Phase Shifting EQs

[AcousticsWizard]

does anyone know what some of those "phase shifting EQs" in the waves masters pack do? I thought phase was about waveforms canceling out each other with regards to their appearance in time, and had no idea EQing frequencies has anything to do with this.

Here is a description in sound on sound magazine:

"Linear Phase Equalizer, Linear Phase Multiband, and L2 Ultramaximiser. As the names suggest, it's the first to use their new 'linear phase' technology, which is designed to prevent the phase shifts that are generated by conventional EQ and filter algorithms and thus sound more natural."

I never knew normal EQs or filters are in any way related to phase shifitng? But i guess one always learns.

 

[mshilarious]

I never knew normal EQs or filters do any phase shifitng? Why would they?

Because that is how they work . . . try some basic research on the physical properties of capacitors and inductors:

http://www.electronics-tutorials.ws/filter/filter_2.html

http://www.soundfirst.com/EQ_Phase.html

So you could actually argue the opposite, that linear phase EQs sound less natural because they do not behave in the same manner as analog EQs. You will see linear phase filters in specialized applications (notably for anti-aliasing and anti-imaging filters), but most people will be using minimum-phase EQs for basic tonal shaping.

 

[SouthSIDE Glen]

Jon, one thing I never quite understood completely; maybe you could help me out here? I think I get why hardware filters can cause some phase shifting due to circuit delays. What I don't get is why software (like plugs) should ever suffer from the same problems, unless they are purposely programmed to emulate a phase delay. With that bizzare exception, why should software ever be anything but phase linear?

G.

 

[Farview]

I think you hit the nail on the head. Plugins are designed to sound like 'real' EQ's. If they didn't, everyone would instantly complain that they don't sound like the real thing.

 

[mshilarious]

With that bizzare exception, why should software ever be anything but phase linear?

G.

Three reasons:

#1: because analog EQs sound that way. Important corollary: because EQs are being used to fix phase response problems. In a minimum-phase device, frequency response is a function of phase response, and vice versa. If you chain two minimum-phase devices, then by correcting the system phase response, you correct the system frequency response.

That said, room response/instrument response/microphone respose problems are not strictly minimum-phase, but they are generally reasonable approximations.

#2: because it's much easier to write an IIR filter (minimum phase) rather than a FIR (linear phase), and it's much lighter on CPU.

#3: because linear phase filters must have latency, the amount of which is a function of desired filter frequency. They also have pre-ringing, which in a badly designed filter (see #2) can be significant.

 

[SouthSIDE Glen]

#2: because it's much easier to write an IIR filter (minimum phase) rather than a FIR (linear phase), and it's much lighter on CPU.

That implies that - correct me if I'm wrong - one filter type (IIR) has delay built into it by pure arithmetic of the physics of waves and not (just) because of physical constraints such as the slowness of capacitor switching and circuit design. Is that true, and if so why does that arithmetic not also exist in linear phase design?

G.

 

[mshilarious]

That implies that - correct me if I'm wrong - one filter type (IIR) has delay built into it by pure arithmetic of the physics of waves and not (just) because of physical constraints such as the slowness of capacitor switching and circuit design. Is that true, and if so why does that arithmetic not also exist in linear phase design?

G.

Not true. With a FIR filter, you don't know what the output is until you have [x] input samples, but the output can be fully defined given [y] subsequent samples. With an IIR filter you know the output right away, but you can't define when the output will rest (theoretically it never does, but practically it is limited by system response).

It has nothing to do with limitations of physical components; the models used for IIR filters are idealized components (unless you purposely model distortions). Note that a basic distortion model doesn't depend on any time based equations at all, whereas all filters do:

f = ax + bx^2 + cx^3

is a very simple distortion model.