I plan to do THD measurement of the audio equipment that I design using REW and a sound card. My ESI Audio U24 XL I have at home didn't perform well during testing of the analog interface and this is why I was looking for better options. In this article I present my findings evaluating the fitness of the Focusrite Scarlett 2i2 Gen3 sound card for THD measurement.
For the measurement, the sound card was connected to my notebook running Debian buster using the provided USB cable.
Each line output was wired to the corresponding analog input using a short, but cheap cable. Audio interconnections were unbalanced.
Noise floor was measured in 192kHz mode and 96kHz mode.
FFT was done in 96kHz mode.
With all controls set to minimum, the noise floor is as good as it gets with 24Bit resolution. What strikes me is the increasing noise floor in the higher kHz range. However, this should not matter for normal audio use.
With the monitor set to maximum position and gain is set to lowest level, the high frequency noise floor now increases earlier in frequency and higher in level. This is still outside the normal audio spectrum.
With the gain set to maximum position and monitor set to lowest level, now the noise floor is around -100dB and falls off at higher frequency with some spikes in the higher kHz range. I assume those spikes are artifacts and they are well outside the normal audio spectrum. However, for FFT of high frequency signals, this affects the measurement.
This is the noise floor with gain and monitor set to 50%. At -130dB, this acceptable, but far from exciting.
Since I experienced issues in 192kHz mode, I repeated the measurements with the sound card set to 96kHz operation:
Above plot shows the noise floor with all controls set to minimum.
This is the noise floor with the monitor set to maximum position and gain set to lowest level.
This is the noise floor with the gain set to maximum position and monitor set to lowest level.
Above is the noise floor with gain and monitor set to 50%. At -135 to -130dB, this acceptable, but again far from exciting.
FFT was done with the sound card set to 96kHz operation.
Above screenshot shows THD with gain and monitor set to 50% level. This is short of the level where the clipping warning of the device turns orange (-6dBFS).
Although this device does not show artifacts at 1kHz, but has a tendency to show such artifacts at 8kHz, above is a plot at 1200Hz for direct comparison with the ESI U24 XL that I tested earlier.
Low frequency behaviour is okay.
Used in 96kHz mode, the sound card performs well. In theory it could work in 192kHz mode, but shows weird effects like very high and slanted noise floor when doing FFT. I assume that this is not having to do with the sound card itself, but the overall setup. I experienced similar issues with the ESI sound card I tried to route into a virtual machine. My assumption is that the issue is related to the USB or management of the USB in my notebook. The noise floor is lower than the one of the ESI U24 XL. In the audio frequency range, the high frequency increase of the noise floor does not matter. Selling for roughly 150€, the sound card offers reasonable value. For me this is not a keeper since the sound card doesn't work in 192kHz mode for me. Using a different cable and connecting to the USB-C port of the notebook didn't resolve the issue.
Here is some background information about the device:
- This is how it identifies on the USB:
- The device works plug and play with Linux. It also registers a a mass storage device. Accessing the mass storage reveals some files.
Unfortunately I wasn't able to have a look under the hood since this would have left marks and I wanted to return the device in mint condition.