PDA

View Full Version : The Sound and Calculations of Digital E.Q's



d-dmusic
02-18-2002, 08:56 PM
Just curious.

If you use a digital E.Q., lets say +,-12 db at a given frequency, would it be correct to say that the computer will re-calculate the waveform with an amplitude boost/reduction at that given frequency-continuously-in real time for the entire length of that waveform ?

If the computer/DAW is very, very busy, with a 50% load on the CPU, is it possible that the computer will miss a few calculations here and there resulting in errors ? What would these errors contribute to the sound ?

In terms of sound quality alone, does it make a difference if you use alot of E.Q.(+-12db) or a
little(+-0.5db) when talking digital E.Q.'s ?

What about the cumulative effect of digitally E.Q'ing alot of tracks simultaneously ? Does alot of active E.Q's degrade the sound quality due to more and more calculations and longer word lengths ?

SteveR
02-20-2002, 09:08 AM
+/-12db is a lot of eq'ing... ;)

When I am really overdoing it in Logic on a G4, it will complain that it doesn't have the horsepower to continue rather than continuing, but dropping out effects and the like. I would be suprised if the Eq degraded before the application complained it was having problems.

Eq is a subjective and personal thing. There are some rules of thumb but everything should be tried. The reality for me is, if I am eq'ing something even close to +/-12db, then there is either something wrong with the sound I have recorded, or I am trying to low or high pass a sound. If you eq anything, be aware that mid frequencies will make the most amount of difference to the sound. If you were to subtract 8db from around 600hz within a finished song, it would be ruined. Take it away from 16khz and you would notice it lose some air.

One last thing. The is a component to eq called Q. This represents the bandwidth of the eq being adjusted. If you chose, 1000hz (1khz) at a Q of .1, you would have a 100hz bandwidth. This means when you adjust the eq up by 5 db. You would be adjusting everything from 950hz to 1050hz. Get the idea.

I have found some eq programs don't subscribe to what I was taught about Q and Eq therefore this .1 business is subjective. Just trying to give you the idea.

Regards

Steve...R

p.s. Eq small amounts only. If you go big on Db you could either make a mix brilliant, or ruin it. Ruining it is most likely ;)

d-dmusic
02-20-2002, 10:28 AM
Hi SteveR :
Actually, I'm very well aquainted with the practical applications of E.Q.
Bandwidth (Q), subtractive vs. additive and their pluses and minuses, as well as the general principles of applying E.Q. to tracks within a mix etc.....

But I am not aware of the technical/theoretical aspects of digital E.Q. and the complex calculations that occurr and various problems that could occur as the CPU gets close to it's limit. It's all very interesting to me.

As an example, my DSPFactory Audio card, which is basically an O2R on a card, exhibits strange behaviour as it reaches its limits of processing power. I can hear one or both of the reverbs dissapear from the mix and then come back again.
This kind of thing is obvious. You can't miss it. But, what about an E.Q. missing calculations during playback of a complex mix. That could be much less obvious.

Does it happen regularly ?
Of course using lots and lots of E.Q.'s on various tracks brings the CPU closer to its limits to re-calculate.
(What is a good "practical" limit for a CPU ? 50% ? 60% ?)

Check out Rogers Nichols forum at eqmag.com
I have the same post there with some good discussions on this.

Thanks again.

SteveR
02-20-2002, 05:50 PM
I didn't mean in any way to belittle your knowledge of Eq. I just wasn't aware whether it was something you had a grip of, so thought I would chuck in what I know.

Your point is therefore more interesting to me. I suspect there maybe some grounds for concern although some extensive tests would have to be done.

Maybe a side by side comparison of a heavily eq'd full range mix would have to take place to the point of CPU shut down and then analyised closely as a wave. This could be done again although bouncing the tracks through the eq's to tape and then side by side the resulting mixdown waves. Just thinking out loud here I suppose. Could work.

I have never really thought of CPU overload as a potential problem but there could be something in it. If you find out more, let me know as it's more than an interesting point..

God Bless

SteveR