0 dBFS 1 kHz sine, ES100 analog volume at 0 dB and +6 dB
It's well into clipping at + 6dB. The onset of clipping happens somewhere in between:
0 dBFS 1 kHz sine, ES100 analog volume at +3 dB and + 3.5 dB
I should have measured the distortion at +3 dB.
With 16-ohms on both channels, clipping happens earlier:
0 dBFS 1 kHz sine, ES100 analog volume at +1.5 dB and + 2.0 dB.
Again, I should have measured the distortion at +1.5 dB.
Of course, you wouldn’t be listening to music this loud on the typical 16-ohm IEM. Here’s what happens with the analog volume control at -6 dB:
1 kHz sine, analog volume at -6 dB, into Focusrite Forte line in (left), into 16-ohm load (right)
From here,
we can estimate the output impedance.
The calculation can also be done for the measurements made with the analog volume at 0 dB. The result is similar—a little lower. Radsone claims that it's around 1 ohm.
J-Test
16-bit J-test signal, USB playback
24-bit J-test signal, USB playback
USB is limited to 16-bit playback, and locked at 48 kHz, so we’re probably not seeing data-induced jitter from the 24-bit file. It might as well be a pure 12 kHz sine. I don't have any AptX HD-capable Bluetooth devices, otherwise I would have tested there instead.
Unlike the equalizer, the choice of reconstruction filter doesn’t have much of an audible effect. I don’t know why they prioritized this over something much more useful, like a parametric EQ. Here’s what their impulse responses look like:
Left to right: Sharp Roll-Off filter, Short Delay, Sharp Roll-Off filter
There seems to be a short delay between the channels. Notice that the peaks of the impulses don't line up. The Apple and Google dongles don’t have this inter-channel delay, which can also be seen here:
Edge of 60 Hz square wave (fake step response): Sharp Roll-Off filter (left), Short Delay Sharp Roll-Off filter (right)
The crossfeed is interesting. It looks like they try to implement some sort of ITD. When it plays back a file with an impulse on only channel, the signal on the other channel starts roughly 2.5 ms later:
Left: Impulse on right channel (top trace), crossfeed level 10 signal on left channel (bottom trace). Right: The same, 2.5 ms per division. Note the difference in the vertical scale for the two traces.
So, is
2.5ms x the speed of sound close to whatever it ought to be? You be the judge.
My scope has more noise than the Focusrite Forte, so I recorded on the Forte the crossfeed signal caused by impulses on the other channel. Here’s what a frequency analysis in Adobe Audition and Audacity look like:
Audacity - frequency analysis of crossfeed signal
Adobe Audition - frequency analysis of crossfeed signal. Blue: spectrum of recorded impulse on right channel. Green: spectrum of resulting crossfeed signal on left channel.
Finally, a word about the noise level. I simply don't hear any hiss on the ES100 with my usual headphones and IEMs. The noise performance is better than most of my other Bluetooth headphones: Bose QC35, Jaybird Freedom/X3 and Skullcandy Grind wireless all have very soft, but audible hiss. I don’t find their noise offensive and I really don’t notice unless I pay attention. The Bose, in particular, has very little self noise that’s almost always better than the outside noise that it replaces with its noise-cancelling circuitry. My Airpods are pretty much dead silent.
With some UE600vi IEMs, I can easily hear the EarStudio’s hiss when its analog volume is at 0 dB. But nobody will be listening to crazy levels like that with sensitive IEMs. The advantage of the analog volume control is that the background hiss is also attenuated whenever you decrease the volume. With pop music, a good level with the UE600vi is around -36 dB on the analog volume control. With classical, it’s around -27 to -30 dB. There is no noise to be heard whatsoever with this kind of attenuation.
Noise level in dBV, into 16-ohm load on both channels
