Another question about phase in equalizers

Getting Stuff Done How do I .... ?
#1

Ok to make this short, I have tons of Linux eq plugins, I’ve done research on IIR filters and FIR filters, when I run tests on ardour using the plugin analyses view, I see some that have the red phase line go all over the place and some have very little movement such as LSP eq using the IIR filter mode. Should I avoid plugins with drastic phase shifting, or am I not educated enough on how this works, people say calf eq plugin is not good and I’ve heard the filters are cheap sounding etc, I also heard about a git version of calf having better filters, I do like eq10q but I saw the phase shifts but not sure if it’s always a bad thing or part of the way plugins work.

I’ve decided to Stick with LSP plugins since they have the option to select different eq filters.

#2

#1 test in audio…

Are you endangering anyone? (This better be no for a home studio)
EDIT: This better be NO period, not just home studio for the record.

#2 test in audio

Does it sound good? If you can’t hear it then don’t worry about it.

If you tried something that people tell you is ‘wrong’ but it sounds good, then maybe it isn’t wrong, just different. I just had a class yesterday where I pulled up one of my student’s work on a song with a music break at the start of the third chorus, with a really bad recording of an acoustic guitar. Many of them tried to ‘fix’ the guitar, one of them said ‘F it’ and decided to put a highly resonant filter sweep on it instead to make it sound intentional and didn’t fix the crappy sound. All of these are right, even if the guitar recording sounds like crap, so long as it sounds good afterwards.

Seablade
1 Like
#3

Thats a great outcome due to technical circumstances.

#4

The phase is that thing which our ear can not detect. So it only makes matter when we’re mixing the same audio source with different phase corrections. The phase difference between two identical sources may yield to frequency boost or fall within some frequency range. Otherwise, with two different sources it doesn’t matter which phase shift per frequency both have.

#5

This isn’t correct. We CAN hear phase issues, we may not identify them as phase issues all the time, but they definitely can be audible, as they will create a magnitude difference in specific frequencies as well when talking about EQ.

What we can’t hear is a singular source, that is phase shifted in totality across the entire spectrum by an identical amount. This is not the case when dealing with EQ.

We can also hear phase issues when dealing with bleed between microphones. This is why setting up your drum kit for instance, with correct polarity inversion or delay/phase between microphones can make such a difference depending on the mic setup used (Won’t be as important if using only a pair of overheads in XY for instance)

    Seablade
#6

This isn’t correct.

Maybe you improperly understood my position. But I’ll try to explain. I’ll give the quote from Steven Smith’s DSP Guide:

Timbre is more complicated, being determined by the harmonic content of the signal. Figure 22-2 illustrates two waveforms, each formed by adding a 1 kHz sine wave with an amplitude of one, to a 3 kHz sine wave with an amplitude of one-half. The difference between the two waveforms is that the one shown in (b) has the higher frequency inverted before the addition. Put another way, the third harmonic (3 kHz) is phase shifted by 180 degrees compared to the first harmonic (1 kHz). In spite of the very different time domain waveforms, these two signals sound identical. This is because hearing is based on the amplitude of the frequencies, and is very insensitive to their phase. The shape of the time domain waveform is only indirectly related to hearing, and usually not considered in audio systems.

F_22_2

Original article: http://www.dspguide.com/ch22/2.htm

All what you’re talking about is true when we’re talking about more than one audio sources which have some correlation one to another. For example, when micing drums it is extremely important to keep all microphones in phase to have clear sounding drum kit. But when we’re mixing guitar with drums, we don’t need to care about phase issues between both since they have no correlation.

So this doesn’t contradict to the following term:

The phase difference between two identical sources may yield to frequency boost or fall within some frequency range.

#7

That is only true if there is no content duplicated between them which your example is not great on, AND there is no phase issue introduced via another method, such as EQ which many styles of EQ actually use phase to introduce a magnitude shift.

So for the first example, a common problem on live recordings, when you might for instance have bleed from drums into Vocal mics, or guitars into vocal mics, etc. The content is not identical, but can still introduce phase issues, even in close micing scenarios. This is why on larger gigs you might for instance delay all inputs to the proscenium line appropriately, or delay to the system, but to fix issues of phase from the system, and phase issues between mics as well, at least as good as can be expected in such a compromise.

This applies in live recordings or in live room recordings where the band may be all together. I think we agree on this in principle but it did need to be clarified a bit.

For the second example, many EQs introduce phase problems as a fundamental function of how they work. You hear these phase problems audibly as a magnitude change, though typically more complex as the are really comb filters across frequency bands dependant on the filters in question. For instance on some EQs merely sweeping frequency bands in particular ways, even if there is no magnitude change in gain, can cause a phase issue to become apparent, even if the EQ sounds decent otherwise. This is also why I stay away from the Calf EQs on linux, as they have audible phase issues introduced as you utilize the EQ.

So as a counter example, when introducing a phase problems you have far more complex signals, and you have introduced a slight timing shift in a particular frequency. The problem is that you have a balance between resonance and overlap of frequency bands. The gentler the slop of the frequency filter, the more overlap with the original signal or nearby bands depending on the design of the EQ. So you have two signals happening, one 100% in phase with the original as it is the original, one something other than 100% in phase, which will cause interference at frequencies dependent on the phase issue and frequency in question, along with the strength of original vs modified phase.

For anyone reading this, I strongly suggest you do tests, and see exactly what I am referring to. Seriously grab the Calf EQs and compare them with other EQs. Look at the measured phase response in a dual-fft, especially in the slope of the filters, and see how much the change is audible. This is part of why some people prefer certain EQs to others. This technically could apply in any frequency specific processor (So full bandwidth processors like a standard compressor or gate generally don’t worry about this).

Or for more reading on the topic(More so you know it isn’t just me that says this), a quick google shows up articles like the following quickly (See #7): https://www.izotope.com/en/learn/12-common-eq-mistakes-mixing-engineers-make.html

    Seablade
#8

How would you do such tests, would looking at the phase meter in ardour plugin edit window help. Have you tested the eq10q or the LSP eq. I try to only use the eqs that are the best for me to use.

#9

Well since you are testing whether phase issues are audible, you would have to confirm the existence of a phase issue which could be done through the plugin analysis, but also whether you could hear it. It is slightly more complex to do this for real with a double blind test, but you could do a basic test that I suspect for many people will be obvious enough just by replicating settings across EQs, measuring magnitude response with plugin analysis to confirm they are similar and phase is the only thing different, and bypassing and enabling them one at a time. If you hear a difference and the magnitude graphs are similar, chances are you are hearing phase introduced. It is about removing all possible variables EXCEPT phase and then comparing.

I haven’t personally tested EQ10Q or LSP myself yet. And again the point isn’t to eliminate phase, but to remove negative sound. I find the phase issues in Calf to be bothersome and a negative to the overall sound, but have used the various EQs in Mixbus and 32c with good effect, along with some from OvertoneDSP (Which IIRC have some definite phase artifacts in some cases, but in most uses work fine) and a small variety of others. And really the point is what gets me the sound, not what has the least phase issues, though there is a correlation there.

    Seablade
#10

I understand, Im not always aware of certain plugin quality or what types of eq that are good or bad. But I definitely learned very fast that calf eq was not the best choice based on issues I immediately had when using it.

#11

If drums are leaking into guitar track, then both tracks have correlation one to another and we need to take into attention how the equalizer does introduce phase shift in the frequency range. That doesn’t go into counter-part to what I’ve written before.

Agree. But again, it doesn’t go in counter-part to what I’ve written before.

That’s right. Since many EQ filters come from the analog world, they’re mosly recursive and thereby introducing phase shifts with magnitude changes. Making a linear-phase equalizer requires to do much more extensive computations (and obtain much more CPU load) which additionally implement huge latency.

BUT lest’s consider we have an all-pass filter which is applied to some single audio track. It doesn’t change magnitude but introduces the phase a huge shift at some frequency. I suppose that no one will be able to detect the presence of the All-pass filter in the chain while doing the blind test until we’ll start mix this track with some another which will have correlations with the original track.

I can not say anything about Calf equalizers now since I’ve moved to using LSP (surprse) as soon as I’ve introduced EQ plugins. But it’s fact that multiband processors have huge phase issues:

#12

The filters are not linear phase, but they should have no more artifacts than any equivalent IIR filter (or for example an analogue) EQ. However they do model the equivalent analogue filter response
which means you can effectively set cutoff frequencies above Nyquist and still expect the correct magnitude response. In fact, all our EQs, have used analogue processing as a starting point. Even the AF210 graphical EQ - the process was to start with the notion of an analogue EQ - that probably couldn’t even exist in the real world due to hardware limitations - and then try to model how that would look in the digital domain.
My feeling is that plug-in analysers are useful for developers (a bit like test equipment in the real world) but often provoke needless anxieties if users are not familiar enough with what the analyser is telling them to draw proper inferences - therefore its questionable if they are of benefit in a DAW.

@seablade: that’s an interesting article. One of the reasons I’ve stayed out of this thread until now is that for every valiant attempt to clarify these issues, there’s always the hazard of inadvertently adding fuel to the fire due to other’s misinterpretation of a technically nuanced explanation. (just look at how many threads there are about sample rate, bit depth, etc etc). I would hesitate to refer to ‘phase problems’ - they exist, but they are only a problem if they cause something you don’t want.

This is all about context. Phase is a contextual issue. It matters most when you start combining sources to construct a mix. Is it worth obsessing about - no. Many outstanding mixes have been created on analogue consoles with plenty of phase issues (and you might hope that plug-in developers have already worried about it for you). Should you be aware of it - most definitely, but the most important thing is to learn how to hear things in the context of the entire mix. You will almost certainly have to EQ a source differently in the context of a mix than in isolation anyway - irrespective of phase issues.

#13

Yep as I mentioned, I thought we both agreed on that point:)

So I will admit I have never tried with just an allpass filter from a critical test, that would be an interesting test. I haven’t looked into the technical details of how allpass filters are implemented, but my guess is they still involves a frequency filter at some point correct? Which also introduces the possibility of an uneven phase shift at the frequency points, resulting in a not quite linear phase and possible comb filtering when mixed back with the remainder of the filter? If this is a correct guess it at least says that in theory there should be a phase shift, whether there is an audible artifact in reality would obviously require further testing (Though I am sure someone has done it, just haven’t looked, all my comments in this thread are based first and foremost off personal experience and testing with my ears:)

Yea the thread you linked to is a good assertion of exactly what I am discussing with regards to the CALF filters, and exactly why there is better options out there (Presumably your EQs and processors, though I haven’t personally tested them yet myself).

 Seablade
#14

Yes I in no way wanted to give an impression they were ‘bad’. I actually like and enjoy using your plugins, so long as I avoid specific non-typical situations with them they work well, certainly much better experience than with the Calf plugins for instance.

And in this we completely agree. I often use the terms ‘phase problems’ or ‘phase issues’ but it would be more correct to say phase artifacts in most cases. An artifact isn’t always a bad thing, sometimes it can actually be intentional even, but in terms of the Calf EQ I will maintain my assertion (Subjective) of phase problems.

And as usual we again agree, in fact you have covered what takes several classes for my students to understand.

Phase can be a tricky issue because if you don’t understand it truly, it is often just heard as inconsistent magnitude. It is one of the more difficult things to teach people to ‘hear’ but can be critical in understand in some contexts. In terms of a mix, as I have said elsewhere, so long as you don’t endanger anyone, and you get the results you want, it doesn’t matter but so much.

     Seablade
#15

Oh… that was in this thread. Point still stands obviously:)

  Seablade
#16

Here is definition and example of allpass filter characteristic:

All-pass filter doesn’t change the magnitude of the signal, so, actually, it has a flat frequency response. But it introduces phase shift at some frequency.

So when you just apply allpass filter to the signal, you won’t notice it’s presence (I suppose). But when you try to mix with another signal which will have some leakage of original signal, then you’ll get magnitude distortions.

#17

Yes I know WHAT an allpass filter is, just not the technical details of how it is implemented. I use them in system optimization to correct for frequency dependant phase issues.

  Seablade

Edit for the record, the last paragraph in that article also suggests that an allpass filter can be audible as well, as I suspected.

#18

They’re implemented in the same way the other filters are. Just chain of digital biquad filters with proper coefficients:

#19

Thanks, I will have to read that later, but even the first article you posted confirmed the gist of what we were discussing, that an allpass filter can be audible due to phase artifacts.

  Seablade
#20

Too simplistic to be correct in all cases.
Listen to the example here of same power spectrum, differing phase. Audio examples at the bottom after the explanatory text.