
Alpha Audio regularly conducts research into the influence of networking on playback quality. We have done blind tests, lab studies and also tested numerous switches live. The conclusion: it’s simply audible. But where do these differences come from?
Frankly, of course, we already know the answer. We have been working on this for years now. However, we have recently acquired a new measuring device, the Aeroflex PN9000. It measures phase noise at a very deep level. Now phase-noise and jitter go hand in hand, so in itself, resultst that come out of the Wavecrest should also come out of the Aeroflex. And fortunately it does.
Small, but consistent differences
The differences between the various power supplies are small. But they are constant and in line with what we measure on the Wavecrest. A while back we did a series around switches where we used the Wavecrest to measure jitter. That was an excellent first attempt where we got a good idea of noise generated by a switch and its impact on playback.
In this case, we take the same Dlink GS108 switch and only swap the power supply. Then we’ll see what this does in terms of jitter.
Measurement setup
The measurement setup is not very complex. We take our Volumio Primo and put an active probe on the 100 MHz clock crystal. This signal goes into the Aeroflex PN9000. As a reference, we use an Axtal 100 MHz crystal. (AXIOM5050ULN) (-80dBC/Hz at 1Hz / -115 dBb/Hz at 10 Hz).
We put the Volumio in the RF shield box. However, with the lid closed, things are acting up and the measurements are not reliable (the noise increases every round and we suddenly lose the lock…). So we had to leave the lid open. We still need to investigate how this is possible.
Phase noise measurements
As you can see, the measurements are seemingly close to each other. However, the scale is 10dBc/Hz. In short: between the top and bottom line is at least 5dBc/Hz. This is significant.
There is also a difference in the course of the curve. It is striking – but explainable – that the IFI supply is higher. This most likely has to do with the fact that it is 12V and therefore the regulators in the switch have to work harder. The switch is actually made for 5V.
Who thought a battery works better: nope…. the bump at 100 Hz is higher and eventually also at 1Hz the noise is highest. Almost 10dB higher than the standard power supply which oddly enough scores best there. On average, the Sbooster power supply seems to score best: it has no real outliers. The lab power supply has some bumps between 10 and 1 Hz. The stock power supply is certainly not bad either.
Who thinks: that blue line is weird! Yes. that’s right. As an additional test, we took an identical Dlink switch and modified it with an external clock input. This is what happens when we let a 25MHz VCXO clock the Dlink. So there is quite a bit of noise added. So be careful with external clocks. Unless skillfully implemented, it will only get worse.
Different noise pattern
What the external clock does in terms of noise, we can also see perfectly on the scope, since the Aeroflex can make the noise visible on a scope. You can see that the complete noise pattern changes.
Conclusion
Power supplies undoubtedly affect music reproduction. Even if they only supply a switch with power. The noise from the power supply simply creeps into the streamer and can affect all sorts of things there. Including the clock that times the dac. And that, of course, affects playback. It’s not rocket-science once you see how the chain is put together and how seemingly unrelated things are nevertheless connected somehow.
And for those who are now thinking: I use a wireless connection! Or: I use a wifi bridge … that too has an impact. Wifi network cards can be noisy. And a Wifi-bridge is still connected with a cable. And also the wifi bridge needs a power supply…. It’s not all that easy.
Jaap,
About 5 years ago a product was reviewed by a magazine with measurements that seemed off. We had the Prism dScope III measurements for that device and they showed a different result. So we tested the Phase Noise on another one at time 0, time 15 mins, time 1 hr and time 24 hours. Seems our problem above was the reviewer was using his reference that was in the system for days tested that then threw ours in cold. Cold the PN (@1Hz) was -61dB but there was a huge bump 6-11Hz, 15 mins -73.8dB, 1Hr -71dB (a bit weird), 24hr -79.2dB.
Just a thought you may want to run it for a couple hours before doing any real testing.
Thanks,
Gordon
We always leave the measurement equipment on for an hour before testing. Especially phase noise is very sensitive.