How do you judge an EQ?

[Frippertronix]

hipass/lopass filter have no ability to boost the frequency range they process (except if the 'q' or resonance setting (e.g. Waves RenEQ) applies a q-boost because of the specific 'q' curve): they simply cut the signal above/below the crossover frequency.

Ok, that's more or less what I thought.
But, wouldn't a high/low pass then not be the same as a shelving EQ, minus the option to boost things?

The high/low shelf doesn't have the same linear drop-off, plus every high or low shelf I've used has a built in "resonance" peak at the cutoff, which high and lowpass don't automatically have, but many EQ's allow you to add one (like GEQ-7). The shelf has a steep curve just after the cutoff, then levels out like a shelf. Hipass/lowpass have a linear dropoff curve of 3,6,12,24, etc. db/oct.

[Christian Budde]

Just a small bump. I've smoothed the filter plots of the filter test (see:
http://www.kvraudio.com/forum/viewtopic ... 39#1940839)

Furthermore:

i think its both, i assume the wiggly low freq response is the short IR but the general numerical inaccuracy comes from the analog relm? you always hear people refering to analog eq X being so powerful, 1db here and it blows away any plugin. i think it has to do with electronics getting inaccurate/non linear in the lows particularly. but i know nothing here, christian enlighten us.

I'm not Jesus
Sometimes this whole business reminds me of religion. I know the whole topic is hard to understand in every detail, but there is no "EQ god". Neither in the digital domain, nor in analog domain.
Basically EQs are only filters. All analog filters are minimum phase, believe it or not. That's because of the linearity of time, they are causal. Everything (ringing) happens after a transient. A classic analog biquad filter can be turned into a digital version (assuming they are linear in terms of level, so absolutely no harmonic distortion). In this case they sound exactly the same. By sampling this filter and convolving it you can achieve the same results (I assume an impulse response length that is enough for all those golden-ears). So well designed biquads will sound equal, regardless of their implementation.

Unfortunately each technology (analog, digital modelling and convolution) might not be perfect:

• - Analog filters have a certain level of (level) non linearities (they usually can't be eliminated, but reduced, so that they can't be heard anymore. Manufactures do a lot to keep this very low).
  - With Digital modelling (IIRs) there might be limit cycles (especially for narrow filters) and they have some problems at the nyquist frequency (nyquist frequency equals +oo in that digital world).
  - The most common error with convolution filters are, that the IR is too short. This will result in a different frequency response (compared to the original), especially for a high Q (narrow band). Additionally if long IRs are truncated without proper windowing, you get errors by hard cutting. These errors can be found in Tritone's products as well as in the Sintefex IRs. They result in ripple (see "Mystery 2" - may also result from a non-biquad realisation = multiple filter stages).

Besides the imperfection of the realisation, there exists different definitions. Especially in case of the Q. On the other hand, it's possible to match every biquad to another (regardles of the definition).

The thing with "coloration" means that there is some artefact which might dominate (can clearly be heard). In posihfopit/Electri-Q the 'Analog' mode has a fair amount of harmonic distortion (odd harmonics (e.g if you input a sine at 1kHz, you also get a bit of 3kHz and a smaller bit of 5kHz and so on...). This harmonic distortion comes from level non-linearities inside the filter realisation (which is basically a passive filter with an op-amp (for gain make-up)). The 'Digital' mode is without this simulation. Within the 'Scream' mode, I used another, more worse op-amp, thus the harmonic distortion is higher.

Other coloration might be the imperfection of IRs (cutted too early). They are still filters and maybe even good filters, but have some coloration.
Other coloration might come by design of the captured original. If this delivers 10dB, although the label on the front reads 5dB, than it's not the convolution, but the original.

These colorations come by using other filter topologies than classic biquads. For example ladder filters has been used very often and there are several interesting topologies (i won't explain here).

So at least you have to decide what you'd like to use. But in my opinion it's important to "know", what you do and not to "believe" what you do. For that reason the 'Electri-Q' tries always to displays the exact filter curve (which is unfortunatly not 100% possible for filters with nonlinearities).

Linear filters are an idea of communications, where it is important to avoid dispertion. Since it is very hard to build them in the analog domain, they only exists in the digital domain (usually applied with rendering & convolution).

But that's another story. Good night,

Christian

[Christian Budde]

Just one thing, the Deconvolver has been updated. It's now donationware (displaying only an 'unregistered' label) and can be downloaded here: http://www.aixcoustic.com/uploads/media/Deconvolve.zip
To remove the label, just donate some money (>5€),

Christian

[kmonkey]

-

Christian

Christian can i ask you one thing (sine it seems that you do know alot about Eqs). By your view is it possible to mimic hydratone IRs with Electri-Q for example? At least to some 75%. I just downloaded demo yesterday and maybe my ears trick me but i can get very very similar results with Electri-q though i would like to know what other more advanced users think about this. I can not avoid my happynes if this is case since electri-q use so little cpu

Then again...maybe i should just trust my ear this time like many times before.

[martian]

I'm not Jesus
Sometimes this whole business reminds me of religion. I know the whole topic is hard to understand in every detail, but there is no "EQ god". Neither in the digital domain, nor in analog domain.

good cover up. but you are christian, saviour of soul, sent to enlighten and heal us mere mortals with your eq. no?

good job with the explanation post anyway.

through my own testing and listening ive come to the conclusion that comparing digital min phase eq is like comparing digital summing, pointless.

[bmanic]

Hi, just got back from vacation so I'll answer a few comments.

unfortunately i find your test flawed. if you equal the freq response from some of those eq settings then you will prob find the phase is almost identical. btw im ignoring sonalksis dither.

Like I said, it was never inteded to be a scientific test at all. That's what we have Christian here for.

i think what you are hearing is not inferioir/superioir sound but simply different settings. smaller bandwidth firstly means you require more gain to equal the apparent level of settings with higher bandwidth. it also mean more phase shift, so particularly in the low end the notes will ring more, subjectively you loose the 'tightness'.

However, experience shows me that there seems to be more to it than what you just wrote (like I also already said earlier, "shape" and phase do not always seem to be the real cause). I'll post some more demos today to illustrate this.

as christian points out with the sonalksis vs Q. Q as i ve mentioned before deflats to a very small bandwidth and will always sound inferoir in the same way any digital can be made to.

Well then, stay tuned for Wavelab Q versus PLParEQ. PLParEQ can be shaped nearly identical to Q in it's non-linear phase mode. Does it sound the same? Nope, not even close.

get christians analyser match up some settings then compare. just my thoughts.

Good idea. I'll see if I can get it working properly. However, I never even tried to set out to do and orange vs orange comparison when it comes to shapes but maybe that is what you guys want? EQ's are a bit like reverbs in my opinion. Density and lack-of ringing is NOT always the deciding factor for superiority.

ps. i think mystery eqs may be hydratone and sintefex.

Good guess but you are wrong.

Cheers!
bManic

[bmanic]

- The most common error with convolution filters are, that the IR is too short. This will result in a different frequency response (compared to the original), especially for a high Q (narrow band). Additionally if long IRs are truncated without proper windowing, you get errors by hard cutting. These errors can be found in Tritone's products as well as in the Sintefex IRs. They result in ripple (see "Mystery 2" - may also result from a non-biquad realisation = multiple filter stages).[/list]

I have to apologise here as I kind of tricked you all with that Mystery eq 2 clip. The reason you get weird plots is probably the fact that I had different EQ on the PLParEQ before the mystery eq 2 clip, simply because it's shelf is much steeper and starts very low (it's fixed position). I never set out to do true scientific comparisons. I've trained my ear quite long and try to be able to hear what happens to the transient instead of just the shape and added energy.

Christian: Would it be better for you with a different kind of clip to do those amplitude and phase plots? Or do you simply need the original, un-treated clip?

I'll post some more examples of EQ's today, including the sintefex and some more freebies. You guys want it blind or labelled?

Cheers!
bManic

[Christian Budde]

Christian: Would it be better for you with a different kind of clip to do those amplitude and phase plots? Or do you simply need the original, un-treated clip?

I'll post some more examples of EQ's today, including the sintefex and some more freebies. You guys want it blind or labelled?

I'd prefer original & EQed (without any limiter or compressor, normalize is ok). And I'd like to have them unlabeled. You can label them later,

Christian

[defjamm]

ears are not much worth and easy to fool, it's sad but it's true.

let's take hydratone. when i cut 12db at 1013hz with a q of 2(in the ff 110 mode), i get in reality this result:

[defjamm]

hydratone has been a bit of a sonic red herring for me, as well as an unstable hog for a year and now a fix that doesnt work on my comp. so ive totally annihilated it from my workflow since i was able to test it properly a few weeks back.

same here. it feels unfinished and this whole pluggo-stuff is annoying. but who cares, there are enough alternatives.

[martian]

ears are not much worth and easy to fool, it's sad but it's true.

let's take hydratone. when i cut 12db at 1013hz with a q of 2(in the ff 110 mode), i get in reality this result:

as far as im concerned that plugin is evil, my ultimate placebo. now that im informed i find everthing i liked about hydratone (punchy lows, smooth highs) was a trick and can be replaced very simpley.

also in use i just cant be completely objective about eq settings with a scale that changes. at any one freq a 1-3db boost/cut may yeild a 5db change then 4-8db gain a 1db change.

for me in a mix session theres a certain amount of eq decision performed on the spot in realtime, but theres also plently of more theoretical decision that comes after longer listening and reference, particularly for perfecting the settings. i need to know exactly what im doing or im kidding myself when getting down to the last (actual) 0.5db or so. since i know perfectly well in isolation double blind i can here level changes of 0.1db in audio, but i also know how i percieve them: warmer, rougher, fuller, brighter etc. i can hear whatever i want at that level.

if i cant tune it objectively then its compromising my ability which i rate as far more important. but im sure others braver and more skilled than myself can still make effective use of it.

[Debutante]

I don't understand the nature of the problem people are having with Hydratone. I've never used it but it's certainly made good reading in this thread....

How is it that the controls are not "responding accurately enough" to the input, a la defjamm's problem illustrated above? Is this something to worry about with other EQ's in gerenral, or what type of EQ IS Hydratone??

[kmonkey]

I don't understand the nature of the problem people are having with Hydratone. I've never used it but it's certainly made good reading in this thread....

How is it that the controls are not "responding accurately enough" to the input, a la defjamm's problem illustrated above? Is this something to worry about with other EQ's in gerenral, or what type of EQ IS Hydratone??

It does audio processing from various impulse responses taken from various(again) outboard EQs. Dunno is it just me but last days i can realy get close with Electri-Q which is waaay cheaper and have so more much power. Though Hydratone got few extras so you may consider it. I love to play with phase in hydratone. And it got nice warming algo. Hope this helps.

[Christian Budde]

It does audio processing from various impulse responses taken from various(again) outboard EQs.

I uploaded a new version of Electri-Q featuring 8 new filters. Guess what EQ i've sampled

Christian

P.S.: The IRs are already corrected a little bit, but they are more or less original. If you don't like them, don't complain.

[living sounds]

I'm curious if you could make Electri-Q sound like the new UAD-1 Neve. It doesn't do anything non-linear (I made impulse responses, and they sound only slightly less impressive and do null, minus the typical convolution artefacts). One EQ I like very much for mixing is the very old, outdated looking, but light on the CPU freeware A0 parametric EQ. This one seems to be very transient-friendly - is there anything special about it?