EQs increase the overall level when cutting out frequencies... WHY!

[Paul Frindle]

Originally Posted by John Suitcase

Well, now that Paul is in this discussion, he can give a better explanation than I could!

My understanding is that DAW meters give an accurate peak reading (value-wise), but the D/A converter can create a waveform that has a higher peak, because of what Paul is talking about with regards to eq. So, the 'weird' effect of a signal level going up because of filtering is neither digital nor analogue, but it's a concern in the digital world because of the clipping that can occur, even when the meters in the DAW show that it's not clipping.

Yes you are right there are 2 issues and they are entirely separate.

The issue about the EQ raising levels even in cut that we were discussing here, applies to any EQ or filter (analogue or digital) and is not a function of the metering.

The other issue about inter sample peaking (from the other quoted thread) is a different problem that applies only to digital systems - and is the result of meters reading sample values only - and not actual reconstructed signal.

[Autotune Prophet]

Originally Posted by Paul Frindle

The issue about the EQ raising levels even in cut that we were discussing here, applies to any EQ or filter (analogue or digital) and is not a function of the metering.

Thanks Paul, John, ryst, T, and everybody for your posts on the "EQ issue"... In the end, I think I just need to stop aiming at the loudest mix possible and leave more headroom so that the "EQ issue" won't be a problem anymore.

However this is hard to do in a genre ("urban-pop music") where loud almost seems to have become THE prerequisit for a "good" mix nowadays. A very good mix being a fair balance between "loud enough for 21st today's average listnener habits" and "pleasant" (vs. harsh).

Originally Posted by Paul Frindle

The other issue about inter sample peaking (from the other quoted thread) is a different problem that applies only to digital systems - and is the result of meters reading sample values only - and not actual reconstructed signal.

Paul, would you also suggest to use a plugin-in meter such as Solid State Logic X-ISM Peak Meter at Kaos Audio in order to bypass this problem?

Furthermore does it mean that I should disregard my daw's own channel meters and only look at the plugin meter?

[Paul Frindle]

Originally Posted by Autotune Prophet

Thanks Paul, John, ryst, T, and everybody for your posts on the "EQ issue"... In the end, I think I just need to stop aiming at the loudest mix possible and leave more headroom so that the "EQ issue" won't be a problem anymore.

However this is hard to do in a genre ("urban-pop music") where loud almost seems to have become THE prerequisit for a "good" mix nowadays. A very good mix being a fair balance between "loud enough for 21st today's average listnener habits" and "pleasant" (vs. harsh).

The trick is to keep levels lower in your tracks and mix buss to avoid clipping (-6dB or so) - and then wind up the gain with your very final limiter at the output of your mix, making absolutely sure that you do nothing else to the signal after the limiter :-)

Paul, would you also suggest to use a plugin-in meter such as Solid State Logic X-ISM Peak Meter at Kaos Audio in order to bypass this problem?

Furthermore does it mean that I should disregard my daw's own channel meters and only look at the plugin meter?

Firstly, normal plug-in meters are not better then the W/S meters. If you keep your levels down around -6dB or so (as above) the problem won't happen in the mixer or plug-ins anyway - so you can relax about your W/S and plug-in meters. The problem then will only occur when you wind up the output of your mix to get high loudness.

The SSL meter (AFAIK) will allow you to see the intersample overs so you can reduce the overall level manually to avoid them. But something like the Oxford Limiter will not only show you the overs, but actually dynamically 'repair' them without loss of overall level or loudness, so this is definitely the best way to go if you can.

[Quinnx]

Hi every one..

Wow this is great info..
But i see some people are still confused..

Here is what i deduce..

It appears the effect people are experiencing is due to phase shift changes.

Example..
Here is a little Workshop of mine for you that may clarify whats going on..

Using your favorite DAW
Take any audio track and place it on track 1
now take the same audio track and place it on track 2
turn the fader down all the way on track 2
invert phase the audio track on track 2

press play

listen to what is being played..
now as you slowly increase the volume of track 2 common sense tells you the overall level combining the two should get louder but
you will notice track 1 starts to become lower in level because of the phase shift effect,
and if you keep going it eventual goes to silence when both tracks are at the same level and eventualy track 2 becomes dominant.

This is how all the different frequencies effect each other in your mix, one canceling out the other and then becoming dominant.

so obviously when you start to decrease the level of one frequency others then become dominant because of the phase shift effect..]

Hopefully this makes sense and clears up the how and why.

[RJHollins]

There is a handy app out there called: VST Analyzer [PC]

You can load in a VST, like an EQ ... and test it to show you its' output curve.

[Paul Frindle]

It's actually much simpler than that - it is due to waveform changes.

All waveforms are a mixture of lots of sine freqs in proportion and phase.

If you lose some of the freqs in the waveform (however they are lost) you risk the peak level becoming higher because of the changed shape of the waveform.

To test this put up a square wave at say 200Hz, then start cutting it with EQ at around 200 - 300Hz.

The loss of the fundamental will increase the peak level by up to 6dB - simply because the top of the waveform is no longer flat!

If you high pass the squarewave the peaks will reach nearly 6dB above the original peak level..

Easy to understand :-)

Originally Posted by Quinnx

Hi every one..

Wow this is great info..
But i see some people are still confused..

Here is what i deduce..

It appears the effect people are experiencing is due to phase shift changes.

Example..
Here is a little Workshop of mine for you that may clarify whats going on..

Using your favorite DAW
Take any audio track and place it on track 1
now take the same audio track and place it on track 2
turn the fader down all the way on track 2
invert phase the audio track on track 2

press play

listen to what is being played..
now as you slowly increase the volume of track 2 common sense tells you the overall level combining the two should get louder but
you will notice track 1 starts to become lower in level because of the phase shift effect,
and if you keep going it eventual goes to silence when both tracks are at the same level and eventualy track 2 becomes dominant.

This is how all the different frequencies effect each other in your mix, one canceling out the other and then becoming dominant.

so obviously when you start to decrease the level of one frequency others then become dominant because of the phase shift effect..]

Hopefully this makes sense and clears up the how and why.

[ramil]

it occurs in both worlds digital and domain. Simple answer = physics. Multiple sound waves with different wavelengths at some frequencies adds up when it has positive value and vice versa and anything in between these two extremes. Basically, when you CUT with EQ certain frequency out of the audio track it reduces its gain at your chosen frequency but it changes phase relationship through entire frequency range from 20hz to 20khz. So when the EQ cut is done lets say for example at 500hz somewhere in high mids different frequencies will be affected as well due to changed phase relationship that may result in gain increase. Hope this helps if unclear let me know I'll try to explain in more understandable way.

[Storyville]

Originally Posted by Autotune Prophet

Thanks John. It makes sense now.

However, the problem seems to be even bigger as what I thought in the beginning... (thank you so much for the link ryst!!! extremely interesting stuff)

It is as if I had only touched the tip of the iceberg with my question.

I also encourage everyone to read the link provided by ryst

Here's an extract... I'm astonished although I have probably only understood half of what's really discussed there:



Does anybody dare to say something on this

That's more about signal recalculation than about EQ. They're related but it's more specific to the way digital mediums work.

[Storyville]

Originally Posted by Quinnx

Hi every one..

Wow this is great info..
But i see some people are still confused..

Here is what i deduce..

It appears the effect people are experiencing is due to phase shift changes.

Example..
Here is a little Workshop of mine for you that may clarify whats going on..

Using your favorite DAW
Take any audio track and place it on track 1
now take the same audio track and place it on track 2
turn the fader down all the way on track 2
invert phase the audio track on track 2

press play

listen to what is being played..
now as you slowly increase the volume of track 2 common sense tells you the overall level combining the two should get louder but
you will notice track 1 starts to become lower in level because of the phase shift effect,
and if you keep going it eventual goes to silence when both tracks are at the same level and eventualy track 2 becomes dominant.

This is how all the different frequencies effect each other in your mix, one canceling out the other and then becoming dominant.

so obviously when you start to decrease the level of one frequency others then become dominant because of the phase shift effect..]

Hopefully this makes sense and clears up the how and why.

By that explanation, I doubt anyone will be less confused.

EQs work on phase shift. If you are hearing anything from using an equalizer is probably the result of phase shift change.

The loss of headroom comes from resonances being produced - it's one of the negative (or sometimes positive) effects of equalizers. The other two being ripple - which is an inconsistency in loss of frequency that blurs transients or can make things sound "fuzzy" - and ring, which is residual voltage powering a the feedback - it causes sustained tones to show up.

Equalization is not strictly the addition or removal of frequencies - it's the calculated re-shifting of frequencies - and there's a number of ways to do it. Using capacitors of different diameters to split signals with a potentiometer on the back end and then resuming them for example - graphical equalization. This results in frequency inconsistencies between the crossover points of the capacitors, and ripple from voltage inconsistencies at the pots. Using time based phase shifts to create peaks and nulls to varying degrees - again creates phase issues simply by the nature of how it works. Turning microscope bandwidths into numerical values and adjusting them mathematically (Fourier equalization) - a lot of noise reduction programs rely on this method, and you can clearly hear the artifacts from that (quantization error).

BUT - if you understand that these are the artifacts, you know what to listen for when assessing an equalizer. Certain EQs have minimal artifacts - the Clariphonic, and the Weiss-Digtial. Some EQs have these effects, but minimize them through using inductors and having a wide slope so you get less phase distortion by distributing it over a wider range (Pultecs!). Some EQs sound awesome because of the resonances and rings they produce (moog).

Also, some eqs handle different frequency areas better. REQ is solid in the midrange, but not the best at the band extremes.

[Paul Frindle]

Originally Posted by ramil

it occurs in both worlds digital and domain. Simple answer = physics. Multiple sound waves with different wavelengths at some frequencies adds up when it has positive value and vice versa and anything in between these two extremes. Basically, when you CUT with EQ certain frequency out of the audio track it reduces its gain at your chosen frequency but it changes phase relationship through entire frequency range from 20hz to 20khz. So when the EQ cut is done lets say for example at 500hz somewhere in high mids different frequencies will be affected as well due to changed phase relationship that may result in gain increase. Hope this helps if unclear let me know I'll try to explain in more understandable way.

This is partially true, changing the phase will indeed change the peak level. But it will still happen with a linear phase EQ due to the loss of a component of the waveform shape. o you need to watch it a bit..

[Storyville]

Originally Posted by Paul Frindle

This is partially true, changing the phase will indeed change the peak level. But it will still happen with a linear phase EQ due to the loss of a component of the waveform shape. o you need to watch it a bit..

That and there's still phase-shift that occurs in a linear phase EQ. It's just smoother because the time-domain shift is even across the whole frequency spectrum, rather than variable at different bands. The same rules apply, but there's no crossover phase funkiness.

[ramil]

you touched very interesting topic. I am trying to "tune" my ears at artefacts and side effects that causes equalisation. Are there any particular audio examples with explanations "where to listen" for artifacts available online? Obviously by trying to eq signal by myself with different eq's would cause side effects that I want hear (just for learning purposes) to be more efficient at spotting it. But still it would be really helpful to find some info with audio examples

Originally Posted by Storyville

By that explanation, I doubt anyone will be less confused.

EQs work on phase shift. If you are hearing anything from using an equalizer is probably the result of phase shift change.

The loss of headroom comes from resonances being produced - it's one of the negative (or sometimes positive) effects of equalizers. The other two being ripple - which is an inconsistency in loss of frequency that blurs transients or can make things sound "fuzzy" - and ring, which is residual voltage powering a the feedback - it causes sustained tones to show up.

Equalization is not strictly the addition or removal of frequencies - it's the calculated re-shifting of frequencies - and there's a number of ways to do it. Using capacitors of different diameters to split signals with a potentiometer on the back end and then resuming them for example - graphical equalization. This results in frequency inconsistencies between the crossover points of the capacitors, and ripple from voltage inconsistencies at the pots. Using time based phase shifts to create peaks and nulls to varying degrees - again creates phase issues simply by the nature of how it works. Turning microscope bandwidths into numerical values and adjusting them mathematically (Fourier equalization) - a lot of noise reduction programs rely on this method, and you can clearly hear the artifacts from that (quantization error).

BUT - if you understand that these are the artifacts, you know what to listen for when assessing an equalizer. Certain EQs have minimal artifacts - the Clariphonic, and the Weiss-Digtial. Some EQs have these effects, but minimize them through using inductors and having a wide slope so you get less phase distortion by distributing it over a wider range (Pultecs!). Some EQs sound awesome because of the resonances and rings they produce (moog).

Also, some eqs handle different frequency areas better. REQ is solid in the midrange, but not the best at the band extremes.

[Paul Frindle]

Originally Posted by Storyville

That and there's still phase-shift that occurs in a linear phase EQ. It's just smoother because the time-domain shift is even across the whole frequency spectrum, rather than variable at different bands. The same rules apply, but there's no crossover phase funkiness.

Ah ok - I understand :-)

You are confusing phase shift in comparison with another source (i.e. relative timing shift), and phase shift with itself (differential timing of freqs).

A phase linear EQ does not exhibit a timing shift between freqs of it's own signal - a conventional one does.

But the point I'm trying to make is that it's the loss of a constituent part of the spectrum that formed the wave shape which is the primary reason for the increase in peak value.

SO it will happen whether you use a phase linear EQ as well..

So for instance a square wave is made of a fundamental and a (theoretically infinite) series of odd order harmonics. If you remove the fundamental and are left only with the harmoncs it's not longer a squarewave.

Since the squarewave has the highest possible energy for peak ratio (same as DC) - then anything else will have a higher peak value..

[ramil]

bob katz covered this topic in his book as far as I remember. Good thread to read

[Storyville]

Originally Posted by Paul Frindle

Ah ok - I understand :-)

You are confusing phase shift in comparison with another source (i.e. relative timing shift), and phase shift with itself (differential timing of freqs).

A phase linear EQ does not exhibit a timing shift between freqs of it's own signal - a conventional one does.

But the point I'm trying to make is that it's the loss of a constituent part of the spectrum that formed the wave shape which is the primary reason for the increase in peak value.

SO it will happen whether you use a phase linear EQ as well..

So for instance a square wave is made of a fundamental and a (theoretically infinite) series of odd order harmonics. If you remove the fundamental and are left only with the harmoncs it's not longer a squarewave.

Since the squarewave has the highest possible energy for peak ratio (same as DC) - then anything else will have a higher peak value..

Right. But isn't one of the principal means of manipulating tone feeding a time-delay signal back into itself, ie negative feedback or positive feedback loops? To my understand, a linear phase EQ still functions on this principal. The only difference is that the time shift is the same so the overall signal doesn't phase with itself, only with the feedback signal.