Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.
Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.
If you turn the Wein network the way it is usually used, here is the gain/phase plot. Note that I removed the extra cap and resistor, and tweaked R2 to get a loop gain slightly higher than 1. Note the well-behaved Gain hump, and only one place where phase is zero.
Closing the loop turns it into a decent oscillator, except that the harmonic distortion is terrible because the opamp runs into the rails...
Winterstone, if you had just posted your circuit with your very first posting to this thread, you could have saved all of us a whole lot of trouble...
I think you can use either one. But with the one more often shown on the web, the resistors R1 and R2 would have to be exchanged in order to get rid of the exponential part (and thus remove the damping required to make an oscillator).
When all the values are the perfect required values the theoretical oscillation is perfect and requires no non linear network. Simulation would show this as a very very long chain of sine waves before they either start to get larger or smaller.
Note that with R4=R3 and C4=C3 the analysis gets much more simple too, which is nice.
I think you can use either one. But with the one more often shown on the web, the resistors R1 and R2 would have to be exchanged in order to get rid of the exponential part (and thus remove the damping required to make an oscillator).
Are you saying that Winterstone's inverted Wien network, the one I posted first, can be made into a useful oscillator? If so, post a working circuit. The only way I could see that it could work is if you added secondary low-pass filtering to the amplifier such that only the first zero crossing would have a loop gain >1 and the others would be attenuated below 1.
Question is: why bother, when the normal circuit works so well?
Oh you are awake too now? I dont know what time it is there but here it is quite early yet
Yes i think it would work. I could double check if you still have doubts, but in the analysis of the ideal circuit gives me:
Vout=-3*sin(wt)
for a step change input which is clearly a sine wave.
The other circuit (more common on the web) gives:
Vout=6*sin(wt)+3 which has a DC offset and is 180 degrees out of phase with the first, but again a sine wave
Note this is pure theory again, but what this means is that in real life this should be a workable design. Are there flaws in doing it this way? I dont know all the ins and outs yet.
You'll have to ask Winterstone why he chose this version.
Almost forgot the circuit:
Make R3=R4=10k, make C3=C4=1.59155e-8 Farads. R1=5k, R2=10k. Remove R33 and C44.
Remove the output voltage source. Add a battery 1v in series with C3 so instead of C3 going to ground the battery negative terminal does.
at first, thank you for all your replies and the "trouble" (Mike`s word) you had. Did you really have trouble? Or is it not good from time to time to make up his mind and to recall basic things?
At second, you all are right. I have made some errors - however, fortunately no principal errors that could put the whole problem into question.
Let me explain:
1.) MrAl - you are right, the time constants are not equal. OK - but this does not touch the main question. I have played a bit with the parts values - and finally, there was a small confusion because of one of the resistors (25 kohms).
2.) This damned resistor is also the reason the loop gain does not reach the 0 dB line. Please, change the value from 25k to 20k and the loop gain will be app. +0.1 dB at the phase crossing frequency.
3.) The ac source - of course - is for determination of the loop gain. I have investigated several methods: With the negative feedback loop closed or open. And in the "open case" it was necessary to connect the ac source with one of its nodes to the opamp output (Middlebrook`s method to keep the bias point). I know that this method causes a small error for very high frequencies (due to the finite opamp output resistance).
However, you can verify that the results are the same for the method as visualized by Mike in post#60.
4.) Quote Mike: "The circuit you posted has the Wein network inverted, which is a horrible mess as far as making a reliable oscillator."
Ha Ha - yes Mike, I know. I never would use such a circuit - but: Exactly THIS is my question. From where do you know? Of course, from your experience only, I suppose.
But remember my question: What are all the conditions that must be fulfilled for a circuit to oscillate as expected? (Experience is good and important, but can it replace a criterion?).
There are 4 other variations of the basic Wien bridge which are able to work as a stable oscillator. The circuit shown is one of the other 4 derivates, which do not work, although they fulfill
Barkhausen`s rule!.
5.) With reference to post#62 and post#63: Yes, it is no surprise that the classical WIEN oscillator works. As I have mentioned in a foregoing post, I am involved in such oscillator circuits since a long time.
And exactly this is the background of some system-oriented problems I have identified during this work.
(By the way: There are some other myth`s connected with harmonic oscillators. If you like I could describe you a circuit with a loop phase of 5...10 deg at the 0dB point - and it oscillates!)
6.) Here are some more information on my circuit: The circuit as described by me has a complex pole pair in the RHP (close to the imag. axis with a pole frequency of app. 2*Pi*130 Hz) and a positive real pole.
Thus, it is obvious that the circuit cannot oscillate as expected (or hoped). But how can we know without doing a pole analysis?
7.) Summary: After a detailed analysis of the circuit (pole location) it is no surprise that the circuit will NOT oscillate at 130 Hz - but my question still is (I repeat from my first post):
Is there any additional requirement which can complete the Barkhausen criterion with the aim to have a more rigorous oscillation criterion?
With other words: Since the Barkhausen rule is based on the loop gain, how can we decide if a circuit will oscillate or not by observing the loop gain only?
(Remember: My circuit - after modifying the 25k resistor - meets this criterion, but does not oscillate as desired).
I hope, you understand my question: I know and I am not surprised about the non-oscillatory behavior of the circuit. However, I like to explain this based on the classical method of loop gain response.
__________________
Thank you for all the effort (and sorry for the trouble)
W.
at first, thank you for all your replies and the "trouble" (Mike`s word) you had. Did you really have trouble? Or is it not good from time to time to make up his mind and to recall basic things?
At second, you all are right. I have made some errors - however, fortunately no principal errors that could put the whole problem into question.
Let me explain:
1.) MrAl - you are right, the time constants are not equal. OK - but this does not touch the main question. I have played a bit with the parts values - and finally, there was a small confusion because of one of the resistors (25 kohms).
2.) This damned resistor is also the reason the loop gain does not reach the 0 dB line. Please, change the value from 25k to 20k and the loop gain will be app. +0.1 dB at the phase crossing frequency.
3.) The ac source - of course - is for determination of the loop gain. I have investigated several methods: With the negative feedback loop closed or open. And in the "open case" it was necessary to connect the ac source with one of its nodes to the opamp output (Middlebrook`s method to keep the bias point). I know that this method causes a small error for very high frequencies (due to the finite opamp output resistance).
However, you can verify that the results are the same for the method as visualized by Mike in post#60.
4.) Quote Mike: "The circuit you posted has the Wein network inverted, which is a horrible mess as far as making a reliable oscillator."
Ha Ha - yes Mike, I know. I never would use such a circuit - but: Exactly THIS is my question. From where do you know? Of course, from your experience only, I suppose.
But remember my question: What are all the conditions that must be fulfilled for a circuit to oscillate as expected? (Experience is good and important, but can it replace a criterion?).
There are 4 other variations of the basic Wien bridge which are able to work as a stable oscillator. The circuit shown is one of the other 4 derivates, which do not work, although they fulfill
Barkhausen`s rule!.
5.) With reference to post#62 and post#63: Yes, it is no surprise that the classical WIEN oscillator works. As I have mentioned in a foregoing post, I am involved in such oscillator circuits since a long time.
And exactly this is the background of some system-oriented problems I have identified during this work.
(By the way: There are some other myth`s connected with harmonic oscillators. If you like I could describe you a circuit with a loop phase of 5...10 deg at the 0dB point - and it oscillates!)
6.) Here are some more information on my circuit: The circuit as described by me has a complex pole pair in the RHP (close to the imag. axis with a pole frequency of app. 2*Pi*130 Hz) and a positive real pole.
Thus, it is obvious that the circuit cannot oscillate as expected (or hoped). But how can we know without doing a pole analysis?
7.) Summary: After a detailed analysis of the circuit (pole location) it is no surprise that the circuit will NOT oscillate at 130 Hz - but my question still is (I repeat from my first post):
Is there any additional requirement which can complete the Barkhausen criterion with the aim to have a more rigorous oscillation criterion?
With other words: Since the Barkhausen rule is based on the loop gain, how can we decide if a circuit will oscillate or not by observing the loop gain only?
(Remember: My circuit - after modifying the 25k resistor - meets this criterion, but does not oscillate as desired).
I hope, you understand my question: I know and I am not surprised about the non-oscillatory behavior of the circuit. However, I like to explain this based on the classical method of loop gain response.
8.) Quote MrAl: You'll have to ask Winterstone why he chose this version.
I hope I could make it clear now: Just because it fulfills Barkhausen - without being able to oscillate!!
__________________
Thank you for all the effort (and sorry for the trouble)
W.
Sorry, instead of editing I have added my own reply as a comment.
Well would you mind if we did the circuit but with two 10k resistors and two 1.59e-8 capacitors instead? Or another combination where R3=R4 and C3=C4 because that makes the analysis much simpler. Is that circuit still a good circuit for what you want to show then?
In general if a circuit wont oscillate and everything else is satisfied then it must be the damping that causes the amplitude to die out. Thus we could state that the exponential damping coefficient be zero or slightly negative. For example in the original circuit you posted we would have something like:
z=C4*R2*R4-C3*R1*R4+C3*R2*R3
and if z=0 then the pole pair is on the jw axis, but if negative then into the RHP, and if positive (not allowed) then LHP which means too much damping so no oscillations. We can look at this in more detail later but for equal caps and resistors it reduces to:
z=2*R2-R1
with the same conditions for z. This kind of simplification is why i suggested equal values for the caps and resistors.
Hello Winterstone,
Well would you mind if we did the circuit but with two 10k resistors and two 1.59e-8 capacitors instead? Or another combination where R3=R4 and C3=C4 because that makes the analysis much simpler. Is that circuit still a good circuit for what you want to show then?
No, of course, I wouldn`t mind. But why ? It was not my aim to build an oscillator. I did some experiments with many different values and - if I remember well - this equal component design was not suitable to demonstrate what I wanted to show. That`s all.
I am not sure if I was able to describe my problem resp. question clear enough (because you have shown that and how the classical Wien oscilator works)
Therefore, let me try it in another way:
If somebody has the following request: Please explain (as simple as possible) under which conditions a circuit with feedback is able to work as an oscillator.
Thanks. That is what we are trying to work out here.
So you dont want to use equal components? So components are right now:
R1=10k
R2=5k
R3=10k
R4=20k
C3=15.9e-9
C4=3.18e-9
C44=100e-9
R33=50k
But i assume you would prefer to see the analysis without the extra R33 and C44 right? As they are only required for the simulation right?
According to my theory about the damping:
z=C4*R2*R4-C3*R1*R4+C3*R2*R3
for your choice of values, z is negative and significant, therefore the pole pair is deep into the RHP so the output might shoot up or down quickly and saturate without some sort of non linear limiting. However, solving that equation for R2 gives us something to work with:
R2=(C3*R1*R4+z)/(C4*R4+C3*R3)
and setting z=0 we get:
R2=14.285k
which would provide a better starting point.
But the short answer is that the circuit will not oscillate because the pole pair is too deep into the RHP and that causes the output to saturate quickly. Limiting might solve the problem though.
I have included C44 to stop positive dc feedback (otherwise problems with a fixed bias point - in simulation and, of course also in reality).
This makes R33 necessary to allow the dc bias current for the p-input of a real opamp model (simulation and reality).
Ok well the short answer is if the gain puts the pole pair too far into the RHP then the output shoots up too quickly and saturates. So with this circuit it appears that it is not a matter of seeing an output that is zero, it looks more like it just saturates quickly and that means it wont oscillate. So we see two kinds of problems, that where the output goes to zero and that where the output goes into saturation.
Howevcer, adjusting R2 would mean the gain puts the pole pair at a better location.
But are you saying that before we adjust R2 that it satisfies Barkhausen?
But the short answer is that the circuit will not oscillate because the pole pair is too deep into the RHP and that causes the output to saturate quickly. Limiting might solve the problem though.
Something in your calculation must be wrong - but I don`t know what (I didn`t go deeply into your formulas).
I have checked the pole location for "my circuit" and - as mentioned before - a pole pair very close to the imag. axis (RHP) and a positive real pole (deep in the RHP, causing the multivibrator behavior)
Here are the poles (for R4=20kohms):
Sp1,2 = 55.9 +- j*975 [rad/sec]
Sp3= 41500.. or larger (dependent on the opamp model used)
Something in your calculation must be wrong - but I don`t know what (I didn`t go deeply into your formulas).
I have checked the pole location for "my circuit" and - as mentioned before - a pole pair very close to the imag. axis (RHP) and a positive real pole (deep in the RHP, causing the multivibrator behavior)
In case you are still interested and motivated to answer my question(s), here is the basic circuit, which the one as posted earlier in my attachment was derived from (see pdf file).
I call it "Wien_invers" because it was derived from the "normal" Wien bridge oscillator (tuned conditions) by exchanging the corresponding parts of the tuned bridge.
Please note, that I have used an idealized opamp model (finite gain 1E5, finite supply rails inherent to the model).
(MikeMi: By the way, I don`t understand the meaning of the given link **broken link removed** can you help me?)
What do you think about the circuit?
Additional information: There is only one single complex pole pair slightly in the RHP . No real pole (strictly second order).
I have a more basic question. Based on the analysis or response plots for the loop gain, how can we really say that we've met the Barkhausen criteria? The linear circuit does not meet it because the gain is about 0.1 dB when the phase is zero. The condition is required to be exact. Don't we need to do a more sophisticated analysis to show that the condition is met at some point by the nonlinear limiting function? In other words, as the limiting occurs, we need to prove that there is a point were we have 0 db and 0 phase.
Once we prove that, then we can look at other issues. Now, I'm not saying that an analysis won't reveal this to be true, because I haven't done it. But, isn't it important to prove this in some way before you can ask your question?
Winterstone,
I have a more basic question. Based on the analysis or response plots for the loop gain, how can we really say that we've met the Barkhausen criteria? The linear circuit does not meet it because the gain is about 0.1 dB when the phase is zero. The condition is required to be exact. Don't we need to do a more sophisticated analysis to show that the condition is met at some point by the nonlinear limited function? In other words, as the limiting occurs, we need to prove that there is a point were we have 0 db and 0 phase.
Once we prove that, then we can look at other issues. Now, I'm not saying that an analysis won't reveal this to be true, because I haven't done it. But, isn't it important to prove this in some way before you can ask your question?
Hi Steve, yes I agree, of course. That is the classical situation.
I will say it with my words:
Everybody knows that loop gain LG=1 is not possible in reality due to known reasons. Even Barkhausen did know this - and he has mentioned explicitly that LG>1 is necessary to start oscillation with rising amplitudes.
More than that, everybody who has designed an oscillator (or only simulated) knows that LG>1 is necessary. That means: If we know that a particular circuit with a suitable amplitude control mechanism theoretically is able to meet the condition LG=1 we can design it for small amplitudes and LG>1.
That is the normal and proofed method to design such a circuit, is it not?
For example, in "my circuit" a pair of diodes across R3 will decrease the positive feedback for larger amplitudes - thereby reducing the loop gain until it is 0 dB.
(Remember the two cases with R4=25 resp. 20 kohms and a loop gain slightly below resp. above 0 dB, LG=1 is in between)
However, the situation changes for circuits with non-linearities that cause rising gain for larger amplitudes. It is not a problem to find such circuits. In this case, you are right again - we never can meet Barkhausen`s rule and the circuit will not oscillate but will stuck at the supply rail.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
