Home HiFi Electronics Audio Test and Measurement for DIY IT Stuff Downloads Bookmarks WaveSpectra and WaveGene mirror

USB noise measurements

I recently performed some noise tests on a USB DAC driven from various USB devices.

The Raspberry Pi was the quietest when running on mains power, but even quieter when powered from a linear PSU and quieter still on battery.

It should be noted that the noisiest scenario was from a mains powered PC into another PC (via the sound card) which is to be expected as a ground loop is formed (the Pi has no ground reference).

Given that most HiFi equipment is isolated from ground, this is unlikely to be an issue - however I plan to measure using the DAC from PC USB into a (grounded) HiFi preamp to see if the problem still exists.

11/02/21 Update: I attached the D10 to a PC and connected it to my HiFi system (the preamp has a ground connection). There was a slight increase in noise. I then connected the PC to my home build DAC (PCM1794 with Amanero USB->i2s interface) and the noise was reduced a lot, but still not as quiet as going directly from the DAC. So as expected, ground loops do raise the noise floor.

I also tried an older Intel NUC (Celeron CPU) and that too was noisier than the Pi.

UPDATE I tested the NUC using a linear bench PSU and it was as quiet as the Pi.

The best scenario would be to use a Raspberry Pi or NUC with a linear PSU. A laptop running on battery would also be a good option but not very practical. When using a Pi - *beware of ground noise* when using the official (or any mains -> USB type) PSU, as clearly some noise was injected, but this will vary according to your individual ground arrangements in your HiFi.

The other alternative would be to use a USB isolator, however if your DAC requires USB3 an isolator won't be cheap.

Unless your music player uses a lot of CPU, the Pi (or the NUC on a linear PSU) is the ultimate player here.


Hardware used for tests:

Raspberry Pi4 Model B 2gb RAM / Intel NUC Celeron CPU (DN2820FYK) / PC: Amd FX6100 / 16GB / Nvidia GFX / Asus Xonar Essence STX sound card / Altor Audio Olivine 2 isolated USB ADC

The linear PSU used to power the Pi is custom built using a toroidal transformer and LM350 3A regulator.

Software used:

WaveSpectra, Sox (for generating tones on Linux), Virtins Multi Instrument (SNR/SINAD measurements)

Conclusion: The USB bus (and its power supply) is not typically noisy - any noise is created by earth / ground loops between the connected equipment and the DAC caused by unequal ground potentials. In some cases it also appears that the ground of a sound card in a typical PC is at a slightly different potential to the USB ground, which then results in PSU / motherboard noise being injected into the resulting analogue signal - this caused a great deal of noise in the loopback test as shown in the images below. However this is probably a rare scenario in a typical computer playback / streamer setup.

ComputerPower sourceTHD+NSINADSNRNoise
Raspberry Pi 4BOfficial PSU-102.75dB102.75dB103.86dB-106.85dBFS
Raspberry Pi 4BLinear PSU-104.23dB104.23dB105.71dB-108.69dBFS
Raspberry Pi 4BBattery-104.52dB104.52dB106.19dB-109.17dBFS
PC (isolated ADC*)Mains-98.49dB98.49dB98.97dB-101.97dBFS
PC (ground loop)Mains-70.32dB70.32dB70.41dB-73.39dBFS

*NOTE the isolated USB ADC I'm using has slightly inferior noise figures to my internal sound card - hence the disparity in noise measurements.

Each of the graphs below is annotated with the specific scenario.

USB noise measurements
USB noise measurements
USB noise measurements
USB noise measurements
USB noise measurements
USB noise measurements
USB noise measurements
USB noise measurements
USB noise measurements

Related items: