Wein Bridge Osc no go

[Orson Cart]

Taking R3 to 10k still gives you a DC gain of 2 which is more than enough for an oscillator, also it lowers the gain effect of the fet a little - worth a try - Orson Cart.

 

[davenn]

Thanks guys
am on overseas holiday for a week, will try the suggested variations when I return and post the results

All in the fun of experimenting and learning

Dave

 

[Einar M]

I was hoping Dave would post some more test results. I looked in my app handbook for ideas and found two circuits that dont involve FETs. I updated them and made some minor changes and tried them out. I think they are a better choice, particularly for low Fo. One is even tunable witrh a single pot. The link is here:https://i1214.photobucket.com/albums/cc485/Einarscircuits/Oscillators/4e0e755f.jpg

 

[FvM]

One is even tunable witrh a single pot.

Not actually tunable in my opinion. Im contrast to a Wien bridge, you can adjust the frequency by a small amount without affecting the performance too much. Tunable would imply a wide frequency range of 1:10 or more. In addition, it should be mentioned, that amplitude stabilization by zener diodes or similar voltage limiters involves more distortion than a variable gain element.

 

[Einar M]

I guess we disagree on the meaning of tunable; I felt it implied limited ( 3 to 1 ) range. Amplitude does vary with Fo and you have to tweak it for min distortion. The claim that clipping involves more distortion is not true. Both circuits can be adjusted for a very small amount of clipping which is then further reduced by the low pass nature of the phase shift networks. Also, you are neglecting the large amount of distortion that is introduced by the nonlinear resistance of the FET.

 

[LvW]

The second of the three presented circuits is based on a multi-feedback bandpass circuit.
If R3=2(R1+Ro) the midband gain is -1 (independent on the frequency). The midband frequency can be varied by R2.
(I have designated the 1k resistor in series with R1 as Ro; the reproduction of the drawing is not easy to read).

Thus, I think the circuit can be regarded as tunable via R2 - of course within a limited frequency band only.
However, the given formula for the oscillation frequency is an approximation: R1 should be replaced by (R1+Ro).

LvW

---------- Post added at 10:40 ---------- Previous post was at 10:36 ----------

Both circuits can be adjusted for a very small amount of clipping which is then further reduced by the low pass nature of the phase shift networks.

Einar, in which circuit do you see a "phase shift network" ?

 

[Einar M]

The varible res R2 is in parallel with R1 , and R1 is 500+ times R2 so is effectively out of the equation. Fo varies as the inverse of the square root of R2. The phase shift commet aplies more to the upper circuit.

 

[LvW]

The varible res R2 is in parallel with R1 , and R1 is 500+ times R2 so is effectively out of the equation.

Yes, and exactly that was the background of my comment: Approximation.

Fo varies as the inverse of the square root of R2.

Not exactly, but approximately.

The phase shift commet aplies more to the upper circuit.

I cannot agree to that. The upper circuit is based on two integrators - and, therefore, exhibits a (nearly) constant phase shift around the oscillation frequency. How can this reduce distortion?
For my opinion, the best circuit - as far as distortion is concerned - is the 3rd circuit if the output of the bandpass is used because it provides good filtering. Of course, this circuit can be equipped with FET stabilization also, which - in any case - is more effective than diode stabilization with clipping effects.
However, if tuning is involved, diode stabilization may be better because the time constant of the FET stabilization must be tuned accordingly.

 

[Einar M]

The upper circuit is a 2nd order low pass filter followed by an integrator for a 3rd order output. only the feedback signal is limited, at the output any harmonics created have been reduced by at least 30 dB. The bandpass circuit reduces harmonics by a factor of Q , 20 to 50 here so again any harmonics are low. The complaint of approximation is unjustified for a variable osc; that is one of the points of tuning it. You could use a FET in either circuit but would see higher distortion unless you kept the voltage across it very low , less than 50mV. thankyou for looking at my circuits, but i think your analysis has been too hasty. that's all I have time for tonight.

 

[FvM]

The claim that clipping involves more distortion is not true. Both circuits can be adjusted for a very small amount of clipping which is then further reduced by the low pass nature of the phase shift networks.

If you adjust the gain very exactly, distortions can be low. If you do the same with a FET regulated oscillator, distortions will be even lower, presuming a suitable circuit design.

Unfortunately, the single resistor "tuned" oscillator needs the gain to be tuned together with frequency variation. So both advantages claimed for the circuit, single resistor tuning and low distortions are mutual exclusive.

It's reasonable indeed, to utilize the double integrator for further reducing distortions, in so far it's clearly superior to a Wien bridge.

Also, you are neglecting the large amount of distortion that is introduced by the nonlinear resistance of the FET.

I'm not neglecting FET distortions, just referring to the fact, that a FET generates less distortions than a zener clipper or similar circuit. I prefer however a linearized FET variable resistor. The FET circuit in your circuit 3 doesn't use this option. A linearized FET can handle higher AC voltages than only 50 mV. Some results have been in reported in a previous thread https://www.edaboard.com/threads/129234/

My preferred solution for a wide range tunable oscillator is the double integrator circuit with a FET amplitude regulator, using a dual variable resistor or other variable gain element for tuning. I once implemented a 1:1000 range with LDRs as variable resistors. Or I should supplement, it's my preferred analog solution.

 

[LvW]

..................
Unfortunately, the single resistor "tuned" oscillator needs the gain to be tuned together with frequency variation. So both advantages claimed for the circuit, single resistor tuning and low distortions are mutual exclusive.

As I have mentioned before, varying R2 does change the pole frequency without touching the gain of the bandpass.
Therefore I still believe that the circuit can be considered as a single-resistor-tuned oscillator (within the limits that apply to each "tuned oscillator").

---------- Post added at 14:10 ---------- Previous post was at 13:52 ----------

Hi Einar,
here are some remarks to your last posting.

The upper circuit is a 2nd order low pass filter followed by an integrator for a 3rd order output.

I don`t think, that this is the correct description/explanation. The stage with the two capacitors is a 1st order lowpass followed by a proportional-integral (PI) stage. For oscillator applications it is tuned by correct dimensioning such that a pole-zero compensation takes place. Then, it acts as a noninverting integrator stage. This forms - together with the other inverting integrator - one of the classical quadrature oscillators.

Quote: only the feedback signal is limited, at the output any harmonics created have been reduced by at least 30 dB.

No, it is not "limited" - however, I will explain it at the end of this posting

Quote: The complaint of approximation is unjustified for a variable osc; that is one of the points of tuning it.

You are right - but only from the practical point of view. If one of two parallel resistors can be neglected it is absolutely necessary to know about this parallel connection - in order to (a) understand/evaluate the correctness of the corresponding formula and (b) to establish the appropriate dimensioning that allows this simplification. It is a good engineering practice to know about all possible influences/parasitics in order to decide if and in which cases simplifications are allowed.

Quote: You could use a FET in either circuit but would see higher distortion unless you kept the voltage across it very low , less than 50mV.

I think, this point has been answered resp. clarified very well by FvM.

Quote: thankyou for looking at my circuits, but i think your analysis has been too hasty.

With regard to my and FvM's comments - do you really still believe that our analyses are to "hasty"?
____________________________________________________________________________________
I have some additional comments to the first circuit (quadrature oscillator). Perhaps you are interested.
There are many different oscillator circuits, which can be amplitude stabilizedwith the help of diodes. That's not new.
With one single exception this method results in amplitude clipping - dependent on the amount of excess loop gain.
The mentioned exception is the double-integrator configuration!
Why this exceptional behaviour? This oscillator type is the only one that exhibits a (nearly) constant loop gain phase around the oscillating frequency (only the loop gain is altered).
Now, what is the task of the diodes? By opening the diodes the integrator reduces to a first order lowpass thereby changing the phase conditions; and there will be only a minor influence (if any) to the loop gain amplitude.
With other words: Only the phase related part of the Barkhausen condition is touched. As a consequence, when the diodes start to conduct the phase is mis-tuned and the oscillation dies out. That means, there will be no clipping of amplitudes because the loop gain magnitude remains practically constant.
According to my knowledge, this behaviour is not yet described in textbooks.

Thank you (sorry for the long answer)
LvW

 

[FvM]

Right. I see my misunderstanding. I was talking about the tuning capabilities of the first (double integrator) circuit.

P.S.: According to a brief analysis, I agree that the second circuit has acceptable tuning behaviour within the intended frequency range. The required excess gain effectively restricts it's usability to medium audio frequencies. But anything is better than nothing.

P.P.S.: Circuit #1 also isn't a double integrator oscillator. Apparently, I tend to see circuits that I would expect in this place rather than those, that are actually there.

 

[LvW]

I think I should slightly modify one of the comments of my contribution above:

......the integrator reduces to a first order lowpass

This applies to the diode stabilization network that normally is used for these kind of oscillators (damping of the integrating capacitor).
The diode network as scetched by Einar is somewhat different and - for my opinion - rather uncommon (and with 5 non-linear elements more complicated than necessary).
Nevertheless, of course it works, but one cannot say that the following integrator stage is "reduced to a first order lowpass".
But, on the other hand, it is true that this non-linear network primarily influences the loop gain phase rather than the loop gain magnitude. This can be easily seen via loop gain simulation. Therefore, my comment is still valid in general.

 

[LvW]

P.P.S.: Circuit #1 also isn't a double integrator oscillator. Apparently, I tend to see circuits that I would expect in this place rather than those, that are actually there.

Hi FvM,

I am afraid, that is a severe indication of my age. You are completely right. The stage under discussion is not a BTC integrator (as I have thought) but a simple S+K lowpass stage with unity gain. That means: I have "seen" something I have expected
(I see S. Freud smiling). However, it looks similar, does it not?
Thank you for the correction.
In this case, everything I have written to Einar still applies to "real" double-integrator circuits, but not to his circuit #1.

LvW

 

[Einar M]

I wish you had studied my schem more closely so more of your comments actually applied to it. I will forgive you as I usually get things wrong on the first read thru myself. I put the zener in the diode bridge as it seemed easier than matching two to get symetrical clipping. I didn't like the extra nonlinearities, but wasn't sure how distortion would be affected.
The data on FETs I have shows that a generic, grab bag part can't easily achieve 1% lin even with the resistotr network in the gate. ( Which is shown on my diag. ) An ideal part can do much better but people just want to use what they have on hand. Also, it is my understanding that FETs create a lot of 2nd harmonic that is much harder to remove than the 3rd of clipping.
The circuit using the BP filter isn't my favorite but I thought others might like it.

 

[FvM]

I wish you had studied my schem more closely so more of your comments actually applied to it.

Yes, this has been granted.

I checked circuit #1 (lowpass + integrator) in a simulation and can confirm, that it's apparently performing quite well. Because the oscillation frequency is set by the bandpass cut-off frequency (-90 degree phase point) it's strictly fixed frequency.

The standard FET linearization circuit, as discussed more detailed in an above linked thread, is effectively cancelling the quadratic term in the FET characteristic. Circuit complexity increases however when implementing an amplitude control loop instead of a simple clipping circuit. And as revealed in recent edaboard discussions, achieving stability for the control loop can become a serious problem for a beginner.

Furthermore, no quality requirements in terms of THD have been given in this thread. Applications and ideas about it can be of course quite different.

 

[LvW]

..........I wish you had studied my schem more closely so more of your comments actually applied to it. I will forgive you .........
.

Thank you. On the other hand, I didn't think it was really necessary to "study your schemes" more closely as all three circuits presented by you belong to the family of classical four-pole oscillators. It was simply a mistake from my side - I overlooked the unity gain feedback in circuit#1. You can be sure if somebody announces a rather new circuit I will be motivated to
study it more closely.

Regards
LvW

 

[davenn]

Hi gang,

WOW, lots of discussions happening in my absence. Had a good time in New Zealand visiting family and even felt a couple of earthquakes whilst in Christchurch ( some of you may be aware of the big events there over the last 10 months) a M5.3 was the biggest one I felt, waking us all up in the early hours of the morning.

OK, I finally gave up on that dual op-amp and FET design as discussed in messages ~ 17 -19 it just wouldnt settle down.
The earlier one that I got working I am happy with, got it down to 10Hz and still stable.
I originally wanted this for testing the preamplifiers of my seismograph system on the "short period" channels 1 - 15Hz area.
It does the job nicely.
Current work is in building a couple of long period seismometers they will be ~ 10sec period (0.1Hz) and will record the big events anywhere in the world. see **broken link removed**
for what I am making copies of

Thanks everyone for your input and suggestion, its been very informative

cheers
Dave

 

[LvW]

Current work is in building a couple of long period seismometers they will be ~ 10sec period (0.1Hz) and will record the big events anywhere in the world.[/I]

Hi Dave,
welcome back to the world of electronics.
If I understood you well, you are going to design a 0.1 Hz oscillator in the near future, right?
As I expect some problems due to the low frequency (rather large resistor and capacitor values) I would be happy if you could report on your first results. Thank you.
Regards
LvW

 

[davenn]

hi LvW,

thanks

naa wont be trying to build an Osc that low in freq. Generally for testing those seismometers we physically let the pendulum oscillate by giving it a gentle push and just time the period. At the same time we count the number of cycles to determine if its over, under or critically dmapened.
This is to ensure that we are recording the motion of the ground (earth) rather than just the free oscillations of the pendulum once the quake has set it into motion.

cheers
Dave