Okay, this is a pointless conversation debating English semantics. A resistor and a zener regulates Vin. As it regulates, is involved in the activity of regulating, it is more than fair to use the noun with the suffix "or" - a regulator. It doesn't have to be the best regulator to be a regulator. However, you are certainly free to decide that your definition of a regulator is a circuit which must contain negative feedback. Now, back to less philosophical pondering... 
Well, regulation involves taken (a sample of) the output, comparing it to a reference, and then control some element to keep the output at the set value. I don't know how to do that without some form of feedback.
Of course, in this post-fact my-opinion-is-as-good-as-anything society, you really can call anything a regulator. I wouldn't be surprised to see a resistive divider being described as a level regulator. So it goes.
Jan
Of course, in this post-fact my-opinion-is-as-good-as-anything society, you really can call anything a regulator. I wouldn't be surprised to see a resistive divider being described as a level regulator. So it goes.
Jan
Another way to skin the cat is to define a minimum set of specs that a circuit must achieve in order to be called a "regulator". Then if a board full of parts meets those specs, it IS a "regulator," no matter what kind of rude and filthy circuits it happens to use. Perhaps something like
- Line Regulation Spec: delta_Vout < 2% of (Vin_max - Vin_min)
- Load Regulation Spec: (Vout@Ioutmin - Vout@Ioutmax) < 2% of Vouttyp
The Zener and Emitter Follower that SGK mentioned does this.
Is the EF a feedback mechanism?
I consider the Zener (or any voltage reference) plus Emitter Follower as a non feedback regulator.
Pass shows a few of these in his various PSUs and Amplifiers.
They are simple and usually good enough for the load/receiver.
The performance is certainly not as tightly specified as feedback style regulators.
But the big advantage is that due to the non feedback there is no stability issue. That's what also makes them simple. One does not need to examine all the loops to find which if any are unstable.
Pass shows a few of these in his various PSUs and Amplifiers.
They are simple and usually good enough for the load/receiver.
The performance is certainly not as tightly specified as feedback style regulators.
But the big advantage is that due to the non feedback there is no stability issue. That's what also makes them simple. One does not need to examine all the loops to find which if any are unstable.
Isn't functionality always defined by specs? It either does what it's meant to do or it doesn't? It either regulates or it doesn't? To avoid the "how long is a piece of string" argument (it regulates just a little bit therefore it is a regulator), you would have to define performance criteria.
Today, however, since it is raining, I prefer to contemplate whether the sky truly is blue...
Can't say I disagree but an emitter follower can very well oscillate so stability is not automagically guaranteed.
Jan
"emitter followers can ...oscillate"
Maybe that only applies when the emitter follower is inside a feedback loop.
When incorrect impedance (negative impedance) is present on an input or output may well give sufficient phase shift that the feedback loop becomes unstable.
Can that be applied to a Zener + EF style regulator?
Can it oscillate due to excessive phase shift inside the feedback loop?
Just one example:
Bruce L. , VP Electrical Engineering at Momentum Dynamics
I believe the emitter follower tenancy to high frequency oscillation is best understood by treating the transistor Beta as complex and a function of frequency.
Consider a Hybrid pi model with only one dominate pole. The transistor Beta = B at low frequency. At the Beta corner frequency the Beta = .707B but the phase shift is 45 degrees lagging. Well above the beta corner frequency the phase angle of Beta approaches 90 degrees. Keep in mind transistors are commonly operated at frequencies above the Beta -3db point.
The key point here is well above the transistor operating frequency the magnitude of the transistor Beta is greater than one--- it still has gain--- but the Beta phase angle is approaching 90 degrees lagging.
Now consider there is a series inductance in the emitter lead. Yes I know there is no physical inductor there but there is lead inductance and the bypass capacitor you put there has considerable series inductance which becomes dominate above the capacitor self resonant frequency.
Next look at the input impedance looking into the the transistor base at high frequencies. If you inject a high frequency base current, the base voltage- with respect to ground will increase in proportion to Beta-- but because the beta is complex the rise in Base voltage with respect to ground does not happen immediately but lags the input test current by about 90 degrees assuming the emitter impedance is purely resistive.
But in the real circuit the emitter impedance at high frequencies is not a pure resistance but instead looks like an inductive reactance. This introduces another 90 phase shift in the emitter current.
Consequentially the rise in Base voltage with respect to ground at high frequencies lags the base current excitation by 180 degrees - producing a negative resistance. Now if you attach pretty much any impedance to ground having a resistive component insufficiently large, the negative resistance dominates and you have a high frequency oscillator.
A small value resistor in the base- or better yet if your circuit will allow it- in the emitter swamps out the negative resistance generated by the phase shift portion of the complex Beta and you are have a stable emitter follower.
A ferrite bead works even better as it introduces resistive loss at high frequencies while being transparent at dc and lower frequencies.
This is a bit of a hand waving explanation but some simple circuit modelling algebra shows the truth with out much grief. The trick to doing this analysis by hand is to avoid the transistor hi frequency hybrid pi model and simply treat the transistor beta as complex and a function of frequency. You can even use a representation of a complex Beta with two break point frequencies representing the second non-dominate pole in the Hybrid pi model. The second pole moves you towards 180 degrees total phase lag even faster than suggested in the hand waving example above.
Oscillation occurs when the barkhausen criteria is satisfied. ie closed loop amplitude or phase margin is violated if you want to think in terms of conventional Bode stability analysis
If the amplitude or phase margins are reduced but not violated you get enhanced transient instability - ringing at a frequency determined largely by the circuit parasitic reactances.
Of course the behavior described above is not limited to bipolar transistors but occurs in any three terminal electronic amplification device and in the emitter or source followers present in most op-amp output stages. I am sure you could come up with a fluidic equivalent of this oscillator as well.
Hope this is helpful
bruce
Bruce L. , VP Electrical Engineering at Momentum Dynamics
I believe the emitter follower tenancy to high frequency oscillation is best understood by treating the transistor Beta as complex and a function of frequency.
Consider a Hybrid pi model with only one dominate pole. The transistor Beta = B at low frequency. At the Beta corner frequency the Beta = .707B but the phase shift is 45 degrees lagging. Well above the beta corner frequency the phase angle of Beta approaches 90 degrees. Keep in mind transistors are commonly operated at frequencies above the Beta -3db point.
The key point here is well above the transistor operating frequency the magnitude of the transistor Beta is greater than one--- it still has gain--- but the Beta phase angle is approaching 90 degrees lagging.
Now consider there is a series inductance in the emitter lead. Yes I know there is no physical inductor there but there is lead inductance and the bypass capacitor you put there has considerable series inductance which becomes dominate above the capacitor self resonant frequency.
Next look at the input impedance looking into the the transistor base at high frequencies. If you inject a high frequency base current, the base voltage- with respect to ground will increase in proportion to Beta-- but because the beta is complex the rise in Base voltage with respect to ground does not happen immediately but lags the input test current by about 90 degrees assuming the emitter impedance is purely resistive.
But in the real circuit the emitter impedance at high frequencies is not a pure resistance but instead looks like an inductive reactance. This introduces another 90 phase shift in the emitter current.
Consequentially the rise in Base voltage with respect to ground at high frequencies lags the base current excitation by 180 degrees - producing a negative resistance. Now if you attach pretty much any impedance to ground having a resistive component insufficiently large, the negative resistance dominates and you have a high frequency oscillator.
A small value resistor in the base- or better yet if your circuit will allow it- in the emitter swamps out the negative resistance generated by the phase shift portion of the complex Beta and you are have a stable emitter follower.
A ferrite bead works even better as it introduces resistive loss at high frequencies while being transparent at dc and lower frequencies.
This is a bit of a hand waving explanation but some simple circuit modelling algebra shows the truth with out much grief. The trick to doing this analysis by hand is to avoid the transistor hi frequency hybrid pi model and simply treat the transistor beta as complex and a function of frequency. You can even use a representation of a complex Beta with two break point frequencies representing the second non-dominate pole in the Hybrid pi model. The second pole moves you towards 180 degrees total phase lag even faster than suggested in the hand waving example above.
Oscillation occurs when the barkhausen criteria is satisfied. ie closed loop amplitude or phase margin is violated if you want to think in terms of conventional Bode stability analysis
If the amplitude or phase margins are reduced but not violated you get enhanced transient instability - ringing at a frequency determined largely by the circuit parasitic reactances.
Of course the behavior described above is not limited to bipolar transistors but occurs in any three terminal electronic amplification device and in the emitter or source followers present in most op-amp output stages. I am sure you could come up with a fluidic equivalent of this oscillator as well.
Hope this is helpful
bruce
Jan, your remarks in post #425 appear to indicate that you object to the phrase "board full of parts." Since that's not a key piece of my gentle recommendation, I am happy to remove it:
Another way to skin the cat is to define a minimum set of performance specifications that a circuit must achieve in order to be called a "regulator". A "regulator" performs these functions and meets these specifications. If a circuit meets those specs, it IS a "regulator," no matter what kind of internal connections it happens to use. Perhaps something like
Naturally you would choose a set of functional specifications that are especially relevant to regulators. Dropout voltage might be another. Ripple rejection at 60Hz might be one more. The proposed specification strawman above merely says that a "regulator" might be defined to be: a circuit whose output voltage is at least 50 times more stable than its input voltage, and whose output voltage does not vary by more than (1/50th) across the entire permitted range of output currents. What's wrong with this definition?
Another way to skin the cat is to define a minimum set of performance specifications that a circuit must achieve in order to be called a "regulator". A "regulator" performs these functions and meets these specifications. If a circuit meets those specs, it IS a "regulator," no matter what kind of internal connections it happens to use. Perhaps something like
- Line Regulation Spec: delta_Vout < 2% of (Vin_max - Vin_min)
- Load Regulation Spec: (Vout@Ioutmin - Vout@Ioutmax) < 2% of Vouttyp
Naturally you would choose a set of functional specifications that are especially relevant to regulators. Dropout voltage might be another. Ripple rejection at 60Hz might be one more. The proposed specification strawman above merely says that a "regulator" might be defined to be: a circuit whose output voltage is at least 50 times more stable than its input voltage, and whose output voltage does not vary by more than (1/50th) across the entire permitted range of output currents. What's wrong with this definition?
Not at all - my objection was that 'a board full of parts' somehow made that board a regulator.
The regulator is not defined by how many parts it has - it is a regulator because it regulates, whether it does that with 3 parts or 300, that's irrelevant.
jan
Quick (basic) question regarding op amp supply voltage: the TL071 is rated +/- 18V; the LTC1150 +/- 16V. Am I right that if V- is grounded, V+ can range as high as 36V and 32V, respectively? I'm beginning to doubt myself with a bunch of things as I struggle with why the apparent drop out voltage is so much greater than the Vgs required to generate the load current.
Last edited:
Jan, I agree with you. I gently suggest that one useful way to decide whether or not a circuit regulates, is to observe its performance. If it performs well, at the tasks a regulator is expected to perform, then that circuit IS a regulator. I gently suggest that one way to decide whether or not a circuit performs well at regulation tasks, is to define a few such tasks and to define a minimum-acceptable-performance at those tasks. Then apply the candidate circuit to those tasks, observe its performance, and see whether or not it meets the requirements. If so, it IS a regulator. This procedure can be carried out on a "black box", a circuit whose internal details are neither known nor visible.
How about a regulator that doesn't do very well?
That's one reason why you might choose to set performance minimum requirements: they help you draw a line on the Venn Diagram which unambiguously divides "Regulators" from "NOT Regulators".
Of course you could add another zone to the Venn Diagram if you wished. You might want to have "NOT Regulators" , "Low Performance Regulators" , and "High Performance Regulators" (to name one possible choice).
For each performance measurement you could establish a minimum requirement and an excellence threshold. Then
Let's leave it.
Okay
That's one reason why you might choose to set performance minimum requirements: they help you draw a line on the Venn Diagram which unambiguously divides "Regulators" from "NOT Regulators".
Of course you could add another zone to the Venn Diagram if you wished. You might want to have "NOT Regulators" , "Low Performance Regulators" , and "High Performance Regulators" (to name one possible choice).
For each performance measurement you could establish a minimum requirement and an excellence threshold. Then
- It's a "NOT Regulator" if performance is less than the minimum requirement
- It's a "Low Performance Regulator" if it meets the min requirement but doesn't meet the excellence threshold
- It's a "High Performance Regulator" if it meets both the min requirement and the excellence threshold
Let's leave it.
Okay