1

Considering all the phase issues that may arise when working with non-linear phase EQs, I asked myself why are the linear ones not the standard? Why do non-linear EQs even exist? It may sound a bit extreme, but there are a lot of cases when using a non-linear EQ might mess things up really bad.

I read that linear-phase eq introduces some latency, but couldn't track down how much, and if it becomes noticeable. This would explain the live scenario, or live monitor when studio recording, but not the mixing/mastering when tracks have already been recorded.

I think I've gained enough knowledge to understand when and why to use one, but because I see no downsides to using a linear one, and I couldn't find an answer to what I'm asking, I turned to this community.

Thanks in advance!

2

I'm gonna throw this in as a first-guess answer, rather than as a comment

I honestly don't know, but I'd be tempted to say they don't make nearly so much difference on a single instrument as a full mix, so people may hold then in reserve for latency/CPU performance until they do become important - on the master bus.

My entire mastering chain is all Linear plugs, EQ, multiband comp/lim etc. Before then… I've never really noticed it to make a difference I can actually hear.

You may as well ask why would anybody use a model of a 1960s Abbey Road EQ… when they're quite clearly "less accurate" than a modern one.

|improve this answer|||||
2

The main problem with linear phase EQs is that if they are perfect, the latency they introduce is infinite. For some applications, that is too much. So basically you have to design them around errors you are still willing to accept. A typical method of implementation is to filter the signal with a normal IIR in windows, once forward in time, and once backward in time. The ringing at the end of input in theory lasts forever, but cutting it off with tapered and overlapping windows introduces a limited amount of error.

In contrast, a good IIR EQ has very little delay except for the transition and stop bands and is simple to implement. It works fine in live situations and when monitoring where the large latency of linear-phase EQ is just not workable.

|improve this answer|||||
  • thank you for introducing the latency thing. You can't select and cut/boost frequencies in real time if you think about it. So LP DSp's work by "predicting the future" which means really adding a delay and then "cheating" nature by feeding out the processed sound from whenever in the past. If yo think about it further this can introduces a raft of problems of its own. You don't get something for nothing. – danmcb Feb 27 at 19:47
2

On single instruments recorded with 1 mic there are no benefits to linear phase EQs. Phase is a relative impact parameter: it produces audible effects only when compared against something else with the same content. Record a guitar with 2 mic, flip the phase on one of them and hear for yourself.

We agree on that. Although not put in words in the question because if LEQs became the standard, single mic takes "wouldn't be harmed either".

Even on multimic'd instruments, the issue is not as likely to occur than one could think. Most of the time, different mics are there to play different roles so that they're set up to pick up different sounds.

That's true on the drum scenario you brought up. Consider now a string section recording, or an orchestral recording, or a chamber music recording, where mics are playing a bigger role that capturing very specific instruments and frequencies. Try using minimum phase eq differently on each mic, you'll notice very, very quickly!

Having said that, Sonar's Everything Equalizers free eBook [1] helped me get some facts straight:

Minimum phase EQs sound great in most instances, but during mastering or when applying delicate EQ adjustments to acoustic instruments it’s often best to use linear phase EQ plug-ins. Linear-phase EQs do not alter the phase of a signal, creating a more natural tonal effect and may also work well for parallel EQ duties. However, linear-phase EQs suffer from greater latency and high CPU usage, which means they can’t be used during tracking or in large, CPU intensive mixing sessions.

It is clear to me now that latency does become an issue, and so des CPU processing, if LEQ is used indiscriminately.

Linear phase EQs also introduce their own type of distortion called pre-ringing, which some very picky people find offensive. In practice, pre-ringing is rarely audible or much of a concern.

To better understand pre-ringing, I found this very useful video:

EQ: Linear Phase vs Minimum Phase by fabfilter

Quoting [1] again:

These equalizers avoid phase shift by analyzing the frequency content and applying gain with FIR filters, a process that takes a lot of time (latency). The processed audio is subsequently shifted earlier to try to keep everything in time. This time shift produces an audible echo, called a pre-ring, that immediately precedes sounds with strong transients, like drums. This pre-ring can smear or weaken the attack of drums and picked or plucked instruments. Getting back to the no free lunch idiom, linear phase equalizers trade phase shift for pre-ringing artifacts. We have now established that both minimum phase and linear phase equalizers have good and bad attributes.

On the use cases for LEQ:

For most applications of EQ, like sweetening instruments or mastering mixes, the actual phase shift is minimal because we use broad bandwidths and gentle boosts and cuts. For extremely narrow cuts or boosts, like when removing a resonant frequency from a snare drum, phase shift will drastically affect the frequencies surrounding the EQ’d frequency. In this case, linear phase EQ would be a wise choice and would remove the offending resonance without affecting the overall sound of the drum. Likewise, a steep high-pass filter that removes rumble from a vocal mic or guitar cabinet could wreak havoc on phase at frequencies well above the cutoff frequency. Phase shift artifacts may cause instruments to sound strange, so linear phase EQ would be a great choice. We can say that narrow cuts and boosts and steep bandpass filters are two situations where you might consider linear phase over minimum phase EQs.

Parallel processing instruments or subgroups in a mix would be another situation for linear phase EQ. Let’s say we have a vocal track and wish to duplicate the track and add some high-frequency shimmer by applying a high pass filter and compression to the duplicate (parallel) track. Minimum phase EQs, especially high pass filters, can induce phase shift that adversely affects the combination of the two tracks, causing unintended tonal changes. You can’t simply “flip the phase” and fix this problem. So, for parallel EQ duties, the linear phase EQ, with its lack of phase shift, may be our new best friend. You should audition both linear and minimum phase for parallel processing since in some cases you might find that phase artifacts sound interesting.

How about blending a top and bottom snare mic? For similar reasons to the above example, a linear phase high pass filter on the bottom snare mic (or hi-hat mic) may help retain the natural body of the snare while removing some low-frequency mud. The same goes for EQ’ing multi-miked guitar cabinets. For stereo miking, however, minimum phase EQ may be just fine since we are typically applying the same eq curve to both mics and both mics contain very similar source audio. Again, try both linear phase and minimum phase to hear what each does in your mix.

As of the latency:

Another consideration when deciding whether to use linear-phase EQ is latency and CPU power. Many linear phase equalizers load your CPU slightly more than traditional equalizers and all linear phase equalizers can impart extreme amounts of latency. Linear-phase EQ latency can range from around 3,000 samples to over 20,000 samples. At a 44.1kHz sample rate, that translates to between 100 milliseconds and more than half a second. While most DAWs provide plugin delay compensation, you may near the compensation limit once you exceed 10,000 samples. Pro Tools, for example, provides about 16,000 samples of delay compensation at a 44.1kHz sample rate.

Many modern plugin equalizers, like those from Izotope and Fabfilter, provide linear-phase, minimum-phase, and even mixed-phase modes. Mixed modes provide some benefits from each type of EQ. You should audition these modes on various sources to find the best uses of each type. Further, many equalizers provide resolution or quality settings to further optimize your preferences of sonic character vs latency.

Finally: should I use LEQ?

To answer the question of “Should I use linear phase EQ?” the answer is, as usual, it depends. Minimum phase equalizers, with their familiar personalities, would be my first choice most of the time. In certain situations, like narrow bells and steep filters, linear phase probably has the advantage. For parallel processing and multi-miked sources linear phase also probably comes out on top.

Bonus:

Some popular Linear Phase options:

  • Blue Cat Liny EQ
  • DMG EQUALITY
  • Fabfilter Pro-Q3
  • Izotope Ozone 9 or Neutron 3
  • Logic Linear Phase EQ (Stock Logic Plugin)
  • Melda MEqualizer LP
  • Nugen SEQ-S
  • T-RackS Linear Phase EQ
  • Waves Linear Phase EQ

[1] https://try.sonarworks.com/everything-equalizers/

|improve this answer|||||
1

In addition to latency, the audible downside of linear-phase EQ is pre-ringing. I cannot explain it better than this page: https://cravedsp.com/blog/linear-phase-eq-explained

Quote:

Pre-ringing is a backwards echo. It sounds like a strange sucking sound. It is most noticeably heard on transients.

If you're familiar with convolution reverbs, and how an impulse response is translated to sound, you might understand what linear-phase EQs (or any audio processing systems) do in terms of pre/post-ringing when you see their impulse responses.

|improve this answer|||||
  • Thanks for commenting on this non-trivial issue whch was unknown to me! – 89f3a1c Feb 27 at 18:24
  • 1
    right! you are better off in fact with some artefacts (phase shifts) which are a natural-sounding consequence of the thing you were trying to do in the first place, than a horrible artefact which is a result of "cheating physics" (so to speak). – danmcb Feb 27 at 19:50
1

On single instruments recorded with 1 mic there are no benefits to linear phase EQs. Phase is a relative impact parameter: it produces audible effects only when compared against something else with the same content. Record a guitar with 2 mic, flip the phase on one of them and hear for yourself.

Even on multimic'd instruments, the issue is not as likely to occur than one could think. Most of the time, different mics are there to play different roles so that they're set up to pick up different sounds. Ex: a kick drum will be picked up by both the kd mic and the overheads but it will sound so different in each (not mentioning the delay introduced by the distance to the mic) that phase is irrelevant and certainly does not grant the use of a PLEq.

On a stereo master, unless each side is eq'd differently, the same is true: no use.

In general, the use cases of LP are very limited. Possibly when EQing individual mics in a Decca tree or in a surround mic setup?

Let's also remember that the phase shift associated with most eqs is a physical constraint of capacitors in eq circuits. They were never "designed" that way. And still, they account for the sound of 99.999% of all recorded music in history.

|improve this answer|||||

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.