Hoping Nika is around...

[Matt Zeiner]

Can you or anyone else who may know tell me something about how time-domain errors (i.e. phase errors, phase smearing, etc..etc) happen inside of a digital eq? To my thinking, these types of things shouldn't be an issue with digital eq, beacuse reactance in the analog sense isn't part of the equation. I've been working on a technique that a friend uses involvong narrow bandwidth boosing. Another friend said that I should avoid boosting anything with a q of less than 2 or so to avoid phase artifacts. Both friends are highly respected in the industry, and both use extensive digital processing (protools,etc). I have enjoyed the results I've gotten with the narrow boosts, but I am curious why phase would even be a concern, since ostensiblty the plugin has all the time in the world to correct any error in the time domain. Maybe someone here could help me understand.....maybe its an artifact of modeling? Maybe it doesn't exist? hmm.....

[Matt Zeiner]

bump....still wondering.................

[Mikey MTC]

The only TDM EQ that does as you request is the Waves Phase Linear EQ. It sounds brilliant on some things but this comes with a price to pay in latency, making it difficult to use in mixing situations unless you compensate. It's marketed more as a mastering EQ though of course you're free to do what you want with it.

In the hardware world I think Weiss have something similar but that's where my knowledge ends.

I can heartily recommend the Waves Masters bundle which includes the aforementioned EQ, a specialized low frequency version of the EQ, Lin MB, their 5 band phase compensated dynamics controller and the venerable L2. It's processor intensive but if you've got the TDM chips spare and you're not working at 88/96K it's fantastic. For higher sample rates you need a powerful host to do the processing for you.

[Matt Zeiner]

Thanks for the reply, Mikey. I am aware of the fact that Waves offers a "Linear Phase" series of eqs and multiband processing. I love them - think they are great. The question I have is, why exactly in the digital domain do we have to worry about analog time-domain errors at all. Why do we need a Linear Phase eq? Shouldn't a digital eq be able to put all the bits back 'WHEN' they belong after it is done screwing around with them? Why wouldn't all eq plugs be designed to do that? Are they? ....I don't know. Guess maybe it doesn't matter all that much. Just curious.

[sdevino]

Matt, there are at least a couple of reasons.

1. Is the accuracy and compensation for errors in trigonometry. What value of PI is used in calculations, how is the Sine or Cos function implemented. What library did the designers use.

2. Working with real audio. Many of the DSP functions that transfor between frequency and time domain can only provide distortion free performance when the data that is sampled or processed is 100% repeatable within a given operating window. For instance the FFT only works on audio samples that are powers of 2 in size and repeatable. So only a Sinewave with a sample length of say 512 or 1024 or 2048 samples works with no distortion. There is no such thing in the real world of audio so modified versions of the FFT and the "windowing" of the waveform to ensure there are no out of band artifacts at the edges of the sample window are applied. All of these come with phase and energy distortion or redistribution.

Some designers anticipate and compensate or avoid some these issues better than others. Others are just aware of better algorithms or methods to avoid math and capture errors. DSP is a relatively young engineering discipline and progress is being made through innovation everyday.

Steve

[Joz]

I just want my EQs to work right [img]images/icons/confused.gif[/img]

[Nika]

Sorry I missed this thread two months ago. Perhaps the paper I wrote on filters would help?

www.cadenzarecording.com/papers


Nika.

[cmaynes]

Equalizers in both the analog and digital domain share basic design models.

In the analog domain, better quality is generally due to more specific "correcting" circuitry being applied to the signal path,
(the reason for some vintage hardware which doesn't use variable pots, opting for stepped setting pots which can be specifically tuned.)

The digital equivalent would be more processing time being needed to provide the modified signal (which is why the Waves matering EQ has its amount of latency that was mentioned.


charles maynes