Caveat #1: I am in no way affiliated with iFi Audio (other than being a satisfied customer of their products). I purchased the iSilencer+ myself, I did not receive a review copy.

Caveat #2: I am a scientist by nature (and profession), and therefore try to remain objective in this strange, dark land of audiophilia. I am well aware that snake oil products exist in abundance, but I remain open to trying new products/technologies/ideas.

- - - - - - - -

I use the iFi iDSD Nano Black Series as a DAC/amp combo at home, as well as a standalone DAC in my office (feeding a Matrix M-stage v.1 amp). I wasn't necessarily experiencing noticeable USB noise issues in either of these two configurations... but you never know how/if things can be improved, right? So when the iFi iSilencer+ was announced at a mere £49, I was certainly intrigued. COVID-19 has kept me a bit bored at home, so I decided to give it a try.

When I received the new iFi iSilencer+ I was having trouble discerning by ear whether it makes a difference on the Nano and Sony MDR-7520 headphones. I don't have access to my (more revealing) office system for the foreseeable future, so I just have this setup to try.

So in my quarantine-induced boredom... I decided to do some objective measurements!

Any other measurements of devices like these that you're likely to find floating around are going to have been conducted with equipment that analyzes the signal itself (rather than how that signal translates to headphone performance). I wanted to do something different here for two reasons: (1) I don't have that fancy equipment, and (2) I wanted something that might be more representative of what I/you can expect in actual use, i.e., how is the noise floor affected as measured from the headphone playback?

So what I did was play a sine wave through the 7520s and record the output on my ThinkPad T470s screen microphone (like I said: no fancy equipment). I generated a 1000 Hz sine wave in Audacity at 0.005 amplitude (0.5% of maximum amplitude). Why so low? I wanted to test the iSilencer+ in its true use case: USB power. To make sure that the Nano powers through USB when I unplugged the cable to add the iSilencer+ in-line, I needed to turn the volume knob all the way down to power it off, so that I could power it back on after plugging the cable back in. This meant that the only way to maintain the same volume across the two recording conditions was to turn the volume all the way up (i.e., hard stop when turning clockwise). The tone at 0.005 amplitude yielded a normal listening level with the Nano turned all the way up, so I used this level to generate the tone. Having such a low signal volume also lowers the SNR, potentially allowing any benefits of noise-reduction to be observed more clearly. Recording and spectral analysis were done in Praat speech acoustic software.

These were the steps of the process:

1. Generate a 1000 Hz sine wave (0.005 amplitude)
2. Play sine wave through iFi iDSD Nano Black Series (iEMatch output) and Sony MDR-7520
3. Place headphone pad over microphone (placement doesn't move between recordings)
4. Record signal at 22050 Hz for 10 seconds
5. Generate FFT spectrum
6. Scale the audio files so that the 1000 Hz spectral amplitude is the same for both (i.e., normalize the signal amplitude)
7. Plot log-transformed frequency spectrum (50 Hz - 10 kHz)

Figure 1 shows the spectrum without the iSilencer+ in red and the spectrum with the iSilencer+ in blue. Figure 2 shows the spectrum with the iSilencer+ in red and the spectrum without the iSilencer+ in blue. What we're looking for here then are areas where the red noise rises above the blue noise in both cases.


Figure 1: log-transformed frequency spectrum without iSilencer+ (red) and with iSilencer+ (blue).


Figure 2: log-transformed frequency spectrum with iSilencer+ (red) and without iSilencer+ (blue).

Although I can't necessarily hear a difference, it seems pretty clear to me that the iSilencer+ is actually reducing noise in a real-world scenario: there is seemingly a lot of red that rises above the blue in the first image, but no red above the blue in the second image.

To empirically test the effect, I applied a 900-1100 Hz band-stop filter (with a 100 Hz bandwidth) to isolate the noise component in the audio; see Figure 3 for the resulting spectra (without the iSilencer+ in red, with the iSilencer+ in blue). I then computed the amplitude in 10 ms windows over the 10-second audio files to perform a Welch Two Sample t-test on the difference between the two.


Figure 3: log-transformed frequency spectrum of band-stop filtered audio with iSilencer+ (red) and without iSilencer+ (blue).

The average amplitudes for the filtered audio are 67.27 dB (without the iSilencer+) and 62.04 dB (with the iSilencer+): over 5 dB reduction in the noise floor! The t-test reveals that the difference (-5.23) is indeed significant: t(1,1936) = -29.564, p < 0.001, 95% confidence interval: [-5.57, -4.88].

In the end, even though I can't necessarily hear a difference, my own tests have convinced me that the iSilencer+ does have some effect. Since the iFi Nano has an extremely low noise floor as it is (in normal use), my hunch is that the noise-reduction would be even more effective in some other systems (but not all).

However, the effect seems to be large enough even in my case, that I'm going to keep the iSilencer+ for piece of mind.

5/5 stars given to the iFi iSilencer+ for actually doing what it claims to do, even in a "real-world" scenario... minus 1 star for not being able to actually hear the effect...