Cirrus / Wolfson sound card power supply noise sensitivity

I had a chance to do some quick measurements on the Wolfson card, I was curious as to how well it measured and also to see if the power supply sensitivity, referred to in the HiFi World review was visible at all.

From the Hifi World article: –

A major issue with this audio board is that it’s very dependent upon the quality of the power supply used to drive it. A noisy 5V switch-mode power supply unit (PSU) resulted in 76dB dynamic range from Line Out, but a quiet one (Apple phone charger) gave an excellent 110dB – an improvement of 34dB gained solely by using a better power supply.

More surprisingly, this also cleaned up the digital S/PDIF output to a Chord 2Qute external DAC, allowing it deliver the full 121dB dynamic range of which it is capable.

A good supply would be a battery of some sort, lead-acid or lithium ion, for top audio results, or a smoothed linear supply, bearing in mind that 5V is needed and batteries deliver 6V. Switch-mode supplies generate rubbish and the audio card is sensitive to this (the Apple charger is switch mode, but clean). See p320 of the WM5102 datasheet for more on this.

The effect of this is quite clear in the following trace – the upper trace is the effect of powering the Pi from an Anker multi-outlet USB charger. The lower trace is using the official and very inexpensive Stontronics Raspberry Pi 2.5A supply (available in White and Black).

There’s between 30 – 40dB of difference across the audio band, which correlates with the Hifi World measurements, a big improvement to be had through careful power supply choice!

Noisy PSU (top) vs quiet PSU (bottom) (Click to enlarge)

Noisy PSU (top) vs quiet PSU (bottom) (Click to enlarge)

As can be seen, the noise floor is lowered across the entire spectrum, resulting in greater dynamic range.

An ideal solution here would be to provide the Wolfson card with its own dedicated low noise supply, but this only makes sense in the context of a packaged unit, for most users the cheap Raspberry Pi supply works well. Alternatively, a genuine Apple charger, as mentioned in the HiFi World article, could be a good alternative, especially if running headless.

The official Pi supply holds up incredibly well, given that in those measurements it’s powering the official 7″ display, a Pi3 and the Wolfson Card.

I also did a quick comparison of THD for the Wolfson card vs. an AudioQuest Dragonfly Red, both measure very low levels, with the Wolfson showing lower levels of higher order distortion harmonics.

Wolfson card (top) vs. AudioQuest Dragonfly Red (bottom) harmonic distortion with 1kHz tone (click to enlarge)

Wolfson card (top) vs. AudioQuest Dragonfly Red (bottom) harmonic distortion with 1kHz tone (click to enlarge)

Wolfson card THD @ 0dBfs (click to enlarge)

Wolfson card THD @ 0dBfs (click to enlarge)

Wolfson card THD @ -10dBfs (click to enlarge)

Wolfson card THD @ -10dBfs (click to enlarge)

Wolfson card THD @ -20dBfs (click to enlarge)

Wolfson card THD @ -20dBfs (click to enlarge)

Both sound great to me, the Dragonfly reveals a little more subtle detail, which may come from its apparent lower noise floor, but it also shows how well the Wolfson card stands up to scrutiny given its very low price.

A few years back, the idea of creating a streaming music player, with 24bit / 192kHz capability at home from parts that can be just plugged together would have been unheard of. I have a number of Slim Devices / Logitech Squeezebox devices around the house, but with the exception of the Transporter, all have limitations with high-resolution files.

A cheap Raspberry Pi, on the other hand, can do a remarkable job, especially when the price of entry is so low.

Never has such high-quality sound been so accessible at such a low price.

This entry was posted in Engineering, Music. Bookmark the permalink.

2 Responses to Cirrus / Wolfson sound card power supply noise sensitivity

  1. John Tindle says:

    Really useful to have some performance measurements of a real system, thank you.

    Two questions on the blue traces:-
    1) which output are the traces measured on:- line out/headphone out …. because the datasheet for the 5102 identifies different performance for the two (if I remember correctly)
    2) you show measurements at 0,-10 and -20dB. Do these levels relate to different input files or different attenuation settings in the 5102? If these are different attenuation settings, what are your feelings on the best level to use ? ie do you compensate for the attenuation with higher gain settings in the power amp ….. I am planning to integrate with Quad 405 amplifiers, which have quite high input sensitivity.

    • Andrew says:

      The measurements were made at the line out and you are right, there are some subtle differences between Headphone and Line out, although they are specified under different operating conditions, which may account for the differences. If anything the Headphone output appears slightly better. I really ought to do some tests with real-world loads…
      The levels represented different input files, I run the line out at 0dB which, IIRC, equates to an ALSA setting of 28. Anything above this is digital gain I believe. The headphones run from 0dB downwards, depending on the sensitivity of the ‘phones I’m using.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.