A Rethink of THD/THD+N (or why both tell us almost nothing)

How does a source play into this? What if the source had odd harmonics? I'm probably asking very basic N. Pass ideas.

 

I think this is where amp/speaker pairings make a lot of sense. And ideal amp is considered to have flat phase but speaker/headphones have phase that goes everywhere. A non-linear phase might actually match a speaker better depending on the profile ‍♂️

 

There are reasons to hold transducers/stages with << 15dB or so negative feedback levels, to different standards than most kinds of active gain stages, based on psychoacoustic principles (see Cheever's thesis, for example). Besides, just because one, or more, element in the auditory chain is less than ideal does not make adding other less than ideal elements more acceptable. Each will add their imperfections/colorations to the whole and will be audible.
(see the necessarily simplistic white paper I posted above).

 

Are you referring to electrical or acoustic phase? Most single driver headphones aren't half bad actually phase-wise (both acoustic and electric). A well-behaved headphone will have a an electric phase shift at the fundamental resonance frequency and acoustic phase will just reflect low and high pass filter behavior at the frequency extremes.

Multi-way speakers and headphones of course are a different ballgame, however a well-designed amp should not care too much about electrical phase swings. As for acoustic phase, I've witnessed good audible results from crossover phase shift correction via all-pass filtering. It's done on the speakers I own and more known companies like ATC and PSI Audio seem to do it to a great effect. Of course, mentioning time domain stuff to the line art appreciation gang is heresy, but then again - 20 years ago directivity would probably get the same reaction too. Need to see their face once they hear about off-axis phase response...

As for non-linear distortion audibility, I have conducted some single blind testing with a small audience and the results were interesting to say the least. We compared two DAC's with opamp output stages, yet significantly different amounts of THD. Here are some points of notice:

1. When voltage matched with a 1kHz sine with a 4 digit precise voltmeter, the more distorting setup would sound 2-3dB louder and was immediately detectable. This forced us to do acoustic level matching.
2. With less spectrally complex tracks many of the listeners would prefer the more distorting setup. The less distorting setup was described as "too naked".
3. Spectrally complex tracks favored the less distorting setup every time with all of the listeners.
4. My go-to test track "Ghost of Perdition" was initially booed away with the more distorting setup with most listeners being unacquainted with the song and blaming poor recording quality for the track being unlistenable. Using the less distorting DAC was a huge eye opener and remained the starkest contrast between the two setups.
5. At low SPL's the differences would appear to be harder to discern with all types of music. Cranking the volume to concert levels (we used a passive 4-way PA setup in free field conditions) largely decimated the more distorting setup as the music would turn into mush.

Most of the psychoacoustics literature I've encountered deals with overall system distortion as detected by the measurement mic and not electronics THD+N. With music audibility thresholds are usually stated around 1 - 2% to 1 - 5% THD, depending on the musical content; listening to test signals can bring that down to 0.1% THD. H3 is 2x times more objectionable than H2 and higher orders usually are 6x to 10x more objectionable than H2 at the same magnitude.

 

They shi

Are you familiar with the BBE Sonic Maximizer? It’s adds phase shift and group delay with decreasing frequency.

http://www.bbesound.com/products/sonic-maximizers/default.aspx

 

No. I have no idea what it's supposed to be doing.

 

This phono stage was specifically designed to keep the high order harmonic distortion and the multitone distortion below the audibility threshold of c. -110dBC at full scale output.
The 1kHz, 10kHz and multitone plots show this.
Below full scale the distortion decreases monotonically.
The RIAA compliance, over production spread, is better than 100mdB p-p.
The two elements of the metric used are satisfied by this.

 

I've always rolled my eyes as people discuss distortion measurements in the thousands of a % as if it means anything. I look at that and think it better be fast or there's too much feedback. (T.I.M.?)