Damping Factor - Interesting article

Another dimension of DF not often discussed is having high DF, and high current through the treble. While conventional speakers tend to have their low points in the mid-bass, where most amps have the highest DF, ESL's are essentially capacitors, and have their lowest impedance in the peak frequency.

If an amp has 'high current' (which is a bit of a myth; current can't exist without voltage) then it will at all frequencies.


The problem with ESLs is that they typically vary by about 9 or 10:1 in impedance from bass to treble, but their efficiency doesn't vary in lockstep as it is supposed to like you see with box speakers. So an amp that doubles power as impedance is halved is typically way too bright on most ESLs. Martin Logan got around this (sort of) by making their ESLs very low impedance in the bass (4 ohms) so they are only 0.5ohms at 20Khz. Even most solid state amps have troubles into that impedance, thus reducing the brightness that would otherwise manifest.


Generally speaking most ESLs don't follow the voltage rules; IOW their impedance curve does not match their sensitivity through their frequency range!

This is too simplistic a view, and I was thinking specifically just related to the basic output stage which does typically behave much like a voltage source, and is usually configured as a voltage follower, and with a light load (lighter than a speaker), behaves as a voltage source, and with load, as a voltage source with an element of constant and variable impedance.

One thing you are not getting has to do with the application of feedback. What I have said in that paper is true if the amp has none- what you say above is true if the amp has enough feedback to allow it to behave as a voltage source.

Again, I find this is too simplistic of a view. Simply saying 35db is too little feedback without taking into account the frequency response of the feed-forward and feedback paths, not to mention what the inherent feedback is in the output stage if you are considering that separately makes any hard number in the sand questionable. The statement makes assumptions about the linearity of the feedback network as well. Ditto for Gain-Bandwidth, which is one number, but gain at frequency is far more relevant. Instrumentation op-amps may have very high gain-bandwidth, but are useless at 20KHz.

This is a bigger deal that it would seem to appear; if the amplifier has too little feedback (less than about 35dB) the consequence is that the feedback itself will introduce distortion, mostly composed of higher ordered harmonics (and some IM). Somewhere in the area of 35dB and north the amp finally has enough feedback such that is can actually compensate for the distortion introduced by the feedback itself.

The bottom line is this is all about Gain Bandwidth Product and the resulting loop gain- both of which have been insufficient in the prior art. The Benchmark amplifier is one of the very few non-class D designs that actually gets the feedback into the ballpark. So if you want really natural sound, you either go with an amp like that or go with an amp that uses no feedback at all- and deal with the simple fact that it won't work on all speakers, which is also true of an amplifier that is a perfect voltage source! So you'll have to audition the speaker and amp combination in any event.

Not sure the justification for this statement. Their response w.r.t. voltage, is fairly flat from mids-highs, with usually a bit of a dip at high frequencies. An amplifier that doubles in power as the impedance is squared will keep the most consistent anechoic output.


The brightness is more a factor of their emission shape and how they will interact with most room, and the resultant room response, which will differ from a "point source" dynamic driver.

The problem with ESLs is that they typically vary by about 9 or 10:1 in impedance from bass to treble, but their efficiency doesn't vary in lockstep as it is supposed to like you see with box speakers. So an amp that doubles power as impedance is halved is typically way too bright on most ESLs. Martin Logan got around this (sort of) by making their ESLs very low impedance in the bass (4 ohms) so they are only 0.5ohms at 20Khz. Even most solid state amps have troubles into that impedance, thus reducing the brightness that would otherwise manifest.

An amplifier that doubles in power as the impedance is squared will keep the most consistent anechoic output.

Buddy, you went the wrong way. I know of no amp that doubles power as impedance goes from 4 to 16 Ohms. That is certainly not an ideal voltage source anymore.


You are also conflating dispersion with relative differences in amp output
vs. impedance.


The feedback loop of Class D amps looks so different from linear amps I’m not at all sure we should be judging by the same criteria, assuming 35db is even correct. Last I looked there were at least 3 different ways in which Class D amps used feedback.


It is not unreasonable to imagine an amp with a 300 DF at 20 Hz but 50 at 20 kHz. At 20 Hz the amp’s output Z is ~ 0.03 Ohms, but at 20 kHz ~ 0.16 Ohms. With a normal dynamic speaker, these are not really significant, but ~ 0.16 Ohms is significant when compared to the 0.5 or less an ESL may present. Some quick math, and you’ll see about 1/4 of the amp’s output voltage is gone.


Of course, this is all hypothetical and math-y. Listening alone will tell you if you’d like it.


Best,
E

Every amplifier has some feedback. Even emitter resistor is a form of local feedback. The problem with global feedback is, that it corrects with a delay (phase shift from input to output). This delay produces overshoot in time domain (odd harmonics in frequency domain). 40dB feedback means, that amplifier has 100x higher gain without feedback. Since amplifier delays signal from input to output, signal fed back and summed at the input is late. It make very little difference for slow sinewaves, but for fast changing input signals amplifier, for a moment, has 100x higher gain and overshoots. Benchmark is trying to time correct it with separate error amplifier (two sets of output transistors). This overshoot shows in some Stereophile reviews at square wave response.