traderbam said:
This is just blind stupidity.
You state you agree with Rodolfo's approach, but are too blinkered to realise that Rodolfo's equation 2 is IDENTICAL to the one derived by the method you ''disagree'' with, if you delete his 1/A' term.
I cannot even be bothered to demonstrate how for the umpteenth time.
Further, do you REALLY know what you referring to when you allude to the so-called ''inverse-transform'' here?
http://www.diyaudio.com/forums/showthread.php?postid=1069266#post1069266
Re: Re: Re: Give SPICE a try (& HF Ingress)
I do not recommend the use of ferrites on the outputs of power amplifiers, partly due to possible nonlinearity. I recommend 1-3 uH air core inductor paralleled with 1-5 ohms. In some cases I use a torroidal air core inductor, as it does a better job of containing the magnetic field created by the output current.
I also recommend R-C shunt Zobels to ground on both sides of the inductor. A 0.05 uF in series with about 8 ohms seems to work well.
Bob
Nixie said:
I do not recommend the use of ferrites on the outputs of power amplifiers, partly due to possible nonlinearity. I recommend 1-3 uH air core inductor paralleled with 1-5 ohms. In some cases I use a torroidal air core inductor, as it does a better job of containing the magnetic field created by the output current.
I also recommend R-C shunt Zobels to ground on both sides of the inductor. A 0.05 uF in series with about 8 ohms seems to work well.
Bob
traderbam said:
While I agree that the Vanderkuy-Lipshitz argument about HEC being representable as NFB, with all of its attendant stability issues, I also support the view that there is something unique going on here, even if only in the nature of the implementation. And I stress strongly the importance of viewing it from both an NFB perspective and an error correction perspective. Indeed, I strongly support the use of the term error correction in describing it. This, even as all of us know, total correction is impossible in any real circuit. This is even true for feedforward error correction.
One needs to be careful in describing the circuit as having 30 dB of feedback based on the ratio of error reduction I achieved on THD-20. Indeed, it seems not quite right to refer to it as 30 dB of effective NFB at 20 kHz, since in order to correct the harmonics to that degree, which is where the distortion lies, the effective NFB factor needs to be 30 dB at those harmonics. Thus it probably needs to be 30 dB up to at least 60 kHz.
Cheers,
Bob
traderbam said:Of course.
By about 30dB.
Sure it is.
Is this the best way to implement NFB around a FET output stage? Maybe it was in 1984. I don't think it is now. I could be wrong, but I want to find out.
NFB around an output stage is not a new idea. Bob's version of Hawksford's circuit is interesting to me not least because it uses positve feedback and it doubles up the function of the bias transistors. What I don't really like on the face of it is that it needs 10 transistors, it needs adjusting, it isn't very symmetrical (electrically), and at its nominal feedback setting it has poor stability into capacitive loads - so it needs an output inductor.
Small signal transistors are cheap. Actually, if you are being fair, I think the net added number of transistors is closer to four or six compared with a quality non-ec output stage.
My circuit does not need adjusting to function exceedingly well. The circuit is dead nuts stable in that regard. Build the prototype with 1% resistors and a pot. Adjust the pot in the prototype. Replace the pot with a fixed 1% resistor in the production models and you are off to the races.
I don't know what you mean by my circuit not being very symmetrical. Please elaborate. If not symmetrical, do you care for electrical reasons or cosmetic reasons?
In my view, one should always use an output inductor on any output stage, with or without EC. Please elaborate on your reasons why you think it is bad to use an inductor.
Cheers,
Bob
Nixie,
I’m sorry if the post order makes it seem as if my post on the previous page could be a criticism of Brian’s, somebody must not be following closely, I can't imagine your excuse, after all we're hardly into the 2nd thousand of posts in this thread...
“traderbam” seems to share my understanding of the issues (although I'm careful to stick to Linear system arguments, with injected "disturbance" standing in for weak nonlinearity)
both of our efforts in this thread have been directed to countering the argument that “error correction” is fundamentally different from “conventional” negative feedback and that error correction could have fundamentally different properties that the control community had been missing for half a century
A recurring problem seems to be that block diagram math shows “perfect correction” - which comes from ignoring real world gain-frequency limits and the problem that simple real number arithmetic can lead to non-realizable infinite gains where complex number calculations that include the gain-frequency limits of real devices show the frequency dependence of the error correction and its ultimate rolloff
john
I’m sorry if the post order makes it seem as if my post on the previous page could be a criticism of Brian’s, somebody must not be following closely, I can't imagine your excuse, after all we're hardly into the 2nd thousand of posts in this thread...
“traderbam” seems to share my understanding of the issues (although I'm careful to stick to Linear system arguments, with injected "disturbance" standing in for weak nonlinearity)
both of our efforts in this thread have been directed to countering the argument that “error correction” is fundamentally different from “conventional” negative feedback and that error correction could have fundamentally different properties that the control community had been missing for half a century
A recurring problem seems to be that block diagram math shows “perfect correction” - which comes from ignoring real world gain-frequency limits and the problem that simple real number arithmetic can lead to non-realizable infinite gains where complex number calculations that include the gain-frequency limits of real devices show the frequency dependence of the error correction and its ultimate rolloff
john
Fascinating!
Hi Jcx
I hope you read this.
Your paragraph above seems to ignore a multitude of posts countering its points in every respect.
For instance i have shown how non-ideality in the summers such as ''your''....''gain-frequency limits of real devices show the frequency dependence of the error correction and its ultimate rolloff'' affects performance here:
http://www.diyaudio.com/forums/showthread.php?postid=1070291#post1070291
Please read ALL the associated and linked posts before talking about ''block diagram math'' in such an uneducated way.
jcx said:
Hi Jcx
I hope you read this.
Your paragraph above seems to ignore a multitude of posts countering its points in every respect.
For instance i have shown how non-ideality in the summers such as ''your''....''gain-frequency limits of real devices show the frequency dependence of the error correction and its ultimate rolloff'' affects performance here:
http://www.diyaudio.com/forums/showthread.php?postid=1070291#post1070291
Please read ALL the associated and linked posts before talking about ''block diagram math'' in such an uneducated way.
traderbam said:
Since my Academia project
is still at the development stage, with plenty of room for improvement, I'd really appreciate it if you could point me to a better way to implement a buffer stage (Av=1, THD<0.01, DF>20, P=50W/8ohm). Transistor count is not critical but I wouldn't want it to exceed Bob's circuit.As for capacitive load, don't worry about that. I am not using electrostats.
Regards,
Milan
Bob wrote:
Hi Bob, I know the implemenation is unusual for if nothing else it relies on a PFB loop. I'm really trying to get to the objective roots of verbal descriptions so that some assessment is possible. I'm afraid I am not one to take people's words for things. My experience of audio has talk me hard lessons.
I'll post a simulation result shortly as a guide. I want to get the part values as close as I can first. But preliminary results agree with your estimate. Do you have any record of what the loop gain vs f was of your circuit actually was?
traderbam said:
Your simulation of loop gain in this loop will be completely worthless for the elementary reason that loop gain is, in fact, EQUAL to error extracted at balance.
Therefore, ALL you will demonstrate, if you sim. your circuit correctly, is the amount of error generated by your output stage, and that circuit stability is compromised in direct proportion to the error produced by the output stage. No Dung!
mikeks said:
Mike, I think I've mentioned this before, but traderbam is talking about the outer loop in the path containing the output stage. This loop looks more like a normal feedback loop, whose loop gain seems to correlate with the amount of distortion reduction. Its prpoerties also lend themselves to normal analysis methods of gain and phase margin for looking at loop stability. The inner loop, with its low loop gain, doesn't seem to lend itself well to these traditional stablity measures. Try it with the loop gain probe. You'll see the high loop gain in the outer loop.
I found it interesting that a circuit with a unity loop gain frequency in this outer loop of about 600 kHz nonetheless had a bandwidth from input to output of 20 MHz or so.
experts who author control theory textbooks like Horowitz, Goodwin, Lurie, and respected audio researchers like Lipshitz, Vanderkooy do seem to agree on these issues (2nd ref: http://www.diyaudio.com/forums/showthread.php?postid=1062416#post1062416
), Hawksford conceeds the point to Lipshitz, Vanderkooy in the later Generalized error correction paper
for myself, I freely admit to considerable intellectual laziness, relying on Horowitz’ argument if I see a Linear control equation operating on the reference command input and the output of a siso plant, with no additional way to affect the output except through the single control input to plant, then I don’t march through all of the block diagram math
the “degrees of freedom” reference may be obscure, the distinction is simple;
“error feedback” control system equations have exactly 1 input, the “error” – the control equation only operates on this single quantity to produce the single control input to the plant
“2 degree of freedom” systems allow general Linear operations on the reference/command input and the plant output, to arrive at the single control input to the plant
these general linear operations obviously include differencing the input and output to create a “error” term and allow additional operations that may be expressed as input prefiltering, command feedforward to the plant input, “conditional control”, “model reference” control, Hawksford/Cordell error correction…
Bob’s fig 11 is Horowitz’ fig 6.1-1)f
maybe I thought people would benefit from Reading Horowitz in my post:
http://www.diyaudio.com/forums/showthread.php?postid=1070802#post1070802
This would Not appear to describe a Linear System; in a Linear System gains, loop transmissions are independent of any signal level, any multiplier of error in a Linear feedback system is a constant of the system, and for real world weakly nonlinear systems it is constant within the limits of the small gain approximation
), Hawksford conceeds the point to Lipshitz, Vanderkooy in the later Generalized error correction paper
for myself, I freely admit to considerable intellectual laziness, relying on Horowitz’ argument if I see a Linear control equation operating on the reference command input and the output of a siso plant, with no additional way to affect the output except through the single control input to plant, then I don’t march through all of the block diagram math
the “degrees of freedom” reference may be obscure, the distinction is simple;
“error feedback” control system equations have exactly 1 input, the “error” – the control equation only operates on this single quantity to produce the single control input to the plant
“2 degree of freedom” systems allow general Linear operations on the reference/command input and the plant output, to arrive at the single control input to the plant
these general linear operations obviously include differencing the input and output to create a “error” term and allow additional operations that may be expressed as input prefiltering, command feedforward to the plant input, “conditional control”, “model reference” control, Hawksford/Cordell error correction…
Bob’s fig 11 is Horowitz’ fig 6.1-1)f
maybe I thought people would benefit from Reading Horowitz in my post:
http://www.diyaudio.com/forums/showthread.php?postid=1070802#post1070802
mikeks said:
…
I think i agree with you here, since, in contrast to classical NFB, the error correction arrangement appears to possess the attribute that, at balance, loop-gain increases with increasing error extracted, as shown here.
This implies, it seems to me, that the system is actually one of active negative feedback.
What think ye?
This would Not appear to describe a Linear System; in a Linear System gains, loop transmissions are independent of any signal level, any multiplier of error in a Linear feedback system is a constant of the system, and for real world weakly nonlinear systems it is constant within the limits of the small gain approximation