Martin

Following on from your comments yesterday

Recently I attended an industry event specifically aimed at how to understand and combat various ingression of conducted and radiated RFI/EMI emissions from virtually DC to 86GHz (vehicle radar systems)

Now one of the lectures addressed low frequency (for RF engineers anyway!) up to 1Mhz and its effects on analogue circuits which was quite enlightening.
While there I was discussing the acquisition of a new piece of test equipment with a large industry leading manufacturer of research equipment, the applications engineer mentioned he had just recently sold a respectable oscilloscope to a very well know UK audio company, what surprised him was the quality of the rest the equipment in their test facility. I quote 'It was like stepping back to the 70's all that was missing were the Open University bearded presenters' The particular company is located due west of Winchester.

Now with regard to actually measuring jitter in circuit before and after the device under test (DUT) pretty much all audio analyzers have a limited measurement bandwidth of 250Khz there is an exception the R&S UPV which has an option to go as high as 400Khz.

So why would you need to go any higher than say 200Khz (192Khz high resolution files) after all BEFORE any fancy up-sampling hardware or software was applied 192Khz is the highest general bit/format rate that is commonly used.

Now Nyquist theorem states that you pretty much require double the sample rate to produce the desired frequency rate, hence 44.1Khz which should be 22.05Khz but is rounded down to 20Khz (human hearing frequency range lol) so if you look at a Hifi News sample rate chart for the music files they review you will see that a 192Khz files produces a 96Khz plot (half the original sample rate).

So why would you need to look at higher frequencies than this? after all most listeners do not have the hearing range to acquire 20Khz let alone 96Khz!

Well in the test equipment world if we are looking at say frequencies around 50Mhz mark the bandwidth requirement to accurately look at this figure would need to be five times that of what were are measuring so a 250Mhz bandwidth scope would be needed. (digital sampling scopes)

Now because we can not actually hear above lets say 15Khz adjusting for the age of the forums users then what's all the fuss about?

Very fair question and this is where it divides the yes it is and no it cannot possible be camps:-

Because the frequencies involved (lets say over 100Khz) are not directly audible then it cannot possible make any difference to the perceived sound quality therefore its all bullshite!

Er no, what I am talking about are the frequencies that interact with the way the equipment actually works and its direct effect on the resultant quality of the sound. What you really need to look at as well as the spectral information shown on audio analyzer plots is what happens far further up the frequency ranges, and how specifically rfi and emi can and do effect clock timing, serial data transmission, power supply noise ingress both ac and dc, internal interconnects (cabling) and any other items that are connected to that reply chain.

So by looking outside the usual parameters it is very straightforward to perform a before and and after analysis of both the audio spectra and those at much higher levels, for instance a great many clocks have harmonic noise which spreads at least 30 sometimes up to 50 fold from the fundamental so in reality way over 2Ghz which in real terms is 2 Billion Hz as apposed to 20 thousand Hz so many orders of magnitude increase in frequency. Yet by helping to remove a lot of the spectra noise (which can also contaminate other critical areas of the circuit board and its associated components which again have a knock on effect with equipment noise and ground plane contamination). It can and does effect the lower frequencies to a sizeable degree which in turn has a noticeable effect on sound quality imho.

After remedial work has been carried out (which could mean a board redesign rather than a simple bridging wire or component change lol) then a retesting of the effectiveness of the rework can be assessed by performing the same measurements again and comparing.