An electrical engineers perspective on the Nordost video.
First, skip the video, the pdf is available on their website.
This isn't scientific, we don't know any of the controls, the nature of the testing equipment, vibrations present in the room, etc. We need the entire test setup, the source file and the ability to reproduce it in other locations to call it scientifically valid.
Ignoring that and moving on to their claims....
First, what they are doing is looking at a raw PCM file, raw audio samples.
They take this raw PCM, burn it to a CD and play it through the player, then sample it with a (hopefully) high end ADC to get it back into the computer, then compare the samples.
The entire thing is really 'blah' they don't mention how they overlay the signals, they also don't mention how the signal was sampled. Presuming this is just a CD player the audio is stored as 16bit 44.1khz audio.
To directly sample the output of the player and compare it to the source with any accuracy they will need to highly oversample the signal with the ADC and down convert it to 44.1khz carefully (with an appropriate algorithm). Otherwise they have no idea how the sampling points line up. The results could be very random if they just try to sample it randomly with a 44.1khz rate for the ADC. One test could be started at the perfect time and the sampling points overlay very well, another test could be started at the wrong time and the sampling points could lie directly between the sampling points in the source in the time domain. Of course there is also the issue of jitter on both involved clocks to deal with as neither will be exactly 44.1khz.
This actually gets more complicated on these 'high end' cd players as they usually take the output of the CD at 44.1khz, massively up sample it (for some unknown reason to me) then spit it out through high sampling rate DAC's. Annoying this usually results is a 44.1khz to 192khz conversion which is just bad as 192 is not a multiple of 44.1 so some tricky has to be done on the samples to make the conversion properly. The signal then goes through a reconstruction filter, gets fed into a line driver and spit out the line level out.
With all the sampling rate conversion going on in this signal chain trying to compare anything to the source samples as a single case is a lost cause.
This offset in sampling timing could account for some of the slope variations they mention in the video, but are curiously absent in the pdf version. Sampling offsets are also quite effective at messing up FFT's. As an aside whats up with all the high frequency noise on the FFT with their equipment hooked up?
At the end of the paper they claim that many of the differences present in the first test are the result of are greater variation in group delay for the stock setup which should be the real focus of study and could also be largely influenced by sampling rate conversion after capture of the signal.
This is really where i have a hard problem with the paper. I can actually think of a few ways that the power conditioner and the isolation table could actually remove noise from the system. I can only think of 1 far fetched way in which group delay could be effected.
All this comes down to needing more information on the test setup as well as running many many tests and averaging the results to get to something conclusive.
The way i would do this test is to use two as close to identical setups as possible and compare the audio output in the analog world with an oscilloscope, just subtract the signals from each other, which any half decent scope can do in real time. Get a base line with both units having the same power/stand, then change one and see if there is any difference.
Keep in kind i have done 0 tests, but from a circuit design stand point the following are at least theoretically possible ways in which their products could provide some benefit, i highly doubt they do or they would just measure this and publish it.
Vibration:
Generally has little to no effect on well designed PCB's and IC's. However most audio equipment i've opened up likes to make large use of inductors (for good reason) and large components with occasionally long leads.
Inductors generate a magnetic flux field, depending on the current sometimes a quite large magnetic field, that is after all, how they store energy.
Even shielded inductors generate a field inside the shielded enclosure. Whenever something conductive moves inside a magnetic flux field it can generate currents. So if have an inductor with a large field and its vibrating, unevenly as is usually the case, it will generate some transients on the signal passing through that inductor as well as currents on other conductive items inside its flux field. There are a lot of variables to this, the nature of vibrations, the rate of change of the flux field, etc.
I would expect currents generated by this effect to be quite small. But this could effect any piece of metal within the flux field of the inductor, this is part of the reason why the inductors on the power supply input are usually shielded and isolated from the audio circuitry but not all inductors are well protected.
If vibrations are introducing currents in other parts of the circuit these could be transfered to the audio signal. In contrast to what an earlier posted said, the DC power generated for use on the PCB of any electronic device is not perfect. It will always contain some higher frequency. Circuits designed for audio use are designed with cleaner power supplies and with high power supply rejection ratios (the amount of noise that couples from the power supply rails to the output of a given device). Numbers around 50-65db, varying with frequency are normal for relatively high end DACs. Thus noise present on the power supply lines for any of the supplies in the system can make it way to the signal path. In well designed systems this is accounted for and designed to be at or under the noise floor. Of course if there is a ton of noise overlaid on your mains line it could be audible. Again thats unlikely, unless your trying to run an audio demonstration at a trade show with a crap power feed and trip the breaker
![]()
))
As for the group delay. The induced current could, completely theoretically and unlikely, mess with the group delay across the frequency spectrum for some filter topologies. Again I highly doubt this would have an impact.
tl:dr
Comparison of the signals to a digital master on a sample by sample basis with so much sampling rate conversion going on is foolish.
Do the comparison in the analog world, run it 100+ times and average the results.
Publish your entire test setup and source track so others can verify your results.
