This is a review and detailed measurements of the Asus noise cancelling audio interface. It is on kind loan from a member and costs US $34 on Amazon including Prime shipping.
This thing looks like a phone headphone adapter:
But unlike those devices, it also has an analog to digital converter with ability to perform "AI based" noise cancellation to your microphone. I had a hell of a time finding out what type of microphone as it is not documented in the horrible "manual." I put "manual" in quotation because it is one of those fold out jobs with 20 languages simultaneously in every section! Worse yet, it tells you absolutely nothing about the device. Anyway, I used an in-line mic from my IEM for voice testing. For instrumentation, I built an adapter for 4-bit 3.5 mm that it accepts to the balanced output of my Audio Precision analyzer.
Asus provides a control app for this device. Upon downloading I ran setup out of the zip folder. It ran but would get stuck at 0% progress. I eventually figured out that you have to extract the files yourself and then run setup after which it finished in an instant with no message.
Testing here has two parts: standard measurements for any audio interface and then objective and subjective testing of the noise cancellation logic.
Asus Ai Mic Adapter Measurements
At first I could not figure what the adapter is doing with respect to input voltage. It seemed to accept up to about 4 volts above which it got severely distorted so I went with that as that is the closes to 0 dBFS input:
Note that I am feeding it one input and it is generating stereo output. On the FFT response we immediately see a filter in action cutting off everything above 5 to 6 kHz. Noise dominates the measurements above distortion bringing SINAD down to 59 dB. Note that this is all with pass through mode with processing turned off using the adapter application.
Looking at our frequency response we get this:
I guess it makes sense that it attempts to get a free pass first with just filtering. Would have been nice if this was not the case when the unit is in bypass mode.
There are three mic settings and three levels of noise reduction in the app. Let's play with inline mic and level set to high:
My tone is continuous so you would think that it would confuse that with background noise and completely mute. It is doing so across most of the frequency range but for some strange reason it lets in a peak around 280 Hz. Same happens around 60 Hz but that may be due to mains hum that it considers to be valid signal.
To get a better handle on what is going on, I changed the input signal to white noise and then compared multiple settings:
Top trace is with the device in bypass mode. We see the facsimile of the frequency response as before. Turning on the mode to either low or medium produces the traces in the middle. Turning it to high, produces the bottom trace with produces highest level of attenuation. Strangely, if you keep it in high mode but select external mic, the response goes back to the middle as noted in red trace.
Next I ran the THD+N versus level and got this unique result:
The level at both low and high ends of the spectrum jumps forward and back indicating some kind of gain control to boost low level signals and some kind of compressor for high level signals. Again, I appreciate them doing this but a) they should document this and b) not have it be active when the unit is n bypass mode. Either way, your noise floor rises quite a bit compared to any professional interface by using this dongle.
To figure out how well the device handles impulsive noises such as keyboard clicking, I created a sine wave burst in the analyzer that jumps up 25 dB for 5 msec and then goes back down for the remainder of the 1 second. Here is what that looks like with the device in bypass mode and not:
One strange thing in bypass mode in blue is that the device seems to start with high noise floor and then goes back down after a couple of seconds. Not sure what is up with that but it was repeatable.
Turning on the mode with high settings, gives us the expected results: the gap in between is recognized as being continuously the same so it is heavily muted. Then when the burst times, it it is let through. There are two issues though: one is pretty high level of latency that is close to 0.2 seconds. This may impact the target market of gamers. Second, the peak levels are modulating up and down. I saw this in my earlier instrumentation where levels were dancing up and down constantly if I had the unit active. With such a steady state signal, it should not do this.
Listening Tests
I created a simple test. I used an in-line mic on my IEM and talked from 20 cm while tapping on the table about 10 cm away from it, again simulating keyboard sound of some sort. With the unit in bypass mode, recording was good quality for voice. Once I turned the unit on, the taps on the table in between my voice notes were reasonably muted but my voice was heavily modulated when the two combined. Levels would also jump up and down just like my instrumentation. This would be quite annoying to anyone listening to me. Maybe tapping would be worse but I should not have to pay a high price for noise reduction.
As a way of comparison, I tested the Nvidia RTX voice the same way. This worked far, far better than the Asus AI mic. While a bit of roughness was added to my voice at times, the filtering had little artifact otherwise. I would give the RTX Voice a grade of 6 to 7 (no noise reduction being 10) while the Asus AI mic would get 2 or 3. RTX voice is free if you have Nvidia graphics card so it is cheaper to boot.
Conclusions
The Asus AI Mic is an interesting concept to add noise cancelling to any microphone for your standard gaming, zoom call or broadcast. Alas, both objective and subjective quality is poor. Yes, you get some noise reduction but it comes at the cost of distorting your voice especially if noise combines with your voice. The fact that the device is crippled even when you are not using the noise reduction, you don't get a decent audio interface out of it either.
Overall, I cannot recommend the Asus AI Noise Reduction Mic Adapter. It needs a hell of a lot more "AI" in it to sound proper to me. And oh the delay needs to be lower as well and programmable.
------------
As always, questions, comments, recommendations, etc. are welcome.
Been busy with gardening and work around the house. Picked 20+ pounds of Italian plumbs from old trees on our property when we bought it:
They are so wonderful. We eat them fresh but then also dehydrate them and eat them throughout the year until the next season comes. I am so appreciative to whoever planted this tree in this property. We get all this wonderful fruit with no effort on our part. Please plant a fruit tree for yourself and future generations if you own your own home. You will make someone like me happy years from now!
Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
This thing looks like a phone headphone adapter:
But unlike those devices, it also has an analog to digital converter with ability to perform "AI based" noise cancellation to your microphone. I had a hell of a time finding out what type of microphone as it is not documented in the horrible "manual." I put "manual" in quotation because it is one of those fold out jobs with 20 languages simultaneously in every section! Worse yet, it tells you absolutely nothing about the device. Anyway, I used an in-line mic from my IEM for voice testing. For instrumentation, I built an adapter for 4-bit 3.5 mm that it accepts to the balanced output of my Audio Precision analyzer.
Asus provides a control app for this device. Upon downloading I ran setup out of the zip folder. It ran but would get stuck at 0% progress. I eventually figured out that you have to extract the files yourself and then run setup after which it finished in an instant with no message.
Testing here has two parts: standard measurements for any audio interface and then objective and subjective testing of the noise cancellation logic.
Asus Ai Mic Adapter Measurements
At first I could not figure what the adapter is doing with respect to input voltage. It seemed to accept up to about 4 volts above which it got severely distorted so I went with that as that is the closes to 0 dBFS input:
Note that I am feeding it one input and it is generating stereo output. On the FFT response we immediately see a filter in action cutting off everything above 5 to 6 kHz. Noise dominates the measurements above distortion bringing SINAD down to 59 dB. Note that this is all with pass through mode with processing turned off using the adapter application.
Looking at our frequency response we get this:
I guess it makes sense that it attempts to get a free pass first with just filtering. Would have been nice if this was not the case when the unit is in bypass mode.
There are three mic settings and three levels of noise reduction in the app. Let's play with inline mic and level set to high:
My tone is continuous so you would think that it would confuse that with background noise and completely mute. It is doing so across most of the frequency range but for some strange reason it lets in a peak around 280 Hz. Same happens around 60 Hz but that may be due to mains hum that it considers to be valid signal.
To get a better handle on what is going on, I changed the input signal to white noise and then compared multiple settings:
Top trace is with the device in bypass mode. We see the facsimile of the frequency response as before. Turning on the mode to either low or medium produces the traces in the middle. Turning it to high, produces the bottom trace with produces highest level of attenuation. Strangely, if you keep it in high mode but select external mic, the response goes back to the middle as noted in red trace.
Next I ran the THD+N versus level and got this unique result:
The level at both low and high ends of the spectrum jumps forward and back indicating some kind of gain control to boost low level signals and some kind of compressor for high level signals. Again, I appreciate them doing this but a) they should document this and b) not have it be active when the unit is n bypass mode. Either way, your noise floor rises quite a bit compared to any professional interface by using this dongle.
To figure out how well the device handles impulsive noises such as keyboard clicking, I created a sine wave burst in the analyzer that jumps up 25 dB for 5 msec and then goes back down for the remainder of the 1 second. Here is what that looks like with the device in bypass mode and not:
One strange thing in bypass mode in blue is that the device seems to start with high noise floor and then goes back down after a couple of seconds. Not sure what is up with that but it was repeatable.
Turning on the mode with high settings, gives us the expected results: the gap in between is recognized as being continuously the same so it is heavily muted. Then when the burst times, it it is let through. There are two issues though: one is pretty high level of latency that is close to 0.2 seconds. This may impact the target market of gamers. Second, the peak levels are modulating up and down. I saw this in my earlier instrumentation where levels were dancing up and down constantly if I had the unit active. With such a steady state signal, it should not do this.
Listening Tests
I created a simple test. I used an in-line mic on my IEM and talked from 20 cm while tapping on the table about 10 cm away from it, again simulating keyboard sound of some sort. With the unit in bypass mode, recording was good quality for voice. Once I turned the unit on, the taps on the table in between my voice notes were reasonably muted but my voice was heavily modulated when the two combined. Levels would also jump up and down just like my instrumentation. This would be quite annoying to anyone listening to me. Maybe tapping would be worse but I should not have to pay a high price for noise reduction.
As a way of comparison, I tested the Nvidia RTX voice the same way. This worked far, far better than the Asus AI mic. While a bit of roughness was added to my voice at times, the filtering had little artifact otherwise. I would give the RTX Voice a grade of 6 to 7 (no noise reduction being 10) while the Asus AI mic would get 2 or 3. RTX voice is free if you have Nvidia graphics card so it is cheaper to boot.
Conclusions
The Asus AI Mic is an interesting concept to add noise cancelling to any microphone for your standard gaming, zoom call or broadcast. Alas, both objective and subjective quality is poor. Yes, you get some noise reduction but it comes at the cost of distorting your voice especially if noise combines with your voice. The fact that the device is crippled even when you are not using the noise reduction, you don't get a decent audio interface out of it either.
Overall, I cannot recommend the Asus AI Noise Reduction Mic Adapter. It needs a hell of a lot more "AI" in it to sound proper to me. And oh the delay needs to be lower as well and programmable.
------------
As always, questions, comments, recommendations, etc. are welcome.
Been busy with gardening and work around the house. Picked 20+ pounds of Italian plumbs from old trees on our property when we bought it:
They are so wonderful. We eat them fresh but then also dehydrate them and eat them throughout the year until the next season comes. I am so appreciative to whoever planted this tree in this property. We get all this wonderful fruit with no effort on our part. Please plant a fruit tree for yourself and future generations if you own your own home. You will make someone like me happy years from now!
Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/