Dave Gamble
January 6, 2010 at 11:10 am
This is a great article!
There ARE a couple of points that I think ought to be mentioned though.
As a designer of EQs (the Sonalksis and Focusrite EQs for example, plus a new one I’m working on at the moment), I can tell you that there are a bunch of different mathematical techniques available for designing digital EQ curves.
The biggest difference you would expect to see is at the very top end, up near nyquist.
I imagine you would find that if you added some air with a bell using something like AirEQ, you might need a couple of bands to accurately replicate that.
It was very unfortunate that your AirEQ example did nothing at the top-end.
I’d actually go so far as to say that the “cramped” top end that the bilinear transform imparts was partly responsible for the “cold” and “sterile” reputation that early digital audio had. Effectively, the top-end air was getting cut out. For a long time since, the Orfanidis technique has been ubiquitous in well designed EQs, although as good as it is, it’s not necessarily perfect.
Eiosis overstate the point in their AMLT demonstration (for instance, in their graphs, the gain at nyquist is incorrect, despite the principle of the Orfanidis technique being that it ALWAYS generates the correct nyquist gain), but it is true that if exactly matching the analogue shapes is your objective (which it probably ought to be, since they’re consistent), the differences between the analogue and digital domains mean that a good approximation is the best that can be done.
For mid frequencies, you’re absolutely right that you shouldn’t expect to see any differences. At the same time, I think creative usage of Q is an excellent skill to develop; and a well-designed EQ (analogue OR digital) will have the gain and Q interact in a productive way – guiding your hand as you work.
A crucial point that’s often ignored is that these vintage analogue EQs that we get excited about are not interesting because they’re old, nor because they’re analogue, but because THOSE are the designs that were good enough to be remembered. What this tells us is that the interactions and ranges, and overall workflow are good. When what you’re looking for is a tool, a well-designed tool has value. If an EQ makes things sound good in 10 seconds, you can fairly judge that it’s a better-designed EQ than one that takes 10 minutes to get a good sound. Sure the maths is still the maths, as you’ve said, but what matters is not the set of shapes it CAN achieve (which, neglecting >10k, is generally the same set), but which set of shapes the workflow guides you towards.
Thanks again for a great article! ?
