Headphones & EQ



It’s simply stunning… Having carefully crafted equalization applied to the headphones makes all the difference in the world. Details in the sounds begin appearing where there were none before. This is mostly in the high treble region, up around 6-8 kHz. The bass response is deeply extended as well, down into the subbass region below 50 Hz where you mostly feel the bass.

After having carefully crafted these EQ sections for my Sennheiser HD650 and Status CB1 headphones, I think the edge still goes to the Sennheiser headphones, but just by a hair.

In the bass region, the Sennheiser was further away from the reference target profile than the Status CB1. The CB1 still feels a bit more bass heavy to me. And that may account for some veiling of the highest faint details in the music.

In the treble region, the Sennheiser was much closer to reference already, and so needed less help with the EQ than the Status CB1. That may also account for its better preservation of details.

The Status CB1s boast 50mm diaphragms, and that may account for its better bass response. But I have to wonder if moving those large diaphragms gets in the way up in the high treble region. Meanwhile, the Sennheiser HD650 with their smaller diaphragms are handling the details quite well, but they need a lot more help down in the bass region.

I’m also finding that the 7 kHz notch is not a good idea for me. When I leave the notch in, those details in the sound disappear on me. I like those details, so I don’t want that notch. Now that may be caused by my hearing, but Crescendo is enabling me to hear up that high. (Without Crescendo, I’m wearing a pair of woofers with 24 dB/octave roll off above 1 kHz!) Maybe other folks with better hearing can hear too much 7 kHz unless that notch is in there. But I’m leaving it in for myself.

But in the end, both are good headphones. I am listening through the Sennheiser HD650 at the moment, using a Lavry DA-11 DAC after the equalization is applied digitally in the MOTU Ultralight audio interface. This costs me zero computing cycles, in my computer, to have carefully equalized monitoring.

It has taken nearly 20 years for the audio industry to look carefully at headphones and how they should sound. Until now, people bought and used headphones without equalization, based on how the listener liked what they heard through them. But I doubt there are any headphones that closely match a good target reference.

My equalized Sennheiser HD650 shows a match to within 0.5 dB all the way from 30 Hz to 9 kHz, and within 1 dB all the way from 20 Hz to 10 kHz. That completely covers my own audible range. The Status CB1 now shows a match to within 1.5 dB over that same range. A bit lumpier than the Sennheiser headphones.

Could a better job be done with EQ? Possibly so. I’m using conventional IIR digital filters. Those impart some phase shifts to the sound, which varies with frequency. I took that phase shifting into account during my tuning procedure, so the dB magnitude plots of excess and deficit include these effects in combination with the phase shifts already inherent to the headphones themselves.

Does that phase shifting matter to our hearing? The answer to that can be complicated, but in general, no. Group delay, or a physical delay to some portion of the sound, can occur wherever you have phase changing with frequency. Usually the frequency rate of change in phase is too shallow to matter to us. But some peculiar steep filters might impart enough slope to cause a noticeable effect on the sound. That isn’t happening here with simple bell filters applied to the headphone.

To do any better than I already am, and without phase shift artifacts, we can resort to phase linear FIR filtering. But that imparts an overall throughput latency on the sound. And to craft sound with FIR filters, you need these filters to extend to more than 3 periods of the lowest / narrowest frequency region of interest. So getting good corrections extending all the way down to 20 Hz means that we need at least 75 ms of throughput delay. That, in itself, becomes objectionable, especially for live performance and audio editing tasks.

I suspect a shorter FIR filter with smaller throughput delay could handle the treble region, leaving the bass region to be handled by IIR filtering as I am now. That offers an interesting design possibility. But at this point, the need is slight and the sound is already so good. How much better can it sound?

But comparing what you hear through different headphones is fraught with difficulties. You need to get them loudness balanced or you will subconsciously prefer whichever one is slightly louder. And that would be nearly impossible to achieve unless they were both closely matched to the reference profile. Otherwise, one pair could be bass heavy, and the other pair treble heavy, yet both would produce the same loudness overall. But they wouldn’t sound at all alike.

Loudness matching can only be done subjectively at this time. It doesn’t matter if everything else is all the same – same digital stream, same processing, same amplifier gains. The fact is that one headphone will be more efficient than another, making it louder for the same electrical signal.

We are also affected by what immediately precedes our listening tests. So if the headphones are not already closely matched, then a brighter headphone listened to first, will make the next headphone seem darker and more muted. We can’t help it. It’s just the way our brains work.

But I can tell you that I am astonished at the combination of Crescendo corrections applied to proper headphone equalization, listened to through a high quality pair of headphones and DAC. Listening has never been this good!

  • DM

Author: dbmcclain

Astrophysicist, spook, musician, Lisp aficionado, deaf guy