Reviewed on: SoundStage! Solo, May 2021

I measured the Bowers & Wilkins PX7 Carbon Edition headphones using laboratory-grade equipment: a GRAS Model 43AG ear/cheek simulator/RA0402 ear simulator with KB5000/KB5001 simulated pinnae, and an Audiomatica Clio 12 audio analyzer. For isolation measurements, I used a laptop computer running TrueRTA software with an M-Audio MobilePre USB audio interface. An Mpow BH259A Bluetooth transmitter was used to send signals from the Clio 12 to the headphones. For wired measurements, the headphones were amplified using a Musical Fidelity V-CAN amplifier. These are “flat” measurements; no diffuse-field or free-field compensation curve was employed. If you’d like to learn more about what our measurements mean, click here.

Frequency response

This chart shows the PX7 Carbon Editions’ frequency response in three modes: Bluetooth with noise canceling on and with noise canceling off, plus wired mode with noise canceling off. A couple of notes here. This is an unusual response—while the bass looks almost Harman curve-ish, there’s an unusual peak around 1kHz, the usual peak at about 2 to 3kHz is muted by a few dB, and the peak up around 8 to 9kHz is maybe 6dB higher than I might usually see. There’s a big difference in response with noise canceling off—a lot less bass, mainly—but very little difference between the wired and Bluetooth connections.

Frequency response

This chart shows the PX7 Carbon Editions’ right-channel response (all modes activated) compared with two other noise-canceling headphones (with NC on) and the AKG K371s (headphones that come very close to the Harman curve response). It looks like the PX7 Carbon Editions will have a deficit of lower treble and a surplus of mid-treble, relative to most other headphones.


Any resonances the PX7 Carbon Editions might have are well-damped and not troublesome—except for that super-high-Q one at 3kHz, but it’s way too narrow to hear.


The total harmonic distortion (measured in wired mode with power on) of the PX7 Carbon Editions shows an unusual 6% peak at 200Hz, which would appear as harmonics at 400Hz, 600Hz, etc., so it might be audible if you’re playing the headphones very loud. There’s a relatively high amount of distortion in the bass, too, but audibility of distortion is much lower at bass frequencies because your ear isn’t very sensitive there, so I doubt you’d notice it.


In this chart, the external noise level is 85dB SPL, and numbers below that indicate the degree of attenuation of outside sounds. The lower the lines, the better the isolation. The isolation of the PX7 Carbon Editions (shown here in Auto mode) is not impressive; it’s about the same as the DALI IO-6es achieve, and those weren’t impressive on this test, either. Maybe the Auto function would work better on an actual airplane? I’m not sure and it’ll be a while before I get on one of those again . . .


In wired mode, the PX7 Carbon Editions’ impedance magnitude runs mostly above 1500 ohms, which is the measurement limit of the Clio 12 analyzer. That’s to be expected of active headphones; as best I could tell, the power always switches on when the cable is plugged in, so there doesn’t seem to be a fully passive mode. I got 98.8dB average when I measured the sensitivity of the wired connection (calculating the drive voltage for the default 32 ohms impedance, which I usually do with active headphones).


Bluetooth latency of the PX7 Carbon Editions used with the Mpow BH259A transmitter is 220ms. That’s typical of SBC, although the PX7 Carbon Editions are said to be equipped with the standard, Adaptive, and HD versions of Qualcomm aptX. The BH259A has standard aptX and aptX HD, so I’m surprised I didn’t get a lower latency here.

Bottom line: Bowers & Wilkins definitely went their own way with the PX7 Carbon Edition headphones. Their frequency-response curve is unusual (although not as weird as the Yamaha YH-E700A headphones I recently tested), and at least in my tests, the noise canceling didn’t seem impressive. But while they’re outside the norm, I’d guess the unusual aspects of the frequency response aren’t so extreme that they’d make these sound bad or weird.

. . . Brent Butterworth