Guest Post by Willis Eschenbach
I’ve been ruminating on the continuing misunderstanding of my position that a governor is fundamentally different from simple feedback. People say things like “A governor is just a kind of feedback”. Well, yes, that’s true, and it is also true that a human being is “just a bag full of organic chemicals, minerals, and water“ … but if you want to analyze humans, that’s far from sufficient.
Similarly, a governor is a kind of feedback … but if you want to analyze it, you can’t assume it’s just simple feedback.
For starters, let me offer my definition of a governor. A “governor” is a control system that uses both positive and negative feedback to maintain some system variable near a “set point” value. A common example in our daily lives is the cruise control on your car. It increases or reduces fuel flow (positive and negative feedback) to maintain the vehicle speed near some pre-set value.
I want to illustrate why you can’t analyze a governed system the same way you analyze an ungoverned system. Suppose I have a car, and I start out from a standstill on level ground. As I gradually add more and more gas by slowly pushing down on the gas pedal, the vehicle speeds up. I record the speed and the instantaneous fuel flow, and I get a graph like Figure 1. As is the usual custom, I’ve graphed the independent variable (fuel use) on the horizontal X axis, and the dependent variable (speed) on the vertical Y axis.
Figure 1. An illustrative graph of automobile fuel use versus speed, with no cruise control. Yeah, yeah, I know the shape of the curve won’t be exactly like that … which is why this is an “illustrative example”. [UPDATE: an alert reader has pointed out that the speed is of course in miles per hour (MPH) and not miles per gallon (MPG) … I can’t be bothered to redo all the graphs, so please just make the appropriate mental substitution.)
Figure 1 shows that the vehicle speed is some kind of function of fuel use. It’s not a straight line because as you speed up, the air drag on the car slows it down. And the drag increases by something like the cube of the speed. So as we add more and more fuel, the effect of each added unit of fuel decreases. You can see that at sixty miles per hour, the car uses 2 gallons of gas per hour.
The takeaway point from this analysis is that in terms of causation, we can see that a change in fuel flow causes a change in speed.
Now, let’s get the car going sixty miles an hour on level ground, turn on the cruise control, and once again measure the speed and the gas. Then, suppose the car goes uphill. It will start to slow down. As soon as it does, the cruise control will react to the reduction in speed by increasing the gas flow to keep the speed near to 60 mph. And when the hill steepens, speed goes down a bit more, and in response the governor further increases the fuel use to keep the speed up.
Then we crest the hill and start down the other side. The car starts to increase speed, and when it gets going faster than 60 mph, in response the cruise control will reduce the fuel flow.
This leads to a very curious situation—vehicle speed goes up as the fuel use goes down, and speed goes down as the fuel use goes up. Figure 2 shows an illustrative graph of the situation .
Figure 2. An illustrative graph of fuel use versus speed, with a cruise control set for 60 MPH.
I’m sure you can see the problem. If we analyze this situation in the exact same way as we analyzed the situation without the governor, we come to the ludicrous conclusion that if we increase the fuel use, it will reduce the speed of the car.
So how can we understand this change? The key to analyzing this system is to understand that the governor (cruise control) reverses the causation in the system. Without the governor, as mentioned above, a change in fuel flow causes a change in speed.
On the other hand, with the cruise control engaged the exact opposite is true—a change in speed causes a change in fuel flow. And as a result of this reversal of causation, our previous analysis method is useless because it incorrectly assumes that a change in fuel flow causes a change in speed.
But wait, it gets worse. Suppose we start up the car from a standstill as in Figure 1, but then when we get up to 20 mph we set the cruise control for sixty mph. The car will continue to accelerate as in Figure 1, but only until the car gets up to sixty mph. When it gets there, it doesn’t speed up any more. Instead, it takes up the pattern shown in Figure 2. So the complete graph of the run looks like Figure 3:
Figure 3. An illustrative graph of fuel use versus speed, with a cruise control that is set to 60 MPH just after the car starts moving.
Here we see that in the blue part of the graph, the change in fuel use causes a change in speed … but once it gets up to the set point speed of 60 mph, the causation reverses, and now the red line shows that a change in speed causes a change in fuel use. Note that this reversal does not require any change in the governor settings. When the situation is far from the set point, the causation goes fuel –> speed. But once it reaches equilibrium, causation reverses such that speed –> fuel.
How does this relate to the climate? Well, the underlying climate paradigm is that the forcing controls the temperature, such that a change in forcing causes a change in temperature.
On the other hand, I’ve proposed that there is a natural governing system regulating the temperature of the climate, a major part of which works as follows.
In the tropics, when it is warm, clouds form earlier and reflect away the sun to cut down the solar forcing. And when the tropics are cool, the clouds form later or not at all, which greatly increases the solar forcing.
What that means is that a change in temperature causes a change in forcing.
And that is why I say that the current method of analyzing the climate is totally incorrect, because it assumes the causation is going the opposite direction from what is actually occurring.
In closing, I can do no better than to post up this marvelous cartoon by Josh from five years ago showing my “Thundercloud” governor at work …
Best wishes to all,
w.
As Usual: If you disagree with someone, please quote the exact words you object to so that everyone can know both who you are addressing and just what it is that you disagree with.
Willis , you really need to find out what the terms mean if you want to discuss this kind of thing.
Positive and negative f/b does not relate to whether the thing being driven is going ‘up or down’. A negative f/b is one which acts to counter a deviation and a positive one is one which acts to increase it further.
If a hill slows the vehicle down, the cruise control acts as a negative feedback by increasing the gas.
Mike, fully agree.
Thanks, Joihn. It was a late-night exercise, I’ve struck out the one incorrect phrase.
All the best,
w.
Well being personally of the electronics persuasion, rather than mechanical, I use the term ” regulator ” rather than governor.
But I agree that there is a basic difference with just ” feedback ” in that with a regulagovernator you compare the system output parameter of interest (speed or other variable such as Temperature) with a fixed value (set point) and that output error variable becomes the source of the corrective input signal (via feedback). that will drive the output towards the set point.
In the case of say climate cloud feedback; there is no set point that we have control over.
The ” system ” has a preferred steady state (not equilibrium) that depends on zillions of variables,
For a given driving force such as TSI there will be some preferred steady state.
If some of the variables change, the system has a new preferred steady state.
Variables such as cloud cover; area, optical density, persistence, location directly alter the drive signal (TSI) and so change the preferred steady state.
Because the system is chaotic, we cannot always know what that adjustment might be.
If we add more CO2 and that reduces the output cooling rate, so Temperatures tend to rise, that Temperature rise, will cause more evaporation and lead to more clouds, which act to reduce that active part of TSI that reaches the surface.
So the net effect of increased CO2, would tend to be an increase in cloudiness, in all of its aspects, and an upward tweak of the Temperature.
That is NEGATIVE feedback, and with 7% increase in evaporation for a one degree C increase in Temperature, that seems like an extremely powerful feedback effect, that makes the effect of CO2 rather puny.
g
But the only set point in this feedback system is what all those myriad variable collectively vote on. Mother Gaia knows exactly what the Temperature and cloud cover need to be at all times.
Ditto.
The Stefan-Boltzmann law is a governor of sorts, increasing or decreasing radiation by the 4th power in an effort to maintain a temperature.
The missing energy may simply be increased outgoing-long wave radiation. As it gets warmer, more radiation/energy is emitted outward. Planck radiative feedback (or governor) is discounted by climate science but appears to be in the -4.0 W/m2/K range from the surface (measured by the missing energy) and -2.0 W/m2/K from the top of the armosphere (measured by Ceres).
Why this known science feedback is ignored and only positive ones like water vapor are counted in the climate change prophesy is beyond me but it reduces the global warming expected to 1.3C per CO2 doubling.
Amen. The use of the terminology is completely wrong.
On the other hand, Willis has the right instinct. Using control system theory to analyze the planet’s climate is seriously dodgy. Some other kind of analysis is probably much more appropriate.
Example: Consider a beaker of boiling water. The temperature will be maintained somewhere near the boiling point (until the water gone).
– Does that mean it’s a negative feedback control system? It is not.
– Might someone try to analyze it that way? They might.
– Would that analysis be valid? It might be “close enough”. Maybe.
– Could the system be more correctly analyzed without the use of control theory? Absolutely.
The climate is not a control system. Analyzing it that way is a gross simplification. People do that because they think they understand the mathematics. They are wrong to do so. We do not understand the climate well enough to justify the use of that simplification.
The climate IS a control system. Otherwise we would not be here to discuss it. ( Unless you believe that some deity is up there taking care the temp for us, and all the natural feedbacks are just noise put there to give us something to lbog about ).
commieBob,
Your analogy is only partway there. The beaker of boiling water would need a special coating that changes reflectivity and emissivity as the temperature of the water changes. If the beaker is close to a radiation source, that special coating will cause net radiative forces to go up or down, keeping the temperature of the water relatively constant for as long as the radiation source remains.
Using control system theory to analyze the planet’s climate is seriously dodgy. Some other kind of analysis is probably much more appropriate.
No., using control system theory is exactly right, that is not the problem, the problem is in using simple linear control system theory when you have a complex non linear climate.
Consider: You have a lump of ice the size of Greenland and you apply a certain excess watts per meter to it. The ice will start to melt, but the temperature of the remaining ice’ and indeed the meltwater, will not change They will hang around 0°C. At the melting point of ice, the relationship between energy input and temperature rise, is completely non linear.
Likewise at the boiling point.
The author has also identified a localised feedback system that works within a certain temperature range in the tropics. The albedo of tropical clouds which are full on negative feedback as well – until the clouds vanish, or its totally overcast.
At this point the cease to be feedback. They cease to govern. 100% cloud cover has the same albedo no matter how thick it is.
Handling these sorts of feedback systems is normal for engineers, and we understand that outside of the linear ranges we are into a very very hard bit of analysis, and there are no simple answer, just a selection of kludges and bodges and approximations that get us somewhere near. And most of our engineering is about designing systems that do NOT go into non linear ranges in normal operation, or if they do, do so in as predictable a way as possible.
And even then we know that its probably easier to e.g. take that race car with the rubber bump stops and the progressive rate shock absorbers out on the test track to optimise it, because calculations rapidly get horrendous.
The gaping hole in climate theory à la IPCC is that they assume a linear relationship between radiative forcing and temperature rise, multiplied by a positive feedback term bodged to fit a very short 20 year period of time.
Its so naïve its pathetic.
Firstly, what they don’t appreciate is the extreme non linearity of climate response to forcing, which makes their equations bound to fail. The water cycle that carries heat from the surface to the stratosphere and forms clouds, dumping terawatts of energy to space at night, and reflecting terawatts by day, is one such extremely non linear negative feedback system.
Secondly, what they don’t appreciate is that negative feedback delayed in time creates instability as much as positive feedback does. Taking the author’s cruise control, if it’s slow to respond, and then over reacts, the speed will oscillate about a mean : at times it will accelerate madly, then slow dramatically as the negative feedback tries to keep things to a mean, but fails to do it fast enough. The multi decadal ocean swings noted in the earth’s climate are just this sort of effect.
Add in some non linearity and you have a chaotic system with possibly multiple attractors. And that is precisely how the geologic climate appears to have behaved.
Sophisticatedsystem analysis can show that that is exactly what you would expect. The problem is, it also shows you how monumentally incalculable the solution may be, which makes nonsense of prediction à la IPCC!
Just because the warmists unbeleivably crude and naïve attempts at linear systems analysis have failed, doesn’t mean that system analysis is the wrong tool. Its just being used in a stupid and desperate way to prop up a pre-ordained commercial and political narrative, for profit and gain, not to actually represent the climate according to best practice
This would also be true if the water weren’t boiling, and it would be more like the actual climate. Geothermal heat is not boiling our oceans away.
@Leopold Danze Smith SPOT ON.
My 2007 GMC had to have a software change to the computerized transmission. The tech must have put in a wrong entry because my Cruse control did exactly like you described. Took it back, they claim they fixed it, but I still “feel” (cant see any change in speed or rpm) the vehicle slow down and then suddenly get pressed into the back of the seat.
You could indeed say that boiling water is a negative feedback, as the rate of boiling prevents the temperature from rising above 100 deg C; as the rate of heating rises, the rate of boiling rises. If you flash heat water, it can get above 100 deg C until it has a chance to convert to steam, suddenly.
Actually, I agree with everything you say except for “using control system theory is exactly right”. Practically, any attempt to do that will be either wrong or intractable. In fact, you eloquently make that point for me. Having said that, I am not proposing any viable alternative. 🙂
Exactly so.
“At this point the cease to be feedback. They cease to govern. 100% cloud cover has the same albedo no matter how thick it is.”
Convective clouds never reach 100% since there are always always areas where dry air is descending.
Also 100% cloud cover does not always have the same albedo. Thin clouds have lower albedo than thick ones. All clouds are translucent to some degree, otherwise it would be completely dark when the sky is cloud-covered.
yes. makes total nonsense of the article.
He’s correct, Willis.
Doesn’t negate your point, but misuse of some terms may turn off certain folks and that would be unfortunate.
Correct. There is no positive feedback in a cruise control. None. Zero.
Thanks, Mike. I wrote this late last night, I’ve struck out the offending phrase.
w.
Good man, Willis. Glad to have you back on board.
Shouldn’t speed in the graphs be MPH (miles per hour) rather than MPG (miles per gallon)?
Good catch, Jeff. I’ve put a note up in the head post, no time for changing it, gotta go.
w.
governor/feedback – seems to be essentially mass-spring behavior = simple to make with complicated inertial-motive responses whether these are thunderclouds or human skeletal muscle tissues
“It increases or reduces fuel flow (positive and negative feedback) … ” That is where you are going wrong. Fuel flow can be increased and decreased and both situations can be negative feedback.
The feedback signal is subtracted (negative!) from the set point – your 30mph, say. If the difference is positive that means that we are not getting enough negative feedback yet so we add fuel. The speed increases so the negative feedback increases until the difference between the feedback and the set point is zero.
If when the feedback is subtracted (negative!) the result is negative then you are getting too much negative feedback. To reduce it we will need to reduce the speed and so a reduction in fuel is required. This will reduce the speed and reduce the negative feedback until the difference with the set point is zero.
Negative, as in negative feedback, just means that the signals are subtracted. The result of that subtraction can be either positive or negative but that is after the feedback loop. The resulting positive or negative signal will be fed into the forward part of the control system.
The spring in a traditional light switch is an example of postive f/b. Once it passes the mid-point in the movement, ( the tipping point as our freinds would likely call it ) the spring acts to accelerate the movement. A small displacement will cause it snap to the on position. It will then meet a strong negative f/b in the form of the rigid body of the switch. This is a very common situation where a positive feedback is eventually bounded by a dominant negative feedback. Any system without this kind of effect would be completely unstable and would only exist fleetingly.
It amazes me how many people do not understand the concept of set point and diff!
The red and blue lines do not represent the same environment for the car. The comparison is false.
What does the blue line look like if you put the same hill into the storey?
I’ll answer my own question: as the car climbs the hill the blue line will deviate to the right of the cubic relationship, having a slower equilibrium speed for a given fuel consumption . As it approaches the brow of the hill it will fall back to rejoin the original blue line. As it begins to descend the other side, the fuel graph will veer off in the other direction accelerating to a higher equilibrium speed than the blue line. Finally returning to level ground it will return to the cubic relationship.
If we explore the graph while on a steady upward climb it will have a similar curvature but lie below the level ground, blue line. Conversely, on a downward slope the curve would be above the original blue line. Mathematically, they would both be linear + cubic as fuel is converted into gravitational potential energy, on one case and in the other some fuel is substituted by recovering gravitational potential energy.
So we see the whole curve is shifted in a way similar to the way the red dot of the cruise control point moves to describe the locus shown as the red line in Willis’ fig. 2.
The comparison is classic apples and oranges.
The blue line without hills compares to the red dot: cruise control point on level ground. The red dot is ON the blue line. The red line should be compared to a 2D surface swept out as the coeff of the linear term ( reflecting the gradient of the hill ) is varied and the curve is shifted to one side then the other.
Willis has confused the issue by introducing a third variable into the discussion but only applying it to one of the cases he was discussing.
The cruise control does not change the behaviour from the blue line to the red line, it locks it to one red dot ON the blue line. If you introduce a hill the cruise control red dot with still be on the NEW blue line corresponding to the same conditions.
The red line represents the set of all possible dots on all possible blue lines at 60 mph.
I do not understand this fuss. I understand perfectly (I think) what is being portrayed.
I consider a control system to use negative feedback but is not itself negative feedback.
I agree that a lag in the control system feedback makes it unstable and it will probably oscillate or even self-destruct (think of “Galloping Gertie”).
http://www.wsdot.wa.gov/tnbhistory/connections/connections3.htm
The question of lag was described by someone else very well below. However, that is irrelevent to the article which considering fuel consomation after settling to an equilibrium speed. This is not discussing transient behaviour.
No, the cruise control introduces an effect with the opposite sense ( negative feedback ) which reduces the original variability. It does remove it totally, nor reverse it.
The red line is due to introduction of a third variable which would also have moved the blue line in a similar way had it been discussed.
The idea that the cruise control changes behavious A ( blue line ) into behaviour B ( red line ) is erroneous. They are not the same thing.
The cruise control will only control the speed when going downhill if it is also coupled to the [braking] system otherwise the car will continue to accelerate (until other mechanical factors contribute).
GIL – depends on your vehicle. Some vehicles apply engine braking to maintain speed on the downhill leg, disengaging over drive and gearing down. Depends on how sophisticated your automobile cruise control is. I have two vehicles that will control speed on both uphill and downhill very well, and one older one that indeed will accelerate downhill unless I drop it our of drive then it works fine, both uphill and downhill.
I’m not sure I’m seeing the complexity here.
Using the analogy the sun forces temp rise and the petrol forces speed increases, and the clouds provide negative feedback on the sun in response to temp and the governor provides negative feedback on the petrol flow in response to speed. .
So in the first case beyond a certain temp the clouds provide negative feedback so that as temp increases the impact of the sun reduces (to the point where temp starts to drop), and if under the certain temp it increases the impact of the sun.
In the second the governor also provides negative feedback so that as the speed increases it constrains fuel flow (to the point where speed starts to drop), and if under the certain speed it increases the fuel flow.
In both cases causality exists between forcing/petrol and temp/speed, it is just that the clouds/governor acts as a damping control over the relationship.
If you treat both as black boxes, continued forcing without a governor causes the systems to be unstable within the systems’ constraints, with a governor they both become stable (but within limits no doubt).
“and the governor provides negative feedback on the petrol flow in response to speed”
Not quite the same thing. A governor enforces a set-point, using a LOT of feedback to achieve “lock”. A phase-locked loop is another example.
Mere negative feedback reduces the effect of a change in input in a linear way. 50 percent negative feeback reduces climate sensitivity 50 percent.
An effective control system changes the negative feedback dynamically. Around the set point, the negative feedback will be minimal or zero; as the velocity (in this example) increasingly deviates from the setpoint, more negative feedback is utilized; but it might not actually be “negative feedback”.
What, exactly, is being “fed back” in the automobile example? Velocity is being measured, not “fed back”. The difference of measurement from setpoint will cause movement of the control input but it must be damped or it will oscillate.
A governor may be thought of as a specialized form of feedback, focused more on maintaining a more or less constant setpoint rather than trying to follow a control signal. Governors also have a connotation of enforcing limits, e.g. an engine governor may be set up to limit maximum speed. Governors also have the connotation of responding to variations in load as opposed to changing the settings on the control knob.
Willis is right in that CO2 is an external forcing to climate, NOT a control knob.
FWIW, the “feedback” from tropical thunderstorms sounds more like a limiter to me. Limiters have a very non-linear response to input. In the case of thunderstorms, the vapor pressure of water roughly doubles for every 20F increase in temperature and this vapor pressure is independent of the partial pressure of CO2.
Micheal 2
I hadn’t really intended to get into the “negative feedback” issue but the problem is there are a number of specialist definitions that means one can argue over whether a governor is or isn’t.
At its heart however a governor is both a feedback mechanism and negative.
It takes a signal (temp/speed) and uses it to modify it (hence “feedback”). It is “negative” because it dampens the signal, accelerating when below the set point or accelerating when below it.
The fact that we use language like “it was the negative feedback that caused the oscillation” shows that if the result is oscillation we don’t still call the process negative feedback.
“What exactly is being ‘fed back’ in the automotive example?” Information about the speed is being used to modify the fuel flow.
Erik Magnuson
I was going to add in my earlier comment that in the analogy the hill is misrepresented (others have said this on this thread). The hill is part of the environment that equally has a separate influence on the signal (speed). In this sense it is the equivalent of the concentration of CO2 in the analogy. Just as the governor will respond to hitting a downhill run and is blind to the fact that the increase in speed wasn’t due to an increased flow of fuel, so the clouds will respond to an increase of CO2 in the atmosphere and be blind to the fact it had nothing to do with the sun hotting up.
Sorry – should read
“…. accelerating when below the set point or decelerating when above it.”
A governor is a governor. Feed back is feedback. Totally different animals. Next thing you know , people will be saying that CO2 is a control knob
A governor is a control device that applies a negative feedback.
“A governor is a control device that applies a negative feedback”
A governor is a control device that applies a negative or positive feedback depending on which side of the set point it is on.
No Roy, you are making the same mistake as Willis did. See comment #1 . You also need to read up on what ‘negative feedback’ means. If the control variable is dropping below the control point, a negative feedback will act to bring it back UP.
I disagree. A governor can produce either a positive or negative feedback, depending on how it is tuned. Excessively turning up the gain or derivative on a PID controller (governor) may produce positive feedback and an unstable system. In some systems such as level control, this is not the case, in fact a very high gain is ideal. Temperature control needs PI, with I, Integral, to prevent reset windup, although with simple systems, P is sufficient.
That’s correct if you want to extend the definition of “governor” to include every kind of control system including PIDs. What was wrong is the idea that positive feedbacks drive things up and neg. f/b drive them down. That is NOT what negative means in this context.
A governor always applies a negative (counter-deviation) feedback.
I have decided a pox on both of your houses. This is not feedback. Negative feedback, often used in amplifier circuits, simply inverts the output and feeds it back to the input to stabilize it. The proportion of feedback is not dynamic.
A control system measures something, then commands a control input in a non linear way to seek the setpoint. By non-linear I mean the strength of the control change will vary depending on distance from the setpoint and how aggressive it is designed which itself takes into account inertia of the system response as otherwise it becomes unstable and will oscillate (I mention Galloping Gertie above).
Feedback is directly applied to the input. A control system measures a thing (velocity) and controls something (fuel flow) seeking a setpoint using an algorithm that might not be, and probably isn’t, a simple mathematical function.
I once built a cruise control for my UV (long before SUV). It was too simple and would surge way over the setpoint, then throttle back to idle nearly coming to a stop, then surge.
Mike is right. A governor always tends to drive the system controlled variable towards the set point.
If the speed / Temperature / whatever, is below the set point, the drive is increased to move it towards the set point, I. e. increase.
If the speed / temperature / whatever is above the set point, the drive will decrease to move it towards the set point I. e. decrease.
The ” restoring force ” is always in the direction opposite to the error from set point.
Now that presumes the system is designed to be stable over all possible variable values.
So if (x) is the displacement from the equilibrium value (set point), the restoring force is given by : fF= -kx where k is a transfer function value.
In a mechanical system, we have that F = m a = m d^2x / dt^2
So we get m d^2x /dt&2 = – kx
This is well recognized as the differential equation of simple harmonic motion and has solutions of the form x = A sin (omega t) + B cos (omega t)
In a practical system, there will be damping that ensures that the amplitudes A and B do not steadily increase, and in fact a designer would seek to get the fastest settling time to a new value, whenever something changes, without overshoots or long time delays to reach within an acceptable error band.
But the restoring forcing is always towards the set point, so opposite to the displacement from the set point. And that makes it negative feedback.
With positive feedback, the feedback forcing, would drive the operation further away from the set point and increase the error.
That does not mean it will explode; but it will be more unstable.
Michael 2:
You’re virtually alone here in correctly recognizing that negative feedback–in the rigorous scientific sense–is the instantaneous addition of the INVERTED output signal to the input. Inasmuch as the input into the planetary climate system is SW solar radiation and the output is the OLWR at TOA, it should be apparent that we have a feedTHROUGH, not a feedBACK, system! Nor is that system any “governor,” in the sense of limiting the system output to the vicinity of some predetermined “set-point,” irrespective of the input.
That quasi-steady response is the consequence of the virtually constant energy of TSI provided to the system as excitation. Variations of surface temperature due to internal changes in trace GHGs are negligible compared to the those produced daily or seasonally by changes in cloud-modulated insolation. All the GHGs, including water vapor, are simply capacitive/inductive components that effect cannot heat storage in the convectively-heated atmosphere beyond their very minor mass.
In the last sentence, it should read “…cannot affect heat storage…”
You guys are missing the point. A governor is a switch, a feedback is an effect.
IMHO, you guys are quite correct, but are all missing real point of Willis’ post, that the cause and effect of the fuel and speed is reversed by the ‘cruise control’, and that this is not understood (or is ignored) by many when discussing AGW. Hence the false assumption (among many I’m sure) that CO2 drives global temperature, when the reverse is the case. Thanks everyone, and especially Willis.
Right! I was waiting for someone to recognize this. A governor limits the allowable range of system. If climate were controlled by a governor the temperature would stop rising when it gets to a certain point (the IPCC turns off the sun when it gets 2C above nominal).
Boyfromtottenham & BobJ
No, Willis doesn’t show anything about the cause and effect of fuel and speed (or sun and temp). In all cases the former causes the latter.
The impact of the introduction of a control is to change the amount of fuel/sun in response to speed/temp.
The point is that an increase in CO2 (say) causes an increase in efficiency of the sun that causes an increase in temp that causes a change in the clouds that causes a decrease in the efficiency of the sun that causes the temp to decline. The observed relationship between CO2 and final temp in any such system is contingent, but in this case there is nothing to suggest that change in temp causes a change CO2.
That is the simple case. Consider now the case where you press the gas pedal and car decelerates. Then you release the gas and the speed starts to accelerate and goes faster and faster even though you use the brakes. Then you release the brakes and the car slows down. Where were the hills and valleys?
Humans have an anticipating central governor: http://runnersconnect.net/running-training-articles/central-governor-theory. Your running slows down before you run out of fuel.
Another BIG difference is Ramp bias ,measured in time and % of rated motor load in response to a deviation of set point
“It increases or reduces fuel flow (positive and negative feedback) to maintain the vehicle speed near some pre-set value.”
As already mentioned, this statement is completely wrong. Servomechanisms are specifically designed to use negative feedback. If the fuel flow is increased in order to maintain the correct speed then it is still negative feedback.
As an electronics design engineer I had a lot of experience with all kinds of electronic and mechanical servomechanisms. Particularly with mechanics, inertia means that, as frequencies increase, the response lags more and more until negative feedback becomes positive feedback. Result: oscillation. The whole point of servomechanism design is to ensure that it remains a negative feedback system over the frequency domain where there is significant system gain.
Chris
Very nicely put: clear, concise and simple. 😉
Yes, unfortunately Willis’ apparent misuse of the term “positive feedback” has caused many to miss his actual point. All he is trying to point out is that the climate is operating in a feedback control regime where it is being controlled within limits by many factors, and that confuses people observing those factors. The cloud factor that he described is a negative feedback control mechanism for the temperature in the tropics. As long as the dominant feedback is negative, the climate system will operate within a stable control regime. However, while operating in that controlled regime things will seem backwards to an observer. If temperature is forced up by something, then the negative feedback adjusts to force it back down. Therefore, one will observe more clouds as well as higher temperatures. Determining what or which came first can become confusing. Simply, in a controlled system the controller uses a manipulated variable to maintain a controlled variable at a set point. While operating in a negative feedback controlled regime, the controller’s actions literally reverse the apparent relationship between the manipulated and the controlled variables, potentially confusing folks on cause and effect. That is Willis’ point.
No Willis has confused the issue by comparing a two variable problem to a two dimensional projection three variable problem. He’s comparing apple to oranges. The relationship appears reversed because there is another factor that is not shown on the graph : gravitational potential energy. One graph is level ground the other is not.It is the new “fuel” which is changing the relationship not the governor.
I explained this is some detail above, but several people just jump in and start commenting without reading.
http://wattsupwiththat.com/2015/08/02/problems-with-analyzing-governed-systems/#comment-1999034
The neg. slope results from the fact that the regulation provided by the governor is not perfect and there is a small residual of the effect or the external input ( gravity )
Mike, I agree with you wholly here. But, how is that different from what goes on in Climate Science anyways? While I appreciate the engineering pedants deeply, it doesn’t detract from Willis sketch of the epistemic problem and the failures that can arise from within it.
Mike, I was discussing Willis’ point about the climate, not cruise control. Although the comments on the use of terminology in his post were correct (and he modified the post to take out those references) the issue that Willis was driving at is that introducing negative feedback in a system changes the system operation. The “Main Stream” climate scientists are intent on ignoring any negative feedbacks and focus instead on every possible or imagined positive feedback. They completely ignore the negative feedback control that is operating daily in the climate system, which is a major reason why their estimates of CO2 sensitivity are so high and that their models are failing so miserably. The entire premise of Catastrophic Global Warming is based on and requires the imagined positive feedback of increased CO2 causing an increase in water vapor in the atmosphere, which has not happened based on actual measurements.
You got it Mike.
Please provide metric graphs for the rest of the world outside of the USA…
It is especially helpful to have Amps and Volts in metric units.
Amps and volts are already metric units.
+1
And, perhaps you just don’t want to do the conversions yourself..
+1
According to alexa 60% of the vistitors to this site is non-US
No ! The metric (SI) units are : amp and volt .
No caps and no plurals. But the symbols are : A and V .
g or G
English is way more fun:
Poppyseed, Line, Barleycorn, Digit, Finger, Inch, Nail, Palm, Hand, Shaftment, Link, Span, Foot, Cubit, Yard, Ell, Fathom, Rod, Chain, Furlong, Mile, League.
🙂
“Suppose I have a car, and I start out from a standstill on level ground. As I gradually add more and more gas by slowly pushing down on the gas pedal, the vehicle speeds up.”
Cringe.
A diesel engine is moderated by fuel, a gasoline engine is moderated by air. Your analogy is fine for a diesel engined car, not gas.
Gasoline engines are controlled by air flow, the fuel system reacts to changes in air flow. Sorry to confuse you, Dean.
Dean says “So I can go faster and faster simply by adding more air, which is free, and no more fuel, which costs money? Please tell me more!”
More is coming… wait for it…
Open the hood, look inside. The “gas pedal” goes to the throttle plate. That’s the only place it goes. You can operate the throttle plate by hand if you wish, grab it and turn it, let more air into the engine.
In most (maybe all) fuel injected engines, an air velocity measuring device measures how much air is entering your engine and squirts the appropriate amount of fuel into the intakes at the appropriate time. In older carbureted systems, the airflow itself pulls fuel at is flows through a venturi, increasing velocity but also reducing pressure.
So the control input in either case is adjusting how much air enters the engine.
PS: Air may soon no longer be free, not the CO2 part of it anyway.
The y axis of each graph says MPG. Should be MPH.
Thats fascinating Willis. Have you got any thoughts on what a governor forcing might look like? Because if it is possible to define a governor “fingerprint”, you could predict the characteristics of that signal, then see if a forcing matched that fingerprint. I know the current thinking is albedo – does that fit the fingerprint?
I think Willis’ governor is the tropical cloud band over the Pacific.
Couple nit pickies here. On the charts, the ordinate label should be MPH rather than MPG, and the abscissa should be Fuel Flow, rather than Fuel Use (fuel use would be in gallons only). In Fig. 3, the blue line should extend beyond where it ends.
The red and blue lines describe different items. The red line describes governor response; its slope is set by the manufacturer, tuned to the engine-vehicle combination. A whole set of roughly parallel red lines would describe settings for different speed settings. The blue line describes flat terrain vehicle performance.
If this is climate, replace y axis ( speed ) by temperature and x-axis ( fuel consumption ) by incoming SW in the tropics. The hypothesised ‘governor’ acting via regulation of the timing, extent and duration of cloud cover.
The negative slope of the fig.2 is the proposed negative cloud feedback. This will act to reduce the slope of the overall δT vs δrad relationship which has a positive slope. That slope is climate sensitivity.
No problem with that part of climatology and feedbacks.
It’s interesting to learn how engineers define and use “positive” and “negative” when discussing feedback but for laymen Willis’ example is clear no matter how feedback is defined: a changing temperature can change forcing to one which acts to reverse the temperature change. At least that’s how I see it.
Paul, perhaps you should read comment-1999034 and state clearly ( ie QUOTE me ) where you think I am incorrect, before declaring me “lost”.
http://wattsupwiththat.com/2015/08/02/problems-with-analyzing-governed-systems/#comment-1999034
Sorry Paul , that was supposed to be a reply to “Cube” below.
Once you start using systems analysis, you are firmly in the domain of engineers, and it behoves you to use the terms as they do, or else time will be wasted arguing definitions.
Negative feedback acts against the sign of the amplitude change. That is the definition. It does not become positive feedback because the amplitude of the change reverses sign.
Nice analogy Willis. Mike, I think you’re lost in the details and completely miss the point.
“Cube”, perhaps you should read comment-1999034 and state clearly ( ie QUOTE me ) where you think I am incorrect, before declaring me “lost”.
http://wattsupwiththat.com/2015/08/02/problems-with-analyzing-governed-systems/#comment-1999034
I think what Mike is trying to say is that the reverse illusion the author refers to isn’t necessarily caused by the cruise control, rather it is caused by the hills COMBINED with the cruise control. If it were level ground there would be no inverse relationship to fuel consumption and speed with or without the cruise control. The author’s point is that it is mostly the cruise control that causes the inverted curve. The way I see it is the hills are the primary cause, and in that particular environment cruise control inverts the curve. Same thing would happen if a person where driving. I don’t see there is much of a point to make about this.
Or I suppose another way to look at it is that a phantom reverse correlation is created between fuel consumption and speed, due to the correlation between the slope of the hill (up or down) and throttle position, i.e. fuel consumption rate, that is forced by the cruise control (or a person driving at a constant speed which is really just cruise control). If you account for this forced correlation and subtract it out you do not get the phantom inverted curve. But if you don’t account for it, as this example doesn’t, then you get the phantom inverted curve. Moral of the story is that failure to account for correlations such as this can lead your results astray. I guess that is a strong point from that perspective.
Tony:
“Temperature control needs PI, with I, Integral, to prevent reset windup, although with simple systems, P is sufficient.”
Most systems these days are PI, however, the ‘I’ can cause windup!
With P only, there is no integral in use, therefore no windup/lockup is possible, but the downside is that without ‘I’, you will always get an offset ( a difference between what you want and what you get).
Governors (PID controller plus actuator) always operate with negative feedback.
Many controlled systems are adjusted to have as much gain (P) and as much integral (I) as they can tolerate whilst remaining stable. We call this Quarter decay tuning.
Willis, should the Y-axis in your graphs show MPH rather then MPG?
A common example in our daily lives is the cruise control on your car. It increases or reduces fuel flow (positive and negative feedback) to maintain the vehicle speed near some pre-set value.
No, that is all negative feedback.
At this point I lost the will to live.
NEGATIVE feedback is feedback that opposes the tendency of the system to do what it start to do. So if the car starts to slow. NEGATIVE feedback opposes the deceleration and speeds it up. Conversely for if it speeds up downhill. NEGATIVE feedback opposes the acceleration.
Please Mr Watts, do not accept articles that start and are based entirely on a misconception of what clear and well define engineering terms actually mean.
Positive feedback is ‘runaway instability’. The tendency of a pencil balanced on its point to fall over. The tendency of an atomic reaction over critical mass to degenerate into an atomic explosion. The tendency of a microphone coupled to an amplifier and loudspeaker to go into full volume howlround.
A governor is a simple negative feedback system.
End of.
“The tendency of a microphone coupled to an amplifier and loudspeaker to go into full volume howlround.”
Which is of course a positive feedback, bounded by a dominant negative f/b: to power available to the amp.
That is exactly what I said. Why do you comment as if it negates it?
Positive feedback is ‘runaway instability. ….The tendency of a microphone coupled to an amplifier and loudspeaker to go into full volume howlround’.
“A governor is a simple negative feedback system.”
I disagree. A governor contains intelligence and maintains a setpoint exactly, negative feedback merely resists a control input producing an output change than is less than would be the case without the negative feedback.
A governor can maintain 60 mph exactly. Negative feedback cannot since there must ALWAYS be at least some error signal for the feedback mechanism to operate.
But a governor measures the velocity and makes corrections to land exactly on 60. It can even have attitude measuring inputs to sense when the vehicle is starting an ascent and add fuel/air in advance of measured speed drop using previously calibrated tables of how much energy is needed for any particular situation. The governor could also use accelerometers with vastly quicker response time than the drop in speed of the vehicle itself.
A governor does not maintain a speed, it limits the speed to some set point. Governor usually have upper and lower droop settings which control how the speed will vary above and below the set point on road grades. A governor is a hard limiter which is a negative feedback system with non linear feedback. In fact, a governor is often called a “vehicle speed limiter“.
To suggest the governor will “maintain 60 mph exactly” is absurd as there will always be some variation in the speed. In fact you admit it does when you state “But a governor measures the velocity and makes corrections to land exactly on 60”. When a vehicle starts down a hill the speed will increase and the governor will reduce the fuel appropriately.
Well Leo you don’t have it correct either.
Positive feedback is certainly going to increase the instability, but that does not mean it becomes ” runaway instability ” That is only if the loop gain exceeds 1.
Early radio receivers used positive feedback to increase the signal gain, but they stopped short of runaway gain increases, which lead to “squegging” where the gain increased until the system ran into one of the stops, either full on or full off. And once at the limit the loop gain drops to near zero and it starts all over again
BDTP beat me to it. Y axis should be MPH. (Ot M/S for you SI bigots. You know who you are.)
The terms “Negative feedback” and “Positive feedback” are misused/misdefined. Traditionally, the +/- notion of feedback defines whether the feedback term is added or subtracted at the summing junction.
Negative feedback will cause an increase in output if the output is less than the set point.
Mike
Hello Willis,
Off topic here, but you have marine experience so—-
Are you or anyone else with access to the historic data interested in attempting to do an analysis of the current flow to find a possible location for the crash of flight 370. I expect that the search area they used was a bit off.
You will also need to account for sea surface movement over the period the floating debris was in the ocean. IIRC the sea surface moves downwind at about 10% of the surface wind speed.
Gerald, the problem is that the ocean is chaotic in the sense that if you go out into the ocean and drop two floats, one on each side of the boat, most of the time they’ll end up in very, very different locations … not good.
w.
Are you sure it’s chaotic and not random? Why/why not? Use math. Thanks.
Still waiting for math. Tell us the difference between chaotic and random. Double dog dare you.
Dinostratus August 3, 2015 at 4:51 pm Edit
Dinostratus August 4, 2015 at 8:21 pm Edit
Thanks for the question, Dinostratus, but I fear that your manner of trying to get your way doesn’t work with me. I don’t respond well to that kind of ragging.
However, your luck is amazing. As it happens I’ve been looking at that very question for a couple of months now. I have the post half written. However, you’ll have to wait, because I plan to finish it on my schedule, not yours.
All the best,
w.
‘I don’t respond well to that kind of ragging.”
Or poorly for that matter.
“As it happens I’ve been looking at that very question for a couple of months now. I have the post half written. However, you’ll have to wait, because I plan to finish it on my schedule, not yours.”
It’s a fairly simple answer. It has a lot to do with f(t+tau) but nothing to do with auto correlations….. be warned.
Dinostratus August 5, 2015 at 4:06 pm
So you know the answer? That’s great! Are you going to tell us?
Or are you going to be a prince among men and dance around the subject and play games and withhold your supposed knowledge for personal aggrandizement?
If you know the answer then why on earth are you asking me? If you think the issue is important then stop playing silly buggers and give us the damn answer. If it’s all that important, I’m sure that I’m not the only one that’s interested, so how about you give us your explanation?
w.
I’m withholding my supposed knowledge for personal aggrandizement. That and chicks. Never forget the ladies.
Okay, new question….. What’s the difference between linear feedback and nonlinear feedback? The answer doesn’t take more than once sentence.
Looking qualitatively at the known current field I would say that the find pretty much supports the choice of search area. The fact that the arrival in Réunion was in 16 months or less, which is slightly less than expected might indicate that the crash site was somewhat further west, where the gyre is stronger.
His analogy may be bad and his terminology wrong. But what he’s saying about climate is that rising temperatures cause rising albedo and other heat loss events, whereas cooling temperatures cause lowered albedo and a decline in heat loss, the result of this being a temperature that is stable within some fairly narrow limits.
He may be right or wrong about the analogy to the car, and right or wrong about his use of negative and positive. The interesting and independent question is whether he’s right about the relationship between rising heat and heat loss events.
If he is wrong, its the other way around. The heat loss, whether through albedo or decline in storms or perhaps from CO2 insulation, causes a rise in temperature and that does not trigger any mechanism to reverse or limit it.
The question is whether that is how it works, or whether he is right that the underlying first cause is the rise. He does then need to explain what causes the rise or fall in the first place. I guess it might be the persistence of one of the correcting mechanisms? We have a rise in albedo, that was caused by warmup, and this rise then causes cooling, so we are always oscillating around a central value with rises or falls in the cooling mechanisms.
Many commenters here are missing the point. Willis is correct, and thinking clearly.
He has a point to make, and its a very good point, but it is not helped by simply redefining standard terms to mean their opposite.
IF he had bothered to study feedback and control systems properly, he would have both discovered that the situation he describes is known, understood, nothing special (to engineers), and had a correct and precise vocabulary with which to express it.
This isn’t skepticalscience.com. Here we try to get things right.
Tropical clouds are a fast reacting non linear localised negative feedback system that act to govern tropical maritime temperatures.
That is ALL he is saying, plus the data to support that statement.
Its a very significant and important statement to make, especially with evidence to support it, and I happen to think its true as well, but that doesn’t make it a very COMPLICATED statement to make.
He may be correct about the dynamics, but if you call a watt a farad, we’re going to have problems.