Concept of Negative Feedback question

[FvM]

Your simulation also shows that an OP without bandwidth limitation isn't stable in closed loop operation. If you change the slew rate limitation to a simple first order low pass, the permanent oscillation will disappear.

 

[dean_winchester]

Well the basic need of Negative feedback is that to provide stability to the system. Suppose a given example that you have an automatic breaking system in the CARS. So, how the feedback wrks there.
Now when a sensor senses that it needs to put a break on the system, it increases pressure by some value. Now, the value is fed back to the input to know that the system does not shoots up to maximum pressure suddenly, so to get that out of picture, what you do is that provide a feedback. It will sense that the Error is more then it will increase pressure accordingly. If it is less, it will increase the pressure relatively. So, rather than just oscillating between the maximum and minimum pressure, you get an idea hw much to increase and whn to increase.
I hope this makes sense to you ?

 

[LvW]

Well the basic need of Negative feedback is that to provide stability to the system.

This statement deserves some comments.
It is a common misconception that negative feedback would "provide stability to the system".
Exact the opposite is true.
An amplifier without any feedback is always stable.
Negative feeedback has many advantages (e. g. it stabilizes the dc operating point) - however, it reduces system stability because in real systems it always moves into positive feedback for rising frequencies.

 

[dean_winchester]

The Negative feedback provides stability to systems when speaking in terms of the Gain... The Gain is controlled and the output varies within the specified range of Voltages, not exceeding Vcc all the times.

 

[FvM]

Stability is a well defined term in circuit theory. "stability ... in terms of gain" is rather referring to common language usage of the term and sounds somewhat screwed if you're actually talking about "stable gain".

Usually, the output voltage of an amplifier won't exceed the supply coltage, with or without negative feedback.

 

[jasonc2]

Perhaps one of you should clarify the misuse of "stability" in the lead paragraph of https://en.wikipedia.org/wiki/Negative_feedback while you're at it.

 

[dean_winchester]

Stability is a well defined term in circuit theory. "stability ... in terms of gain" is rather referring to common language usage of the term and sounds somewhat screwed if you're actually talking about "stable gain".

Usually, the output voltage of an amplifier won't exceed the supply coltage, with or without negative feedback.

It won't exceed but there will always be saturation. It will either be at +Vdd or at -Vdd... Stable Gain means user can have control over the Gain....

- - - Updated - - -

Perhaps one of you should clarify the misuse of "stability" in the lead paragraph of https://en.wikipedia.org/wiki/Negative_feedback while you're at it.

Yes. that link would help out...

- - - Updated - - -

If a system has overall a high degree of negative feedback, then the system will tend to be stable.

 

[LvW]

If a system has overall a high degree of negative feedback, then the system will tend to be stable.

This sentence (second line of the referenced wiki contribution) is rather questionable and misleading.

Look at the statement in the last paragraph (electronic amplifiers, fifth line) of this wiki-page. For my feeling, this sentence is in contradiction to the beginning of the page.

Quotation dean_winchester: The Gain is controlled and the output varies within the specified range of Voltages

The fact that the gain value can be controlled and set by external resistors has nothing to do with "system stability". Stability is a term that is defined in the time domain - and has no relation to effects that may be caused by parts tolerances.

 

[FvM]

This sentence (second line of the referenced wiki contribution) is rather questionable and misleading.

And unclear, too. E.g., what's meaned with overall?
- a paraphrase of loop gain ("gain over all circuit blocks")?
- for all frequencies?

Negative sign of DC loop gain is a necessary but not sufficient condition for stability.
Loop gain can't be negative for all frequencies in a real circuit. As a consequence, the loop gain magnitude must roll-off before the sign changes to positive.

 

[D.A.(Tony)Stewart]

I recall power brakes to be open loop ,where the gain on pedal force is amplified by the position of the vacuum valve that amplifies the pedal force by the vacuum pressure over the diaphram area to exert a force to the piston which allows a constant pressure to scale by the area exposed in the valves for the master and slave pistion.

I could be wrong, but I was not aware of any peddle pressure feedback or fluid pressure feedback control.

Stable gain means to me that the gain provides minimal ringing or overshoot and also that the gain does not change much with degraded SNR, temperature effects on gain, ESR/load ratio variations, spurious resonance effects, saturation effects on loss of feedback gain etc etc,.

This stable gain has a margin to becoming unstable, as we know by several criteria and analysis methods, is the threshold where the phase becomes positive feedback. ie. the 0dB loop gain phase margin. or the 180 phase shift point with 180deg feedback is attenuating by a margin factor instead of amplifying. Margin is everything, as the previous suggested examples may erode this margin.

(Proportional Integrator, Differentiator) PID loops are an example of feedback for optimizing the amplitude and phase so that this gain margin is maximized, but bandwidth and response time are both improved and remain stable..

Wiki says "tends towards stability" without details.

The ringing effect of saturation in feedback loops is from the reduction in proportional feedback gain with some second order effects when capacitance increases both affecting the margin. more or less...

Hysteresis, stiction and dead-zone are all characteristics of no feedback that affect the results with overshoot from lag in response or loss of phase margin. These sometimes show up in edge controlled mixers used in PLL's and amplify phase noise which requires moderate attention with small filters. They also show up as dither in driver steering with eye-hand coordination going over bumpy roads with loose steering rods. They show up even more with delayed reaction to vision from drunk drivers who introduce a loss of gain margin from reduced feedback and phase shift from reaction times increasing to weave over the road. I like to tune my brain's PID loop when I drive by doing Integrated Accumulated Error Correction (IAEC) position error calculations to see how centred my car is positioned relative to the centre of the highway, if I get bored. Big truck drivers do this all the time without thinking ;p then I go back to auto-pilot.

The body fluid pressures all have transfer functions which I remember calculating in my BioMedical experiments at the university pub, where we had to calculate based on heart rate, weight, fluid volumes, consumption of beers , reservoir capacities and fluid transfer functions so we had to predict how long it would take before Nature Called. Women have a much smaller reservoir.

 

[LvW]

Hey, SunnySkyguy,

Sorry, I cannot comment on your example of power brakes as I have no experience in this area.

However, I like to remind you that the discussion on stability has started in post#24 with the statement
"The Negative feedback provides stability to systems when speaking in terms of the Gain...the Gain is controlled and the output varies within the specified range of Voltages"

That means: Subject of the succeeding discussion was the gain of an amplifier (not loop gain).

Therefore, one short comments to your long and interesting reply:

Quotation: Stable gain means to me that the gain provides minimal ringing or overshoot

I cannot agree to this definition. An amplifier with feedback and with an absolute fixed and constant gain value can exhibit ringing and overshoot (in the time domain) as well as peaking (in the frequency domain).
Rather, these effects are caused by the feedback effect only. If the gain "peaks" at higher frequency it can be described as "frequency-dependent", but - for my opinion - not as unstable.
Instead, when a gain is considered as "unstable" it's value fluctuates - for example due to changing environmental conditions. Do you agree?

 

[D.A.(Tony)Stewart]

Common emitter amplifiers usually have feedback regulated by the collector to emitter impedance ratios.
Common collector amplifiers usually have degenerative feedback for unity gain. (which ring with capacitive loads)
Common base amplifiers usually have feedback by impedance ratios again.
Op Amps without feedback are just saturating comparators.

The same applies to open drain,source amplifiers, although I have never seen a common gate amplifier.

Which open loop gain amps are you referring to? Just the current gain amplifier in a Bipolar device? or the transimpedance gain in FETS?

yes I agree open loop amplifiers are inherently stable, but even semiconductor amplifiers have internal feedback from stray and junction feedback.

What did I forget?

Oh right.. "Stable Gain" i was inferring to as closed loop stability with gain margin
and otherwise "stable gain" can be from the perspective of stable hFE or gm of a discrete device or sensor.
I suppose stable gain could also mean the gain tolerance for any stress factor whether it is open or closed loop or even a passive attenuator, depending on the context of the discussion.

For example A Path Loss study on Wifi telemetry has a series of blocks each with a gain or loss factor and a variable depending on stress factors such as Raleigh Fading or Return Loss or atmospheric attenuation. Where most of these are open loop gain or attenuation factors in a one way communication path.

I digress from the transistors but sometimes I apply simple terms of gain in discrete devices to large networks with feedback, to apply the the common characteristics of cascaded stages.

 

[LvW]

What did I forget?

Nothing.
And, of course, I agree to everything. Perhaps there was a misunderstanding between us (due to my limited ability to express myself in english).
Let me explain it with other words:
For my understanding the "gain" (that is a fixed number) of a device cannot be unstable - perhaps it can fluctuate or deviate from its nominal value or can have tolerances. But I wouldn't say it is "unstable".
The reason behind this "conflict" of nomenclature is that the term "stability" is a well defined term (see FvM's post#25).
Only a feedback system (which can consist of several blocks and components as in control loops) can be "unstable" in this sense
This can be observed in the time domain, for example, by oscillations (decaying or not) of the output signal. But it is not the "gain" that oscillates and is "unstable"..
And to be exact: It is also not the "loop gain" that is unstable - insofar it does not matter at all if there is a hidden internal feedback (which always is present) or not.
So - what/who is unstable? It is the complete system in case it delivers a signal which does not fulfill the existing definitions for "stability" .

Final remark: I am aware that this discussion has deviated from the original problem as mentioned in post#22 and #23.
But I consider this as normal - it happens very often that an answer creates new questions.

Another remark: Somebody may think that we argue only about "words" and "definitions". On the other hand, how can electronic engineers discuss without using common terms and definitions?

 

[FvM]

Which open loop gain amps are you referring to?

Open loop gain is a system level term. It shouldn't be hold against the fact, that an amplifier can already implements internal feedback, although it complicates the analysis.

In the system level analysis, amplifier blocks can be considered as black boxes, described by a set of parameters. You don't necessarily need to know how the behaviour is created internally.

LvW already discussed different meanings of "stability" and I don't have much to add, except for a remark on "stable" gain.

Gain "stabilization" fights both systematical and random deviations from nominal value. Systematical like voltage and temperature dependent variations, random like process variations. Stability is used here in the common language sense in contrast to the specific meaning in system theory (e.g. Nyquist stability criterion).

 

[jasonc2]

Negative feeedback has many advantages (e. g. it stabilizes the dc operating point) - however, it reduces system stability because in real systems it always moves into positive feedback for rising frequencies.

Can you clarify what "it moves into positive feedback" means?

 

[LvW]

Can you clarify what "it moves into positive feedback" means?

For rising frequencies the phase of the loop gain function (starting at -180deg) get's more and more negative until it reaches at a certain frequency a value of -360 deg which is equivalent to positive feedback.
When the loop gain magnitude at this frequency is equal to or larger than unity (0 dB) the system is unstable. If the gain already is smaller than 0 dB the circuit is stable and can be used as an amplifier.

 

[D.A.(Tony)Stewart]

@LvW<

I agree your comment, that we are discussing the meaning of technical words and it is a pleasure to share different views with Engineers around the world with different backgrounds in this form vs other forums that I have encountered. I think we agree on the technical theory, but perspectives between block gain or loop gain with "gain stable" require care to avoid misunderstanding. It can mean several things.

In a "theoretical" sense, if we define the border between a system that can reach a steady state and one which grows with oscillations for ever, then we have the 0dB threshold but no margin for stability but it is theoretically stable with no disturbances.

Yet another Control Systems might not be considered "stable" unless the loop gain was < -15dB or "Gain Margin" > 15dB and a sine wave oscillator might need more gain to ensure it remains oscillating. and use non-linear feedback to control the amplitude. Oscillators with high gain stability on the other side, have high Q to reduce phase noise and frequency drift and then rely on logic gates to act as non-linear limiters with high, yet fixed gain to drive the noise into a square wave but if the gain is loaded excessively, might not oscillate, so high Q but low gain can create an instability and too much gain can create an instability for harmonic spurious oscillations.

So "gain stable" can have different meanings in both time response control systems and linear oscillator control circuits and may be somewhat an abbreviated term depending overall gain or loop gain margin is being considered.

In practise, these two diverse negative feedback systems need an acceptable margin criteria, to remain stable over a wide source of stress factors so an "operating point" near 0 dB margin might not be stable.
A general practise for 2nd order control systems is to use 15 dB as the acceptance criteria for adequate design margin.

I recall that we used this to define the servo phase margin tested in every 14" disk drive in production at Burroughs in Winnipeg, Mb, Canada using a Bode Plotter in the early 80's. Anything less than this margin was rejected, corrected with various components or worst case, another new massive baseplate and servo mechanical parts were put in. Consider if you had more than one overshoot pulse > 10% and servo said it had arrived "ONTRACK". If data position was ready to write to a sector that was just coming around on that track AND lack of loop gain margin continued to move the heads off-track with ringing {in this negative feedback control system}, it is possible that this data written offtrack might never be recovered, since it occurs over a fraction of a sector. This design used a large >1Hp linear voice-coil motor with >3 " stroke like a big woofer to drive a stack of heads up to 30 inches/sec and onto a track with 0.0005" error within 30 milliseconds. It was impossible to achieve the "optimal" value of Zeta or dampening factor of 0.7 but a gain margin > 15dB was possible, but any less gain margin was "unstable".

Now for RF amplifiers in a CATV repeater network gain is controlled in a local feedback loop and monitored remotely by a network manager who can remote control the gain and slope of the line amplifier feeding your home to <<1dB so that attenuation vs frequency may be compensated on coax to the home. This closed loop system has stable {loop} gain in the control characteristics but what is more critical is the forward gain downstream.

When we think of "stable gain in amplifiers" with closed loop gain , we tend to have plenty more than 15 dB margin so it is the stability of gain that is more critical, rather than gain margin of the AGC loop.

But in 2nd and higher order Control Systems, we are much more concerned about the a "stable gain margin" measured by the lack of ringing or minimal overshoot to a step input or namely the "step response", than the stability of the actual forward gain or feedback gain.

 

[LvW]

Hi SunnySkyguy,

Of course, I can follow your statements – and I can say nothing against it.
Now you bring the stability margin into the play – and this makes the discussion even more complicated.
Of course, there are applications which require a certain stability margin – and if this margin is not available
one might consider the system as „unstable“ because it may show an oscillatory step response.
But this would lead to a situation in which each application requires its own definition for stability.

Remember: Even the definition of the stability margin is based on the stability threshold where we have unity loop gain (real).

For example, what shall I answer if somebody describes a system with feedback and continues with the yes/no question: :
„Is this system stable or not stable?“

I think, there is no other choice than to rely on the commonly agreed definitions for stability (global stability, BIBO) and to apply the well known criterion (Nyquist).

 

[FvM]

But this would lead to a situation in which each application requires its own definition for stability.

This should be clearly avoided. Besides requiring stability, applications have specific performance criteria, e.g. overshoot, gain peaking, settling time for linear amplifiers, or integral of absolute error value for control systems, and different performance numbers.

Although they are often related to stability margin, they shouldn't be discussed under stability. I prefer to see stability as such as a binary quantity.

 

[D.A.(Tony)Stewart]

My perspective is that even binary logic is analog in many ways during transitions.
But I agree if you have to decide on steady state, its binary.

If I have to evaluate the stabillty of a negative feedback block then stability is analog with user preferences for criteria for stable and unstable vs full amplitude oscillations at extreme instability.

You won't many accidental linear systems with gain precisely at 0dB forward gain and 180 deg phase shift that make perfect sine wave oscillators without some non-linear feedback, unless specially designed as such.. So I agree in that sense too it either oscillates or it doesn't (ie binary state), like some high power class AB without an RB snubber or the same in vacuum tubes with the plates turning blue from RF plasma emissions like my Bogan stereo did after non OEM replacements.

But for Control System design stability for me, it is a question of analog margin to instability with various attributes that FvM indicated.. (thanks for reminding me on IAE name , BTW) So it can be either for me depending on how you analyze it.. Steady state or dynamic step response.. Lovely discussion.

I might also add that the famous criteria for stability (binary) are based on steady state, but in the real world dynamic conditions require margin so it becomes analog. e.g. Routh–Hurwitz stability criterion