What is your go to EQ?

[Nokenoku]

Ah, I see now. It was a while ago that I did the test with Waves SSL and API trials and I had just starting doing nulltests so my guess is that once the level got loud enough, the noise level increased enough that it came off as distortion rather than noise. My mistake on that one and thanks for giving me some relief on what had perplexed me (as it had been advertised and presented as just noise). It was an error in my testing methodology for that particular plug-in and my lack of thinking, rather than an overabundance of it.

I assume this is meant somehow ironically and you came to different results doing the nulltest?
So, could you please describe, what exactly you have done?

I just sent a full song (so every important frequencies in there and limited to 0dbFS) parallel through two instances of the Waves SSL EQ and made a maximum boost with the high band on both. For one EQ I disabled the Analog-mode and flipped the phase.
And the result was just the pure noise of the one with the analog-mode enabled. There was absolutely nothing else.

Moving along, however, there is the issue of artifacts. Some EQs oversample, others don't and some give you an option. This can, potentially, affect the sound. Do a null test between the oversampled and normal versions of the Stillwell EQ plug-ins across a variety of material and you'll see what I mean.

Of course it changes the sound. Because it changes the curve!

I had a chance to speak with George Massenburg, one of the originators of the parametric EQ, about the topic of digital oversampling in conjunction with digital processing such as EQs and compression and he talked a great deal about how it could change the plug-ins behavior.

Yes, and that's actually true. It's because -if you are working in 44,1kHz- the curve is heavily warped near 22,05kHz. Oversampling works against that effect.
But oversampling is not the only way to achieve that. There are other ways to avoid the curve-warping. I don't know how exactly it's done, but it's used for example in the AirEQ if I recall correctly. I think Bootsies plugins do as well.

Nearly every of todays EQ plugins has an option against the curve-warping, so it's not really an issue. Though there are a few plugins, which introduce additional phase shifting through oversampling. This is theoretically not so good, but practically the difference in sound is still smaller than any 0.1db boost/cut you would do.

Similar experience with software samplers, where I found that even playback of a sample at root frequency without filters enabled could vary from sampler to sampler.

Well ... a good sampler should not do something like that.

IMO you have to bear in mind that if you apply any minimum phase EQ on a an individual track by track basis and then further on a stereo bus all these 'minimum phase' artifacts soon become a big issue and can totally skew a mix.
That is how i see it anyway and hear it and so i always try to use EQ as little as possible that i can get away with

Like I said, every normal EQ and every analogue one is suffering from that. But I personally don't think it's that bad.

Listen to this:
http://rapidshare.com/files/142639719/test.rar

Which one of those is "totally skewed" by phase artifacts in your opinion?
Both files are the same six tracks summed together. But on one of those I added additional phase shift (without EQing) on every single track.

[Per Lichtman]

Ah, I see now. It was a while ago that I did the test with Waves SSL and API trials and I had just starting doing nulltests so my guess is that once the level got loud enough, the noise level increased enough that it came off as distortion rather than noise. My mistake on that one and thanks for giving me some relief on what had perplexed me (as it had been advertised and presented as just noise). It was an error in my testing methodology for that particular plug-in and my lack of thinking, rather than an overabundance of it.

I assume this is meant somehow ironically and you came to different results doing the nulltest?
So, could you please describe, what exactly you have done?

Nope, I wasn't being ironic. The emoticon was simply lampooning my error.

[Nokenoku]

Oh, ok. *g*

[martian]

Similar experience with software samplers, where I found that even playback of a sample at root frequency without filters enabled could vary from sampler to sampler.

Well ... a good sampler should not do something like that.

have to agree. ive tested a few software samplers and they all null against one another. identical sound.

[Per Lichtman]

when i compare eqs with oversampling the sound is different to those without, i percieve it only as a slight brightening in the highs though, so i can match that by turning the non-oversampled eq up slightly. the phase response gets adjusted quite radically with oversampling i think, which may account for the sound differences. but really if you only eveer percieve a sound difference as 'brighter' or 'darker' what difference will this make when using it? since youd just keep changing the parameters till you get the sound you want, so end up with the same result with either in a mix situation.

In another part of my post I mentioned scenarios where I was unable to achieve a similar sound, even with continued tweaking. When the sound diverges beyond a certain point it becomes difficult to compensate with a limited number of bands if the curve is sufficiently complex. When working a limited resource environment, or in a host with a limited number of inserts, adding sufficient instances of the second EQ to duplicate the curve of the first becomes prohibitive or, under real world circumstances, potentially close to impossible.

It seems unfair that I am concurrently being criticized for using techniques such as adjusting by ear and using frequency analyzers to check tonal balance at the same time as I am told that the differences in curves don't make a difference because I'll automatically compensate by ear. I'd expect that from two people, or at least two posts by the same person, but both in one post? I'm only human: I do tend to have to choose on possibility at a time.

[EDIT: Added missing commas.]

[Per Lichtman]

Similar experience with software samplers, where I found that even playback of a sample at root frequency without filters enabled could vary from sampler to sampler.

Well ... a good sampler should not do something like that.

have to agree. ive tested a few software samplers and they all null against one another. identical sound.

You might want to take a look at this test of a wide variety of samplers. It emphasizes resampling but some of this carries over. Also, keep in mind that the issue I reported was .SFZ and .WAV files whereas this test used .SF2 files.

http://www.maz-sound.de/index.php?show=mpcs&mpc_id=34

Keep in mind that various samplers use techniques to reduce aliasing that can cause changes. Try loading sounds as oscillators in Rapture and mess with the settings to get it neutral with a good level. Do the same in GVI. Some can sound different. The situation only gets more pronounced if you start resampling as shown in the tests above, that I am not affiliated with. I did however use the test files for my own comparisons on my computer.


If you are getting nulls, then I am very happy for you. That would have been a great relief on one of the projects I was working on this summer but alas... ah, but I must have a sense of humor about the extra work that caused.

Btw, did the sampled files all null with the original as well? And did it vary at all by the frequency content of the original sample used?

[Per Lichtman]

Moving along, however, there is the issue of artifacts. Some EQs oversample, others don't and some give you an option. This can, potentially, affect the sound. Do a null test between the oversampled and normal versions of the Stillwell EQ plug-ins across a variety of material and you'll see what I mean.

Of course it changes the sound. Because it changes the curve!

You could argue that this is simply a change in the approximation of the curve due to the accuracy of the sample rate but I would posit that the presence or absence of artifacts and oversampling or other techniques such as forms of "smoothing" (forgive the non-technical term) can make a difference when ostensibly applying the same curve.

I think I already addressed that point. Plus, if you're talking about the sudden presence, absence or change in a related shelf or filter than that isn't really changing the direct part of the curve that the band is ostensibly addressing, is it? In other words, you suddenly have to add a second band to compensate. So if you are in fact dealing with a limited number of bands, at some point it becomes impossible to replicate, right? [Edit for clarity: impossible to replicate in a second EQ with the same limited number of bands.]

[Nokenoku]

if you're talking about the sudden presence, absence or change in a related shelf or filter than that isn't really changing the direct part of the curve that the band is ostensibly addressing, is it?

I don't understand exactly what you mean here (might be because english is not my mother-language). But it just alters the curve. That's where the sound difference is coming from.

In other words, you suddenly have to add a second band to compensate. So if you are in fact dealing with a limited number of bands, at some point it becomes impossible to replicate, right? [Edit for clarity: impossible to replicate in a second EQ with the same limited number of bands.]

Jo, that's the reason why every modern EQ-plugin has an option for oversampling (or something different) against curve-warping.

Good night everyone.

[Per Lichtman]

Also I'd like to point out, once again, that I have nothing to gain by showing these differences. My previous school of thought was that such differences were nonexistent or insignificant. However, forgetting the more accurate null test for a moment, lets try a more mundane test: normalizing the ouptut signal of each plug-in, looking at the waveform, putting it into a list and having somebody blind test you after making sure that each one is exactly the same number of samples. Can you consistently hear a difference?

Now if you can hear a difference, than that should be reflected in the null test (though obviously our ears play tricks on us sometimes). But do you ever not hear a difference and have it show up in the null test? Those are what I sometimes think of as "statistically insignificant" differences when I am feeling cavalier but they do add up generation by generation.

The differences I mentioned between samplers were not, what I would call, statistically insignificant. So if you have any advice on how to get GVI, Rapture, SFZ, Kontakt, Dimension LE, HighLife, XS-1 and the trackers to all null up with each other, I'd be genuinely happy if you could provide it. My experience has been that some null and some don't, and the ones that null often (but not always) have higher CPU usage than the ones that don't.

[Per Lichtman]

if you're talking about the sudden presence, absence or change in a related shelf or filter than that isn't really changing the direct part of the curve that the band is ostensibly addressing, is it?

I don't understand exactly what you mean here (might be because english is not my mother-language). But it just alters the curve. That's where the sound difference is coming from.

In other words, you suddenly have to add a second band to compensate. So if you are in fact dealing with a limited number of bands, at some point it becomes impossible to replicate, right? [Edit for clarity: impossible to replicate in a second EQ with the same limited number of bands.]

Jo, that's the reason why every modern EQ-plugin has an option for oversampling (or something different) against curve-warping.

Good night everyone.

The version of RenEQ I've been using in RTAS form at Dreamtown Music has no evident option of that kind. Many EQs either have it enabled or disabled and don't give the user control. Several of the UAD plugs automatically resample. Several older digital plug-in EQs (still modern) have no resampling option. That is all I was referring to: a differentiating feature that affects the sound.

I think that maybe I'm too old school in my vocabulary here so I apologize if I'm creating unnecessary confusion here: when I hear talk of EQ curves, I think of the frequencies that are affected each time a band is modified. If another EQ cannot, using a single band, create that same sound, regardless of the reasons for that, then two of the many options seem most likely: either one of the EQs has the ability to go to a narrower/wider Q than the other, or they have have different colors.

If the new definition of curves is that they mean whatever frequencies a given EQ band affects for any reason, whether directly or through distortion or the presence or absence of artifacts, or presence or absence of phasing, then I will conceded the point. If that's what's meant by curve than I can think of few variables that would not be included. Why even bother differentiating between EQs that introduce harmonic distortion and those that do not? Or Linear Phase EQs from minimum phase or whatever variants there may be?

I hope I don't sound defensive, this is more just honest confusion. I've found that some EQs I can practically replicate with another and others I can only get a very close approximation. Are you saying that you've found you can replicate every EQ in your collection with every other EQ simply by using curve analysis?

If so, then I'll have to invest more time in it, right after I finish the 3 albums I'm currently mixing, composing or producing. It sounds like an interesting exercise at the very least, even if I ended up going from EQ to EQ just for the the GUIs varying ease at certain tasks.

[Per Lichtman]

Good night Nokenoku: Thanks for challanging my assertions rather than letting me grow lazy in my complacency.

I wish I could fall asleep too but my insomnia has been pretty nasty the last couple days. Oh well, worth another shot.

[Dean Aka Nekro]

you expect me to tell you the difference with crappy 192k mp3 encoded files! you idiot, send them at 16-Bit 44.1Khz and then i try will try and tell you!

[BASSDRIVE]

IIEQ PRO

[martian]

Now if you can hear a difference, than that should be reflected in the null test (though obviously our ears play tricks on us sometimes). But do you ever not hear a difference and have it show up in the null test? Those are what I sometimes think of as "statistically insignificant" differences when I am feeling cavalier but they do add up generation by generation.

yes, all you have to do is slightly adjust the phase of one signal across a part of its freq range and it wont null with the original, but you most likely wont hear any difference either when you blind A/B them (double blind listening).

[martian]

In another part of my post I mentioned scenarios where I was unable to achieve a similar sound, even with continued tweaking. When the sound diverges beyond a certain point it becomes difficult to compensate with a limited number of bands if the curve is sufficiently complex. When working a limited resource environment, or in a host with a limited number of inserts, adding sufficient instances of the second EQ to duplicate the curve of the first becomes prohibitive or, under real world circumstances, potentially close to impossible.

with most eqs the curves are easily exactly matchable until you get to the highs, where bell shapes vary near nyquist (half the sampling rate). this is were it will be hard to match some eqs and where small differences will sound.

http://www.eiosis.com/amltproducts

It seems unfair that I am concurrently being criticized for using techniques such as adjusting by ear and using frequency analyzers to check tonal balance at the same time as I am told that the differences in curves don't make a difference because I'll automatically compensate by ear. I'd expect that from two people, or at least two posts by the same person, but both in one post? I'm only human: I do tend to have to choose on possibility at a time.

yes im propably not making much sense, sorry

my 2nd comment when i said youd compensate by ear i was really refering to the sound of oversampling, which to me only sounds like things get slightly brighter in the highs, (and im not talking about the change in freq response, i think the change comes from the heavily warped phase response). but i also brought in the word darker to mean generally duller, and that you should be able to compensate that. but your use of the word 'darker' is to describe the more complex slight differences in sound of bell shapes near nyquist? which will result in different freq content around 10k+, and be very hard to match up by ear. unless you use something like buddes plugin analyser.

btw, all ive been arguing is the difference in sound is most likely caused by the different freq responses. but if one gives you the exact response you want then of course you will use it. for me personally there is no right curve, depends on the source, but at the same time im happy to use one flexible eq for all. i find analytically i can tell difference in eq bell shapes way beyond the precision of my mixing, which may only be to about 0.4-0.5db, and varies depending on the sound, but thats just me, everyones different.