Audio Asylum Thread Printer

Ralph

In Reply to: RE: A list of sonic properties "un-measurable" (or, not easily measured)? posted by genungo on July 28, 2014 at 20:15:20

The human ear/brain system is very sensitive to higher ordered harmonics (5th and above), particularly the odd orders.

If these harmonics appear as the result of distortion of the signal, the result to the human ear will be that the signal will sound louder than it really is because the ear/brain system uses the higher orders to calculate how loud a sound is.

Add to that the fact that the ear/brain system is tuned to be the most sensitive at birdsong frequencies and the result is that trace amounts of odd ordered harmonic distortion can be very difficult to measure, but can be heard as a brightness/hardness in the resulting reproduction, because the ear/brain system translates distortion into tonality.

It is this reason that a 2nd order harmonic is interpreted as lushness or warmth.

Imaging is a different matter but is a function of bandwidth and resulting phase linearity. We can't hear phase of individual tones, but we can make out phasing information over a band of frequencies, if that phasing information is part of the soundstage in a recording.

These facts were not well understood 45-50 years ago, and there are still some who dispute them today although the ear's sensitivity to odd ordered harmonics is well known and easily proven with very simple equipment.

Now the use of Global Negative Feedback in an amplifier or preamplifier design imposes issues that relate directly to the above comments. There is a certain amount of time that it takes for a signal to propagate through any audio circuit. Not surprisingly, this time is known as Propagation Delay. The fact of Propagation Delay causes timing errors when GNF is used: the higher the frequency, the later and less useful the feedback becomes. In fact one has to be careful of GNF in wide bandwith circuits, as its use can lead to oscillation as the phase of the feedback signal can be so altered relative to the input (due to propagation delay) that it is actually positive rather than negative feedback.

So how this plays out is that a high frequency event that travels through the circuit is then applied back to the input, out of phase with original signal, as a correction voltage. But because it is arriving slightly later than it really should, the input signal has changed so the correction voltage is not correcting what it should. This results in trace amounts of odd ordered harmonic distortion, which can go up as high as the 81st harmonic(!). In addition, IM can be added as well due to intermodulations at the feedback node. This has been known for some time; Norman Crowhurst wrote about this phenomena over 50 years ago.

It is for this reason that some designers have eschewed the use of GNF in their designs. The idea is that while greater overall distortion is the result, most of it is less objectionable and less audible to the human ear, while the types that are more audible (higher orders) are not present at all.

Of course the ear translates to the lower orders to tonality as mentioned earlier, the trick here is to use all the techniques known to keep such distortions to a minimum. And example might be to use differential circuits to eliminate the even ordered harmonics through cancellation.

Distortion can mask detail so it is important to keep it down. So a low distortion circuit should also be smoother and more revealing to the ear as the harmonics that irritate the ear are not present- this allows the music to presented in a more intact state.

General Asylum

here's one: odd ordered harmonic distortion

Post a Message!