[jcx]

quote:
--------------------------------------------------------------------------------
Originally posted by ppl since the Op-Amp is receiving it's feedback from it's own output rather than from the output stage the Phase shift of the output stage will not impact Phase margin as much as with the conventional method of one overall feedback loop. In most cases this removes the requirement of using Phase lead capacitance around the op-amp and allows the use of modest bandwidth Output stages.
--------------------------------------------------------------------------------

Of course the above quote touches on one of the more misleading concepts floating around these forums about the inner feedback loop; for the values in the PPA schematic the capacitive feedback from the parasitic capacitance of R6 is ~ 4X larger than the resistive term at the closed loop corner frequency of the amplifier without even adding in the larger layout related capacitance of the left ends of R5,6 bringing op amp output and negative input to adjacent pins of a socket – so yes R6 has aided stability by reducing phase shift but due to parasitic capacitance and not by any virtue of high inner loop gain set by resistive divider ratios (I’ve assumed GBP AD8610 = 25 MHz, loop gain = 11, parasitic cap ~0.3 pF; you should add ~1 pF for the socket and maybe 1-3pF more for the lacquered metal end caps of R5&6 touching)

Additional consideration of the AD8610 8&15 pF diff&common mode input capacitance makes it hard to know how much credence to place in sweeping statements about noise gain compensation when in PPL’s published circuit the effects of local capacitive feedback are of the same magnitude as the resistive feedback at the closed loop corner frequency. The capacitive effects quickly become dominant when the loop gain is lowered to 2-4X unless you lower feedback resistance as gain is increased – which then moves even further from Jung’s recommendation to match input and feedback Z to minimize the influence of the input capacitance nonlinearity

I have no problem with the idea that the Jung multiloop with unity gain output buffers may be a “sweet spot” among op amp based amplifiers for audible and DIY tweakablity but I think optimizing this topology or moving (hopefully advancing?) on to other circuit topologies can be aided by a good engineering understanding of the circuit

 

[morsel]

(deleted - I posted to the wrong thread.)

 

[aos]

Wow... I guess you're right, it would seem that the open loop is gain is constant only to about 100Hz. I was looking at that strange diagram which starts at 1MHz only and probably either mistook the phase curve for the gain curve, or guessed them right but thought that the gain curve is more or less flat as often it looks much steeper. But I didn't read the actual numbers which show clear 20dB/decade drop. And the gain is minimum 100dB (10**5) from the datasheet. So it's about 100Hz bandwidth for open loop gain. If I have read it correctly I'd never have told ppl about this chip. But since it really sounds wonderful I'm glad I did.

On the other hand I have to say that another favorite of mine - OPA2228 - has open loop gain of only a few Hz (or even sub-Hz). And it sounds bettwe without multiloop than with it. Perhaps it's not the fact that the open loop gain is constant in audio range but rather that it changes linearly. So if you set corner frequency to be either above audio range or below it, you'll have constant THD over audio range.

 

[Vdubjunkie]

thanks again ppl. That is exactly what I was hoping for. I'm sure I will keep quite consumed this evening!

 

[ppl]

I have implemented Walt's exact purposed method to balance input impedances as close as possible with a volume control. however this was further refined in the PPA by using 10K resistors on the op amp inputs rather than the 2.2 -3.3k as recommended by Jung. I do not believe Walt nor do I have an erroneous understanding of Input error reduction as enlightened to by Dr.T.I.M.


In your headwize post you give theory as to why the OPA-637 would not be stable in Multi-loop Amps and claim you do hear that Input error reduction and Improved stability offered of Walt's concept are in some way flawed. Several members of this Forum have used the OPA-637 in Multi-loop gains of 5 or less without any stability problems as posted in that very headwize traded you referenced to.

aos> I was aware of the low Conner frequency of the AD-8610 from the Preliminary data sheet after you mention that this device was just introduced and looked interesting. This was ok since I could easily see that when installed in a multi-loop amp the Open-loop band width of the complete Amplification system would be about 100KHz.and that implies all else being equal the THD would remain constant over the 20-50KHz. bandwidth and probably would remain constant up to 100 KHz. since op amp manufactures often error on the side of more GBP that the minimum specified in the data sheet. yes you are Right I am glad to have had the pleasure of how wonderful this chip can sound and doing such wonders on 3 mA is indeed a gift from God to us Audiophiles. "It don’t get no better than this".

 

[jcx]

PPL,

I find that in Jung’s Op Amp Audio series [ see the Walt Jung Archive: http://www.elecdesign.com/Articles/Index.cfm ] the juxtaposition of Gilbert’s articles analyzing input stage tanh distortion and Otala’s phase distortion paper with the comments on the inner feedback loop providing a flat gain over the audio frequency range creates a strong presumption that the inner loop feedback reduces the discussed conversion of input stage static nonlinearity to phase modulation nonlinearity – which is not true and easily seen in the unity gain buffer case, if you don’t like the conceptual step of replacing the unity gain buffer with a wire, try moving the inner feedback resistor to the buffer’s output and tell me how the circuit’s loop gain with repect to a signal (or distortion) at the op amp input has changed (assuming we are discussing a input op amp with high open loop gain and low open loop corner frequency such as the AD8610 or the AD823 used in Walt’s example multiloop circuit) – the input stage distortions are inside both feedback loops with the op amp’s integrating gain and 90 degree phase shift over nearly all of the audio frequency range and are subject to Otala’s conversion to phase modulation distortion

You even have half of the story correct, the distortion of the output buffer is modified by the inner feedback loop causing the loop gain around the output buffer to be flat over the audio frequency range and therefore the “benefit” of the inner loop is the reduction of phase modulation distortion from nonlinearity of the output buffer – not the input op amp input distortion

Where is your technical critique of my response over on HeadWise? Does it invalidate my argument that “not recommended for gain less than 5” not = “guaranteed to oscillate at gain 2”?

If you find technical problems with my arguments or presentation I want to know if I’m making a mistake or where I’m not communicating the point clearly

 

[aos]

jcx you do have a point that the parasitic capacitance of the feedback resistor is helping phase margin and stability of the opamp. But to prove or disprove multiloop effect on stability of the opamp, it's best to consider a different opamp. In this configuration OPA627 is driving a high impedance and low capacitance load (buffer) - further isolated by the resistor between them to insulate opamp from that capacitance (this is a filter by the way) - and parasitic capacitance helps already satisfactory phase margin so this is not a good example to discuss at all. Rather choose an opamp which is definitely not stable even under these ideal conditions and then see if the multiloop helps. All you can prove with your arguments is that the OPA627 would be stable even wihtout the multiloop. Which doesn't mean that the multiloop has no validity, only that in this case it doesn't matter (or doesn't matter as much).

I will admit that the multiloop mystifies me in some respects - e.g. supposedly the gain of the opamp is 100 but if you measure signal at its input and its output, it's going to be multiplied by the actual gain of the overall feedback, not 100. But I dont see why one of the ppl's arguments - that multiloop excludes phase shift introduced by the buffer from the opamps's feedback - would not hold. I can see why you'd want to replace buffer with wire, as really gross simplification is the way circuits are analyzed by hand and conceptually - but another thing we learn is that while you're doing that you must be careful not to exclude effects that do matter. Perhaps if you're just analyzing how the multiloop changes the behaviour of the input opamp, you may be able to ignore the buffer - if you assume ideal output stage of the opamp and ideal buffer with much higher bandwidth (so there's no additional phase shift). Then what you'd get is just a classic feedback around an opamp. High valued feedback resistor becomes irrelevant (except through parasitic capacitance) as the outer loop features far lower resistance... and suddenly it looks like inner loop has no funnction at all. This is why I don't get it completely.

I wish somebody would simulate this - it's not like opamp and buffer models for chips that we're using are not available. I would assume you professionals have access to those expensive simulators...