We are developing audio hardware for the Raspberry Pi. One issue here is the quality of the power supply on the 5V and 3.3V voltage rails. A high quality power supply is a critical aspect in audio circuits. The Raspberry Pi hardware is quite challenging for a power supply, let’s see what happens with different power supplies.
We will look at 3 different 5V power supplies. We use a cheap Nokia AC-10E 5V/1.2A switching mode power supply, an Apple 10W USB power adapter and an inexpensive Manson NSP3630 lab power supply. There are much better lab power supplies available than the Manson, but for our tests this will be good enough. If the Raspberry Pi is just powered on and booted, the 5V rails looks like this:
|Nokia: 4.90-5.54V (0.64V pp)
Noise average: 265mV
|Apple: 4.67-4.81V (0.14V pp)
Noise average: 71mV
|Manson: 4.94-5.05V (0.11V pp)
Noise average: 62.5mV
You can easily see, that the quality of the Nokia power supply is really bad. The Apple power supplies have a good reputation and looking at the oscilloscope picture you can see, that the signal is a lot cleaner. The Quality of an external lab power supply is still better – but it is also much more expensive and very bulky. Also the voltage is different. But this is not a problem, if the voltage itself is stable.
But what happens, if the Raspberry really runs under load. To stress the system, we transferred files to the Raspberry using SSH. This brings load to the CPU, the network controller and the SD card.
|Nokia: 4.63-5.42V (0.79V pp)
Noise average: 266mV
|Apple: 4.60-4.82V (0.22V pp)
Noise average: 80mV
|Manson: 4.94-5.05V (0.11V pp)
Noise average: 66.1mV
The cheap Nokia adapter is worse than before, the voltage drops and peak-to-peak swing further increases. The Apple power supply looks good. But having a closing look at the data, you can see, that the voltage dropped a bit. The Manson lab power supply have no problems at all driving this load.
Conclusion: If you want to produce high quality sound on the Raspberry Pi, you should not use a cheap 5V/1A power supply.
Update 14.10.: We did some more research on noise sources. Check out what we found out about networking.
Pingback: HiFiBerry Mini – distortion and noise measurements | Crazy Audio
Interesting. Since you have the setup ready, would you please also do another test?
Play a 1 kHz sine wave at a good volume and measure the spectrum and the distortion caused on the output by the different power supplies, instead of measuring the noise on the input supply.
Of course, to get a good reading you should connect both channel of the analog output to a proper resistor, to create a regular load (10 kOhm for the analog line out of the DAC, suitable for an audio amplifier).
This test I suggested you has never been done elsewhere (I searched) and I think it could be very useful.
we did this kind of test already with our own HiFiBerry DAC. However, we did not document it here. There was a very small impact on THD+N. When we get some time again, we will publish those results. Even with the worst power supply, THD+N was still below 0.006% (usually 0.004%). However, other sound cards might perform different. Therefore it is risky to say that in general there is only little impact.
I’m trying to make a little “music box” using a 5v amplifier and a couple of small passive speakers.
When I use the same power supply for the raspi and the amplifier, the noise introduced is huge.
Could your dac be a solution? of I must use independent power supplies for the amplifier and the raspi?
there are many possible noise sources. The DAC itself will not solve this problem. You should have an amplifier that is relatively resistant to different noise sources. Therefore, even an additional power supply might not fully solve your noise problems.
There is no easy way to say what will work best. It needs an analysis of all parts and the potential noise sources.
If you’d be inclined to test it and post your results, I could send you a Pi Power evaluation sample. I sell them on Tindie, and I believe they’re an excellent solution to the problem. It has a 2.1mm barrel connector and accepts 6-15 VDC and will generate up to 2A @ 5V with 25 mV P-P ripple (when the input is > 9V. Below that the ripple rises up to 45 mV @ 6V in).
While this does mean you still have to come up with a DC power supply, the problem of finding one with such a wide range and with reduced current requirements (at 12 volts a Pi with WiFi takes less than 300 mA) and – most importantly – no voltage regulation requirements (apart from 6-15 VDC) vastly simplifies the problem.