanging with a practical look at some of the methods used to achieve them.
By now, everyone must be familiar with the story of how George Martin christened flanging when he told John Lennon that the effect being used on his voice was in fact a bifurcated sploshing flange, but the effect of phasing goes back much further than that. Ever since early experiments in electronic music, phasing has been known and exploited in various ways but, not to put the cart before the horse, let's take a look at the physics behind the effect.
If two identical waveforms are added together 180° out of phase, and provided that they are of equal magnitude, total cancellation will result. This rather obvious fact on its own is of limited use, but consider a periodic waveform added to a delayed version of itself. If the waveform is a regular sinusoid with a period of P and the delayed version is delayed by a time equal to P/2, then we have the situation where total phase cancellation occurs, if this delay time, call it T, is kept constant but if the frequency of the waveform is changed, then total cancellation will not occur until the frequency is increased in pitch by a factor of three (and all the odd numbers above that), and all this can be made clear by looking at Figure 2. When the frequency is increased by a factor of two (or any even number), the wave forms are in phase and consequently add up. As the delay time is increased, more of these points at which addition and cancellation occur are established and if you were to look at the response of such a circuit over a wide range of frequencies, it would consist of a series of well defined peaks and troughs, a characteristic that has earned it the name of a comb filter.
Now, unless you subscribe to a previously undiscovered school of minimalism, you will be working with signals that are more complex than a single pure tone at a fixed frequency so how will this be affected by our comb filter? As you have probably worked out by now, the filter will emphasise the harmonics within the signal that correspond to the peaks in the filter response and diminish those that coincide with the dips. By varying the delay time, the comb filter frequencies shift and so different harmonics are emphasised and by sweeping the delay time, the characteristic phasing effect is created.
On early recordings, tape phasing was used, though the effect was created much earlier than that by simply placing two mics in front of a loudspeaker and then moving one mic relative to the other so that the sound arriving at the two mics has to travel different distances. As the mics are moved, the sum of the signals from the two mics undergoes comb filtering, the nature of which is determined by the time difference between the sound arriving at the two mics. As sound travels at around one foot per millisecond, it is fairly easy to calculate what mic spacing is needed to cancel which frequencies. A spacing of one foot would mean that the lowest frequency to be cancelled totally would have a period in the order of two milliseconds (1mS corresponding to half a cycle at the point of total cancellation) which is 500Hz, and of course higher frequencies would also be affected, right up the audio band and beyond. By sweeping the delay time, either by moving the mics, or in the case of electronic delay devices, sweeping the clock frequency, the phasing effect that we all know and love is created.
The phasing effects that became popular on early sixties pop music were produced by the tape method due to the fact that digital and analogue delay devices were beyond the scope of the technology of the day, at least outside the R&D lab.
To create tape phasing is not difficult; you need two tape machines, two identical recordings and some means of mixing the outputs from the two machines. Both machines are loaded up with their copy of the recording and set such that the output levels at the point where mixing occurs are identical. The tapes are then positioned so that they can be started in sync and you are ready to roll. By varying the speed of one of the machines slightly, it can be made to run just ahead of or just behind the other machine and as the relative delay changes, the phasing effect is created.
Most modern machines are fitted with vari-speed which makes control fairly simple, but in the early days, the speed change was accomplished by applying hand friction to the tape reels themselves to slow the machine down slightly. As the machine passes from leading, through the point of zero delay to lagging, the effect is very intense and it is this crucial portion of the process that is most difficult to simulate digitally.
"Ever since early experiments in electronic music, phasing has been known and exploited in various ways..."
Because phasing could not be produced live, the Phasor pedal was invented in an attempt to emulate tape phasing but, because of the limited phase shifts that could be introduced using conventional circuitry, the comb filtering effect was mild and contained few peaks and notches. The result was generally held to be artistically pleasing but in no way replaced or even successfully emulated tape phasing. Nevertheless, the Phasor became an accepted effect in its own right and is still used to this day.
The advent of the low cost analogue delay line spawned a host of time related effects devices including the chorus unit, the delay and of course the flanger. These so called bucket brigade devices could generate a delayed audio signal without the need for mechanical devices such as tape loops and, because the delay time was controlled by a single clock frequency, it could easily be varied over a wide range. To go back to the original concept of phasing, it soon becomes apparent that the delayed and undelayed signals required to create the desired effect can be provided by such devices but there is one major restriction which has been somewhat neglected. With tape phasing, the time delay between the two signals can be reduced to zero at the point where one tape machine catches up with the other before overtaking it, but not so the analogue delay line. This has a maximum and minimum practical delay time over which it can be swept and so the flange ratio, (the ratio between the longest and shortest delay time) is necessarily restricted. Digital delay circuitry now produces these effects with an extended bandwidth and less noise but the restrictions still remain.
In order to produce a more dramatic effect in the face of this fundamental shortcoming, a feedback arrangement is provided on flanger units which, as its name implies, feeds some of the output back to the input reinforcing the effect of the comb filter. This has caused the accepted flanger sound to move away from the old tape phasing sound, even though that was the original design aim, which means that if you want the original tape sound, you have to use the tape method - or do you?
There is one method that can be used to simulate the effect of true tape phasing and this relies on a bit of electronic sleight of hand (or should it be sleight of ear)? It is very difficult for the human brain to detect delays of less than 10mS and this fact can be used to our advantage. In the Infinite Flanger project recently described in H&SR, Paul Williams used a short offset delay on the normally direct signal so that when the main delay was swept, it could cover a range stretching from something less than the offset delay to something above it. This means that the sum of the two outputs can actually achieve zero delay and actually sweep through it in the same way as the old tape method. One disadvantage of this method, however, is that you can't apply feedback to peak up the filter response because the frequencies emphasised don't coincide with the frequencies highlighted by the comb filter output. If you give the subject a little thought, the reason is obvious. By applying positive feedback to a delay line, you emphasise those frequencies whose periods are related to the delay time. So far so good, but in this case, our comb filter works on the difference between the delay times of the two delay lines and so the frequencies that it emphasises will be different.
There's another difference between tape flanging and the effect produced by off-the-shelf delay units and that's concerned with timing. A standard delay line is usually modulated by a sinusoidal or triangle wave oscillator and as such has a readily perceived cyclic component. Tape flanging on the other hand, being such a hit and miss affair, is almost random so what can be done here?
After considering the above problems, I decided to spend a few hours locked in my studio with a couple of DDLs to find out what could be achieved and whether or not it was worth the effort. The results follow.
"By varying the delay time, the comb filter frequencies shift and so different harmonics are emphasised..."
Firstly, to try and reduce the cyclic effect and to introduce a pseudo-random element, I used two DDLs with their oscillators set to different frequencies. However, before any further setting up can be done, it's necessary to ensure that both delay lines are being mixed together at the same level and this is achieved by bringing the DDL outputs into two adjacent mixer channels and panning these to left and right so that their levels could be monitored separately on the main meters. White noise from a synth is ideal for setting up this effect but any regular, harmonically rich sound can be used. Once this has been achieved, the two channels are panned to mono as the phasing effect disappears if the two signals are separated; in fact the brain uses the delay information to estimate the distance between the two sound sources so we're back to square one.
The choice of delay time is fairly important as it needs to be as long as possible without being so long that the brain receives a delay; a practical figure seems to be between five and ten milliseconds. These can be set up to be exactly equal by manually adjusting the delay time of one of the units so that the signals sound in phase. Any deviation from the in-phase position will be evident by the colouration of the sound so the correct setting is obvious. By shifting the delay away from this point, there will be a position where the low frequencies all cancel out leaving you with a thin nasal sound and you now need to set the oscillator frequency and depth so that these points are swept through. If too much sweep is employed you will find that too much of the sweep is spent away from the effect area and so the depth should be set as low as possible whilst still sweeping between these two points. Try a modulation rate of about 1 Hz on this unit and around 0.4Hz with the same modulation depth on the other delay unit, but take care when setting up, as most delay lines have an annoying habit of changing their delay times when you start to set up the modulation depth. If your DDL has any indication as to the sweep depth in milliseconds, a starting point of around 2mS won't be far wrong. Final adjustments should be made so that the heavy phasing occurs in the middle of sweeps rather than at the end and then you are ready to inject a real signal.
Not all DDLs are kind enough to give you an output that's in phase with the input and if yours falls into this category, you'll find that when you come to locate the in-phase point where both signals are supposed to add, they will all but disappear completely. In practice, the effect still works very well when set up like this as slight level differences and disparities in frequency response conspire to prevent the signal disappearing completely at the null point. If however you have a phase invert switch on your mixing desk, you can choose which system works best in a given context, but if you haven't, you just have to use what you get.
Phasing, by its very nature, works best on a signal with a high harmonic content and a long bright reverb is ideal material as is a distorted guitar or the cymbals in a drum kit, but it's worth trying on all kinds of programme material just so that you can get a feel for what it can do. It is a richer and more subtle effect than that created by dedicated flanger units and if you have the time to set up an effect which is sympathetic to the music being recorded, it is definitely worth the effort.
Another area of experimentation for the more ambitious is to use a DDL which permits you to feed in an external modulation voltage and this could be fed from an envelope follower tracking all or part of the input programme. This technique lets you experiment with triggered phase sweeps and phased reverb triggered from the snare drum is a typical example.
Good taste usually dictates that you don't use solid phasing all the way through a track, so the easiest way of controlling it is to remove one of the delayed signals until needed, which means that the effect can be faded in subtly rather than just turned on and off. If you need some kind of stereo image, you can pan the two delays very slightly to the sides without losing the effect (but only slightly). Your ears will soon tell you what you can get away with. Doing this has the advantage of giving the bass end of the sound more chance to project if you feel that the effect is robbing you of punch so it's worth trying.
Well that's all on phasing for now but I hope that it will inspire some of you to experiment with this wondrous and ancient effect. I think you'll find it a lot more rewarding than the type of flanging that comes in packets.
Feature by Paul White
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!