[JohnR]

Following are spectrograms of around a second of silence followed by 9 seconds of a 1kHz -20dB sine wave streamed to a Linux streamer board at 96 kHz. This is captured at the analog output of a DAC (Topping NX4 DSD) connected to the streamer by USB. Guess which is which: high-power charging port on USB hub, 5V4A switching power supply (Meanwell); linear power supply (Teradak U9VA) with good USB charger cable; linear power supply (Teradak U9VA) with cheap USB cable.

This is not a Raspberry PI but the same type of thing.

(If you're wondering what the graphics mean, the purple horizontal bands are the 1 kHz test signal. A pure white background would mean noise less than -130dB, any color on the background means there is noise, with the noise increasing as the color goes from white to red to yellow to green - per the scale on the right. BTW ignore the spike at 1 second, that's just the change from no signal to signal.)

A



B



C



D

[JohnR]

I'm wondering now if this is partially an instrumentation issue, as I can't hear the noise even in the worst case with my headphones plugged into the amp Topping... Still, the only difference between the four is the power supply to the streamer (and cable)... any thoughts?

[randytsuch]

Looking at the graphs, I'm having a hard time figuring out the frequency of the noise.
My guess is either noise is outside of audible band, or too low amplitude to hear or below the noise floor.

Since different PS's give different measurements, seems unlikely to be measurement error.
Are measurements between the same PS repeatable?

I'm also going to guess the meanwell switcher was quiet.

Randy

[art]

Without reading closely.....................my guess is the middle two are switchers. But, now that I read the text more closely, they are the same supply, but with different cables. I'll pass on guessing which cable is the worst. The baffling part is what is causing the cyclical pattern.

That leaves 2...................so, I am going to guess the truly dreadful one is the USB hub.

That means the Meanwell is the quietest. Which I find...................uh.............. ......reluctant to believe.

Can't easily explain this, especially in layman's terms, but I can tell you a certain very popular brand of 5 V wall wart really screws with some of my SW radio stuff. It creates 60 Hz sidebands on the oscillator. So, we try to avoid SMPS at all costs, which is almost impossible. The SMPS on our computer monitor makes 60 Hz crap on our phase noise plots. Depending on its physical location, wrt to the phase noise rig and what is being tested. Don't ask how much time was lost trying to solve that one.

[JohnR]

Yeap, the Meanwell is the quietest. The linear supply is the two in the middle, with the "good" cable (beefy cable that came with my Moto phone) first then the cheap cable. Although strictly speaking this is not measuring noise from the supply but as measured at the DAC output. It could be coming through the ground somehow...

I think the cyclical noise is actually on my mains supply and is getting through into the audio differently for the different supplies.

Randy, the freq scale is on the left, 0 to 48 kHz. The noise is audible when I plug headphones into my "measurement system" aka ADI-2 Pro and crank up the gain. It's probably about time I got some isolation transformers to try and isolate what's what in these situations. This is actually a sideline from what I was trying to measure, I initially was using the USB hub but wasn't expecting to see so much noise in the analog output. I thought the results were interesting...

Oh, another thought, the Meanwell has all kinds of certification, whereas neither of the others do. Datasheets etc here.

[audioengr]

1kHz test signal is inadequate for a power supply test like this.  Digital needs high di/dt, so you need a 1 amp transient in the load to see if the supplies are different or even any good for digital.  Most LPS will droop.  Switchers may do better, but more noise.

[art]

Yeap, the Meanwell is the quietest. The linear supply is the two in the middle, with the "good" cable (beefy cable that came with my Moto phone) first then the cheap cable. Although strictly speaking this is not measuring noise from the supply but as measured at the DAC output. It could be coming through the ground somehow...

So, I got it right! Shocking.

Oh, another thought, the Meanwell has all kinds of certification, whereas neither of the others do. Datasheets etc here.

That is good to know, as I just bought some for a proto of yet another project that will never become a product. BTW, that series is no longer made, and was replaced with the GST series.