EQ plug-ins... I just don't hear it

[jdtrbn]

Originally Posted by Nishmaster

There is also another weird myth that I've seen propagating lately around here, and that's that somehow some eq's can exhibit more or less phase change than others. The way eq works is by altering the phase relationship between the original signal and the derived signal and adding them back together, thus producing a frequency specific change in gain.

Well, different EQs can have different phase behaviour. You can always have an all-pass filter or push the low frequencies forward like some exciters do. But after that you had better not add it to the original signal, of course.

[Paul Frindle]

Originally Posted by Nishmaster

Ridiculous. The very fact that something cancels completely in a null test means by definition that it cannot possibly sound different.

Yes - if you arrive at exactly the same response, if all else is equal it cannot sound different. This is a logical reality. The point is how you reach your setting during creative use.

There is also another weird myth that I've seen propagating lately around here, and that's that somehow some eq's can exhibit more or less phase change than others. The way eq works is by altering the phase relationship between the original signal and the derived signal and adding them back together, thus producing a frequency specific change in gain.

EQ processing falls into 2 main categories infinite impulse response (IIR - where signal is fed back through a few weighted delays) and finite impulse response (FIR - where signal is fed through a great number of weighted delays but never fed back again). Combinations of both are possible to save processing under some conditions

The sort of EQ that most closely matches analogue (and the real world) is the IIR feedback versions. These produce the expected phase shift wrt freq response and gain similar to analogue EQ and other freq sensitive stuff in the natural world.

A more recently offered variant of EQ is the FIR feed along type model, which can produce arbitrary phase/freq responses (any phase relationship is possible) - with a correspondingly unnatural/novel impulse response over time. So called phase corrected (or phase linear) EQs are a version of this, most often consisting of a conventional IIR followed by an FIR to correct the phase shift (kind of doing the IIR in reverse). Such EQs must have a very significant latency because gaining a correct freq response at LF and high Q requires an impulse response that lasts a very long time - with the peak ages after the signal stimulus. They can also sound distinctly odd because the sound of the preamble precedes the sound that caused it - and of course no amount of delay compensation can fix this softening effect on the events and timing within the program.

Given two equal eq settings (Q, Center Frequency, Cut/Boost), any two eqs will produce an identical phase response.

That depends on what you actually mean; same settings, or same responses? If you follow the links I posted you can see that for the different types of EQ on offer the actual response for any gain and Q setting is in fact different between the EQ types. In fact what the 'Q' value actually is presents a bit of a conundrum for an EQ that passes all freqs, since a 'Q' value is related to the bandwidth of a natural filter that aims for no output at DC and extreme HF.

The styles of gain/Q dependency are intentionally provided to give character to the operation of the EQ and provide the correct 'feel' and facility depending on what the job in hand actually is. For instance the type 2 EQ with it's sharper response in cut is ideally suited for drum EQ where you want to tightly control resonances, but still want to accentuate overall responses in a single EQ style.

One of the significant changes I made to the SSL E-series EQ to form the G-series version was to change this gain/Q dependency, so that a larger bandwidth was boosted (or cut) at lower gain settings. This gave the impression of a gentler more forgivingly musical EQ when operated by the user and encouraged a different artistic style. This was more in line with the changing fashions of the time which were moving away from the harsher sounds popular in the early 1980 - and the most significant cause of the characteristic sound of the console was in fact the gain/Q dependency of the EQ section and how people were using it.

Of course, in a digital world, the steep sampling filter will introduce ringing and pre-echo, but with many digital eqs oversampling internally, I would say that this problem is mostly academic.

The ringing has nothing to do with sampling rates - it is a direct result of the freq/phase response and the associated impulse response. An analogue EQ with the same response will do the same thing - as would a collection of swinging pendulums or acoustic cavities coupled to give the same roll-off :-)

[Paul Frindle]

Originally Posted by jdtrbn

Well, different EQs can have different phase behaviour. You can always have an all-pass filter or push the low frequencies forward like some exciters do. But after that you had better not add it to the original signal, of course.

If only this were so :-) You cannot use a conventional all-pass section to correct for an EQ phase response, since an all-pass section is in fact a polarity inverter that simply progressively 'uninverts' as the signal freq changes.

Inverting the signal makes it sound quite different before you EQ anything at all :-(

The only way to make a phase linear EQ is to produce the correct impulse response using large amounts of time in the process. This involves synthesising the inverse of the normal exponential decay of a ringing filter over time, using lots of delays with their contributions to the output appropriately weighted.

[Mr.HOLMES]

When I started mixing all I knew was the internal Plug In stuff of Logic 6.
A friend of mine who is a great engineer came over and advised me to buy the waves SSL stuff. If I wont do it he will hit me

I bought it and I said from the first moment on WOW this EQs are sounding nearly always musical.....

So I love them and use them in nearly every mix.
And you know when you drink Champagne all the time you forget about that light bulb moment and you ears are so much blessed.

I first have to switch back to an Logic poor AU EQ to feel and hear again AHHHHHH...AMEN my SLL simulation GOD

For real maybe you can find EQs which are nearly the same.
The SONNOX EQ vs Sonalksis in my ears sounding the same.

[Ethan Winer]

Originally Posted by Paul Frindle

if you arrive at exactly the same response, if all else is equal it cannot sound different. This is a logical reality.

Indeed. Yet you will find people who claim they "hear" a difference anyway, and some even believe different hard drives sound different playing the same Wave files. Go figure.

--Ethan

[CoteRotie]

Originally Posted by Ethan Winer

Indeed. Yet you will find people who claim they "hear" a difference anyway, and some even believe different hard drives sound different playing the same Wave files. Go figure.

--Ethan

A little off-topic, but my all-time favorite is the claim that plugging the DAW in to the AC socket with a special AC cable will make everything sound better.

Forget the yards of cheap Romex and all the utility transformers the power passes through before it gets to the magic cable.

Forget the power suppy that regulates and filters the AC down to low ripple DC. (The ripple at this point is never AC line ripple but switching frequency ripple.)

Forget all the balanced circuitry and filtering to reject anthing that gets through the power supply. (Not to say you can't have AC hum, but it doesn't get in through the DAW's AC cord.)

Buy this $90 AC cable and everything will sound better.

I'm in the wrong business......

Regards,

John

[Nishmaster]

Originally Posted by Paul Frindle

Yes - if you arrive at exactly the same response, if all else is equal it cannot sound different. This is a logical reality. The point is how you reach your setting during creative use.

I agree 100 percent. My point was more that given two waveforms that null perfectly against each other, you cannot possibly arrive at any other conclusion than they must by definition also sound identical. This comment was more aimed at that particular myth, which also had seen a lot of attention in the recent DAW sound thread. The aesthetics and subjective "feel" of the processor are oftentimes the most important aspect indeed.

...some very informative information from Paul Frindle...

Yes, and I am aware of the various approaches to digital EQ, although I could have never myself explained them so succinctly.

The odd transient behavior of minimum/linear phase FIR feed along type eqs is a large amount of the reason why I try to avoid them. There is no analogue equivalent, yet albums have been mastered for years without digital linear phase eqs.

That depends on what you actually mean; same settings, or same responses?

In an attempt to not sound overly technical, I think I may have been misleading. My point was that given an identical transfer function between two eqs, you will necessarily have the same phase response. One cannot be "better desgined" to be "less phasey," as it were. Once again, corrective reduction of errant phase response in the aforementioned digital linear phase eqs is the exception.

One of the significant changes I made to the SSL E-series EQ to form the G-series version was to change this gain/Q dependency, so that a larger bandwidth was boosted (or cut) at lower gain settings. This gave the impression of a gentler more forgivingly musical EQ when operated by the user and encouraged a different artistic style. This was more in line with the changing fashions of the time which were moving away from the harsher sounds popular in the early 1980 - and the most significant cause of the characteristic sound of the console was in fact the gain/Q dependency of the EQ section and how people were using it.

Fascinating. I had (apparently erroneously) assumed that the vast majority of high quality modern parametrics were based mostly on a state variable filter approach, thus retaining identical Q regardless of gain. Was the G series more of a single variable gyrator per band type topology? Makes me rethink my current equalizer design.

The ringing has nothing to do with sampling rates - it is a direct result of the freq/phase response and the associated impulse response. An analogue EQ with the same response will do the same thing - as would a collection of swinging pendulums or acoustic cavities coupled to give the same roll-off :-)

Right. In the quote, I don't reference sampling rate but the steep Nyquist filter which would introduce ringing. A properly designed digital eq with oversampling and a first order filter far above audio would reduce these artifacts. These artifacts are of course coupled directly to the frequency/phase response and bandwidth, as you state. An analog eq doesn't have such steep filters (not to say analog is better or worse).

[Ethan Winer]

Originally Posted by CoteRotie

Buy this $90 AC cable and everything will sound better.

$90 eh? How about up to $5,000:

Shunyata Research

Of course they receive only excellent reviews!

--Ethan

[Paul Frindle]

Originally Posted by Nishmaster

In an attempt to not sound overly technical, I think I may have been misleading. My point was that given an identical transfer function between two eqs, you will necessarily have the same phase response. One cannot be "better desgined" to be "less phasey," as it were. Once again, corrective reduction of errant phase response in the aforementioned digital linear phase eqs is the exception.

Yes this should be true unless added phase correction processing is employed.

Fascinating. I had (apparently erroneously) assumed that the vast majority of high quality modern parametrics were based mostly on a state variable filter approach, thus retaining identical Q regardless of gain. Was the G series more of a single variable gyrator per band type topology? Makes me rethink my current equalizer design.

In fact whatever kind of architecture you use for the filter section you can still manipulate the effective Q depending on the gain setting.

The E and G-series EQ were in fact Wein bridge designs. These are far more sensitive to potentiometer tracking, but are quieter than a state variable design. The EQ could have just as easily been realised in either topology, but noise was a big issue with a console of this size and quality.

In fact as a point of history, an argument did break out when people at the time worried that a state variable should be used basically because other EQs at the time employed it - but of course the architecture itself had nothing to do with the sound apart from noise performance.

And even more significantly, one of the other manu's highly popular design overloaded 6dB prematurely at low Q settings (a result of the state variable architecture), and if fixed this would have raised the noise for their whole EQ by another 6dB above what it already was :-( Something definitely to be avoided.

Right. In the quote, I don't reference sampling rate but the steep Nyquist filter which would introduce ringing. A properly designed digital eq with oversampling and a first order filter far above audio would reduce these artifacts. These artifacts are of course coupled directly to the frequency/phase response and bandwidth, as you state. An analog eq doesn't have such steep filters (not to say analog is better or worse).

Ok - this filter is required to down sample after an oversampled filter? There is absolutely no advantage in up sampling for a filter - particularly when you have to down sample again to pass the signal on. Such a design is too complex and not optimal. None of the EQs I have helped to make for the OXF-R3 or Sony plug-ins use this kind of architecture.

[Nishmaster]

Originally Posted by Paul Frindle

The E and G-series EQ were in fact Wein bridge designs. These are far more sensitive to potentiometer tracking, but are quieter than a state variable design. The EQ could have just as easily been realised in either topology, but noise was a big issue with a console of this size and quality.

In fact as a point of history, an argument did break out when people at the time worried that a state variable should be used basically because other EQs at the time employed it - but of course the architecture itself had nothing to do with the sound apart from noise performance.

And even more significantly, one of the other manu's highly popular design overloaded 6dB prematurely at low Q settings (a result of the state variable architecture), and if fixed this would have raised the noise for their whole EQ by another 6dB above what it already was :-( Something definitely to be avoided.

Thanks for the anecdote. It's always fascinating (to me, anyway) to hear how classic devices came about and why they ended up the way they did. I can imagine having the audio proceed through another 2-3 opamps (per band) would not be especially beneficial, especially distributed over the whole console. I had not considered the possibility of inadequate headroom at low Q settings. Looks like some more bench testing is in order. I do like the idea of constant-Q regardless of gain, though. I'm not seeing how that would be possible in a Wein bridge filter, really (then again, I'm a comparative amateur).

Ok - this filter is required to down sample after an oversampled filter? There is absolutely no advantage in up sampling for a filter - particularly when you have to down sample again to pass the signal on. Such a design is too complex and not optimal. None of the EQs I have helped to make for the OXF-R3 or Sony plug-ins use this kind of architecture.

You're absolutely right, it would make no sense. I hadn't really thought that line of thinking through much, and looking at it now, what a complete "duh" moment.

I'm wondering now, though, how it is handled in the corner case of having the bandwidth of a filter band exceeding Nyquist (i.e. low Q boost at 20k)

[dannygold]

Originally Posted by Nishmaster

Ridiculous. The very fact that something cancels completely in a null test means by definition that it cannot possibly sound different.

I know that. You know that. Others don't. In this thread someone claimed DP sounded different when you bounced thru a master fader, not thru a master fader and thru and AUX. I tried the exact test he recommended (and recorded via a bus in realtime for good measure) and the results were the same as proved via phase inversion tests. He "doesn't believe" in null tests I think. Says it still sounds different.

Will Pyramix be the end of my DAW Holy Grail Quest ?

[CoteRotie]

Originally Posted by Ethan Winer

$90 eh? How about up to $5,000:

Shunyata Research

Of course they receive only excellent reviews!

--Ethan

PT Barnum would be proud!

This way to the amazing egress.....

John

[Paul Frindle]

Originally Posted by Nishmaster

T

I'm wondering now, though, how it is handled in the corner case of having the bandwidth of a filter band exceeding Nyquist (i.e. low Q boost at 20k)

Well running at 44.1KHz it can't output signal frequencies above 20KHz or so - but it can have the response of an EQ set with centre freqs above 20KHz - and produce the correct response in the allowable range you can hear.

The GML emulation we made does this as the centre freq max for the GML box was 26KHz.

Sonnox Oxford Plugins

[Nishmaster]

Originally Posted by dannygold

I know that. You know that.

I misunderstood what you said and took it to mean that you didn't believe in null tests.

Please accept my apologies.

[jdtrbn]

Originally Posted by Paul Frindle

If only this were so :-) You cannot use a conventional all-pass section to correct for an EQ phase response, since an all-pass section is in fact a polarity inverter that simply progressively 'uninverts' as the signal freq changes.

Inverting the signal makes it sound quite different before you EQ anything at all :-(

The only way to make a phase linear EQ is to produce the correct impulse response using large amounts of time in the process. This involves synthesising the inverse of the normal exponential decay of a ringing filter over time, using lots of delays with their contributions to the output appropriately weighted.

I don't quite understand. How exactly does the sound change when you invert it? Should this apply to mic amplifiers' phase buttons too?

Tritone Digital has a plug-in called Phasetone that changes the phase but not the frequency response unless you mix it with the original signal. Couldn't you put this kind of a filter after any EQ? Or does it have to be some FIR tweak?

By the way, "inverting the signal makes it sound quite different" alone doesn't matter, EQs are allowed to sound different.

[dannygold]

Originally Posted by Nishmaster

I misunderstood what you said and took it to mean that you didn't believe in null tests.

Please accept my apologies.

No, I definitley do! Please go to this thread an "vote" for their validity. So far it's two to one against the validity of phase tests, me being the one "for" their validity.

Will Pyramix be the end of my DAW Holy Grail Quest ?

[3rd&4thT]

Friedemann Tischmeyer on EQs:

"The differences and limitations of native and DSP-aided EQs are most obvious when these devices have to cope with significant increases between 14 and 20 kHz. You should take the trouble to compare your EQs with each other and also possibly with quality external devices. You will be surprised at the differences. Especially in the higher regions, even a mediocre tube EQ simulation can sound harsh, digital, and quite simply, awful. A further indication of the quality of an EQ is its sound behavior at extreme bell filter settings. Comparisons will help you to achieve better results, because you will then be able to restrict yourself to a few well-chosen tools."

3rd&4thT

[jdtrbn]

I've watched the DVD . Unfortunately I haven't been able to hear the differences when the frequency responses are matched. If you have experiences of this, I'd like to know.

Originally Posted by 3rd&4thT

Friedemann Tischmeyer on EQs:

"The differences and limitations of native and DSP-aided EQs are most obvious when these devices have to cope with significant increases between 14 and 20 kHz. You should take the trouble to compare your EQs with each other and also possibly with quality external devices. You will be surprised at the differences. Especially in the higher regions, even a mediocre tube EQ simulation can sound harsh, digital, and quite simply, awful. A further indication of the quality of an EQ is its sound behavior at extreme bell filter settings. Comparisons will help you to achieve better results, because you will then be able to restrict yourself to a few well-chosen tools."

3rd&4thT

[AnthonyRochester]

So, I'm only using the Digirack EQ plugin in ProTools LE. Is that bad?

[amenbrother]

Originally Posted by Paul Frindle

Yes - that is possible. But the point is that nulling in this way tells you little if anything about the product itself.

A whole discussion like this once broke out concerning actually nulling the OXF-R3 against ProTools or some such! This was also a pointless exercise that told people nothing of value to them. The whole concept of these applications could not have been more different.

As a reverse example - the fact that our latest DSM plug-in will null to itself completely to silence was actually thought by some people at one point as a potential fault! The point being that almost all other apps these people had would not even null with themselves (apparently) - and it was therefore thought that the DSM could not have actually been working!

But the deeper point with all that is - what are people actually expecting to find out?

For instance if you took a couple of LA2As or Fairchild tube compressors and did the same test, you would be hard pressed to get even a 12dB null. What would this tell you about these units? Would they be any the less classic, somehow invalidated and have their long standing reputations for creativity quoshed?

Word

You dont buy gear cause it is able to null I hope. Youll buy it cause it sounds good, has charachter, are precise, do what you want or what ever.

On a sidestick.
Its strange that the most lousy singers can sound really good, whats up whith that.

Cheers

[Companda]

The only way to make a phase linear EQ is to produce the correct impulse response using large amounts of time in the process. This involves synthesising the inverse of the normal exponential decay of a ringing filter over time, using lots of delays with their contributions to the output appropriately weighted.

Not entirely true to my knowledge. You can use 2 tracks of the same content(duplicate), apply the EQ to one then reverse the other and apply the same EQ too. You then reverse it back and there's your phase linear EQ. However you can't really adjust too much.

[tamasdragon]

Originally Posted by beebay007

So, I'm only using the Digirack EQ plugin in ProTools LE. Is that bad?

No, they are fine and very underrated.
Tamas Dragon

[StudioKing]

Originally Posted by Clem Snide

Mmm, padded with foam: nice! Those low frequencies will be quakin' in their boots.

My room is also padded with auralex, what people can't see is the huge bass trapping behind and in the ceiling.

[Nu-tra]

I think the eiosis air eq is a good candidate

[3rd&4thT]

Originally Posted by jdtrbn

I've watched the DVD . Unfortunately I haven't been able to hear the differences when the frequency responses are matched. If you have experiences of this, I'd like to know.

Just for the helluvit, using the same source material, I tried setting +10 at 20K with minimum Q, and some EQ plugins came through fine and others sounded like ass, so I pulled the latter off my hard drive and haven't missed them at all.

I don't know how scientific this whole thing is, but it sure helps speed up the decision making process.

3rd&4T

[Paul Frindle]

Originally Posted by Companda

Not entirely true to my knowledge. You can use 2 tracks of the same content(duplicate), apply the EQ to one then reverse the other and apply the same EQ too. You then reverse it back and there's your phase linear EQ. However you can't really adjust too much.

Ha - if you think about it, this is exactly the same thing! Two impulse responses of a normal EQ that are overlapped - but with one in time reverse wrt the other - their freq responses add but their phase shifts cancel (because one has the reverse impulse response wrt music from the other) :-)

Thats the exactly same concept as we use we make a digital phase linear EQ - it's just that we can't reverse the music playing in real time (obviously) - so we construct the reverse impulse response of the first conventional EQ over time - and tack it on in advance :-)

So we end up with two sections - the first one is an FIR with the impulse of the EQ in reverse - and the second one is the conventional IIR EQ with the normal impulse response. Each effectively does half the freq response you want to get in the end (cos they are in series) - and one has the reverse phase characteristic of the other - so the overall phase cancels :-)

[Companda]

if you think about it, this is exactly the same thing!

Yeah, that's what the guy who explained it to me said. I was just saying that the exact programming wasn't the only way.

I feel intelligent though for regurgitating somebody else's advice .

BTW thanks for posting all that stuff about IIR and FIR. That's exactly what this engineer told me but it's not very easy information to store being told it once. Now I've got regular access to it(don't edit or delete it!!!!)
Cheers.

[ProducerBoy]

I mixed a song using just sonalksis plugs, the just sonnox plugs, and then other plugs from other companies.
There is a very noticeable difference between them all.
Between sonalksis and sonnox, I felt the sonnox were better equipped. However, they both sounded great. Sonalksis have their own subtle color, but I found the sonnox to be more transparent.
To my taste, I preferred the sonnox. They were much easier for me to work with and gave me the results I was looking for. (Big tip of the hat to the designers)
I'd say if you can't hear the difference that is there, then try them based on how well they fit into your workflow.

[jdtrbn]

Originally Posted by ProducerBoy

Sonalksis have their own subtle color, but I found the sonnox to be more transparent.

Seems to be a popular opinion. Maybe it has to do with that Sonnox allows bigger Q values.

I wonder if the so-called colour can be seen in measurements or if it's just that the controls points the user to the direction of certain kind of settings. For example, URS N-series has a filter shape that looks like a pancake with small Q and loses it's "peakiness" very quickly when decreasing it from the maximum. A-series leaves the area outside the peak practically untouched when the gain is increased (if I remember right). Without an analyzer the frequential change itself can surely be perceived as colour.

[ProducerBoy]

Originally Posted by jdtrbn

Seems to be a popular opinion. Maybe it has to do with that Sonnox allows bigger Q values.

I wonder if the so-called colour can be seen in measurements or if it's just that the controls points the user to the direction of certain kind of settings. For example, URS N-series has a filter shape that looks like a pancake with small Q and loses it's "peakiness" very quickly when decreasing it from the maximum. A-series leaves the area outside the peak practically untouched when the gain is increased (if I remember right). Without an analyzer the frequential change itself can surely be perceived as colour.

Hmmm... maybe instead of color, character would have been a better choice of words?