Misplaced Pages

Talk:Negative feedback: Difference between revisions

Article snapshot taken from Wikipedia with creative commons attribution-sharealike license. Give it a read and then ask your questions in the chat. We can research this topic together.
Browse history interactively← Previous editNext edit →Content deleted Content addedVisualWikitext
Revision as of 18:33, 5 December 2014 editBrews ohare (talk | contribs)47,831 edits →Conclusion: WP link← Previous edit Revision as of 19:08, 5 December 2014 edit undoGliderMaven (talk | contribs)Extended confirmed users15,503 edits →Dicklyon's further reversions: collapse thread per agreementNext edit →
Line 169: Line 169:


==Dicklyon's further reversions== ==Dicklyon's further reversions==
{{hat|By consensus, further discussion here would not be helpful for the article.] (]) 19:08, 5 December 2014 (UTC)}}
In Dicklyon removed many different topics raised in the section ] with no Talk page discussion but only the erroneous one-line Edit Summary: ''Tangential bloat, not supported by the cited sources''. Before these reverts contained the observation: In Dicklyon removed many different topics raised in the section ] with no Talk page discussion but only the erroneous one-line Edit Summary: ''Tangential bloat, not supported by the cited sources''. Before these reverts contained the observation:

:The difference signal ''I''–&beta;''O'' that is applied to the open-loop amplifier is sometimes called the "error signal". The output is the open-loop gain times this error signal, added to any disturbance ''D'' that may be present. In the absence of a disturbance ''D'', this signal is given by:<sup>1</sup> :The difference signal ''I''–&beta;''O'' that is applied to the open-loop amplifier is sometimes called the "error signal". The output is the open-loop gain times this error signal, added to any disturbance ''D'' that may be present. In the absence of a disturbance ''D'', this signal is given by:<sup>1</sup>


Line 262: Line 264:


::::The simple guideline is that discussion of ''what sources say'' is liable to be less controversial, less ego-engaging, and less subject to ] than discussion of projected or even accurate perceptions of what WP editors happen to think. It also is what is germane to the WP project. ] (]) 17:45, 5 December 2014 (UTC) ::::The simple guideline is that discussion of ''what sources say'' is liable to be less controversial, less ego-engaging, and less subject to ] than discussion of projected or even accurate perceptions of what WP editors happen to think. It also is what is germane to the WP project. ] (]) 17:45, 5 December 2014 (UTC)

{{hab}}


==How do you separate an application of negative feedback from an approach that could be applied to it? == ==How do you separate an application of negative feedback from an approach that could be applied to it? ==

Revision as of 19:08, 5 December 2014

This is the talk page for discussing improvements to the Negative feedback article.
This is not a forum for general discussion of the article's subject.
Article policies
Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL
Archives: 1, 2, 3, 4Auto-archiving period: 30 days 
This article has not yet been rated on Misplaced Pages's content assessment scale.
It is of interest to the following WikiProjects:
Template:Wikiproject MCB Please add the quality rating to the {{WikiProject banner shell}} template instead of this project banner. See WP:PIQA for details.
WikiProject iconSystems High‑importance
WikiProject iconThis article is within the scope of WikiProject Systems, which collaborates on articles related to systems and systems science.SystemsWikipedia:WikiProject SystemsTemplate:WikiProject SystemsSystems
HighThis article has been rated as High-importance on the project's importance scale.
Taskforce icon
This article is within the field of Cybernetics.

This is the talk page for discussing improvements to the Negative feedback article.
This is not a forum for general discussion of the article's subject.
Article policies
Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL
Archives: 1, 2, 3, 4Auto-archiving period: 30 days 

Dicklyon's comments on Rashid

According to Dick:

1. "Please see Rashid's section 10.3.2, which treats A as an unknown and fluctuating quantity. It is therefore not possible to treat the error signal as "fixed" or a function of system constants." Dicklyon (talk) 04:05, 23 October 2014 (UTC)

Section 10.3.2 Gain Sensitivity is not different from other standard treatments of gain fluctuations in which A may become (A+δA). The desensitivity analysis proceeds using the same formulation identified in Rashid's Eqs. 10.1-10.3 that lead to his Eq. 10,7:

S e = S i − S f = S i − β S o = S i − β A S i 1 + β A = S i 1 + β A   . {\displaystyle S_{e}=S_{i}-S_{f}=S_{i}-\beta S_{o}=S_{i}-{\frac {\beta AS_{i}}{1+\beta A}}={\frac {S_{i}}{1+\beta A}}\ .}  Rashid's Eq. 10.7

Rashid's fluctuation analysis uses the formulation in which A is known and the gain subject to departure (A+δA) also is known in order to find how much such a change in A affects the standard closed-loop gain expression A/(1+βA). The variation in gain is found using the d/dA derivative of the standard gain expression and therefore assumes the the standard gain expression holds for all values of A., including (A+δA). Nothing here suggests that "it is therefore not possible to treat the error signal as "fixed" or a function of system constants." That is exactly what Rashid (and every other source) does. The change δA is attributed by Rashid to "changes in temperature and operating conditions of active devices". These changes result in a gain (A+δA) that is determined, just as is A, by the system components at their new bias points and temperatures. That means A changes, not in a manner affected by feedback considerations, but for reasons that are equally applicable when the open-loop amplifier is not in the feedback circuit at all.

2. "The feedback ratio Beta is more often treated as a constant, because it is created by stable passive components, but Rashid also looks at that as fluctuating or unknown, in 10.3.3. Dicklyon (talk) 04:05, 23 October 2014 (UTC)

Section 10.3.3 Feedback Factor Sensitivity, proceeds in the same fashion as the treatment of gain variation, calculating the d/dβ derivative of the standard gain expression. The analysis therefore assumes the the standard gain expression holds for all values of β.

3. "And it doesn't really work in general to treat A as the instantaneous ratio of output to input in such a way as to include noise, since the input may be zero." Dicklyon (talk) 04:05, 23 October 2014 (UTC)

Dick, you appear to be thinking of the noise generated by the open-loop amplifier itself independent of any input signal. That is, A fluctuates to become (A+δA). This situation is described by Kal, §6.3.1 Gain Stability and employs the standard formulation of Rashid's Eqs. 10.1-10.3. Even if the input signal actually is zero, the amplifier gain does not depend upon the physical presence of a signal but is the gain that would be seen if there were a signal, no matter how small. It is defined as A = output/input regardless of the actual physical presence of an input signal.

4. "Distortion perhaps one could fold into gain fluctuation, as Rashid does in 10.3.5, but that's not usually how it's handled (especially not workable for crossover distortion). Dicklyon (talk) 04:05, 23 October 2014 (UTC)

Section 10.3.5 treats a nonlinear open-loop transfer characteristic, an S-shaped curve of output versus input depicted in Rashid's Figure 10.5. He breaks the curve up into approximate straight line segments and shows that the standard gain expression applied in each region with gains A1, A2... leads to a closed loop gain of 1/β in each segment provided the gains all are large, and renders the closed-loop gain independent of the nonlinearity in A. Again it therefore assumes the the standard gain expression holds for all values of A.

5. "Rashid pretty much ignores noise in his feedback treatment. He certainly never treats the A as fixed when showing how the error signal relates to the feedback factor, nor suggests that the error signal count be generated any way other than by monitoring the output. Dicklyon (talk) 04:05, 23 October 2014 (UTC)

To see the standard treatment of noise reduction for noise introduced (like δA) inside the feedback loop, look at Kal, §6.3.2 Noise Reduction. It also uses the standard formulation.

The remark: "He certainly never treats the A as fixed when showing how the error signal relates to the feedback factor" contradicts Rashid's derivation of the error signal. What Rashid and everybody else does is to use A as an algebraic variable that takes on the value of the open-loop gain defined as the gain relating the input to this amp to its output when the amp is considered in isolation. This approach leads to the relation between the error signal and the feedback factor given by Se = Si./(1+βA). See Rashid's Eq. 10.7 reproduced above.

And Dick's remark: Rashid "certainly never suggests that the error signal cou be generated any way other than by monitoring the output." is vague indeed. Rashid 'never suggests' a great many things. However, it is clear that the standard gain expression A/(1+βA) results because feedback is present. It is equally clear that the output is a consequence of amplifying the error signal Se = Si./(1+βA) that is determined by the three parameters: Si., β, and A; all of which have values set (respectively) by the operation of the signal source, the feedback network, and the open-loop amplifier, just as they would perform in isolation, outside the feedback amplifier.

Summary

Although I have dealt in detail with Dicklyon's commentary here, the basic issue remains as stated in the thread above:

Using Rashid's formulation of his Eqs. 10.1-10.3 leads to his Eq. 10.7 (reproduced above) stating the equivalence of all forms for the error signal Se. Therefore, a justification of Dicklyon's claim that the input-output form of the error signal Se=Si–βSo is more general than the input-only form Se=Si./(1+βA) requires a formulation different from Rashid's (and every other textbook) using different definitions that no source has provided. Brews ohare (talk) 17:01, 23 October 2014 (UTC)

Almost everything you are saying is either nonsense or a misinterpretation of what I've said. Go back to what started this discussion. You claimed that the 'error signal' amplified by the open-loop amplifier can be calculated in advance of operation (in principle), with no need to monitor how the operation is going, a contrast with error-controlled regulation which (in principle) measures and corrects a gap calculated as operation takes place. This is still nutty, to claim to be able to characterize a feedback amplifier without feedback from the output, which is where the unknowns and non-idealities show up, just as in the systems that feedback controllers control. Your disclaimer of "in principle" really means if you ignore noise and distortion and unknown fluctations. OK, fine, if you there's no noise, no distortion, and no uncertianty in the forward gain, then you can calculate the error signal without monitoring the output. So what? Anyway, have fun going on about it; I'm going to be off the grid for a three weeks, so someone else might want to play. Dicklyon (talk) 04:51, 24 October 2014 (UTC)
Dick: Calling my remarks nonsense is not helpful and is itself nonsense. Here are the facts as simply as they can be stated:
Rashid's formulations is his Eqs. 10.1-10.3. They state:
S o = A S e   . {\displaystyle S_{o}=AS_{e}\ .}   (1)
S e = S i − S f   . {\displaystyle S_{e}=S_{i}-S_{f}\ .}   (2)
S f = β S o   . {\displaystyle S_{f}=\beta S_{o}\ .}   (3)
Now simple algebraic manipulation leads to:
S e = S i − S f = S i − β S o = S i − β A S i 1 + β A = S i 1 + β A   . {\displaystyle S_{e}=S_{i}-S_{f}=S_{i}-\beta S_{o}=S_{i}-{\frac {\beta AS_{i}}{1+\beta A}}={\frac {S_{i}}{1+\beta A}}\ .}   Rashid's Eq. 7
Your statement is that Se=Si–βSf is a more generally valid statement than Se=Si./(1+βA). Inasmuch as Rashid's formulation shows them to be equivalent, your remark is not compatible with Rashid's formulation (or, you don't understand algebra, if that is your preference). Brews ohare (talk) 05:17, 24 October 2014 (UTC)
The algebra is fine. The formulation (equation 1) is limited to the no-noise, no-distortion case. Equation 2 is valid more generally than that. And your interpretation of not needing to monitor the output relies on A being known exactly. I still don't see why you're trying to describe feedback while pretending you don't need to feed back the output. Dicklyon (talk) 06:03, 24 October 2014 (UTC)
Yes, the algebra is fine. The rest of your remarks are incorrect. You have no source for any of them. Without sources to back up your formulation of operation, which directly contradicts the traditional textbook approach as itemized in my comments above, these ideas of yours remain your own. Brews ohare (talk) 13:03, 24 October 2014 (UTC)
Your remark upon the limitations of Eq. 1 is unsourced imagination that would seem to apply equally to Eq. 3 and invalidate your (erroneous) claims for the greater generality of the input-output formulation of the error signal. Your remark that Eq. 2 has greater generality than Eq. 1 also is unsupported. It could be taken as the definition of a summer, just as Eq. 1 is the definition of gain. Hard to argue over definitions. Eh? And your interpretation of my comments is wrong, a strawman portrayal. Brews ohare (talk) 13:19, 24 October 2014 (UTC)
Look, basically, so far as I am, or as far as I can tell, everyone else is concerned, you've misunderstood this class of circuit. Your rants on the talk page are not persuading anyone.
Given that, we are not letting you write any of your non consensus views into the article, and if you do so, they will be reverted without mercy.
Given that, you are completely wasting your time on the talk page, nobody buys a word you say. Please stop wasting your own, and everyone else's time.GliderMaven (talk) 20:23, 24 October 2014 (UTC)

GliderMaven: There are 3 equations here defining gain A, feedback factor β, and the difference or 'error signal'. These 3 equations are combined by Rashid to provide several equivalent forms for the error signal. But you find this simple matter too complex and prefer to rant that "nobody buys a word" of it. Too complex, too well referenced, and too contrary to your unsourced and incorrect prejudices. The issue, as Dick sees it, is that Rashid's (and every other textbook's) 3 definitions don't apply to the 'real' negative feedback amp, but only to a severely compromised abstraction of it. There is no doubt that it is an idealization, but Dick's notions about a better model are not supported by sources. Your approach to resolving this matter is simply to spit. Brews ohare (talk) 22:17, 24 October 2014 (UTC)

|}

Collapsing discussions

I collapsed this section, and was reverted:

What do others think? Of course it is unusual to collapse discussions, but the activity on this talk page is already unusual. At 06:00, 14 October 2014 above, I noted that this talk page has had 1375 edits since 6 June 2014—just over 10 edits per day for 130 days. There have been 248 edits since then—over 20 edits per day for 11 days. Here are some numbers:

Edits to talk page since 6 June 2014
Editor Number of edits Percentage
Johnuniq 10 0.6%
Lowercase sigmabot III 11 0.7%
JohnBlackburne 21 1.3%
Nigelj 24 1.5%
DaveApter 31 1.9%
GliderMaven 78 4.8%
Trevithj 114 7.0%
Dicklyon 153 9.4%
Brews ohare 1181 72.8%
Total 1623 100%

If anyone wants to collapse the parent section, including this subsection, please do so. Or, should there be a meta-discussion about how discussions should occur? There is no reason this topic should require this much discussion that I can see. Do people want this much talk? Is there any likelihood it will conclude in a reasonable time? Johnuniq (talk) 03:21, 25 October 2014 (UTC)

I completely agree, and I intend to collapse any further pointless discussion by Brews Ohare or anyone else for that matter, unless more than one person complains.GliderMaven (talk) 03:54, 25 October 2014 (UTC)
I would object if there was any sign that the discussions were converging towards any sort of consensus. But they clearly aren't. So I also support the collapse. Trevithj (talk) 09:31, 25 October 2014 (UTC)
The discussion of Rashid is between Dicklyon and myself as others here have not participated. So this action is really just a closing of discussion with no attempt at clarification. The issue involved is the simple one of Dicklyon's unsourced challenge to the generality of Rashid's approach, which happens to be the universal approach of all textbooks. Apparently all but Dicklyon and myself feel unmoved to comment upon his original research, and consensus is to collapse this discussion as a substitute. Brews ohare (talk) 14:24, 25 October 2014 (UTC)
The purpose of the talk page is to discuss things that may actually make a contribution to the article. Where any discussions are clearly not reaching consensus, as here, they will be collapsed.GliderMaven (talk) 15:49, 25 October 2014 (UTC)
It is not WP policy for non-administrators to collapse discussions, whether or not it is the supposed consensus of non-contributing bystanders that a discussion is pointless. I don't agree that it is pointless to emphasize the universal textbook view of the negative feedback amplifier in the face of WP:OR, and oppose the disdain of the those present for sources in favor of personally held intuitions. Brews ohare (talk) 17:23, 25 October 2014 (UTC)
On the contrary, we're all over those sources; we just don't agree with your interpretation of them. We can certainly get an administrator here if you prefer. The likely outcome will be you getting blocked. I'd like to remind you that this is not a WP:FORUM. If you prefer we could refactor the discussion to a subpage off your own user page where you can continue it (if anyone else wants to, which I doubt), but I think, without wishing to put words in anyone else's mouth, that everyone else is agreed drawing a line under this here.GliderMaven (talk) 17:46, 25 October 2014 (UTC)

Recent suppression of Talk page contributions

By consensus, further discussion here would not be helpful for the article.GliderMaven (talk) 03:54, 25 October 2014 (UTC)
The following discussion has been closed. Please do not modify it.


The Misplaced Pages:Talk page guidelines cited in connection with collapse of talk-page discussion by Johnuniq contains the paragraph Off-topic posts, which does not suggest deletion, but collapse of off-topic material with a show/hide facility. That approach was used by Johnuniq to collapse two entire sehctions of Talk-page discussion of Rashid's derivation of the 'error signal' in the negative feedback amplifier, with the reason By consensus, further discussion here would not be helpful for the article Whether or not these editors think discussion of Rashid's formulation (typical of all textbooks on the subject) is 'not helpful', there is no doubt that it is on topic, and its collapse is unwarranted according to the guidelines. The same applies to the repeated collapse of this discussion by GliderMaven, another violation of WP policies and guidelines.

The purpose of these actions, IMO of course, is that these two editors, who have not contributed to the content of the discussion of Rashid, wish either to suppress this examination or possibly just to interfere with my attempts. Brews ohare (talk) 18:52, 31 October 2014 (UTC)

In essence, the issue that these editors do not wish to see discussed is the interpretation of Rashid's Eqs. 10.1-10.3. They state:

S o = A S e   . {\displaystyle S_{o}=AS_{e}\ .}     (10.1)
S e = S i − S f   . {\displaystyle S_{e}=S_{i}-S_{f}\ .}   (10.2)
S f = β S o   . {\displaystyle S_{f}=\beta S_{o}\ .}     (10.3)

and their simple algebraic manipulation leads directly to:

S e = S i − S f = S i − β S o = S i − β A S i 1 + β A = S i 1 + β A   . {\displaystyle S_{e}=S_{i}-S_{f}=S_{i}-\beta S_{o}=S_{i}-{\frac {\beta AS_{i}}{1+\beta A}}={\frac {S_{i}}{1+\beta A}}\ .}   Rashid's Eq. 10.7, p. 645

Eqs' 10.1-10.3 are definitions of gain A (Eq. 10.1), the error signal Se input to the open-loop amplifier (Eq. 10.2) and the definition of the feedback fraction β (Eq. 10.3). They lead in Eq. 10.7 to several equivalent forms for the error signal Se. Brews ohare (talk) 18:54, 31 October 2014 (UTC)

Some controversy has arisen here on this Talk page over the interpretation of the "input-only" form (the last version in 10.7). This form of the error signal shows that it is determined by the input signal Si from the external signal source, the open-loop gain A determined by the amplifier construction, and by the feedback factor β determined by the feedback network, all of which parameters are therefore fixed by their own stand-alone construction, entirely apart from their incorporation in the feedback amplifier. The formula for the error signal depends upon the presence of feedback, of course, but its value is set by external parameters and is not changed during operation. This fact is found disturbing by some assembled here, who hold the (unsourced) belief (contrary to Rashid's Eq. 10.7) that the 'error signal' is not set by external parameters that are determined apart from, and independent of, feedback operation. Rather, these editors believe, the error signal plays the role of a 'performance gap' that is measured and minimized by feedback during operation, as occurs within error-controlled regulation, a different use of feedback integral to the operation of such error-minimization systems. Brews ohare (talk) 18:55, 31 October 2014 (UTC)

Dicklyon has proposed that various considerations result in "Se=Si–βSo being more general than Se=Si./(1+βA)", limiting the generality of the 'input-only ' form of the formula and potentially undermining the conclusion that outside parameters fix the value of the error signal. As no-one denies that Rashid's 'input-only' form for the error signal is equivalent to all the other forms as is shown by Rashid's Eq. 10.7 displayed above, such controversy is resolved by establishing under just what circumstances the defining equations 10.1-10.3 are suitable. In particular, Dicklyon has suggested that Eq. 10.1 defining the gain is unsuitable in the presence of noise, making the various forms of the error signal in Rashid's 10.7 equivalent only under restrictive conditions (that remain to be elaborated upon and sourced). Although Dicklyon has not suggested it, his ideas seem to fit a modified version of Black's circuit proposed by Marc Thompson that includes an external noise disturbance inside the feedback loop. Brews ohare (talk) 18:57, 31 October 2014 (UTC)

Rather than engage in this discussion, GliderMaven and Johnuniq wish to cut this examination short. These editors may have missed the point of this discussion, or perhaps they are uncomfortable with its implications for their own views of the negative feedback amplifier as a form of error-controlled regulation. In either eventuality, the high-handed interference with a discussion pertinent to the literature on negative feedback is out of place and not suggested by WP:TPG: Off-topic posts . Brews ohare (talk) 16:29, 27 October 2014 (UTC)

This and the other use of collapse above both are actions contrary to WP guidelines. Brews ohare (talk) 16:48, 27 October 2014 (UTC)

In view of the simplicity of the item under discussion as explained above, it is no longer possible to think these editors have missed the relevance of this issue, which puts their actions in suppressing its discussion beyond the rational. Brews ohare (talk) 16:59, 27 October 2014 (UTC)

Reverts by Dicklyon

Dicklyon has said in other remarks that the feedback amplifier is a form of feedback control. However, contrary to Dicklyon's own expressed view, in this massive revert Dicklyon removed material that supports and sources this claim. This removed material outlines the key difference between this type of control and the more elaborate methods identified by the sources. The deleted explanation is as follows:

When a system, sometimes called a load or a plant, is characterized by a simple input/output model, it can be controlled by a single variable like the output of a negative feedback amplifier. Then the negative feedback amplifier can be used as a type of feedback control, maintaining a stable control signal for the load or plant. Such a controller is more rudimentary than those that employ measurements of the internal variables governing the plant, the so called full-state feedback or state-space approach.
Notes
Boris Lurie, Paul Enright (2000). Classical feedback control with MATLAB. CRC Press. p. xiii. ISBN 9780824703707.
A simple example is the case where the 'plant' presents itself as a load resistor and the input signal to the amplifier is a voltage. The feedback amplifier then serves as an interface between the input signal source and the load that tends to keep the voltage across the load constant (presumed to be a 'plant' requirement). If the feedback amplifier were not present the voltage across the load would vary with changes in either the load resistance or in the output impedance of the signal source because they would constitute a simple voltage divider.
Boris Lurie, Paul Enright (2000). Classical feedback control with MATLAB. CRC Press. p. xiii. ISBN 9780824703707.
Shimon Y. Nof (2009). "§4.7: The emergence of modern control theory". Springer handbook of automation. Springer Science & Business Media. pp. 63 ff. ISBN 9783540788317.
D Roy Choudhury (2005). "Chapter 11: State-variable formulation". Modern control engineering. PHI Learning Pvt. Ltd. pp. 522 ff. ISBN 9788120321960.

Dicklyon has removed this material with the one-line Edit Summary: Tangential bloat, not supported by the cited sources. It is obvious that the material is sourced and is hardly tangential. At a minimum, Talk-page explanation of this peremptory action is needed. Brews ohare (talk) 18:30, 30 November 2014 (UTC)

Dicklyon's further reversions

By consensus, further discussion here would not be helpful for the article.GliderMaven (talk) 19:08, 5 December 2014 (UTC)
The following discussion has been closed. Please do not modify it.

In this massive revert Dicklyon removed many different topics raised in the section Negative feedback amplifier with no Talk page discussion but only the erroneous one-line Edit Summary: Tangential bloat, not supported by the cited sources. Before these reverts the original subsection contained the observation:

The difference signal I–βO that is applied to the open-loop amplifier is sometimes called the "error signal". The output is the open-loop gain times this error signal, added to any disturbance D that may be present. In the absence of a disturbance D, this signal is given by:
I − β O = I ( 1 − β O I ) = I 1 + β A   , {\displaystyle I-\beta O=I\left(1-\beta {\frac {O}{I}}\right)={\frac {I}{1+\beta A}}\ ,}
which is determined by the independent parameters I (set by the signal source), A (set by the open-loop gain) and β (set by the feedback network), and therefore is set by these external variables, unlike more sophisticated feedback control of errors using internal state variables of the controlled system. This error signal tends to be small if the open-loop gain A is large.
Notes
Muhammad Rashid (2010). "§10.3.1 Closed-loop gain – Equation 10.7". Microelectronic Circuits: Analysis and design (2nd ed.). Cengage Learning. pp. 644–645. ISBN 9781133007487.
Boris Lurie, Paul Enright (2000). Classical feedback control with MATLAB. CRC Press. p. xiii. ISBN 9780824703707.
Shimon Y. Nof (2009). "§4.7: The emergence of modern control theory". Springer handbook of automation. Springer Science & Business Media. pp. 63 ff. ISBN 9783540788317.
D Roy Choudhury (2005). "Chapter 11: State-variable formulation". Modern control engineering. PHI Learning Pvt. Ltd. pp. 522 ff. ISBN 9788120321960.

Comments

Dicklyon claims that this material is not supported by the cited sources. I find that claim invalid. Dicklyon also claims this material is "tangential bloat", again an unsupportable contention. Brews ohare (talk) 18:30, 30 November 2014 (UTC)

I thought it was tangential and unnecessary, more appropriate in a sub-article. I agree with the removal of this material from this summary article. Binksternet (talk) 19:29, 30 November 2014 (UTC)
Also concur. Since both sections have links to dedicated pages, the extra detail here serves no purpose. Trevithj (talk) 20:47, 30 November 2014 (UTC)
  • Binksternet, of course. saying you thought it tangential is your view, and without any argument to support it, it becomes just a 'me too' alongside Dick's view, also unsupported. Of course, it is always gratifying to have 'me too' support, but WP is not based upon a 'me too' count, but is a presentation of sourced opinion, with topics chosen according to sources. The reasoning as to why this is relevant, beyond the fact that published authors think it worth including in their treatments, is that the use of the feedback amplifier involves an 'error signal' that is not determined in the same way as the 'error' in controllers that determine the system errors using internal state variables of the controlled system. This difference underlies the reason why the feedback amplifier can be used as a controller only for systems with simple input/output characterization. The sources cited elaborate upon thsese points. Brews ohare (talk) 20:55, 30 November 2014 (UTC)
  • Trevithj: You suggest that links to dedicated pages obviates any need to point out the relevance of these topics, or to source them. And with the reversions of Dicklyon, there are in fact no links left to the respective pages on full state feedback and state-space feedback. As a general rule, the existence of more extensive articles is not a reason to avoid pointing out the relevance of a topic. Brews ohare (talk) 20:55, 30 November 2014 (UTC)

In these edits that I reverted, Brews has introduced the state-space approach, apparently because he thinks it supports his thesis that amplifiers are somehow fundamentally different in concept from control uses of negative feedback. The logic of twisted interferences from sources is so bizarre and flawed that I can't begin to point out its errors. Reverting was the only easy way to contineu to protect the article from the abuse it gets from Brews when I go away for a while. Dicklyon (talk) 03:58, 1 December 2014 (UTC)

Brews ohare, if you are trying to tell the reader that the negative feedback amplifier uses a control mechanism more suitable to relatively simple I/O than other methods of feedback, why don't you say it in so many words, without resorting to a proof of the concept? However, I don't see the literature supporting this view, that it is such a different beast. Binksternet (talk) 05:26, 1 December 2014 (UTC)
  • Dicklyon: The process by which I arrived at the sources provided was explained to you at length on your talk page, which you seem to have failed to read. The basic issue you raised yourself was that the negative feedback amplifier was a form of controller. You dug up the Lurie & Enright source that refer to this as "classical control" and say "Classical design does not utilize the plant's internal variables and/or their estimates for compensation, unlike the full-state approach."1 They point out further that in classical control "the object of control (plant) is characterized either by its measured frequency response over the range of effective feedback or by a rather simple input/output mathematical model".2 It appears from their examples where a negative feedback amplifier acts as a controller, it controls one of these "plants" with a "simple input/output mathematical model", an example being a simple resistor. I fail to see anything hard to follow here, and am inclined to think you have not looked at the linked sources. Brews ohare (talk) 06:32, 1 December 2014 (UTC)
  • Binksternet: You comment: " if you are trying to tell the reader that the negative feedback amplifier uses a control mechanism more suitable to relatively simple I/O than other methods of feedback, why don't you say it in so many words, without resorting to a proof of the concept?" I did not aim at a "proof" of anything. My object was to direct the reader to sources where the differences were noted. For example, here you will find the comment:
"The transfer function approach confines to input-output behavior of linear systems only. On the contrary the state-space representation gives information about the internal behaviour of the system as well as information about its input-output behaviour...Hence, through the state-variable approach, perfect control of the system is possible"
Here you will find:
"Classical design does not utilize the plant's internal variables and/or their estimates for compensation, unlike the full-state feedback approach. ...These are the reasons this book with extensive plant modeling."
Here you will read:
"The state variables allowed for a more sophisticated representation of dynamic behaviour than the classical single-input single-output system modelled by a differential equation..."
and later
"Digital adaptive control, however, offers much greater possibilities for (1) Identification of relevant system parameters (2) Making decisions about the required modifications... (3) Implementing the changes."
Binksternet, as presented in this thread, I attempted to summarize the discussion of these sources and lead the reader to them. I do not think pointing out the extreme limitations of the negative feedback amplifier as a controller and leading the reader to a broader understanding is out of place here, nor is it an expression of my ideas, but rather that of numerous sources, only three of which are noted here.
So, what I am trying to say in broad terms is: "the negative feedback amplifier as a controller is suitable only for control of relatively simple single-input single-output systems, and much more limited in its use of feedback than other feedback approaches in common use that use internal variables and complex modeling of the controlled system and the response to its variations." If the presentation is unclear, perhaps you could assist in improving it? Brews ohare (talk) 16:17, 1 December 2014 (UTC)
Binksternet, perhaps your reference to a proof is to the thread immediately above? That thread describes the "error signal" for the feedback amplifier and identifies its reliance upon external variables A, β, and I, as a contrast to the use of internal variables referred to by the sources just mentioned. You may regard this discussion as overkill, but the other editors on this page are very anxious that it not be presented because they do not admit there is any fundamental difference between controllers using negative feedback. Brews ohare (talk) 17:30, 1 December 2014 (UTC)

Conclusion

There being no attempt here to support the idea that placing the negative feedback amplifier in the context of modern control theory is "tangential bloat", and it being perfectly obvious that all statements in the material reverted by Dicklyon are supported by reputable sources, I propose to revert Dicklyon's action. Any objections, or perhaps, suggested reformulations? Brews ohare (talk) 13:53, 3 December 2014 (UTC)

It seems like an elaborate setup for an odd point of view; for example, your conclusion "therefore is set by these external variables, unlike more sophisticated feedback control of errors using internal state variables of the controlled system". Is there a source that draws such a conclusion? How is the simpler case unlike the more complex case, other than having a 1D state variable? Dicklyon (talk) 15:33, 3 December 2014 (UTC)
Dick, I don't know how you can miss what has been said and sourced over and over. The error signal employed by the negative feedback amplifier as derived by Rashid shows it to depend upon A, β , and the input signal I, all of which are external variables completely unrelated to the state of the controlled object or "plant". The three quotes above from three separate sources all show that, in contrast, the full-state or state-space controllers all make use of internal variables identifying the state of the controlled object. Do you get this? Brews ohare (talk) 15:43, 3 December 2014 (UTC)
In addition, the sources all make a distinction between "classical" control, in which they include control using the negative feedback amplifier, and "modern" control (post 1956 according to Choudhury) based upon the development of systems that incorporate internal variables of the controlled object, also known as essential variables. Brews ohare (talk) 15:57, 3 December 2014 (UTC)
There is no difficulty in treating the negative feedback amplifier as a control system using either classical or modern methods; I don't see you producing a source that suggests there would be. It is just a simple case, for which the modern methods offer no particular advantage. There is no source that suggests that the error signal does not depend explicitly on the output. The algebra is just for computing a gain under the model that the output is purely an open-loop gain times the input; that's a model, not a suggestion of how it works. Dicklyon (talk) 16:20, 3 December 2014 (UTC)
Your concept "the negative feedback amplifier as a controller" is also very odd. We can describe the negative feedback amplifier as an example of a feedback control system, but "as a controller" is a weird twist, the motivation for which escapes me. Are there sources that do this?
Your claim that the error signal is "completely unrelated to the state of the controlled object or 'plant'" remains bizarre, and contradicted by sources that I pointed out in which the 'plant' is identified with the amplifier output stage and/or load; it's feedback from this plant that makes the error signal.
And I don't have time to waste on continuing this silly circus of yours. Just give it up. Dicklyon (talk) 16:05, 3 December 2014 (UTC)
Dick, you have this all wrong, and simply refuse to understand the sources. If you want to come to grips here, try addressing the quotations from sources above. If you can construe them differently, do so. Otherwise your remarks are just bluster and avoidance. Brews ohare (talk) 16:13, 3 December 2014 (UTC)
I have no quarrel with any of the quotations from sources. I have quoted a few parts of your text that don't follow from them. Enough. Dicklyon (talk) 16:22, 3 December 2014 (UTC)
No , Dick. It is not enough. You have not quoted parts of my text that do not follow from the sources, and for a simple reason - there aren't any. Looking at the text here and the text here you have not identified anything that is contrary to sources. You have identified your own confusion about what the sources say, which you could clear up by comparing your remarks with these sources.
For example, you say "Your concept 'the negative feedback amplifier as a controller' is also very odd." And yet that is exactly what is done by Lurie and Enright who illustrate a voltage regulator in which they say "the plant is the load resistor RL", and the voltage across it is controlled by a negative feedback amplifier (in the form of an op amp), regulating the output voltage. "The command is the 5V input voltage (when the command is constant, as in this case, it is commonly called a reference, and the control system is called a regulator)." Brews ohare (talk) 18:50, 3 December 2014 (UTC)
To continue, Dick, you say "Your claim that the error signal is "completely unrelated to the state of the controlled object or 'plant'" remains bizarre, and contradicted by sources that I pointed out in which the 'plant' is identified with the amplifier output stage and/or load". Now, the two parts of this sentence contradict each other. There is no doubt that it may be that the "load" is a simple resistor, as in the example above from Lurie and Enright. A resistor exhibits the properties of a single-input single-output system as described by these authors (what they refer to as a plant having a "simple input/output mathematical model" ), and identified with these words in the reverted text. So there is nothing contradictory found here. However, the first part of your sentence, "Your claim that the error signal is 'completely unrelated to the state of the controlled object or ‘plant’' remains bizarre" is simply out to lunch. The error signal in the negative feedback amplifier, as defined by Rashid's Eq. 10.7, is determined by the externally determined variables A, β, and I, none of which have any connection with the load resistor at all. For example, the load resistor can fluctuate in resistance, and the 'error signal' I/(1+βA) doesn't change.
Dick, please put aside your overwhelming desire to squelch all mention of differences between feedback controllers and look at the sources. Brews ohare (talk) 19:12, 3 December 2014 (UTC)

Brews, you better find someone who supports your analysis to explain it better, as I'm still not getting it. Everything you say seems wrong to me. I have no desire to "squelch all mention of differences between feedback controllers", though it's not clear to me what kind of differences might be relevant to this article. Maybe someone else can point out differences that you're finding that would be relevant to the article on negative feedback. Dicklyon (talk) 19:29, 3 December 2014 (UTC)

To me it looks like Brews ohare is trying to use Misplaced Pages to prove a point which is either not important or not at all supported. In the first case the matter is not widely discussed in the literature, so it doesn't seem like Misplaced Pages should pay disproportionate attention to it. In the second case I would recommend publishing a treatise elsewhere. Binksternet (talk) 20:07, 3 December 2014 (UTC)
If Brews Ohare was an acknowledged expert on this stuff, but in fact not so long ago he didn't even know what an opamp was, and he is taking positions that literally nobody else agrees with. There is zero chance that others will agree, and this talk page consists entirely of Brews Ohare arguing, entirely unpersuasively, with everyone else. Given that, it's not profitable for it to continue. I propose to collapse this thread.GliderMaven (talk) 20:33, 3 December 2014 (UTC)
Yes collapse this and the one above. I have stayed out of the content dispute as I do not have the time for it but ignoring me there are four editors rejecting Brews ohare's changes with good reason and so a clear consensus against them. This is about changes to the article, his changes have been rejected. That's not going to change and this should end.--JohnBlackburnedeeds 23:12, 3 December 2014 (UTC)

The point here is not who is an expert here, but what the sources say. I am the only one here using sources. Some of you want to say the quotes I have used to express the content of these sources are defective, but none of you has provided a source-based alternative. All you have is some fuzzy conceptions of your own, that you are unwilling to support with links to sourced text. Maybe you all have persuaded yourselves that something is amiss here, but it appears to be a case of groupthink, detached from published material. Brews ohare (talk) 22:30, 3 December 2014 (UTC)

It seems to me that if you read this reverted text and the notes accompanying it you will find it is a nearly verbatim version of what these sources say. Assuming that is the case, the only question is whether it is pertinent. I'd say setting the negative feedback amplifier in its use as a controller in the context of full-state and state-variable control is entirely pertinent and that is all that is done here. So far no-one has addressed this matter directly. Brews ohare (talk) 22:57, 3 December 2014 (UTC)

If you read this second block of reverted text it consists of Rashid's formula for the error signal he defines for the negative feedback amplifier, and the observation that A, β and the external signal I are external variables (which no-one can dispute) while the full-state and state-variable approach involves internal variables, as stated explicitly by the three cited sources. There is no dispute about these matters, and they separate the negative feedback amplifier from the so-called "modern" state-space approach.

Dicklyon has raised questions about the limitations of Rashid's analysis when fluctuations are present, and these are dealt with by the introduction of the disturbance D in the Negative feedback amplifier section. They do not impact Rashid's formulation in the present context of separating "classical" from "modern" feedback control.

This entire brouhaha is due to failure to concentrate upon what is being said and failure to directly address what the sources have to say. That problem would be avoided if critics would stick to sources, not using proxy summaries about personal concepts of what the sources say, but actually using the explicit text from sources to support their assertions. If this were done, it would soon evolve that any issues here are actually over other matters not raised in these two reverted contributions. Brews ohare (talk) 23:30, 3 December 2014 (UTC)

There is so much to disagree with here that one can't even start. If you'd like to propose some points in smaller increments, we can discuss. For example, we've been all around on the use of Rashid's formula and the weird unsourced conclusions that you draw from it (like where you say "There is no dispute about these matters" followed by another bit of unsupported synthesis that "they separate the negative feedback amplifier from the so-called 'modern' state-space approach"). How do you separate an application of negative feedback from an approach that could be applied to it? The meaning and logic there escape me; no source supports these conclusions. Dicklyon (talk) 02:12, 4 December 2014 (UTC)
Brews ohare, you are exhausting the community's patience and wasting everybody's time. Please stand down on your own terms or I will initiate a topic ban discussion. Binksternet (talk) 02:53, 4 December 2014 (UTC)
I do not understand this hostility. It would seem that you all think you have explained yourselves blue in the face, yet not one of you has addressed the three sources cited in the reverted material. And no-one but Dick has addressed Rashid's standard formula for the error signal, which he agrees with under Rashid's assumptions, as well he might as his is the standard treatment in every textbook for the last 60 years. I am left thinking that none of you knows how to address the points raised, and certainly not by reference to sources. So be it, I guess. Brews ohare (talk) 03:27, 4 December 2014 (UTC)
I'd like to point out that there is next to no connection between the collapsed discussions on this talk page that concern how the negative feedback amplifier itself operates, and in particular, the adequacy of Rashid's formulation for the ideal amplifer when faced with reality, and the present discussion that concerns the use of the ideal amplifier as a controller, a point I am confident that none of the critics here have grasped. Brews ohare (talk) 13:34, 4 December 2014 (UTC)
The hostility could have something do with 6 years of stupid arguments, in which you have been banned for a year more than once for disruptive non-cooperative editing style. It's no surprise that you don't understand this. Dicklyon (talk) 02:15, 5 December 2014 (UTC)
Dick, your referring to "years of stupid arguments" seems to suggest their stupidity originates with myself. However, the true origin, as demonstrated by this instance as one example, is a refusal by yourself and others to address the goal of WP to present sourced material, and the refusal to use Talk pages to discuss what sources say, rather than to pontificate over self-congratulatory imagined personal expertise. Your approach to the negative feedback amplifier doesn't address sourced topics, revealed by a simple mouse click on provided Google book links and supported by verbatim quotes, but relies upon your own (unsourced, and often off-topic) opinions and prejudices. Brews ohare (talk) 15:50, 5 December 2014 (UTC)
For example, if this material were addressed directly by considering the references cited rather than a one-line cryptic edit summary Tangential bloat, not supported by the cited sources it would assist me to understand what you are talking about. When taken to the Talk page you say Brews has introduced the state-space approach, apparently because he thinks it supports his thesis that amplifiers are somehow fundamentally different in concept from control uses of negative feedback. This comment seems to say the state-space approach is my idea, when the cited sources bring it up. It also suggests that I am "supporting his thesis", which again suggests this is my idea, and then completely misconstrues the point as (in your words) amplifiers are somehow fundamentally different in concept from control uses of negative feedback. This is not what the cited sources say, and not my representation of what these sources say. The "thesis" presented by these sources is that the use of negative feedback amplifiers (as controllers) is restricted to control of simple input/output objects, while more general state-space controllers use internal variables as well to achieve control over more sophisticated objects. If you directed your comments to the sources instead of searching for some pretext to throw brickbats at me, this matter could be engaged much less heatedly and to better end. Brews ohare (talk) 17:32, 5 December 2014 (UTC)
The simple guideline is that discussion of what sources say is liable to be less controversial, less ego-engaging, and less subject to idées fixe than discussion of projected or even accurate perceptions of what WP editors happen to think. It also is what is germane to the WP project. Brews ohare (talk) 17:45, 5 December 2014 (UTC)

How do you separate an application of negative feedback from an approach that could be applied to it?

This question posed by Dicklyon about the distinction between the negative feedback amplifier and state-space control is followed by this remark: The meaning and logic there escape me; no source supports these conclusions. The confusion here is born of impatience interfering with understanding.

The logic of Dicklyon's remark is that "modern" control theory could be applied to the negative feedback amplifier, so how can it be a separate matter? That seems to be so. But the meaning of my remark, which is perhaps badly expressed in this one sentence, is very clear from what is said in the material Dicklyon reverted: the meaning is that the negative feedback amplifier belongs to the "classical" approach that ignores the internal variables determining the state of the controlled object, while the "modern" approach to control employs these variables.

This division is remarked upon by all three of the state-space sources. To quote two of them:

  • As linked above, Lurie & Enright define the "classical" approach as one where the object of control is characterized by "a rather simple input/output mathematical model." In particular, in their Figure 1.5(a) they show a negative feedback amplifier controlling the voltage across a resistor, which obviously fits the "classical" model where the controlled object is described by the input/output equation of Ohm's law: I=V/R.
  • The same point is made by Choudhury who divides the evolution of controls between the "classical approach (up to 1956) and the "state variable approach" (post 1956). He characterizes the "classical" approach as one based upon using transfer functions for the system components, and the negative feedback amplifier falls into this category. About the comparison with the state-space approach he says: "The transfer-function approach confines to input-output behavior of linear systems only. On the contrary the state-space representation gives information about the internal behavior of the system as well as information about its input-output behavior."

I think with this in mind it is clear that what I am trying to express is that the negative feedback amplifier, in its focus entirely upon single-variable external control of its controlled object, is to be distinguished from state-variable control that uses information about the internal state of the controlled object. That claim is supported by both these sources, and is obvious. Brews ohare (talk) 15:04, 4 December 2014 (UTC)

The amp is a simple case of state-variable control. But what does this distinction that you want to draw have to do with the topic of the article? Dicklyon (talk) 02:13, 5 December 2014 (UTC)
What makes it a simple example? It doesn't use internal variables of the controlled system. Different controllers do.See this. What are the consequences? It can control only systems governed by simple input/output models. Different controllers can control more complex objects and can control them more precisely.See this. Why should these differences be part of this article? Because the article is about negative feedback, and hence about how it is put to use. An article about motor vehicles might point out differences between cars and trucks. Brews ohare (talk) 03:10, 5 December 2014 (UTC)
Categories: