Bubba Oscillator and THD

[sameerdhiman]

Hi,

My questions are
1. How much low THD we can get with Bubba Oscillator ?
2. Methods to get lowest THD possible?

I am interested in Bubba Oscillator because it does not require complex circuit for amplitude stablization.

Regards,
Sameer Dhiman

 

[chuckey]

Try this Texas Instruments link :-http://www.calvin.edu/~pribeiro/courses/engr332/Handouts/oscillators.pdf
Frank

 

[FvM]

2. Methods to get lowest THD possible?

Adding output filters and using high performance OP.

 

[sameerdhiman]

Thanks FvM,

What about the use of LME49740 (THD=0.00003% typical) ?

Does it lower the THD if we use 6-Stages (30 deg) in place of 4-Stages (45 deg) ?

Regards,
Sameer Dhiman

 

[LvW]

Thanks FvM,
What about the use of LME49740 (THD=0.00003% typical) ?
Does it lower the THD if we use 6-Stages (30 deg) in place of 4-Stages (45 deg) ?
Regards
Sameer Dhiman

Hi sameer, you should be aware that the THD primarily is determined by the amplitude stabilization circuitry and NOT by the opamp(s). By the way: Who told you that this kind of oscillator requires the least complex circuit to stabilize the amplitude?
I see no reason why other oscillators should need more complex circuitry.
What kind of stabilization do you have i mind?
More than that, the BUBBA oscillator needs 4 opamps, and it´s advantage - if compared with some other basic types (1 or 2 opamps) - is that it has two quadrature outputs. Do you neeed a quadrature oscillator?
If not, you should consider also other topologies.

 

[FvM]

Does it lower the THD if we use 6-Stages (30 deg) in place of 4-Stages (45 deg) ?

No, the low-pass filter effect of the phase shift network will be reduced. My suggestion was to use a post filter outside the loop. It can effectively have any harmonics suppresion.

 

[sameerdhiman]

LvW : Actually i am intrested in making an audio band oscillator with lowest possible distortion. I am touching electronics field after more than 12 years of inactivity so please forgive my mistakes. I simulated Wien-Bridge but slight variation in Gain puts this oscillator into clipping and it requires PTC or complex amplitude stablization circuit. Then I simulated Bubba oscillator and I found it was not suffered from the amplitude clipping problem as well as the frequency was extremly stable. So, these features attracted me towards bubba oscillator.

FvW : I drawn a circuit around LME49740 with two BPF and the simulation shown me lowest THD around 0.010%. Can I improve this figure any more if yes please guide and show me the correct path to achieve this.

Please see the attached circuit.

Thanks & Regards,
Sameer Dhiman

 

[LvW]

LvW : Actually i am intrested in making an audio band oscillator with lowest possible distortion. I am touching electronics field after more than 12 years of inactivity so please forgive my mistakes. I simulated Wien-Bridge but slight variation in Gain puts this oscillator into clipping and it requires PTC or complex amplitude stablization circuit. Then I simulated Bubba oscillator and I found it was not suffered from the amplitude clipping problem as well as the frequency was extremly stable. So, these features attracted me towards bubba oscillator.

Hi sameer, it´s OK - if you are happy with the Bubba type, choose it - if the opamp numbers is no problem for you.
Just because of correctness: Of course, also the Bubba oscillator - that means the main op with a gain of 4 - suffers from clipping (if there is no special limiting circuitry).
However, because there are some other opamps whose output you can use, you have the advantage of harmonic filtering inside the oscillator (which is not the case for the Wien-type and all other single opamp topologies). And please note, that the amount of clipping depends on the amoint of "over gain" - that means it depends on your dimensioning, because each linear oscillator needs a loop gain slightly above unity at the target frequency.
Your approach is a good and typical example for the compromize to be found between performance and expense (costs).

---------- Post added at 19:34 ---------- Previous post was at 19:30 ----------

One additional question: How did you determine the "extreme" frequency stability of the oscillator? What do you mean with "extremly stable"? Short-term? Long-term? Phase noise? Stability against opamp gain variation?

 

[FvM]

Because the oscillator uses clipping, you would of course further reduce the THD by implementing an AGC circuit. Alternatively adjusting the gain to the minimum required amount would improve THD.

The high performance OP comes into play, if the distortions generated by the filter won't allow further THD reduction. Your result suggests, that there are still several orders of magnitude left. Assuming a 20 dB attenuation for 3rd harmonic each, your two filter stages should perform factor of ten better, if the TI numbers for the oscillator are correct. I guess, that you should rather use a low-pass than a Q=10 bandpass filter for the final filter stage, because the it has a higher feedback factor and thus less distortion.

You should also consider, that the standard SPICE parameters, particularly allowed iteration error and timestep may be inappropriate for the intended THD range, and are already generating simulation non-linearity.

 

[sameerdhiman]

LvW :

Your approach is a good and typical example for the compromise to be found between performance and expense (costs).

Yes you are right generally performance and expense are two proportional quantities. The number of OP does not matter to me because my target is lowest possible distortion. I have not built this oscillator physically but probably on coming Sunday or next I'll breadboard {if my 2.5yrs old daughter will allow me } the oscillator using TL074 (Rs40/- each, Rs50=1$US) because NE5534 or LM4562 (LME49740) they are not available in local market.

Of course, also the Bubba oscillator - that means the main op with a gain of 4 - suffers from clipping (if there is no special limiting circuitry).

Do you mean that the first OP in loop or the separate gain controller in my circuit?

One additional question: How did you determine the "extreme" frequency stability of the oscillator?

Under simulation Wien-Bridge oscillator shown frequency variation with the time on the other hand Bubba oscillator shown me constant frequency over the Long-term duration.

FvM :

if the TI numbers for the oscillator are correct

What is the meaning of TI ? I am sorry, I could not understand.

I guess, that you should rather use a low-pass than a Q=10 bandpass filter for the final filter stage

Which order will you prefer for LPF and cut-off from fundamental frequency?

You should also consider, that the standard SPICE parameters, particularly allowed iteration error and timestep may be inappropriate for the intended THD range, and are already generating simulation non-linearity.

Yes I am aware of that because when I simulated Mr. Dale Eagar's Ultrapure Oscillator (Linear Technology AN-67 page 62) the simulator shown me around 0.031% THD (original had around part per billion) due to models and simulator limitations. I also simulated Mr. Cordell's THD Analyzer's Oscillator section (original had 0.0003% 3ppm) and simulator shown around 0.018%.

I do not believe or rely on simulator numbers. I'll breadboard my circuit this coming Sunday or next and I'll test THD using soundcard with RMAA. The built-in soundcard of my PC can measure upto 0.003% (-90dB) THD.

Could you please suggest me any PCI-based soundcard which can measure -120dB THD.

Thanks & Regards,
Sameer Dhiman

 

[LvW]

Quote: Do you mean that the first OP in loop or the separate gain controller in my circuit?
What is the "first" OP in a closed loop? Sorry for this unnecessary comment.
As I`ve mentioned, it is the gain-of-four amplifier that suffers from clipping (always the amplifier with the highest gain in the loop). But - as shown in your diagram - it is the best choice to use the output of the (last) unity gain amplifier (in front of the gain setting amp) because of the filtering action of 4 low pass stages.

Quote: Under simulation Wien-Bridge oscillator shown frequency variation with the time on the other hand Bubba oscillator shown me constant frequency over the Long-term duration.

Although it is not important for your design, I am interested to learn how you have observed "frequency variation with the time" for the WIEN oscillator during simulation. Did you vary the temperature or something else? Or are you referring to the start-up process prior to the steady-state operation? Otherwise, i cannot understand why and how the frequency could change over time.
By the way, I am pretty sure that you could use a lot of other oscillator topologies with (a) the same or (b) even better or (c) perhaps worse properties (I don´t know, it´s a matter of experience) if you are able to spend 4-6 opamps in total.
In your case, you perform a good portion of filtering within the oscillator and, thus, only a relatively small amount of post-filtering is required to arrive at the required THD. For other designs (perhaps WIEN or something else) the oscillator itself comprises only one or two opamps; but a corresponding amount of post-filtering (3...4 opamps) is required for the same THD.

 

[FvM]

What is the meaning of TI

TI is Texas Instruments, I was referring to the 0.1% THD values said in the above linked literature.

I have no particular suggestion for a low-pass filter, I wanted only to point out the problem of possibly higher THD of high Q bandpasses.

I'm not aware of audio ADC with THD better than about -110 dB. So you would need additional filters to measure lower values.

 

[sameerdhiman]

Thank you LvW and FvM

I am into learning stage. So, please treat me as a student.

What is the "first" OP in a closed loop? Sorry for this unnecessary comment.

It's OK. For me the first OP-AMP is which connects to gain controller's output.

How you have observed "frequency variation with the time" for the WIEN oscillator during simulation.

Simply watching at virtual oscilloscope where amplitude got stable and looking at virtual frequency counter. Might be the frequency variation is due to simulator's internal processing because I did not make any changes to temperature or any thing else.

Could you please give me the link where I can read about topologies I would like to give them a try as well.

FvM: please tell me the ADC which can go around -110dB.

Regards,
Sameer Dhiman

---------- Post added at 15:06 ---------- Previous post was at 14:57 ----------

One more question

Why the THD figures gets low when we use half value resistor and double value capacitor in alternate sections (in 2nd and 4th section). I have also seen this configuration in Wien-Bridge where positive feedback path contains half value resistor and double value capacitor.

Regards,
Sameer Dhiman

 

[LvW]

Sameer, there is a lot of information to be found in the internet. Some links can be found here:
https://www.edaboard.com/threads/202415/

 

[FvM]

ADC which can go around -110dB

E.g. Cyrrus CS5381 (-110), recent types from Texas Instruments (-108)

 

[sameerdhiman]

Why the THD figures gets low when we use half value resistor and double value capacitor in alternate sections (in 2nd and 4th section). I have also seen this configuration in Wien-Bridge where positive feedback path contains half value resistor and double value capacitor.

Regards,
Sameer Dhiman

 

[FvM]

Why the THD figures gets low when...

You're talking about keeping the time constant of a decoupled (by OPs) RC circuit and reducing the impedance? Obviously, effects can exist either related to OP parameters or simulation behaviour, but not as a principle circuit property. So I don't see a general answer.

The question sounds mysterious related to Wien bridge oscillators, because there are no capacitors outside the positive feedback path.

 

[LvW]

Why the THD figures gets low when we use half value resistor and double value capacitor in alternate sections (in 2nd and 4th section). I have also seen this configuration in Wien-Bridge where positive feedback path contains half value resistor and double value capacitor.
Regards,
Sameer Dhiman

Hi sameer,

you are right, sometimes the positive feedback path of the WIEN oscillator consists of unequal components:
R1=a*R2 and C1=C2/a and T1=T2=T.
This has two effects:
(1) the quality factor Q of the RC bandpass is slightly larger with a small phase slope improvement (steeper) due to a small selectivity improvement and
(2) the reqired gain of the opamp is somewhat larger.

That means: a slight improvement due to (1) as well as the opposite due to (2).
Example (a=2): Q=2/5 (instead of 2/6=1/3) and opamp gain=1+Rf/Ro=5 (instead of 3).

In summary: The advantages are very small (if any) and this version (also for other values a) is not used very often.

 

[sameerdhiman]

LvW: Absolutely right. This is the answer I was looking for.

Yes, advantages are small. But I was curious to know WHY (?) and you answered very well.

I learnt from a very old man who tought me Astrology. He said "No body learns until he puts ? (WHY)"



Regards,
Sameer Dhiman

 

[LvW]

.................
But I was curious to know WHY (?) and you answered very well.
I learnt from a very old man who tought me Astrology. He said "No body learns until he puts ? (WHY)"
Regards,
Sameer Dhiman

Yes Sameer, I agree. Very often such a simple question (WHY?) is the source of new findings. Therefore, don`t stop to ask again and again "why ?".
I like to give you an example:
There are some special oscillator circuits consisting of two integrators (inverting/non-inverting) connected in a closed loop.
These circuits oscillate at a frequency very close to the cross-over frequency of the loop gain T(jw) (i. e. T=0 dB).
That`s no surprise, but - as you can prove very easily - the phase at this point is NOT zero deg (sometimes -5...-10 deg). Nevertheless, the circuits oscillate. Does this mean that the Barkhausen criterion is violated?
Up to now - there is no textbook that puts the question "why ?" or even gives an answer.
(By the way: I have asked myself "why ?" - and I have an explanation).
Regards
LvW