If you believe Benchmark's marketing then the answer's a definite 'no' - https://benchmarkmedia.com/blogs/ap...audio-myth-switching-power-supplies-are-noisy
However if you read further down, they do admit that 'many switching supplies are noisy'. They put this down to older designs using lower switching frequencies and 'low cost' designs like USB chargers. So they recommend that switching supplies be optimized for audio applications.
Nowhere in that article is there any mention of common-mode noise. That could be because its not really an issue when the inputs of an amp are balanced (via XLRs) as they are in Benchmark's AHB2 amp, which is their exemplar of 'switching PSUs done right'. With balanced ins, pin1 (direct to chassis) on the XLR takes care of any 'leakage' current and so the conductors carrying audio (pins2 and 3) are unaffected. However if the inputs to an amp are unbalanced then there are only two conductors available to carry leakage (common-mode) currents, both of which are in the audio circuit. Since practically all equipment I have has only unbalanced ins/outs the issue of CM noise is a major concern to me.
Common-mode currents arise because the oscillator in the SMPSU is connected to the primary of the power transformer and so inherently capacitively couples to the secondary via transformer parasitics. Here's a technical treatment : https://www.eetimes.com/power-tip-47-tame-conducted-common-mode-emissions-in-isolated-switchers-part-1. CM currents are ultrasonic, they can be seen on FFTs which look over a wide enough bandwidth or where the noise floor is low enough in the audio band. For example Exasound characterize their DACs both with and without galvanic isolation on the USB input and show the difference this makes on the FFT within the audio band : https://www.exasound.com/portals/0/Images/e32-Measurements/exaSound-e32-noise-no-isolation-800.png. With USB isolation there's no 'grass' above ~-168dB : https://www.exasound.com/portals/0/Images/e32-Measurements/exaSound-e32-Noise-Floor-800.png. These two plots indicate there must be some CM->DM conversion taking place as CM (presumably coming from the PC as USB host) won't show up on a typical DM measurement.
Suppressing CM noise currents isn't very straightforward and the EETimes article points out why - they are sourced from a very high impedance. So high impedances are needed to make a dent in their magnitude. Ferrite beads (for example) are useless as their impedance typically tops out around 1kohm. The first line of defense against CM noise is a CM choke and a typical insertion loss plot looks like this :
The different lines represent different current ratings, the one with the highest loss has the lowest current rating. Notice above roughly 1MHz all these different current rating chokes behave about the same - their parasitic capacitance determines their insertion loss. They all have SRFs (self-resonant frequencies) between 100kHz and 700kHz, the frequency where their stray capacitance resonates with their inductance.
To improve the HF performance its a standard technique to segment the windings - with two segments on each side the insertion loss plot looks like this :
More segments are possible of course but I've not seen any commercial CM choke with more than two. These two plots are from Murata devices designed for power lines - signal line chokes with high insertion losses close to the audio band aren't so easy to find. However on signal lines, transformers can do a reasonable job of increasing the CM impedance.
Update :
The EEtimes article has a substantial number of images with broken links, here's a document which helps make up for the lack of schematics and waveforms : http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.123.2941&rep=rep1&type=pdf. Page 1-8 and onward are most relevant in the treatment of CM noise currents.
However if you read further down, they do admit that 'many switching supplies are noisy'. They put this down to older designs using lower switching frequencies and 'low cost' designs like USB chargers. So they recommend that switching supplies be optimized for audio applications.
Nowhere in that article is there any mention of common-mode noise. That could be because its not really an issue when the inputs of an amp are balanced (via XLRs) as they are in Benchmark's AHB2 amp, which is their exemplar of 'switching PSUs done right'. With balanced ins, pin1 (direct to chassis) on the XLR takes care of any 'leakage' current and so the conductors carrying audio (pins2 and 3) are unaffected. However if the inputs to an amp are unbalanced then there are only two conductors available to carry leakage (common-mode) currents, both of which are in the audio circuit. Since practically all equipment I have has only unbalanced ins/outs the issue of CM noise is a major concern to me.
Common-mode currents arise because the oscillator in the SMPSU is connected to the primary of the power transformer and so inherently capacitively couples to the secondary via transformer parasitics. Here's a technical treatment : https://www.eetimes.com/power-tip-47-tame-conducted-common-mode-emissions-in-isolated-switchers-part-1. CM currents are ultrasonic, they can be seen on FFTs which look over a wide enough bandwidth or where the noise floor is low enough in the audio band. For example Exasound characterize their DACs both with and without galvanic isolation on the USB input and show the difference this makes on the FFT within the audio band : https://www.exasound.com/portals/0/Images/e32-Measurements/exaSound-e32-noise-no-isolation-800.png. With USB isolation there's no 'grass' above ~-168dB : https://www.exasound.com/portals/0/Images/e32-Measurements/exaSound-e32-Noise-Floor-800.png. These two plots indicate there must be some CM->DM conversion taking place as CM (presumably coming from the PC as USB host) won't show up on a typical DM measurement.
Suppressing CM noise currents isn't very straightforward and the EETimes article points out why - they are sourced from a very high impedance. So high impedances are needed to make a dent in their magnitude. Ferrite beads (for example) are useless as their impedance typically tops out around 1kohm. The first line of defense against CM noise is a CM choke and a typical insertion loss plot looks like this :
The different lines represent different current ratings, the one with the highest loss has the lowest current rating. Notice above roughly 1MHz all these different current rating chokes behave about the same - their parasitic capacitance determines their insertion loss. They all have SRFs (self-resonant frequencies) between 100kHz and 700kHz, the frequency where their stray capacitance resonates with their inductance.
To improve the HF performance its a standard technique to segment the windings - with two segments on each side the insertion loss plot looks like this :
More segments are possible of course but I've not seen any commercial CM choke with more than two. These two plots are from Murata devices designed for power lines - signal line chokes with high insertion losses close to the audio band aren't so easy to find. However on signal lines, transformers can do a reasonable job of increasing the CM impedance.
Update :
The EEtimes article has a substantial number of images with broken links, here's a document which helps make up for the lack of schematics and waveforms : http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.123.2941&rep=rep1&type=pdf. Page 1-8 and onward are most relevant in the treatment of CM noise currents.
Last edited:
