Last Friday evening I found myself again having to EQ our main sound system at work, due to a combination of what I believe to be environmental factors and physical changes to the speaker setup, namely the replacement of some faulty speaker drivers in a bass cabinet that needed taking into account – itself a blog subject for another time I’m sure!

My usual experience with setting system EQ has usually been centred around one of two things:

  1. Making the system sound as good as possible with CD-sourced playback material, in the hope that this will provide a known starting point for the sound of any mix we create on said system during a live event, or…
  2. …putting a key microphone into its usual position (such as a lectern for a church) and having someone speak into it, making their voice sound as natural as possible (without resorting to desk EQ beyond a simple high-pass-filter). Once this is done I’d then slowly turn up the gain for that microphone (keeping the fader level constant) and using some form of EQ to pull out any frequency bands that feed back.

Both methods have been “good enough” for rock-n-rolling into a venue and making something more than reasonable come out of the speakers, but neither method is terribly scientific, nor does it lead to consistent results.

More recently I’ve been playing with using pink noise and an RTA to show me what the system’s doing, then EQ’ing the system so that pink noise played out of it and measured with a flat-response microphone is shown on the RTA as being as close to the original pink noise as I can get.  This has lead to more consistent results than either of my previous methods, and has cut down the time spent on the task by something like 50-70%, but still the resulting system sound is somewhat variable to say the least.

So at a pinch on Friday evening, I happened upon what seemed to be a better method, and one I’ve not tried since my earliest days of sound mixing/system engineering:

  1. Make a CD with test-tones at a fixed level (usually -20dB), centered at the typical frequencies found on the faders of a 31-band (1/3 octave) graphic equaliser.
  2. Starting with 40Hz (the lowest audible frequency in most mixes/systems I deal with), get the tone playing through the system at about 65dB on a typical SPL meter.  A or C weighting doesn’t matter at this point – what does matter is that I get it set in my mind how “loud” that tone sounds/feels.
  3. Then go to the next tone up.
  4. Is this playing at the same perceived level as the 40Hz tone that preceded it?
  5. If yes, move on to the next tone.
  6. If no, then set the system EQ (I had both parametric and graphics to hand) to compensate.  Keep comparing and adjusting until the level sounds comparable.
  7. Repeat steps 3-4 until all frequencies that you can hear, either due to the system itself or your hearing (!), are pretty much perceptively even.
Note 1:   If you have a parametric EQ, and you find that frequencies progressively become more or less prominent than those preceding them, you can set an EQ curve centered at the point where the smallest difference occurs between adjoining tones, boosting or cutting accordingly.  The width of the filter is  roughly defined by the number of tones you find to be different.  It’s hard to explain in text, but becomes more obvious the more you play with the EQ parameters.  Using a parametric EQ here gives more precision and control over what you do to the signal, without the distortions of graphic EQ, which essentially is a chain of 31 or more audio filters run in series.
Note 2:  This is best done as an iterative process, so it’s worth playing through the test tones up and down the scale and adjusting until you feel you can’t make any more positive adjustments.
Note 3:  On our system, I was able to accurately do this up to around 16KHz, as with the combination of my hearing and our system I wasn’t able to discern anything beyond around 17KHz.  Not bad for a tired near-30-year-old engineer, working late at night on a combination of Bose 802/402/302’s!

Having applied this method to our system I played a couple of favourite songs through it from my laptop, which has a pretty good quality sound output (equivalent to most “hifi” grade CD players when fed with CD-quality content), and the system sounded immediately more musical, more involving and less “PA-like”.

Out of interest I measured the pink-noise response of our system What I found with this method was that my system curve had a notable reduction in the 2-6KHz range than would be obtained by using the pink-noise method above, which might be seen by many engineers as a significant disadvantage.

I then had a couple of our other engineers use the system in live services with this new system EQ curve, and their feedback was that the system sounded so much better than they’ve been used to.  They were making much more subtle (And arguably more accurate) changes to desk channel EQ for both speech and music, and the usual issues we have with feedback or tinny-sounding speech microphones were much reduced.

On reflection, I wonder whether part of the success story here is that my chosen reference level of 65dB (SPL, A-weighted) is pretty close to someone talking passionately to another in a quiet lounge – and given that the sensitivity of human hearing at specific frequencies changes depending on the overall sound level, this coincides quite nicely with our main material, speech reinforcement with some louder music that doesn’t often get much louder than 90dB.

I’m sure I’ve done many things wrong by working this way, but it was quick, easy and seems to have worked out well for us – our engineers are happier working with the system set up this way than they have been for a long, long time.  I’m sure I’ve missed a few crucial things out in my explanation here, so I might re-visit the topic in the future.  But meanwhile I hope this stands as a demonstration of another way of using EQ to get more out of your sound system.

As always, your mileage may vary – and your needs might be very different to ours!