Indicator of what? Phase margin is an indicator of how much feedback you can use before it gets instable. In fact, that's where the creativity of the designer comes in. Finding ways to increase feedback and still keep it stable.
There are whole threads here about how to wring the last drop out of an amp, in this respect, by clever and creative compensation topologies.
Bruno is a real artist in this - his class D amps use very large amounts of feedback, in very novel ways, and are invariably praised for their sound quality.
jan
Exactly.
One needs to make sure one's amp can do at least 30 kHz, and preferably higher, under full power open loop conditions, and remain stable and clean. With as little overshoot as possible. AND into real world loads, not lab resistors.
I've seen too many great amps ruined once you put a 1 uF cap in parallel with an 8 Ohm load and ask for a 10 kHz square wave. And you get something which resembles modern hip-hop art high on some bad weed instead of a square wave. I admit that's a cruel test, but it does show up a thing or two.
On the subject of time, I must also admit I tend to be old fashoned in many ways. For example, I would never dream of making a power amplifier based on op amps, although people are doing it, and some wondefully too (e.g. my own aforementioned Karan amp is based around Burr-Brown's 2604 op amps and still manages to sound devine, Dan d'Agostino - Mr Krell - has designed a power amp for Aragon based on op amps, etc). Somehow, I feel much more comfortable using discrete devices, with which I keep complete control over the resulting sound. I can tweak it any way I want.
As an indicator of sound quality.
I don't remember ever seeing a discussion about this.
I only see phase margin discussed in terms of stability. Obviously, too little phase margin hampers stability. The engineering dictum is to have enough phase margin to prevent instability, and then add some for good measure.
Obviously, too little phase margin sucks regarding sound quality. Given a continuum of phase margins, increasing it is better until it is at least up to a certain value to prevent instability. Therefore, does it keep getting better, sound wise, with further increase? While the amount of FB and phase margin go hand in hand, often discussion is about the amount of FB, ignoring what phase margin that amount of FB gives as an added variable.
I'm saying that I don't think that oscillation could be good for sound quality.
This is all random speculation. The answer is DEEPER than either the amount of feedback or the phase margin. IF just the amount of feedback made 'clearer' sounding circuits, then we would have switched to IC op amps, decades ago.
Halcro made the 'best' high feedback discrete circuitry. Where are they, today?
It is true that zero negative feedback can sound 'soft'. Is that because that is the most natural sound, or a problem? It is anybody's guess.
However, for the Blowtorch, zero feedback was the best overall compromise. For power amps, even I cannot be sure.
Halcro made the 'best' high feedback discrete circuitry. Where are they, today?
It is true that zero negative feedback can sound 'soft'. Is that because that is the most natural sound, or a problem? It is anybody's guess.
However, for the Blowtorch, zero feedback was the best overall compromise. For power amps, even I cannot be sure.
Phase margin is just one measure of how close the amp is to loop oscillation. There is also gain margin, another measure of stability. Neither of them tell you much about sound, provided they are good enough. You could plot the loop gain and see what shape it is, especially near the (1,0) point. Or plot the closed-loop frequency response gain and phase.
Well, sorry, you were definitely not saying that 
If you mean that, I'm all in agreement....
The phase margin is the difference between the phase shift and 180 degrees at the frequency where the loop gain has dropped to 1 (0dB).
You can have all the phase shift you want when the gain has dropped below one, because you need gain > 1 to sustain oscillation.
You also don't want to cut it too close because even if it doesn't oscillate, you can get ringing with fast signals. So that's where the rules of thumb come in.
But I don't know what the link is to sound quality here?
Edit: DF96, we xposted...
jan
If you mean that, I'm all in agreement....
The phase margin is the difference between the phase shift and 180 degrees at the frequency where the loop gain has dropped to 1 (0dB).
You can have all the phase shift you want when the gain has dropped below one, because you need gain > 1 to sustain oscillation.
You also don't want to cut it too close because even if it doesn't oscillate, you can get ringing with fast signals. So that's where the rules of thumb come in.
But I don't know what the link is to sound quality here?
Edit: DF96, we xposted...
jan
Last edited:
just nothing related to audio or audio amp stability with dynamic driver loudspeakers
reputedly the test originated from concerns about driving ESL, which are a tiny fraction of the market- but it is flawed even for that purpose
most ESL systems use a step-up xfmr when driven from conventional audio power amps - their impedance doesn't look like a 1 uF cap at any frequency - there is winding series R, leakage inductance - typically there is a resonance ( = pure resistive load at the resonant frequency) just an octave above audio - at higher frequencies the amp sees the xfmr parasitics - its practically unconnected to the ESL panel C anywhere near its loop gain intercept frequency
so a 1 uF Cap isn't related to any load a typical audio power amp will ever see
with load isolating series L the ringing is way below amplifier gain intercept - it isn't telling you anything about amp stability
if the designer wants the amp to "look good" on this artificial test power output Zobel damping networks can be added
and please spare us the "resistive load" Strawman - any characterization of the sorry state of "conventional" audio amp engineering has to at least include the content of Self, Cordell's books - multidriver loudspeaker + XO complex load, I,V phase angle, peak currents from special test signals that "pump" the load resonance’s are discussed, plenty of paralleled output Q in the example circuits
Last edited:
Give us a break John. Calling a reasoned technical discussion 'random speculation' is a load of horse manure.
If you think there's an error, just say so.
jan
Is it incorrect that phase margin determines stability? Is it incorrect dat whatever the feedback, the amp itself always works open loop? Is it incorrect that feedback in itself does not determine sound quality, but that things like (internal) overload, slew rate limiting, non-zero Zout, non-linear Zin, low PSRR, underpowered supplies etc determine sound quality?
Whole tropical forests have been cut to write engineering books that explain these issues, that back them up with facts and figures. You have these books. My foundation sits on your bookshelf, and you know it.
Random speculation, my foot.
jan
no electrostatic panel has uF capacitance - the n^2 reflected C appearing at the primary of the 1:n step up transformer may be aproximated by uF capacitance, in series with xfmr wiring R, leakage L
care to give us the step up xfmr frequency dependent specs?, up to MHz (typ modern audio pwr amp gain intercept) - remember I am talking about the unreasonableness of applying a pure uF C directly to a conventional audio power amp output
loading the amp directly with a high Q Cap and applying a 10 kHz square wave (disabling any input filter the designer included) was popular for a while in audio mags late 70's-80s
but when using a conventional audio power amp ESL are always driven with a step up transformer with substanial parasitics, limited coupling bandwidth
direct drive ESL amps are special purpose amps designed for the load
care to give us the step up xfmr frequency dependent specs?, up to MHz (typ modern audio pwr amp gain intercept) - remember I am talking about the unreasonableness of applying a pure uF C directly to a conventional audio power amp output
loading the amp directly with a high Q Cap and applying a 10 kHz square wave (disabling any input filter the designer included) was popular for a while in audio mags late 70's-80s
but when using a conventional audio power amp ESL are always driven with a step up transformer with substanial parasitics, limited coupling bandwidth
direct drive ESL amps are special purpose amps designed for the load
Last edited:
I guess my point is that I get so tired of the one-variable thinking of the subjectivist camp; or any camp, for that matter. Someone plugs a new power amp into a system and declares it worse than the prior amp. Is this because the amp is worse, or because the new amp is so much better that it exposes other weaknesses in the chain, like ringing tweeters or crappy coupling caps? I really don't intend to cause this thread to spin off ad nauseum into a debate we've all seen before. But declaring, as a generality, that X amount of feedback is better or worse than Y amount of feedback, without considering other variables, is totally pointless.
No, it does not work such way in reality. Plugging better component always improves quality.
It does not work like something better exposes something wrong. Instead, after enjoyment of results even better results are desired.
Lucky you... I am getting tired of arguments of theoreticians who don't have experience in measuring and improving sound quality by search for correlation between sound quality and measurements...
Last edited:
An externally hosted image should be here but it was not working when we last tested it.
jcx's description of what an ESL load looks like is way more accurate than a 1/2u cap that's a fact.
The picture shows measurements including the impedance & phase of the original Quad electrostatic.
Well, I would have to disagree with that as a blanket statement. A lot of times, a better component makes some aspects sound better, but some worse. How many times has your favorite music changed after a component change; i.e., a crappy recording now sounds good, but a preferred recording is not as special? But I didn't intend to go up that alley any farther. I was merely trying to use it as a symptom of one-variable thinking. Everything is a balancing act. This involves dealing with many variables, and simply stating that X is better than Y is counter productive without stating conditions.
- Status
- Not open for further replies.