The Status SM-CB1 headphones are closed-back studio monitor headphones, at a pretty good price. They compare favorably with the Sennheiser HD650 open-back headphones, widely regarded as a standard among audio professionals. But both can use a bit of equalization to bring them to a flat spectral standard.
The Status CB1 exhibit a bit more bass response than the HD650, a well known deficiency in the HD650. And the treble response of the CB1 is a bit more chaotic than the HD650.
What this means is that both headphones need more help in the subbass region below 50 Hz. But the HD650 needs very little elsewhere. It has a mild excess around 200 Hz, but only by a few dB. And the HD650 needs one measure of boost between 4-6 kHz.
In comparison, the CB1 needs more work in the treble. My MOTU Ultralight audio interface has a 5-band Parametric EQ that I can put to use. And while you might think you just line up and match each peak in the deficit curve, we can do a bit better with some higher mathematics.
I set up a computation to twiddle the 4 bands above 100 Hz, jostling each of their center frequency, Q, and gain boost/cut, all simultaneously. Moving one filter in any manner affects adjacent filters to some degree. So we want the settings in all four filters where their interactions are taken into account. It’s like having a dozen hands on EQ knobs to move them all at the same time in various directions. We let the computer do that for us.
The routine I used is called the Levenberg-Marquardt optimization algorithm, which can deal happily with nonlinear systems of interacting parameters. I start it out with gains, Q’s, and frequencies for each band that are in the right neighborhood, place limits on the parameter ranges, and let’er rip.
Each band is given frequency limits in which it must stay, gains are limited to a range of -10 dB to +10 dB, and Q ranges from 0.01 to 3. My Ultralight imposes these limits on Q.
In order to deal more smoothly with running up to a boundary limit on any parameter, I used a hyperbolic tangent mapping, so that the computer can try values from minus infinity to plus infinity, and those values map nicely into my restricted parameter ranges.
So how’d we do? After about 1,000 iterations (about 2-3 minutes), the answer popped out. I set up the Ultralight EQ with the nearest available values to those shown in the computed answer, and it sounds pretty good.
Here’s the graphs showing how we did.
The top graph shows the frequency response of the CB1 headphones alone, in light green. The red curve is the modified Harman target profile, corresponding to a flat room and near-perfect speakers.
This target curve shows the measured sound intensity at the eardrum. Our goal is to have our headphones produce that same sound field at our eardrums. And you can see that the untreated CB1 misses that mark by a bit.
The dark green curve shows our CB1 headphones after applying Ultralight Parametric EQ. Now you see we have a pretty close response to the target curve.
The bottom chart shows the original headphone deficit in light green, the individual equalizers in blue, and the resulting equalized residual deficiency in magenta. The goal is to get that magenta curve close to zero across the audible frequencies.
I restricted the fitting to consider only the frequency range from 20 Hz to 10 kHz. I did not bother with anything above 10 kHz because that is all fiction up there. Above 10 kHz the wavelengths of sound are so short that we are seeing resonant responses between the earphone cup and the side of our head and pinnae. It is very difficult to make decent measurements up there. (And frankly, I can’t hear beyond 10 kHz anymore anyway, so why bother with them?)
The curves for the bare headphone and the Harman reference profile were manually digitized from published graphs, available at InnerFidelity.com. Once digitized, these samples were converted into spline interpolations for use in the computations. The fitting considered the difference between the headphone + EQ and the reference curve over a logarithmic frequency grid stepped by 0.01 (= 2% steps in frequency), from log10 Hz = 1.3 to 10 (representing 20 Hz to 10 kHz).
The fitting routine takes into account the magnitude and phase of the headphones and individual EQ filters. The reference target curve is a magnitude only specification. Using phase as well as magnitude in the fitting makes a small difference, compared to just fitting dB magnitudes.
The headphones are nice and bassy, but not too much so. The detail in the music is nice and crisp, and the high percussive sounds are delicate and not overly present. Sibilance stands out clearly enough that I can actually discern many lyrics. I have long ignored lyrics because I could never hear them, even when I was young and had perfect hearing.
At lowered listening levels, the sound is very nice, and helped along by some playback EQ shelving boost of about 2 dB: low boost starting at 105 Hz, and high boost starting at 2.5 kHz. That’s to help overcome the lessened sensitivity of our hearing at deep bass and high treble frequencies as the sound intensity decreases. (the Fletcher-Munson equal-loudness contours)
The end result is a very comfortable pair of accurate studio monitoring headphones for about $80 USD. It took a bit more work to get them into perfect shape than with the $400 USD Sennheiser HD650. But the end result is just about the same between them.
[ And here’s the solution when we leave the 7 kHz notch in place by not using the highest frequency filter. I don’t really hear much difference, but this is the recommendation…