We all know that linear phase is important in the masterchain, but how important is it on the individual tracks?

Any thoughts on this?

Thanks

/R

I'm curious where you get the idea that LP is "important?"


DC

Just a victim of the hyped Linear Phase EQ's advertisement i guess Maybe its not that importand after all?

thanks
/R

I go through entire projects, record to master without even using a digital EQ, much less linear phase.
Actually, too much linear phase EQ is a bad thing because it has a pre-ringing effect that appears to soften transients, which is already an epidemic in the loudness war.

I'll say though, linear phase EQ can be great when you need a narrow cut (Q8 or so) to reduce resonance on a track.

Would you mine explaining more about the "pre-ringing" effect? Ive never heard of that one When should i most noticably hear this effect?


And as i tend to do a lot of surgical cuts then a linear phase EQ would be better? Why is that?

Sorry for stupid questions but if you never ask.....


thanks again
/R

OK, most digital EQ is based on delay. Minimum phase EQ uses just a few delay lines with feedback and this is called "infinite impulse response". There's phase shift involved, which is a function of how steep the filter is and there's also ringing. However, this ringing takes place after a given sound. So if there's a snare drum hit or something like that, there will be some ringing but the decay of the snare drum will cover most of the effect of it and sounds fairly natural. Most linear phase EQ uses finite impulse response. Instead of a few delays with feedback it manually calculates out every single delay needed within a given time window and applies it bidirectionally. FIR filters require hundreds or even thousands of individual delays in extreme cases. For every delay that takes place after an event, there's another that takes place before the event. To further explain, if the first "echo" in the filter is at -60dB to the original event, there with be a -60dB pre-echo. This cancels out the apparent phase shift but now you have a pre-echo which doesn't happen in nature and can thus give the appearance of softened transients & strange ambiance.

this was the best explanation ive seen in a long time! easy to understand, even for me

In another forum someone said that Linear Phase EQ's has a pre and post ringing. Is this true then?
They also said that generally minimum phase EQs are better at sharp surgical cuts/boosts than linear phase EQs, which do better with wider Qs and gentler settings.

Is this some general rule to have? Kind of the opposite to what you said earlier in the thread



thanks again for taking time!!
/R

Correct. Minimum-phase has only post ringing.

I think it's impossible to generalize other than the pre-ringing of linear phase is more audible the lower in frequency you use it, so HPF is probably the worst application of LP.

How often are 'surgical' changes done in mastering anyway? Anything like a Q of 3 would make me wonder unless it's to notch out hum or something.

Q of 10 should require a lead apron.


DC

Well, it's important enough for the anti-alias filter on the A to D converter, but...

Most Lp eq's create large amounts of latency and can use a lot of the computer processing power. They're not that feasible for individual tracks.

so in that case, LP is worse? I thought that LP cancelled all the ringing out.
But i still dont get it, if the post-ringing is a cause of a phase problem, and a linear phase EQ should correct this phase problem, shouldnt the post-ringing disappear?
Hard stuff for a layman like me


thanks for taking time!!
/R

It doesn't disappear, linear phase has twice as much. Because it rings both before and after the signal.

Ringing is from the Q or filter slope, higher = more ringing, it does not come from the type of filter used.


DC

fwiw - I nearly always use minimum phase eq's - but in the cases where I do use linear phase it's either to do very gentle slope broad shelf cuts or boosts to tilt a mix darker or brighter for things like jazz or acoustic chamber music where the "sheeniness" and "plasticness" of the pre-echo effect is not a disadvantage and where maintaining the existing stereo image is requested

- or to do very narrow Q notch cuts for things such as removing feedback from a live concert recording - as for this purpose it often sounds to my ear much less obvious and intrusive than using a minimal phase eq.

As always with these things ymmv - but it's funny that one of my uses is opposite the advice given to you by some others.

Best regards,
Steve Berson

Interesting! So why use LP EQ's at all then? I mean if they are twize as bad?

/R

Minimal Phase EQ's are all the ones that are NOT Linear Phase or?

So LP i guess is required to maintain the stereo image and not letting it fall "out of phase"? Could this be true on individual tracks as well?

But as i read, shouldnt a higher q setting bring more ringing with a LP EQ?


Great info in this thread for me!!! Thanks!

/R

thumbsup.

If you want to hear what pre-ringing sounds like in exaggeration you can try the Hum Removal filter in iZotope RX:
Set it to 'Linear-phase filters' and apply it at least twice with realy sharp Q and high attenuation per band.
Especially with something like a slow piano solo piece you'll start to hear the notes 'fade in' , kind of like reversed reverb.

bests,

The total amount of ringing is actually the same, it's just distributed to the beginning and end of a transient, instead of just the end.

Some people like them, but I've never been a fan. The stuff about improved imaging in linear-phase I don't get either, fwiw.


DC

I would say that's the worst application for a linear phase. High Q = lots of pre-ringing.

I only like linear phase eqs with low Q to slightly adjust the overall balance of a source.

Sample? .

I don't have that track at hand now...
But you can easily demo RX, just download it from iZotope.com. It's great for a lot of things, so good to know anyway!

bests,

So Minimal Phase EQ's are all the ones that are NOT Linear Phase or?

Are for example the URS series and the Waves REQ's "Minimal Phase"?


thanks
/R

The DDMF LP10 EQ is a quick easy demo for this.

You can switch the filters between linear and minimum phase with the click of a button, so load it onto to a busy track, put a sharp Q on a low band with a lot of gain and listen to the differences.

"---- Donk" / "sssSWWoosh"

Remember, phase shift and ringing have nothing to do with each other. Digital EQ is based on delay so there HAS to be ringing somewhere. LP EQs just cancel out the phase shift by using bidirectional delay instead of just post delay. The side effect is that there's now pre & post ringing.


Here's an example I slapped together. The top is the raw square wave (1KHz) demonstrating how the image should look with no filtering. Below that is a 12dB cut at 3KHz using a minimum phase EQ. You see how the waveform is no longer symmetrical? This shows there's a phase shift present. There's much better ways of showing phase shift but for simplicity, I'll leave it at that. At the bottom is the same EQ done with linear phase filtering. This one is a bidirectional IIR filter FYI. That means it applies half the filtering as minimum phase going forwards, then again going backwards to cancel out the phase shift. My scaling is very crude but you can see the phase shift is no longer present and the ringing is now half its amplitude but takes place both before and after the event. BTW, if I slid the filter down to 1KHz, the ringing was 3x the time of the 3KHz filter.



Well, I should clarify that I mean extremely high pitch ringing like 10Khz. There's much less delay involved with high frequencies so less ringing. But you're right, if there was a single note that rang on an acoustic guitar or something (usually 200-250Hz) I would generally want a minimum phase EQ with a Q of maybe 3.

Hope that helps.

It's "minimum" phase, not "minimal." Non linear phase EQs are indeed minimum phase. The "why" of the term "minimum phase" is beyond the scope here, so lets stick to linear/non-linear phase EQ.

Minimum phase, or IIR filters (Infinite Impulse Response) suffer from group delay, where certain groups of frequencies lag behind others. Not all the signal arrives at the same time. This phase distortion is a necessary part of the equalization in these types of filters.

Linear phase, or FIR filters (Finite Impulse Response) fix this problem so that all frequencies arrive at the same time. Phase distortion is not eliminated, it is simply made linear, meaning that all frequencies are delayed by the same amount. In other words, the signal has a small delay. However, with these filters you suffer from the added problem of pre-echo, aka pre-ringing.

As some less common examples, you technically could create a minimum phase FIR filter, and you can use a special dual-pass IIR process to make linear phase EQ (one pass for the EQ, and a reverse pass to correct for the phase response), but these are the exception, not the rule.

I did some testing on a bass with an EQ that could switch from Linear to non-Linear. Yikes, what a difference Totally different sounds.

When i used a HP filter it sounded quite weird but the bass was in pitch in Linear mode. When i switched to non-Linear the pitch was altered and lowered just a tiny tiny tad, but it sounded better. Well, out of pitch is never a good thing IMO, even if its just a tad. Using a notch filter in the low area gave me the same result. At 10k in Linear mode it sounded not that bad but also here, in non-linear mode, it altered the pitch of the bass :(

Its hard to describe the nosie that came from the HP and lowcut in Linear mode but i could call that ringing. It was like it was just a tad before the sound and after whilst in non-linear mode only after. It was more obvious at certain frequencies.

so where do i go from here?


Just read at meldaproduction about EQ's:

"Minimum phase equalizers basically sound the same and differ mostly in their user interface. Linear-phase equalizer algorithms are generally much more complicated and this can usually change or influence the sound. Whether or not this is a good thing in a particular case, is up to you."

This cant be the truth, can it?


many thanks again for taking time with my annoying questions

/R

It does always depend on what you want to do, that's for sure. I'd still say that linear phase sounds cleaner as long as you don't go to high Q values. The detuning effect that you heard in minimum phase mode is a consequence of the different frequency distributions being played back at slightly different times. Of course it would be nice if we could have the best of both worlds (no frequency smearing and no preringing). Every now and then I try to develop something like that but so far without success... for the time being, as mentioned above, LP10 gives you the maximum flexibility that's currently available (sorry for mixing in this little piece of advertisement heh ).

Some of it is true. Linear-phase EQs do tend to be more complicated, but to state that minimum-phase EQs are basically all the same is pure bollocks. Take the Mackie D8B. The EQs are worthless! Same as the EQs in iPods and Sansas. However, on a proper professional plugin or hardware unit, they can be quite good. Bear in mind also that analogue parametric EQs are minimum-phase and they can vary vastly from one to the next.

I will state that poorly implemented FIRs sound MUCH worse than poorly implemented IIRs.

Have you ever heard this effect on an analog equalizer?


DC

yeah, ive tried waves API and SSL EQ's today along with URS. The same. Ive noticed that Linear phase also tend to smear the attack :( but it sounds better pitch-wise Much more natural.

/R