Digital EQ Fact & Myth.

[Aleksey Vaneev]

It's pink noise. Far from stationary. The same is true if i play music as i keep on saying. Hard to trick a nulled file..

Noise signal is statistically a stationary one, at least for dynamic processes. It has a stable RMS power. Beside that you can't really measure harmonic distortion with white/pink noise files without going into complexities of Volterra kernels. So, this aspect is beyond your test.

What is a 'nulled file'? You haven't shown one except frequency response/phase plots. It's not the same as bit-null file.

[vieris]

So by getting a complete null while playing any content doesn't = same to you?

What I am wondering here is what you mean with 'complete null'. I could not see a prove for that. Again, I do not want to mistrust you, when you claim, that several EQs can produce a near equal output, but from a programming point of view there should be at least some, probably very tiny difference. You only need to have the internal topology not to be equal, or if one uses SSE or just have the initial states initialized differentially. Not to mention the denormal prevention, which is for sure different in every plugin.
You might get a "null", that will be equal to no difference played with any DA converter (if the difference is somewhat larger than -100 dB). It might be inaudible, sure, but I doubt it is zero.

Don't get me wrong, I do not want to critisize your work exploring all these EQs, but you need to improve your measurements. In that business you have to be more than accurate, especially if you want to convince other people. Believe me or not, but several years ago I was on the same path, doing the same tests. Last year I have written my diploma thesis about measuring nonlinearities (e.g. harmonic distortion) and I can tell you there is so much more to evaluate, once you know what to look for.

Honestly I wish you good luck and success for your research.

Christian

PS: The only problem with these kind of research is the fact, that you can not save the planet with that, you can only make it worse by wasting energy. But then I shouldn't write these lines either...

I really respect your views. & I will take your

The only problem with these kind of research is the fact, that you can not save the planet with that, you can only make it worse by wasting energy.

statement to heart. I know this will always remain heated & there will always be those poking the fire out of context, pointing out the exceptions "like dynamic eq". I leave it up to the user to decipher internal differences, flexibilities, ect. I'm going to post my findings and let everyone take what they desire from it.

P.S. Let me take this time to thank you for the analyzing tools. They have been priceless for my compressor designs.

[Vitaly]

mustgroove Could you do me a favour? Could you publish here a screen shot of Harmonic Distortion of 1176 when GR meter is at -10dB and Ratio is 4:1. It would be much appreciated, thanks.

I'm not sure why you need this, but if you are trying to recreate the 1176 by this plot only you will surely fail. You do not cover any dynamic behaviour with that nor the frequency dependent harmonic distortion. You need at least to plot a series of THD plots for every single frequency (just to be sure it's equal over the whole range). Also the THD depend on the attack/release time. So there is no single plot for -10 dB with the ratio 4:1 if you do not specify the attack/release. Also I can already tell you, that the emulation is far away from the original, but that's another story...

Yes, I know all this, Christian, and I didn't want to recreate 1176, I don't think it's necessary. I wanted to get a basic impression regarding quality of that algorithm that's all.

[living sounds]

I myself am coming closer and closer to the same conclusion you've come. Yet I prefer using my favorite EQs from BTE Audio because I get used to their sound.

How do they sound? How did you get them? It says on their website they do only offer test versions to "well-established" companies...

[Vitaly]

I do beta-testing for BTE Audio.

I don't want to use any hype words to describe their sound but it is definitely one of the best vintage eq emulation around. I compared BTE's AEQ3 to API's 560 and, with oversampling on, it was very hard for me to tell who is who (because I knew who was who). In a mix, I wouldn't find the difference.

[living sounds]

I do beta-testing for BTE Audio.

I don't want to use any hype words to describe their sound but it is definitely one of the best vintage eq emulation around. I compared BTE's AEQ3 to API's 560 and, with oversampling on, it was very hard for me to tell who is who (because I knew who was who). In a mix, I wouldn't find the difference.

The real thing API or Waves API?

[Vitaly]

The real API.

[birrbits]

results of a sinusoidal sweep would probably be more revealing than pink noise..

Again, this wouldn't capture the dynamic behaviour. In case of a so called 'logarithmic sweep', you could at least evaluate the harmonic distortion for the fixed input level, but only if you apply a deconvolution with windowing all the HIRs.

agreed, my point was that even this simple approach would be an improvement over a noise sample. You have my curiosity though, while deconvolving the chirp or sweep is intuitive enough, why would you have to window the HIRS? Couldn't you just compare the different frequency responses of the whole IR (between different EQs)?

[Kenton]

Don't get me wrong, the UAD plugins definitely have an x factor of some kind, they give an instant "wow" whenever you use them... all I've discovered is that it isn't distortion or harmonics or anything of that kind.

The whole distortion/harmonics and "warming" thing needs to be rethought IMHO. I think that whole theory arose at the dawn of the digital audio era, when some people were scratching their heads as to why all this digital gear didn't sound as "good" as the old analogue stuff... But to me it seems like just a theory - these UAD EQs kick total ass over everything else I've ever used... and none of them have any harmonic distortion to speak of. Clearly something else is going on.

When I was doing my assessments of various plugins, I found the UAD 1073, Pultec and 88RS the most difficult to null to any degree. (The Cambridge seems to be a standard BiQuad)
My suspicion is that there is something like Allpass filtering going on somewhere in these UAD plugs which produces phase shift without attenuation across the spectrum. I could never get the bass end and the treble end to both cancel to my satisfaction at the same time - sometimes I couldn't get the bass end to cancel at all...........

K.

[fader8]

So by getting a complete null while playing any content doesn't = same to you?

vieris,
How are you determining that your results are actually a complete null file? Are you looking at the raw data in Matlab or something? See my post HERE on gearslutz. Hopefully that will help you understand how limited audio editing apps are at determining a "null" file.

[vieris]

My in host method. I have a prism D-scope but there is no need. I can analyze a larger dynamic range in the box.
It's simple. Pink noise sent to two buses with 2 different EQ's panned hard left & right. I use my analyzer (nugen visualizer with it's dynamic range of -195)on the master. I tweak the EQ's so all analyzed measurements match. Matched frequency response & matched phase characteristics. I can then invert one of the buses & pan them up the center mono. This is when they will null. I can then look at the -160 & lower noise differences in plugin denormanal or internal dither design, if any. I do my tests & 64bit/96k to avoid upsampling aliasing & internal dither stages.

[james0tucson]

You're lying. If only I could place THIS one on my tracks:

http://www.genwaveaudio.com

Then I would be multi-platinum baby!

Brent

Oh, yeah, forgot about this one. Yes, when cost is no object and reality has lost its grip, this is the only way to go. I think ultra-multi-platinum is more appropriate -- and still understated. It's more than an EQ. It's...it's...I think I'm at a loss for words.
[/quote]

"Danceable" is the word you're looking for.

[james0tucson]

9x = 2

45x = 2

What is x?

Modular?

[acoosticzoo]

I love this post! It's been irking me digital eq's and minimum phase VS Linear blah blah finally a great discussion. Gospel to my ears. Thanks heaps.