The discussion I think John finds interesting (as do I) is the fact that based on the measurements one would be inclined to think that -150dB is the noise floor.
When there is no signal only some (inaudible) hum is probably measured and the DAC would appear to have no noise at all.

Good analog stage and high FS output voltage can bring these levels.

What I meant was that the 'poles' of just one frequency are already plenty and very low. Below audible levels when viewed as a single tone.
In music there are MANY frequencies (I see them more as 'organised' pink noise) which all will produce 'poles' and these will effectively 'lift'  the noise floor.
Some of these poles may cancel out and others may amplify in a similar way as paralleling noise signals does.

Say all the poles stay below -120dB (which may arguably be the case).
All these poles will 'create' a noise floor of around -120dB on the FFT BUT some poles will add and others cancel out so the peak levels of the poles may even be a few dB higher.

When you were to measure those -120dB poles you would measure a much higher level as it is the energy of each pole that is added.
As there are a lot of them the noise floor would thus appear higher than -120dB WHEN a signal is present.

IMO it is still moot as the noise floor of a recording is usually MUCH higher than what the DAC produces WHILE it is playing music.

Opinions on this thesis are welcome.
Quite sure there are people here that know more about this than I think I do.

Yes proly not the best way. However, noise in there would yield sort of a ball park or sanity check number for the effective number of bits (after compensation).

The problem I have with the Stereophile claim is one of the few things that often bug me.
To me they are skating on thin ice as they could not measure any differences so the added noise floor obviously doesn't increase.
The remark 'I heard an improvement in sound quality that I can attribute only to the JitterBug' is therefore not a proven fact, just a subjective finding that does not correlate with the facts.
This 'phenomenon' that he hates when it happens' is something I just view as it is .... a personal opinion.


The noise floor does vary with signal levels and thus is also recording dependent.
With this I mean the DR, which I really don't view as an absolute rating for perceived SQ, just as an indicator with certain genres at best, also has some influence when the average music SPL is lower to the noise floor.
Even more so it is very DAC dependent as all DAC's have different amounts of distortion at different levels depending on the tech used.

Noise in the lower part of the frequency is less objectionable than 'hiss' when at the same level.
Those who have listened to a white noise recording and have swept a peak filter from bottom to top of the frequency range will know this.
This silly bugger actually had been looking for his own personal hearing threshold and really doesn't care about other peoples findings as much as my own.
Detecting your own thresholds is very enlightning by the way.

Here is the thing (at least for me) is that when noise is added to the signal, on purpose, it only becomes audible in those parts of the spectrum that contain little 'musicial content' and are high up in the freq range (hiss).
Add noise to 'busy' rock and pop music and the 'added' noise floor can be unexpectedly high before you hear it.
A simple way to test your hearing for this is controling bit depth.
Yes, quantization noise is not the same as added white noise as the noise is more present in the higher freq (treble range) but is does give an idea of audibility.

Still, for the better (measuring) DAC's out there the 'real life increased noise floor' I expect to be magnitudes lower than the noise floor of recording equipment.
Just have a look at the noise spectrum of a silent passage in a well recorded classical recording where mics are relatively far away from instuments.

So for me measurements are usefull if they have a real meaning and show values exceeding my own audibility threshold.

I often silently smile and refrain from comments when some people make claims of audibility when interpreting plots or make claims that (acc to me) have no basis on well performed technical measurements.

Still the 'hunt' for accurately linking measurements with subjective SQ is interesting to me.
Start with testing ones own thresholds though.

Just my opinion of course.

There's absolutely no way to prove or disprove JA's subjective experience.

Yup, agreed.
Depending on who's side of 'the fence' you stand one chooses to find the subjective evaluation more valuable or the (meagre ?) measurement results.
So it's a personal choice, based on what one thinks they know (for certain), whether one chooses to ignore the measurements or the subjective findings or take parts of both for granted.

Even then ... when one chooses for the measurements, it is essential to understand what the implications of a few plots, made with steady state signals at specific levels only, can have on the audible side when music is reproduced which is nothing like any solid state test signals.
This is what John touched and I responded to.

The other way around, if one chooses to find the subjective found differences to be of more importance than this is a choice based on personal experience.
Those vary and may be equally inadequate as passing judgement about measurements with a lack of understanding them.

Test signals are only used in the way they are now (by those pesky test/engineer guys) to show the 'edges' of performance with a certain set of signals that has been agreed upon by engineers to show the limits of equipment as far as 'they' consider the parameters essential to signal integrity, and not so much the implications to the audibility of it.
Artificial signals that do not exist in real music are used to make the 'borders' more visible.
It takes a certain amount of knowledge about test signals and the relevance of it to make an informed guestimate of the impact the electrical performance has on music signals.

I found the 'what do you do for a living' thread quite valuable.