Does power supply matter for DAC?

Many audiophiles believe that the power supply affects sound quality. However, it may or may not. So what about the power supply to the DAC? We powered the Raspberry Pi 4B from several different power supplies and compared the residual noise emitted from the RCA terminals of the I2S DAC HAT (HiFiBerry DAC+ Pro). In addition, we fed power to the USB DAC (TOPPING D50) from several types of power supplies and compared the residual noise.

We compared the following four power sources

1. 2.4A switching AC adapter purchased for 300 JPY (about $2.20 USD) at DAISO dollar-store. (white AC adapter on the right in the photo below)
2. 3.0A switching AC adapter purchased for 1700 JPY (about $12.60 USD) at a part store called Switch-Science. (Second black AC adapter from the right)
3. Allo's low noise switching power supply Nirvana SMPS. $59.00 USD. (third from right)
4. L.K.S. Audio's Linear Power Supply LPS-25. $169.98 USD. (fourth from right)

Power the Raspberry Pi 4B from these four sources and compare the outputs from the HiFiBerry DAC+ Pro's RCA jacks. In addition, we will also verify the difference between the power supplies for the TOPPING D50 USB DAC.

In addition to these, we also compare the isolated power supply of the Raspberry Pi and the HiFiBerry DAC+ Pro. The HiFiBerry DAC+ Pro is powered via the Raspberry Pi's GPIO. However, by removing the HiFiBerry DAC+ Pro's R14 dummy resistor and feeding power through the P3 pin header, the analog part of the DAC+ Pro can be powered independently from the power supply of the Raspberry Pi and the digital part of the circuit. The Raspbery Pi is powered from the Allo Nirvana SMPS and the HiFiBerry DAC+ Pro's P3 header pin is powered from the L.K.S. LPS-25, making this the fifth comparison.

A high-performance AD Converter, Cosmos ADC,  was used for the measurement.

Measurement accuracy can be further improved by using the Cosmos APU (center of photo) in the stage before the Cosmos ADC.

First, the Cosmos APU can be used as an input buffer for measurements, eliminating the low impedance problem of the Cosmos ADC, since the Cosmos APU has an input impedance of 20kΩ (balanced) / 10kΩ (unbalanced).

Second, the Cosmos APU has the ability to amplify the input signal to +34dB or +60dB, so if you send a signal of -60dBFS (1/1000 of MAX) from the DAC and amplify it to +60dB (1000x) in the Cosmos APU before passing it to the Cosmos ADC, noise other than the signal is also amplified. This allows more accurate measurement of the dynamic range.

The third function is a 1kHz/10kHz -30dB Notch Filter, which creates a -30dB notch at 1kHz or 10kHz in the input signal. 1kHz signal is sent from the DAC, the Cosmos APU attenuates the signal by -30dB before passing it to the Cosmos ADC. If you compensate +30dB before applying the FFT, you can measure THD+N more accurately. In fact, Audio Precision, the standard for audio analyzers, also uses this Notch Filter for highly accurate measurements.

For more details, please refer to the link below:

• Cosmos ADC (E1DA)
• Cosmos ADC Review: EARLY LOOK: E1DA Cosmos ADC - affordable high performance measurements for the audio hobbyist!
• Cosmos APU (E1DA)
• Cosmos APU Review: EARLY LOOK: E1DA Cosmos APU - high performance, low-noise pre-amp and 1/10kHz notch filter.

Measurements were made using Cosmos APU/ADC and REW (Room EQ Wizard) RTA as a spectrum analyzer to compare the residual noise caused by the power supplies. We compared the noise during playback of a silent WAV file (all data chunks are 0) on the HiFiBerry DAC+ Pro with power supplied to the Raspberry Pi from each power source. Noise is amplified by +60 dB in the Cosmos APU and passed from the Cosmos ADC to REW at 96 kHz/24 bit, with REW's RTA displaying a frequency band from 20 Hz to 40 kHz (see above). 

NOTE: If a 0dBFS (2Vrms) signal were sent with this setting, it would exceed the input range of the Cosmos ADC.

Let's start with a comparison of DAISO and Switch-Science AC adapters. Both show leakage from 50Hz household AC power and its harmonics (100Hz, 200Hz), but DAISO has more noise (150Hz, 9kHz, etc.).

Next, we also compare it to Allo's switching power supply and L.K.S.'s linear power supply. These also have 50 Hz leakage, but overall have lower noise than the AC adapter; the Allo Nirvana is a switching power supply, but it is very good with noise suppression.

Comparing noise with the TOPPING D50 USB DAC, the difference due to the power supply is barely noticeable. The left image below shows DAISO and the right image shows L.K.S. LPS.

The comparison seems to indicate that the L.K.S. LPS has slightly less noise, but it is hard to tell. (The images are exactly the same, so I will omit them.)

Next, let's compare the difference in dynamic range by power supply: the DAC produces a -60dBFS signal (1kHz Sin wave), which is amplified by the Cosmos APU by +60dB before AD conversion. (see above) Since we are measuring a -60dBFS signal, the signal-to-noise ratio is measured to be 60dB worse. Adding 60dB to the inverse of the measured THD+N, we can obtain the dynamic range. There is little difference in dynamic range depending on the power supply: the DAISO AC adapter shows a leakage at 50 Hz, but there is no other significant difference when compared to othr power supplies. Below are the measurement results of DAISO and L.K.S. LPS. The other images are almost the same, so they are omitted.

Next is TOPPING D50, but this one did not show any difference at all depending on the power supply. The following is the measurement result of DAISO and L.K.S. LPS. The other images are exactly the same, so I will omit them.

When powered from any power supply, the results were as follows

HiFiBerry DAC+ Pro's THD+N at -60dB = -50.5 dB / Dynamic Range = 110.5 dB

TOPPING D50's THD+N at -60dB = -60.8 dB / Dynamic Range = 120.8 dB

Finally, THD+N and SINAD (the inverse of THD+N) are measured: 0dBFS 1kHz Sin wave from the DAC, -30dB attenuation at 1kHz by the APU notch, AD conversion at 48kHz/24bit with Cosmos ADC, +30dB compensation of the signal after import into the Mac before applying FFT (see above). Again, only the DAISO AC adapter shows a small leak at 50Hz, but the others, including the Switch-Science AC adapter, show no difference at all. SINAD=91.9dB and THD+N=0.0025% with any power supplies.

Next is the measurement of TOPPING D50. Here, too, there is no difference at all depending on the power supply. SINAD=109.0dB / THD+N = 0.00035% for both power supplies.

Verification results:

1. For the HiFiBerry DAC+ Pro with I2S DAC, there was a difference in the noise from the RCA terminals depending on the type of power supply feeding the Raspberry Pi.
2. Separate power supply for Raspberry Pi and DAC showed a slight noise reduction compared to the linear power supply, but there was no significant difference.
3. As for the USB DAC TOPPING D50, no difference was observed between the types of power supplies feeding the DAC.
4. For both DACs, no differences were observed in dynamic range, SINAD, or THD+N depending on the power supply.

For DACs, there is almost no difference in sound quality depending on the type of power supply used.

However, when powering the Raspberry Pi, we do not recommend using an AC adapter like DAISO's AC adapter for charging smartphones. If you do not use a power cable with low DC resistance, the resistance of the power cable will cause a voltage drop, and the Raspberry Pi may frequently give the warning "Under-voltage detected!" It is recommended to use an AC adapter designed for Raspberry Pi like Switch-Science AC Adapter or this one.