There is some confusion about the output level in the THD measurements (or in any of the measurements in the video actually), and this was asked about in comments on YouTube. Any confusion about this is entirely my fault, and I really should have been clearer about some procedural changes we've made.
- We are now setting the output voltage level in the Audio Precision analyzer using white noise, setting the analyzer to auto-set the voltage to the level at which 80 dBSPL of white noise is output from the headphone.
- Once we reach 80 dBSPL of white noise, we initiate the sweep.
- In the case of the Elite with hybrid earpads, that corresponds to an output of just over 93 dBSPL at 1 kHz.
- With the Alcantara earpads, it corresponds to an output at 1 kHz of around 93.8 dBSPL. I really should have been clearer about this, and am sorry for the confusion.
If you're curious what the dBSPL output level is at any given frequency for which we're using white noise to set level, just look at the frequency response (as long as it has an absolute (versus relative) y-axis) and see what the dBSPL is at that frequency. Again, for the Elite with hybrid earpads it's ≈93 dBSPL @ 1 kHz, and for the Elite with Alcantara earpads it's ≈93.8 dBSPL @ 1 kHz. (You can see this for the Elite with hybrid earpads in Fig.3 below, but please read the rest of this post first.)
Again, I really should have been clearer about this, and I am sorry for any confusion.
Obviously, the next question is
why I've started doing this. For years now I've felt strongly that setting the voltage in consideration of only one frequency (usually 500 Hz or 1 kHz) for every headphone is less than ideal, especially when you want to compare more than one headphone. We used to set levels at 1 kHz, typically at 90 dBSPL. (Tyll Hertsens at InnerFidelity used 90 dBSPL @ 1 kHz, and we followed suit.) Also commonly in use by others is 1 Pa (≈94 dBSPL) at 500 Hz. Let's take a look at the Meze Audio Elite's frequency response compared to the Bowers & Wilkins P9 matched up at 500 Hz:
Fig.1: Frequency responses of Meze Audio Elite (hybrid earpads) and Bowers & Wilkins P9, normalized at 500 Hz.
I think you can see the problem we have matching these two headphone up at 500 Hz in Fig.1 above.
Here are those same headphones compared, normalized at 1 kHz:
Fig.2: Frequency responses of Meze Audio Elite (hybrid earpads) and Bowers & Wilkins P9, normalized at 1 kHz.
Normalizing at 1 kHz in Fig.2 above looks better than the comparison in Fig.1 at 500 Hz. However, if you've heard these two headphones, I think you would agree this isn't really representative of a good comparison relative to what you hear either, but it's closer.
Here are those same headphones compared
unnormalized (Fig.3 below), the measured output for both being 80 dBSPL of white noise:
Fig.3: Frequency responses of Meze Audio Elite (hybrid earpads) and Bowers & Wilkins P9, unnormalized. The output level for both was set with white noise at 80 dBSPL.
If you're familiar with these two headphones, I think you'll agree that Fig.3 is a more meaningful comparison of both headphones, relative to what you hear. And this comparison -- because the levels were set using
white noise versus just a single frequency -- did not require the arbitrary selection of a frequency to normalize at, for what I think/feel constitutes a good comparison of these two headphones. Of course, nothing's perfect, and certain comparisons may require intervention from time to time.
We may further refine and change these procedures, but this is what we're doing now, and we like what we're seeing so far.
What can I do better to make this clearer? I can find a way to state this more clearly in each measurement. Since we log the output voltage for every sweep, I can also consider including the average output voltage (average of all sweeps) to indicate sensitivity for comparison to other headphones similarly measured.
Anyway, again, I'm very sorry if I've caused confusion here -- that's completely on me. I hope for anyone reading this post, it's clearer now what I'm doing, and I hope you can appreciate the reasons why I'm doing it this way. Of course, I'm open to any critique or feedback about this, but let's keep that discussion for a separate thread. (I can create a separate thread later with a copy of this post to start it. I'll update this post with a link to that thread when I do.)
NOTE: The Bowers & Wilkins P9 measurement in the three graphs above was from just a single seating of the P9 (averaging both channels of that single seating), as I just did a quick P9 measure for the purpose of this post. We'll eventually do a multi-seat, averaged frequency response for the P9.