I was interested in using sound cards to plot pairs of signals in XY mode,1 and so measured the performance of a couple of popular USB sound interfaces.

## Basic set up

I fed a 1V peak-to-peak sine wave into the input of the units, then grabbed a 10s sample with SoX,2 and looked at the statistics:

\$ sox -d foo.wav trim 0 10 stat

Input File     : 'default' (coreaudio)
Channels       : 2
Sample Rate    : 48000
Precision      : 32-bit
Sample Encoding: 32-bit Signed Integer PCM

In:0.00% 00:00:10.01 [00:00:00.00] Out:480k  [    |    ]        Clip:0
Length (seconds):     10.000000
Scaled by:         2147483647.0
Maximum amplitude:     0.383392
Minimum amplitude:    -0.383575
Midline amplitude:    -0.000092
Mean    norm:          0.243714
Mean    amplitude:    -0.000053
RMS     amplitude:     0.270708
Maximum delta:         0.001953
Minimum delta:         0.000000
Mean    delta:         0.000744
RMS     delta:         0.000835
Rough   frequency:           23
Volume adjustment:        2.607						

From this I extracted the ‘RMS amplitude’ at various frequencies.

I also used the sum of the ‘Maximum amplitude’ and ‘Minimum amplitude’ to get some measure of the interface’s DC-offset. This isn’t perfect, and is likely be noisy because we are using extremal values.

Also, our logging window is not synchronized to the incoming signal so we are unlikely to see an integral number of cycles. The incomplete cycle will have a non-zero DC component, and at low frequencies this might be a significant contribution.

If there are $$n$$ cycles the worst-case fractional contribution will be $$2 / \pi$$. At 20Hz, $$n = 200$$ which implies a fractional noise level of about 0.16%.

I made no attempt to measure the phase changes at different frequencies.

## Results

It is simplest to just see the frequency responses plotted on a log-log scale:

### Behringer UCA202

This sells for about £23 and seems well-regarded by Amazon reviewers. It offers a stereo input on a couple of phono plugs, and stereo outputs on phono plugs, headphones and optical.

Internally it is based around the Burr-Brown/TI PCM2902.3

The data above were collected at a sampling rate of 48kHz.

Key results:

• Excellent low-frequency response: -3dB point at about 4Hz.
• Reasonably flat pass-band.
• Very little DC offset. The measured fractional offset is about $$\pm 6 \times 10^{-5}$$ in the mid-band, and this might just be noise coming into the interface.

Other experiments show the UCA-202 inverts the signal. They also suggest that things behave oddly at the high-frequncy end—I don’t understand this and so can’t explain it succinctly.

### Generic CM6202 interface

These sell for about £7 and offer 5.1 analogue outputs and stereo inputs on 3.5mm jack sockets, and optical input and outputs.

Internally, the main chaip is a C-Media CM62064

The data above were collected at a sampling rate of 44.1kHz.

Key results:

• Rather poor low-frequency response: the -3dB point is about 42Hz.
• Visible pass-band-ripple.
• Significant DC-offset: roughly 2% across the entire frequency range.

Other experiments show that this CM6206 interface is not compatible with an iPhone or iPad using the Apple’s Lightning to USB cable because it draws too much power.