[Ghoostknight]

I am pretty sure my ears would condemn music containing that much energy in the 2.5 kHz band unlistenable. I suggest knowing what you are doing. The signal peaks are constructed from the signal energy on the whole frequency range. If we look at the highest peak, part of it is from 2.5 kHz band. If our music was 10 octaves wide band (20 Hz to 20 kHz) of spectrum that goes down 3 dB/octave (as pink noise does), the each octave would contain 10 % of the whole energy. Of course the peaks in each band are somewhat random, but you would need to increase the level in one octave band massively to compensate the overall level drop of 3.4 dB. The tricky part here is the crest factor of the signal, which makes the math complex.

The point is music doesn't live only in the 2.5 kHz band. Only a fraction of the overall energy lives there. That's why even if your music has peaks of 0 dBFS, the peaks in the 2.5 kHz band are much lower (below -3 dBFS). Dropping overall level by 3.4 dB seems enough safety margin to me.

kinda unsure about this
if you EQ 2,5khz +5db and only reduce overall levels by -3db then it can of course clip with the right kinda music, you are right that most stuff is already "EQ`d" with a slope going down to higher frequencys, but that doesnt mean that there isnt material out there that "would" clip, so the whole "energy" talk is kinda pointless

the only excuse for not turning overalls levels down by atleast the amount boosted is when filters overlap, or you use a low shelf filter before the boost for example

there is also little wrong with turning overalls level down even more (by -1 to -3db) to avoid intersample clipping with certain dacs...

 

[VNandor]

well SPM sounds a bit different then the others, FIR and FFT sound kinda close (but not exactly the same) but there is a clear difference between all of these and IIR which leaves me to think the phaseshift is the obvious difference

Could you upload the output of the EQ with the FIR and the FFT setting (with everything else being the same) using a short snipped of music? These two settings are supposed to not only sound identical but the outputs are supposed to be a bit perfect copy of each other.

 

[71 dB]

kinda unsure about this
if you EQ 2,5khz +5db and only reduce overall levels by -3db then it can of course clip with the right kinda music, you are right that most stuff is already "EQ`d" with a slope going down to higher frequencys, but that doesnt mean that there isnt material out there that "would" clip, so the whole "energy" talk is kinda pointless

the only excuse for not turning overalls levels down by atleast the amount boosted is when filters overlap, or you use a low shelf filter before the boost for example

there is also little wrong with turning overalls level down even more (by -1 to -3db) to avoid intersample clipping with certain dacs...

I don't know what 2.5 kHz music you are talking about. I made a quick example:

Anyway, if people are not willing to believe me that be it. Do what you want and I do what I want.

 

[Ghoostknight]

I don't know what 2.5 kHz music you are talking about. I made a quick example:

of course this doesnt matter if the peak level is at -3db anyway ... but imo that will not be always the case

 

[71 dB]

of course this doesnt matter if the peak level is at -3db anyway ... but imo that will not be always the case

The peak level in my example is -0.1 dB. This whole argument is about whether or not it is enough to drop the overall level "only" 3.4 dB to not cause clipping when you boost 2.5 kHz band by 5 dB. My claim is it is enough and I made the example as an illustration as to why this is the case.

 

[gregorio]

Affected by what?

The “Q” setting when correcting HP freq response curves, which is what we’re discussing.

Of course. Q factor is defined mathematically. When I talk about "ringing" related to Q factor, I talk about mathematical/physical ringing, not audible ringing.

And is it going to be audible, given the amount of gain and the Q setting used to correct HPs?

I am pretty sure my ears would condemn music containing that much energy in the 2.5 kHz band unlistenable. I suggest knowing what you are doing.

The question isn’t what your personal “ears would condemn” and I don’t know what “knowing what you are doing” has to do with it?

If our music was 10 octaves wide band (20 Hz to 20 kHz) of spectrum that goes down 3 dB/octave (as pink noise does), the each octave would contain 10 % of the whole energy.

The pertinent point there is “If”! Sure, when we analyse the freq content of whole tracks we’ll generally see a wide freq band with a pink’ish distribution, although commonly peaking in the mid bass rather than at 20Hz. However, this is not necessarily true of all the individual sections/parts of the track. Some parts will almost certainly have a much smaller frequency range and within that range the distribution could be white’ish, brown’ish or pretty much anything and have peak levels almost anywhere, although peaking near 0dB in the high mids/high freq ranges would be extremely rare. On the other hand, heavy limiting/full-band compression is hardly rare and generally significantly raises the level of the mid freqs relative to lower freqs.

G

 

[71 dB]

The “Q” setting when correcting HP freq response curves, which is what we’re discussing.

Q factor and damping are connected. Higher Q factor means less damping.

And is it going to be audible, given the amount of gain and the Q setting used to correct HPs?

Probably not audible, as long as the Q factor is in reasonable range.

The question isn’t what your personal “ears would condemn” and I don’t know what “knowing what you are doing” has to do with it?

Whether you know what you are doing or you don't. Simple as that.

The pertinent point there is “If”! Sure, when we analyse the freq content of whole tracks we’ll generally see a wide freq band with a pink’ish distribution, although commonly peaking in the mid bass rather than at 20Hz. However, this is not necessarily true of all the individual sections/parts of the track. Some parts will almost certainly have a much smaller frequency range and within that range the distribution could be white’ish, brown’ish or pretty much anything and have peak levels almost anywhere, although peaking near 0dB in the high mids/high freq ranges would be extremely rare. On the other hand, heavy limiting/full-band compression is hardly rare and generally significantly raises the level of the mid freqs relative to lower freqs.

G

I have NEVER encountered music that contains even short sections with spectral content ONLY in the 2.5 kHz range at near 0 dB level. The day I encounter such music samples, I need to update my knowledge. Care to mention one such sample for me to inspect? Thanks!

 

[gregorio]

Q factor and damping are connected. Higher Q factor means less damping.

Which doesn’t answer the question!

Probably not audible, as long as the Q factor is in reasonable range.

When wouldn’t it be in the “reasonable range” when correcting the freq response of HPs? The post I was responding to included this:

Parameters of each filter (which have to be set) include frequency in Hz, amplitude in dB, and Q (damping factor)….
My AMT headphones require, to compensate for this slump in midrange response, one filter with a considerable boost or 4-5 dB in the midrange around 2.5 kHz, and with a moderate Q damping factor of 0.83.

So, “Q (damping factor)” if anything should be Q = Inverse Damping Factor and again, how much is the damping factor going to be affected by a “Q damping factor of 0.83”?

Whether you know what you are doing or you don't. Simple as that.

If it’s as “simple as that”, then why don’t you answer? What does “knowing what you’re doing” have to do with how much 2.5kHz content a track contains at any particular instant?

I have NEVER encountered music that contains even short sections with spectral content ONLY in the 2.5 kHz range at near 0 dB level. The day I encounter such music samples, I need to update knowledge. Care to mention one such sample for me to inspect? Thanks!

What does having content “ONLY in the 2.5kHz range” got to do with it? That’s irrelevant, what’s relevant is if adding an additional 0.6dB-1.6dB of uncompensated 2.5kHz content could cause clipping, however much freq content is in the rest of the track. The answer is “possibly”, especially in the case of heavily compressed/limited tracks and it’s common in many popular and electronic genres to very heavily compress/limit to -0.1dB.

G

 

[71 dB]

Which doesn’t answer the question!

Sorry.

When wouldn’t it be in the “reasonable range” when correcting the freq response of HPs? The post I was responding to included this:

Parameters of each filter (which have to be set) include frequency in Hz, amplitude in dB, and Q (damping factor)….
My AMT headphones require, to compensate for this slump in midrange response, one filter with a considerable boost or 4-5 dB in the midrange around 2.5 kHz, and with a moderate Q damping factor of 0.83.

So, “Q (damping factor)” if anything should be Q = Inverse Damping Factor and again, how much is the damping factor going to be affected by a “Q damping factor of 0.83”?

I wasn't answering to his question. I was commenting on your answer to it. Your answer was good, but I wanted to refine it. Somehow this triggered this bizarre back and forth which I rather ended already. To answer your question, Q factor 0.83 doesn't "affect" damping much, hardly at all in audible sense. I mentioned in my earlier post that Q factors up to 0.8 are generally considered "safe" in audio and 0.83 is hardly at all bigger than that. Hopefully this question is now settled.

If it’s as “simple as that”, then why don’t you answer? What does “knowing what you’re doing” have to do with how much 2.5kHz content a track contains at any particular instant?

It is about knowing that having near 0 dB levels of only 2.5 kHz content is not something you encounter in music.

What does having content “ONLY in the 2.5kHz range” got to do with it? That’s irrelevant, what’s relevant is if adding an additional 0.6dB-1.6dB of uncompensated 2.5kHz content could cause clipping, however much freq content is in the rest of the track. The answer is “possibly”, especially in the case of heavily compressed/limited tracks and it’s common in many popular and electronic genres to very heavily compress/limit to -0.1dB.

G

I have tried to explain why. I made an example illustrating this. If you have content only in the 2.5 kHz range at 0 dB then you get clipping by 1.6 dB or so, but as I have said, this is not something you encounter in music and if you do, it is one crazy Japanese freak track that doesn't suffer much from the 5 seconds of mild clipping this causes. If you have content on other bands, the energy is distributed and the signal level in the 2.5 kHz range is below 0 dB and -1.6 dB (-2 dB to be on the safe side) is enough to prevent clipping.

 

[emba4]

im not a massive fan of EQ myself but thats because i seems somewhat crap at getting it right and the presets often dont agree with me, probably because i think you should likely EQ to your ears not a particular curve.

i have however wondered, generally with most things audio we will prefer analog hardware over software, even down to say a volume control knob, and yet its pretty rare you see hardware equalizers these days (i remember them being common in the 80s) its always done with software now. is there really no benefit to a hardware EQ vs a software one?

 

[VNandor]

Hardware EQs are still sometimes used by producers, mainly because working with it can have a substantially different workflow compared to digital EQs. Less often, it is also used to add "character" (for example low level noise, gentle non-linear distortion, small channel imbalances to add width) to the sound. These could be so tiny that they only add up to something appreciable only if it's used at multiple places in the signal flow. The last thing I can think of is that when working on an analog synth patch, sometimes it can be worth to avoid the low-passing that is required if a digital EQ is used because certain pitch bend effects don't work well with low-passed audio.

By the way, seeing that most mediums are digital, most producers use DAWs to make music and generally use digital effects and digital synths, digital mixers are way more used than analog mixers, I'm not sure "we" actually prefer analog over digital. It looks like people prefer digital over analog.

 

[gregorio]

I wasn't answering to his question. I was commenting on your answer to it. Your answer was good, but I wanted to refine it.

But you didn’t just “refine it”, you contradicted it.

To answer your question, Q factor 0.83 doesn't "affect" damping much, hardly at all in audible sense.

“Hardly at all” in the audible sense or not at all?

I mentioned in my earlier post that Q factors up to 0.8 are generally considered "safe" in audio and 0.83 is hardly at all bigger than that. Hopefully this question is now settled.

Pretty much it “is now settled” and therefore that there was in fact no reason for you to refine/correct my response! In practice, I’ve often used several notch filters at the same time, with very high Q settings and very rarely ever encountered audible ringing.

It is about knowing that having near 0 dB levels of only 2.5 kHz content is not something you encounter in music.

Again, you do not need near 0dB levels of ONLY 2.5kHz band content, you could get clipping distortion (by adding an additional 1.6dB to that band) with near 0dB levels of 2.5kHz band plus some content in other bands. This scenario of having near 0dB levels in the 2.5kHz band can occur if you know what you’re doing! A rimshot for example will produce very significant content at 2.5kHz, more content lower in the spectrum but with heavy compression/limiting, which is likely in some of the genres where a rimshot would be used, we could easily have near 0dB in the 2.5kHz band and an overall -0.1dB level spread across relatively few bands.

If you have content only in the 2.5 kHz range at 0 dB then you get clipping by 1.6 dB or so, but as I have said, this is not something you encounter in music.

But of course we only need to reach +0.1dB to get distortion, I’m not suggesting we would get +1.6dB levels in music. I stated we could get distortion with an uncompensated additional 1.6dB in the 2.5kHz band, if we consider certain sections of some music types/instruments, high levels of compression and factor in ISPs. Lowering the input gain to how much gain we’re adding (or even slightly lower) guarantees no clipping distortion from employing the EQ with any content.

G

 

[71 dB]

I’ve often used several notch filters at the same time, with very high Q settings and very rarely ever encountered audible ringing.

You are correct, notch filters can have very high Q factor setting without audible ringing issues. Try the same "Q"-settings for narrow band boosts and it is another story!

Again, you do not need near 0dB levels of ONLY 2.5kHz band content, you could get clipping distortion (by adding an additional 1.6dB to that band) with near 0dB levels of 2.5kHz band plus some content in other bands. This scenario of having near 0dB levels in the 2.5kHz band can occur if you know what you’re doing! A rimshot for example will produce very significant content at 2.5kHz, more content lower in the spectrum but with heavy compression/limiting, which is likely in some of the genres where a rimshot would be used, we could easily have near 0dB in the 2.5kHz band and an overall -0.1dB level spread across relatively few bands.

But of course we only need to reach +0.1dB to get distortion, I’m not suggesting we would get +1.6dB levels in music. I stated we could get distortion with an uncompensated additional 1.6dB in the 2.5kHz band, if we consider certain sections of some music types/instruments, high levels of compression and factor in ISPs. Lowering the input gain to how much gain we’re adding (or even slightly lower) guarantees no clipping distortion from employing the EQ with any content.

G

Jeesus. I give up...

 

[dmagnan]

But you didn’t just “refine it”, you contradicted it.

“Hardly at all” in the audible sense or not at all?

Pretty much it “is now settled” and therefore that there was in fact no reason for you to refine/correct my response! In practice, I’ve often used several notch filters at the same time, with very high Q settings and very rarely ever encountered audible ringing.

Again, you do not need near 0dB levels of ONLY 2.5kHz band content, you could get clipping distortion (by adding an additional 1.6dB to that band) with near 0dB levels of 2.5kHz band plus some content in other bands. This scenario of having near 0dB levels in the 2.5kHz band can occur if you know what you’re doing! A rimshot for example will produce very significant content at 2.5kHz, more content lower in the spectrum but with heavy compression/limiting, which is likely in some of the genres where a rimshot would be used, we could easily have near 0dB in the 2.5kHz band and an overall -0.1dB level spread across relatively few bands.

But of course we only need to reach +0.1dB to get distortion, I’m not suggesting we would get +1.6dB levels in music. I stated we could get distortion with an uncompensated additional 1.6dB in the 2.5kHz band, if we consider certain sections of some music types/instruments, high levels of compression and factor in ISPs. Lowering the input gain to how much gain we’re adding (or even slightly lower) guarantees no clipping distortion from employing the EQ with any content.

G

My experience differs. With my Toneboosters parametric EQ, in the range 1kHz - 10kHz, there is very appreciable ringing at Q's over 1.3 or so. The audible effect is a sort of smearing and overbrightness especially noticeable on sharp transient sounds like guitar plucks.

 

[bigshot]

Consumer analog EQs are often subject to ground loop problems and lots of spill compared to digital ones.