Download as pdf or txt
Download as pdf or txt
You are on page 1of 39

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281830241

WIEN BRIDGE OSCILLATOR WITH REAL AMPLIFIERS

Research · September 2015


DOI: 10.13140/RG.2.1.4906.2880

CITATIONS READS

0 7,234

2 authors:

Josef Punčochář Jitka Mohylová


VŠB-Technical University of Ostrava VŠB-Technical University of Ostrava
109 PUBLICATIONS   139 CITATIONS    44 PUBLICATIONS   155 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

"Passing" of courses View project

All content following this page was uploaded by Josef Punčochář on 17 September 2015.

The user has requested enhancement of the downloaded file.


WIEN BRIDGE OSCILLATOR WITH REAL AMPLIFIERS

Josef Punčochář and Jitka Mohylová

VŠB-Technical University of Ostrava


Faculty of Electrical Engineering and Computer Science
Department of Electrical Engineering (420)
17. listopadu 15/2172,
708 33 Ostrava - Poruba, Czech Republic

1. Abstract

The Wien – bridge oscillator is probably the most widely used type of RC

oscillator. In a standard analysis it is assumed that the used amplifier is ideal. In this

paper is presented the Wien – bridge oscillator with different real amplifiers. It would be

better to talk about “half – bridge”. The half – bridge (R1C1 + R2C2 “Wien – divider”) in

the positive feedback loop produces a maximum response at its characteristic frequency

ω0 and gives us zero phase shift. The remaining two arms are used in a negative

feedback loop and provide necessary gain of the amplifier. Thus Wien oscillator is

formed from the “half – bridge” and amplifier – with vacuum tubes, transistors,

operational amplifiers and other modern amplifier structures.

2. Keywords

Wien oscillator, Real Amplifier, VFA, CFA, CCII+

1
3. Introduction

Basic rules of the generalized nodal voltage analysis have been presented in the

literature [1], [3], [4], [5]. This method is very useful in investigation of circuits with

real amplifying structures, thus for investigation of the Wien oscillator we have different

amplifiers. The literature on the Wien oscillator with classic voltage feedback amplifier

(VFA) [6], [9], [11], [12], [15], [26], [27] provides a number of interesting results in this

area, as well as literature describing oscillators with current feedback amplifier (CFA)

[17], [28] and with current conveyors (CCII+) [18], [20], [23], [24], [25]. But problems

are often solved somewhat academically regardless of the needs of practice.

The aim of this paper is to derive the basic relations applicable to estimate the

behaviour of the Wien oscillator with different real amplifiers – realized by means of

VFA, CFA and CCII+. Some relationships are verified experimentally, some using data

from the literature.

A free running harmonic oscillator can be considered as a linear system as long

as its amplitude is not limited by the power supply. The oscillator properties can

therefore be studied by means of a linear circuit theory. The node equations give us a

systematic and rigorous way to reach the right mathematical model of the oscillator –

they give us the admittance description of the linear (linearized) electronic circuit.

Such an admittance model (matrix) does not need a feedback theory

construction. But the right determined model always includes “feedback properties” of

the investigated circuit – automatically [1]. We can solve the same problem by means of

the feedback theory, of course, and we must get the same solution.

The Barkhausen criterion [to provide oscillations at frequency ω0, A(jω0) – an amplifier

gain, β(jω0) – a transfer of selective feedback network]

2
A( jω0 ) ⋅ β ( jω0 ) = 1 (1)

moves in the equation

D( jω0 ) = 0 (2)

where D is the determinant of the admittance matrix of the investigated structure.

In fact we examine the circuit poles, which are the roots of eq. (2) –

characteristic equation [2].

4. Matrix description

The classic Wien configuration of the oscillator is in Fig. 1. Basic rules of the

generalized nodal voltage analysis are for example in [1], [3], [4], [5]. An excitation

current I1 is connected only formally. Oscillators operate without externally–applied

signals.

Figure. 1: The classic Wien oscillator

After numbering of nodes (1, 2) we can determine an admittance model of the

circuit (p = jω for steady state solution), we assume the amplifier K input resistance to

3
be infinite [1], [4]:

(1) a (2) b
(1) a  Y1 + Y2 − Y2  U 1   I 1  (3)
 ⋅ =
(2) b − Y2 − KG0 Y2 + G0  U 2   0 

where Y1 = G1 + pC1, Y2 = pG2C2/( G2 + pC2)

K is the voltage gain of the amplifier K

G0 = 1/R0 is the output conductance of the amplifier K.

Solving this matrix set of equations give us nodal voltages

U i = Di D (4)

where Di is the determinant according to the Cramer’s rule and D is the matrix

(circuit) determinant. If

D→0

then

Ui → ∞

and we get oscillations – see eq. (2). We may interpret this condition as the

condition for the onset of oscillation – start up condition.

Then, due to the noise signals (in general; for example transients generated when

the power supply is turned on) a growing sinusoidal voltage appears [6]. Now it is clear

that the current I1 is not needed; the circuit behaves as a generator.

Form eq. (3) and eq. (2) we easily determine

D = G0 [(Y1 + Y2 )(1 + Y2 G0 ) − Y2 (K + Y2 G0 ) ] = 0 .

We suppose that R0 → 0, thus G0 → ∞, so we need to solve only characteristic

equation

Y1 + Y2 (1 − K ) = 0 (5)

4
4.1. Notes to AGC

If D exceeds zero, oscillation will grow in amplitude. It is well known, however,

that the parameters of any physical system [2] cannot be constant (in time, temperature

…) – we need a mechanism for forcing D to remain equal to zero (Aβ to remain equal to

unity) at the desired value of output amplitude. This is done by providing a nonlinear –

AGC – for example see [1], [2], [6], [7], [8].

The AGC mechanism ensures that oscillation will reliably start, because when

you turn on the oscillator, then there must be issue

β A 〉 1.

In this case the poles of D are in the right half of the p – plane. When the

amplitude reaches the desired level, AGC reduces loop gain (βA) to unity, thus D → 0.

The poles (of D) will be “pulled back” to the “jω axis.” If the AGC network is omitted,

the amplifier K will saturate .

This matter is not addressed in this paper.

5. Ideal amplifier

Let us consider the classical case

R1 = R2 = R

C1 = C2 = C

K = K 0 is frequency independent

We substitute

Y1 = G + pC

Y2 = pGC (G + pC )

5
and eq. (5) takes the form

p 2 C 2 + pGC(3 − K 0 ) + G 2 = 0 (6)

Now it is easy to modify eq. (6) into

2
 1   1 
p 
2
 + p   ⋅ (3 − K 0 ) + 1 = 0
 RC   RC 

thus

s 2 + s(3 − K 0 ) + 1 = 0 (7)

where

s = p ω0 (8)

is the normalized frequency (relative ω0) and

ω 0 =1 RC (9)

is the characteristic frequency of the Wien divider.

We need just zero real part of eq. (7) root (Laplace domain, Routh-Hurwitz

criterion), thus circuit must ensure

3 − K0 = 0

thus

K0 = 3 .

The circuit would then oscillate at the frequency ω0 because

ω1,2
s2 +1 = 0 ⇒ s1,2 = ± j = (10)
ω0

The poles are placed on the imaginary axis.

Let us investigate our problem in the frequency domain (harmonic steady state).

6
Then p = jω and

s = j (ω ω0 ) = jΩ (11)

We get from eq. (7)

− Ω 2 + jΩ(3 − K 0 ) + 1 = 0

thus

− Ω02 r + 1 = 0 ⇒ Ω0 r = ω0 r ω0 = 1

Ω 0 r (3 − K 0 ) = 0 ⇒ K0 = 3 .

We get the same solution as above.

5.1. Ideal amplifier but K0 ≠ 3

The steady state solution (condition) does not tell what happens when K0 ≠ 3. We

have to solve eq. (7) again, the roots are

(K 0 − 3) ± (K 0 − 3)2 − 4
s1,2 = (12)
2

The transient response of the oscillator is thus described as

U n exp(sn t )

with n = 1, 2. When K0 < 3 the oscillator response is damped sinusoid. On the other

hand, the oscillator output increases when K0 > 3.

The oscillator will start only if one of the roots lies in the right half of the

complex plain. This leads to the following conditions [9].

K0 〉 3 
 ⇒ 3 〈 K0 〈 5 (13)
(K 0 − 3)2 − 4 〈 0 

The actual starting frequency Ω0S = ω0S/ω0 is determined by the imaginary part of eq.

7
(12)

(K 0 − 3)2 − 4
Ω0S = (14)
2

But AGC keeps K0 → 3 when the amplitude reaches the desired value, so ω0 →

1/RC, still.

5.2. Ideal amplifier but non ideal “Wien divider”

Let us suppose now that (Fig. 1) the amplifier K is ideal while “Wien divider” is

non ideal:

G1 = G + ∆ G , C1 = C + ∆ C

G2 = G , C2 = C

Form eq. (5) we get now a new value Ω0∆:

ω02∆ 1 + ∆G G
Ω = 2 =
2
(15)
ω0
0S
1 + ∆C C

and also needed

K 0 ∆ = 1 + R f R g = 3 + ∆G G + ∆C C (16)

6. Real amplifier

In general, the frequency response of the amplifier K is determined by many

poles and zeros; however, in order to assure stability in closed-loop feedback

configuration, mostly the amplifier voltage gain is

K = K0 ⋅ Ω3 (s + Ω3 ) (17)

where

8
Ω 3 = ω3 ω0

is a normalized 3 dB frequency (dominant pole).

Substituting from eq. (17) into normalized version of eq. (5) [see eq. (7), K0 →

K], we get characteristic equation

s 3 + s 2 (3 + Ω 3 ) + s[Ω3 (3 − K 0 ) + 1] + Ω3 = 0 (18)

Solving eq. (18) in the frequency domain we get [10]…

2
ω  1
Ω =  0 r
2
 = (19)
 ω0 1 + 3 Ω3
0r

K0 = 3 +
1
Ω3
(
⋅ 1 − Ω 02r ) (20)

For completeness, let us derive needed K0 and Ω 0r


2
once more – by means of the

Routh-Hurwitz stability criterion [30]. In the case of a third order system we have

s 3 + a2 s 2 + a1 s + a0 = 0

a0, a1 and a2 must be positive. The stability boundary is given by

a2 a1 − a0 = 0

thus

(3 + Ω3 ) ⋅ [Ω3 (3 − K 0 ) + 1] − Ω3 = 0
and needed

3
K0 = 3 + ⋅
Ω 3 (3 + Ω 3 )

If we substitute from eq. (19) into eq. (20), we get

9
1 1 3
K0 = 3 + − = 3+ (20a)
Ω3 Ω3 + 3 Ω 3 (3 + Ω 3 )

It is the same result as before.

We need to determine Ω 0 r , too. The characteristic equation must contain an

undamped second – order factor. Hence we have

(s 2
)
+ Ω 02r ⋅ (s + b ) = 0

thus

s 3 + bs 2 + Ω 02r s + bΩ 02r = 0

It is evident that

a1 = Ω 02r

thus

 
Ω 02r = Ω 3 (3 − K 0 ) + 1 = 3Ω 3 − Ω 3 3 +
3 1
 +1 =
 Ω 3 (3 + Ω 3 )  1 + 3 Ω3

see eq. (19).

Figure. 2: Possible realization of the amplifier K

10
6.1. Amplifier K realized by means of VFA

The conventional realization of K with one voltage feedback op amp (VFA ≡

voltage controlled voltage source) is shown in Fig. 2. We can determine an admittance

model [1], [4]:

(a ) + (b ) o (c ) −
(a ) +  Gi 0 0  U a   I 1 
 
− G f + AG A  ⋅ U b  =  0 
(21)
(b ) o − AG A G f + GA
(c ) −  0 −Gf G g + G f  U c   0 

From the admittance model [eq.(21)] we get

1
K = K0 ⋅ (22)
1 + (R f + RA ) Rg
1+
A

where A is the gain of the op amp

RA is the output resistance of the op amp

K 0 = 1 + R f Rg (23)

r = R f Rg (24)

The op amp is ideal when A → ∞, thus the circuit in Fig 1 will oscillate if

K0 = 1+ r = 3 ⇒ r = R f Rg = 2 .

We suppose an infinite output conductance GA → ∞ (RA → 0). Most modern op

amps voltage gain is

A0ω1
A( p ) = = ≈ ωT p (25)
p + ω1 ω 〉〉 ω1

where A0 is the op amp dc gain

ω1 is the 3 dB frequency (dominant pole)

ωT = A0ω1 is a gain – bandwidth product.

11
Let us normalize eq. (25):

ωT ω0 ΩT
A(s ) = = (26)
p ω0 s

Now we can easily determine that [from eq. (22)]

ω3 ωT K 0
Ω3 = = = ΩT K 0 (27)
ω0 ω0

Substituting eq. (27) into eq. (19) and eq. (20) we get

Ω 02r = ΩT (ΩT + 3K 0 ) (28)

and needed

(
K 0 = 3ΩT ΩT − 1 + Ω or
2
) (29)

The signal amplitude calibration (K0) disturbs the oscillation frequency (Ω0r) and

vice versa. The degree of coupling is proportional to ΩT = ωT/ω0, then at higher

frequencies ω0 the inaccuracy in the Ω0r and K0 can be intolerable [11].

Substituting eq. (29) into eq. (28) we get equation

Ω 04r + Ω 02r (7 + ΩT ) − ΩT + 1 = 0 (30)

thus

− (7 + ΩT ) ± (7 + ΩT )2 − 4(1 − ΩT )
Ω 2
0 r a ,b =
2

We know that Ω 0r
2
is always possitive, thus the right solution is

7 + ΩT  4(ΩT − 1) 
Ω 02r = − ⋅ 1 − 1 +  (31)
2  (7 + ΩT )2 

It is evident that for ΩT > 1 is

4(Ω T − 1) (7 + Ω T ) 〈〈 1
2

12
and we can use the approximate relationship

1+ x = ≅ 1 + 0 ,5 x .
x 〈〈 1

Therefore we get a very simple expression

Ω 02r ≈ (ΩT − 1) (ΩT + 7) (31a)

thus

Ω 0r ≈ (ΩT − 1) (ΩT + 7) (31b)

and from eq. (29) needed

3ΩT 3
K0 = = (29a)
ΩT − 1 + (ΩT − 1) (ΩT + 7) 8
1−
ΩT (ΩT + 7)

ΩT
Figure. 3: The calculated ratio and K0
Ω0 r

13
Table 1 shows numerical results of the oscillator with real VFA. It is clear that

one can use the simplified relations (31b) and (29a). The penultimate column contains

the calculated ratio ΩT/Ω0r. Its ideal result is ΩT because ideal Ω0r = ω0r/ω0 = 1. We can

see that reality for small values of ΩT is significantly different, see Fig. 3. The last

column states Ω3 = ΩT/K0; K0 – eq. (29a).

Table 1 Numerical results of the oscillator with VFA

- - - - - - -
1,00 0 0 ∞ ∞ ∞ 0
1,01 3,53·10-2 3,53·10-2 269,400 269,40 28,590 3,75·10-3
1,05 7,88·10-2 7,88·10-2 56,040 56,040 13,330 1,87·10-2
1,10 0,1110 0,1111 29,380 29,370 9,910 3,74·10-2
1,20 0,1559 0,1562 16,050 16,040 7,684 7,48·10-2
1,40 0,2176 0,2182 9,386 9,383 6,416 0,149
1,50 0,2417 0,2425 8,058 8,053 6,206 0,186
1,80 0,3000 0,3015 6,068 6,061 6,000 0,297
2,00 0,3318 0,3333 5,407 5,400 6,028 0,370
2,50 0,3942 0,3974 4,531 4,524 6,342 0,553
3 0,4429 0,4472 4,098 4,091 6,774 0,733
4 0,5160 0,5222 3,674 3,667 7,752 1,091
10 0,7169 0,7276 3,153 3,148 13,950 3,177
20 0,8284 0,8389 3,048 3,045 24,14 6,568
40 0,9031 0,9109 3,014 3,013 44,29 13,28
100 0,9578 0,9619 3,0025 3,0022 104,4 33,31
200 0,9782 0,9805 3,00065 3,00058 204,5 66,65
400 0,9889 0,9901 3,00017 3,00015 404,5 133,3
1000 0,9955 0,9960 3,00003 3,00002 1004 333,3
Eq. (31)
NOTES Eq. (31) Eq. (31b) Eq. (29a) Eq. (29a)
Eq. (29)

Let us use eq. (31b), then

ΩT ΩT2 (ΩT + 7)
≈ (32)
Ω0r ΩT − 1

14
We can specify that the function “32” has a minimum 6.0 at a ΩTMIN = 1.828, thus

corresponding Ω 02r MIN = (1.828 − 1) (1.828 + 7) = 9.379 ⋅10−2 and Ω 0 r MIN = 0.3063 thus

ΩT MIN Ω 0 r MIN ≅ 5.969

This means that the limit oscillator frequency is ω0r = ωT /5.969 which represents

16.67 % of ωT [11], [12].

Figure. 4: Qualitative description of conditions for different ΩT

15
For ΩT < ΩTMIN rapidly grows the needed value of K0 – “hard coupling Ω0r and

K0” – see Fig. 4a. Likewise holds true that derivation of ΩT/Ω0r is negative – see Fig. 3.

For ΩT > ΩTMIN goes K0 to the ideal value 9.54 dB (K0 = 3) it is “soft coupling Ω0r

and K0” – see Fig. 4b. The derivation of ΩT/Ω0r is positive now.

6.2. Experimental results with VFAs

A simple Wien oscillator using a filament lamp is shown in Fig. 5. The cold

resistance of the used lamp is 52 Ω. The nonlinearities associated with lamp resistance

provide the control of oscillator amplitude [6].

Figure. 5: Simple Wien oscillator with transistor buffer and “LAMP AGC” (T1 –

BC546B, T2 – BC556B, UCC = ± 15 V)

The feedback resistor Rf is 294 Ω. For gain of 3, the lamp filament resistance

must be Rf /2 = 147 Ω, and the op amp must be able to provide sufficient current to

supply the feedback circuit total resistance 294 + 147 = 441 Ω. Therefore a simple

transistor buffer (T1, T2) is used (not very good crossover distortion).

Table 2 gives collected data from the tested oscillator. We use resistors and

16
capacitors with a 1 % tolerance and suppose fT = 1 MHz (LM741). The oscillator

amplitude is U0 ≈ 11 V.

Table 2 Collected data from the tested oscillator

MEASURED CALCULATED
*
DIST NOTES
µs Hz
Hz % - - -
1517 104,89 105 1,1 1,001 9533,8 0,9996 1
757,8 209,77 209 0,8 0,996 4767,1 0,9992 1
182,1 874,10 870 1,1 0,995 1144,1 0,9965 1
106,2 1498,4 1490 1,5 0,994 667,39 0,9941 1, 2
33,97 4684,5 4520 2,3 0,965 213,47 0,9817 1, 2, 3
30,35 5244,3 5060 2,8 0,965 190,68 0,9795 1, 2, 3
17,66 9010,9 8670 3,5 0,962 110,98 0,9655 1, 2, 3
16,19 9830,0 9410 7,8 0,957 101,73 0,9625 1, 3, 4
17,19 10800 10180 11,0 0,943 92,593 0,9590 1, 3, 4
*
used eq. (31b)

NOTE 1 – The accuracy of the oscillator is limited by the precision of the components

NOTE 2 – Op amp starts to work in a nonlinear mode of operation – “slew rate

distortion”

NOTE 3 – The accuracy of the oscillator is limited by large of the components

NOTE 4 – Hard “slew rate distortion”, is measured

For the oscillator, slew rate (SR) becomes the dominant nonlinearity for a

frequency operation beyond the limit given by (power frequency)

SR
fp ≈ (33)
2π U o

where U0 is the amplitude of the output voltage and SR is amplifier slew-rate

factor [1], [13], [14]. If U0 = 11 V and SR = 0.5 V/µs, then

f p = 0.5 ⋅10−6 (2π ⋅11) = 7234.3 Hz

17
We must always design oscillator frequency

SR
f o 〈〈 f p = (33a)
2π U o

In fact, the term “slew–rate distortion” is somewhat improper. Distortion and

slew–rate have a common beginning only – limited current in an input stage of a

classical operational amplifier (VFA).

We can find another interesting experimental results with the famous “741”

(expected and SR = 0.81 V/µs – too optimistic) in [15] – Table 3 – for resistors with a 1

% tolerance. The transistor buffer is not used. The output voltage amplitude U0 is

diminished with increasing oscillation frequency. However, this approach is not suitable

for tunable oscillators!

Table 3 Experimental results [15] and theoretical results – Esq. (31), (31b) and (33)

DIST
µs Hz V % KHz - KHz KHz KHz
9,52 16,72 8,2 0,99 15,6 59,81 15,59 15,59 15,72
7,02 22,67 6,1 0,79 20,5 44,11 20,65 20,82 21,13
2,70 58,95 2,7 1,58 46,9 16,96 47,47 48,11 47,75
1,80 88,42 2,1 1,88 64,1 11,31 65,38 66,35 61,39
1,08 147,40 1,4 1,88 93,2 6,79 94,33 95,49 92,08
NOTES [15] [15] [15] - Eq. (31) Eq. (31b) Eq. (33)

It is evident that even a very simple eq. (31b) gives us a good guideline for the

design with a real op amp. Furthermore, it is clear that it is suitable to choose f0 in

accordance with eq. (33) to keep down the nonlinear distortion (“slew–rate distortion”).

6.3. Amplifier K is realized by means of CFA

The current feedback amplifier (CFA – transimpedance amplifier) offers

18
practically constant ω3 at low – medium gains and its slew – rate is from 10 to 100

times higher than for classical VFA, for the same fT [11].

An implementation of the amplifier is the same as in Fig. 2. We can easily

determine [4], [16] that for zero output resistance

1
K = K0 ⋅ (34)
1 + R f Rg

where K 0 = 1 + R f Z T

and
RT
ZT = (35)
1 + pRT CT

is a transimpedance of the CFA with dc resistance RT and parallel capacitance CT.

Substituting eq. (35) into eq. (34) we get

1 ω3
K = K0 ⋅ = = K0
1 + R f RT + pR f CT R f RT 〈〈 1 p + ω3

where

1 RT R
ω3 = = = T ω1 (36)
R f CT R f RT CT Rf

where

ω1 = 1 (RT CT )

is a dominant pole of ZT.

Now ω3 is not a function of Rg. This means that the amplitude calibration by means of

Rg does not affect f0r – it is very interesting in frequency variable oscillators. We can use

eqs. (19) and (20) directly.

CFAs allow operation at higher frequencies and larger amplitudes than in the

case of the VFA based oscillator. Table 4 shows numerical results of the oscillator with

real CFA. An equivalent ΩTE of the amplifier is determined as ΩTE = K0· Ω3.

19
Some experiment results are included in [11]. The BW (≡ ω3) the AD844 has

been adjusted at the value of 8 MHz (Rf = 4,3 kΩ). For f0 = 4,42 MHz the CFA exhibits

an oscillator frequency f0r = 2,71 MHz. It is evident that Ω3 = ω3/ω0 = f3/f0 = 8/4,42 =

1 f
1,810, thus: Ω 02r = = 0,3763 ; Ω 0r = 0 r = 0,613 ; f 0r = 4,42 ⋅ 0,613 = 2,709
1 + 3 / 1,81 f0

MHz.

Attention must be paid to comparison of VFA and CFA properties. Authors [11]

do not distinguish accurately the meaning of ΩT and Ω3 – see ΩTE in Table 4. If the

amplifier K is implemented by means of VFA, τ = 1/ωT corresponds to inverse value of

ωT (≡ gain – bandwidth, GBW). If the amplifier K is implemented by means of CFA, τ =

Rf CT =1/ω3 (bandwidth only, BW) which may slightly affect the conclusions in [11].

Table 4 Numerical results of the oscillator with real CFA

- - - -
0,041 202,7 1,013
0,005
0,01 0,058 102,7 1,027
0,03 0,100 36,00 1,080
0,05 0,128 22,67 1,134
0,10 0,177 12,69 1,269
0,2 0,250 7,688 1,538
0,3 0,302 6,030 1,809
0,5 0,378 4,714 2,357
0,7 0,435 4,158 2,911
1 0,500 3,750 3,750
2 0,632 3,300 6,600
5 0,791 3,075 15,38
10 0,877 3,023 30,23
20 0,933 3,007 60,13
50 0,971 3,0011 150,1
100 0,985 3,00029 300,0
500 0,997 1500
Eq. 3,000012
NOTES Eq. (20)
(19)

20
The main advantage of CFA is still a large slew – rate and considerably larger ω3

(thus ωTE) in comparison with the VFA. The SR affects f0r (Ω0r) slightly, it mainly acts

on the distortion level. The main contribution to the frequency error is due to ω3 (ωTE)

limitations.

The other type of Wien oscillator with CFA can be found for example in [17].

6.4. Amplifier K is realized by means CCII+

The first practical implementation using a single CCII+ [18] is shown in Fig 6a,

the second one is in Fig.6b.

Figure. 6: Two practical implementations of the amplifier K with CCII+

Let us determine an admittance model of the amplifier K only in Fig 6a:

(a )Y (b ) X (c )Z
(a )Y  YY 0 0  U a   I a 
(b ) X  − K V YX 0  ⋅ U b  =  0 
(37)
Gg
(c )Z − K V K C Y X K C YX G + YZ  U c   0 

where transfer properties of the CCII+ are (ordinary one pole models):

21
UX ωV 3 ΩV 3
KV = = KV 0 = KV 0
UY p + ωV 3 s + ΩV 3
(38)
IZ ωC 3 ΩC3
KC = = KC0 = KC0
IX p + ωC 3 s + ΩC3

and

YY is an input admittance of the node Y (ideally zero)

YX is an input admittance of the node X (ideally infinite)

YZ is an output admittance of the node Z (ideally zero).

The solution of eq. (37) gives us

YX → ∞
U KV K C Gg R
K= C = = YY → 0 = K V K C (39)
Ua G g (G + YZ ) Rg
G + YZ + YZ → 0
YX

Using Thevenin's theorem we find that the output resistance of this amplifier K is just R.

This output resistance is in series with C – see Fig 6a – it is a part of “Wien divider”.

In other words, oscillator in Fig 6a is functionally equivalent to classic Wien

type oscillator. It is appropriate to note that the output impedance at the node C is now

nonzero.

Ideally KV = KC = 1 and needed K0 = R /Rg = 3, thus Rg = R /K0 = R/3 and Ω0r = 1.

Realistically we use eq. (38) and we get for KV0 = KC0 = 1

R ΩV 3Ω C 3
K= ⋅ 2 (39a)
R g s + s (Ω V 3 + Ω C 3 ) + Ω V 3 Ω C 3

A “simple Ω3” is not available now.

Substituting from eq. (39a) into normalized version of eq. (5) (see eq. (7),

K0→K) we get characteristic equation

22
s 4 + s 3 (3 + Ω V 3 + Ω C 3 ) + s 2 [3(Ω V 3 + Ω C 3 ) + Ω V 3 Ω C 3 + 1] + L
  R   (40)
L + s Ω V 3 + Ω C 3 + Ω V 3 Ω C 3  3 −  + ΩV 3 Ω C 3 = 0
 R g  
 

In case of a fourth order system we have [30]

s 4 + a3 s 3 + a 2 s 2 + a1 s + a 0 = 0 .

a0, a1, a2 and a3 must be positive – see eq. (40). The stability boundary is given by

relation

a3 a 2 a1 − a12 − a0 a32 = 0
We need to determine Ω0r, now. The characteristic equation must contain an

undamped second – order factor. Hence we have

(s 2
)( )
+ Ω 02r ⋅ s 2 + as + b = 0

thus
s 4 + as 3 + (b + Ω 02 r ) s 2 + a Ω 02 r s + bΩ 02 r = 0 .

Now it is
a3 = a

and
a1 = aΩ 02r .

It is evident that
aΩ 02r a
= 1
a a3
and we can write
Ω 02r = a1 a3
then

Ω V 3 + Ω C 3 + Ω V 3 Ω C 3 (3 − R R g )
Ω 02 r = = =
3 + ΩV 3 + Ω C 3 Ω C 3 = mΩ V 3
(41)
(m + 1)ΩV 3 + mΩV2 3 (3 − R Rg )
=
3 + (m + 1) Ω V 3

23
From the stability boundary equation it is evident that

2
a  a
a 0  3  = a 2 3 − 1
 a1  a1

We can easily prove that a2 a3 /a1 > 10, always, and therefore it applies

2
a  a
a 0  3  ≅ a 2 3
 a1  a1

a1 ≈ a0 a3 a 2 .

Now we can determine needed ( Ω C 3 = mΩV 3 )

R 1+ m 3 + (1 + m )Ω V 3
= 3+ − (42)
Rg mΩ V 3 1 + 3(1 + m )Ω V 3 + mΩ V2 3

Substituting eq. (42) into eq. (41) we get

1
Ω 02 r = (43)
3(1 + m ) 1
1+ +
mΩ V 3 mΩ V2 3

Let us assume m = 1 (ΩV3 = ΩC3), then

R 2  Ω V 3 + 1,5 
≈ 3+ 1 −  (42a)
Rg ΩV 3  ΩV 3 + 6 + 1 ΩV 3 
m =1

and
1
Ω 02 r ≈ (43a)
m =1 1 + 6 Ω V 3 + 1 Ω V2 3

Table 5 shows some numerical results for m = 1. In this case, the amplifier K has

a double – pole of the transfer function. This is reflected in eq. (43a) “as number 6” –

compared with “the number 3” in eq. (19). On the other hand, we must remember that

ΩV3, ΩC3 (of CCII+) may be from 2π·106 ·20 to 2π·106·200 rads-1.

24
Table 5 Numerical results of the oscillator (Fig. 6a) with CCII+ ΩV3 = ΩC3 (m = 1)

- - -
0,1 0,0788 21,01
0,5 0,2430 6,059
1 0,3536 4,375
2 0,4851 3,588
4 0,6247 3,232
10 0,7881 3,057
20 0,8762 3,018
40 0,9323 3,0049
100 0,9712 3,00085
200 0,9853 3,00022
400 0,9926 3,000055
1000 0,9970 3,000009

Now we determine an admittance model of the amplifier K in Fig 6b:

(a )Y (b ) X (c )Z
(a )Y  YY 0 0  U a   I a 
 
− G f  ⋅ U b  =  0 
(44)
(b ) X  − K V YX G g + G f + YX
(c )Z − K V K C YX − G f + K C YX G f + YZ  U c   0 

The solution of eq. (44) gives us (YY = YZ → 0, YX → ∞)

UC  KC Gg 
K= = K V 1 + ⋅  (45)
Ua  1 + K G 
 C f 

ideally
Gg Rf
KV = K C = 1 and K = 1+ = 1+
2G f 2 Rg

thus for K0 =3

Rf Rf R
1+ =3 ⇒ Rg = = =
2 Rg 4 R f = 2R 2

If we use Thevenin's theorem again, we find out that the output resistance of the

25
amplifier K in Fig. 6b is just Rf /2. Therefore, we must choose Rf = 2R to obtain correct

“Wien divider”.

Substituting eq. (38) into eq. (45) for KV0 = KC0 = 1, we get

ΩV 3 s + 2Ω C 3 + Ω C 3 R f R g
K= ⋅ (46)
s + ΩV 3 s + 2 ΩC3

It is evident that the feedback (Rf ) leads to zero in the transfer function.

Substituting from eq. (46) into normalized version of eq. (5), we get

characteristic equation (ΩC3 = mΩV3):

[ ]
s 4 + s 3 [(2m + 1)Ω V 3 + 3] + s 2 2mΩ V2 3 + 3(2m + 1)Ω V 3 − Ω V 3 + 1 + L
  Rf   (47)
L + s 6mΩ V2 3 − mΩ V2 3  2 +  + (2m + 1)Ω V 3  + 2mΩ V2 3 = 0
  R g  

Using the same procedure as in the previous case, we get relations

1
Ω 02 r = (48)
3m + 1 1
1+ +
mΩ V 3 2mΩ V2 3
and

Rf
= 4+
1 + 2m

(1 + 2m )ΩV 3 + 3 (49)
Rg mΩ V 3 2mΩ V2 3 + 2(1 + 3m )Ω V 3 + 1

Let us assume again ΩV3 = ΩC3 (m = 1)

1
Ω 02 r = (48a)
m =1 1 + 4 Ω V 3 + 0 ,5 Ω V2 3

and needed

R 3  ΩV 3 + 1 
= 4+ 1 −  (49a)
Rg ΩV 3  2Ω V 3 + 8 + 1 Ω V 3 
m =1

see Table 6.

26
Table 6 Numerical results of the oscillator (Fig. 6b) with CCII+ for m = 1;

- - - -
0,1 0,1048 32,187 17,093
0,5 0,3015 9,182 5,591
1 0,4264 6,455 4,227
2 0,5657 5,140 3,570
4 0,7016 4,519 3,259
10 0,8436 4,183 3,092
20 0,9124 4,084 3,042
40 0,9533 4,040 3,020
100 0,9806 4,015 3,0077
200 0,9901 4,0076 3,0038
400 0,9950 4,0038 3,0019
1000 0,9995 4,0015 3,008

It is evident that configuration in Fig. 6b has a slightly better values of Ω0r

depending on ΩV3 – thanks to zero of its transfer function.

We must note, however, that the oscillators in Fig. 6 have a non-zero output

resistance. Therefore there is usually used a buffer. But at the moment we are actually

working with the CFA – “single – stage”, AD844 with externally – accessible

compensation pin for example [19]. Conveyors are not available in a “two – stage” CFA

– AD8011 for example [20], [21].

There are many other structures of Wien oscillator with current conveyors

described in the literature. Numerous modern integrated circuits contain CCII+ (or

another type of conveyor), as CCII+ is a versatile element that allows us to build all the

necessary controlled sources [22]: VCVS, VCCS, CCCS and CCVS.

27
Figure. 7: Wien oscillator with “VFA based on CFA”

A structure [23] with two CCII+ and voltage follower (buffer) creates in fact

“voltage feedback amplifier based on CFA” – Fig. 7. From a formal point of view, it is

VCVS (as well as classic VFA), but with dynamic properties as CFA (large ft and slew

rate) – we can use eqs. (31b) and (29a). The amplifier K gain is

 Rf  1
K = 1 + ⋅
 R g  1 + R f Rg
 1+
A
where
A = ZT RX

ZT – see eq. (35)

thus

28
RT 1 R ω1 A0ω1 ωT
A= ⋅ = T ⋅ = = .
R X 1 + pRT C T R X p + ω1 p + ω1 p + ω1

Now the transimpedance ZT determines the dominant pole – not the properties of

conveyors. The same argument is true for the CFA.

Another solution of the Wien oscillator with CCII+ can be found for example in

[24] and [25].

Figure. 8: A composite amplifier [26]

7. How to improve the amplifier K with VFA?

The main source of oscillation frequency error is the additional phase shift of the

amplifier K. By properly shaping to phase response of the amplifier it is possible to

extend the frequency range of the oscillator. Fig. 8 [26] shows a composite amplifier

with a transfer function (s = p/ω0)

α −1
s+ ΩT
K = ΩT ⋅ α
Ω α −1 2
s2 + s T + ΩT
α 3α

which can be easily determined from the admittance model of the circuit. We expect the

same amplifiers (VFA) with gain

ωT ωT ω0 Ω
A≈ = = T .
p p ω0 s

29
This composite amplifier (with zero in its transfer function) is optimized [26] for

α = 3 . The zero reduces the phase shift of the amplifier at low frequencies – see Fig.

6b and eq. (46), too. Another similar composite amplifier is described in [27] – the

authors draw attention to the possibility of double oscillation (basic ω0 plus parasitic

one), if the amplifier is set improperly.

However, this type of amplifier does not solve the slew–rate problem. We may

achieve a higher oscilation frequency, but slew–rate limits us the more.

Figure. 9: The composite amplifier with transistor voltage amplifier A2


(INVERTING); KF507 – equivalent BD135; KF517 – equivalent BD136; connected in
testing as inverting structure [29].

A very useful means of increasing op amp performance can be obtained by

blending the performance advantages of a standard op amp and discrete transistors [28].

30
However, whenever an output circuit is added to an op amp that provides a voltage gain,

we need to ensure frequency stability of the composite amplifier – and very well. Hint

to the possible solution may be in [29] – Fig. 9.

Within this circuit T1 and T2 are the complementary buffer transistors that

provide rail-to-rail output swing and generate additional voltage gain (inverting)

proportional to the ratio RLOAD/RE – see Fig. 9.

The inverting amplifier A2 changes “meaning of 741 inputs”. Frequency

properties of A2 are controlled by a frequency dependent divider R + 1/jωC, so that the

gain of A2 (its modulus) at higher frequencies is larger. C1 reduces the gain of the 741.

For greater closed–loop gains, C1 may be quite small (not to be used at all).

For Rf =Rg =10 kΩ (feedback circuit transfer is 0,5) and RLOAD = 1,8 kΩ the

optimum (found experimentally) is: R = 10 kΩ, C = 1,5 nF, C1 = 330 pF. The measured

slew rate of the composite structure is SR = 4,5 V/µs.

For Rf = 100 kΩ, Rg = 10 kΩ and RLOAD = 1,8 kΩ (feedback circuit transfer is

1/11) the optimum is: R = 10 kΩ, C = 330 pF, C1 = 56 pF. The measured slew rate is

SR = 3,3 V/µs.

For Rf = 1 MΩ, Rg = 10 kΩ and RLOAD = 1,8 kΩ (feedback circuit transfer is

1/101) the optimum is: R = 4,7 kΩ, C = 1,5 nF, C1 = 15 pF. The measured slew rate is

SR = 2 V/µs.

At the same time 741 itself has SR = 0,6 V/µs. It is evident that (Fig. 9)
u out = A2 ⋅ u OPA
thus
u OPA = u out A2 .

The voltage amplifier is much faster than the actual op amp. The slew rate of the

operational amplifier is defined as

31
du OPA
SROPA = .
dt MAX

The slew rate of composite amplifier is defined as

SR =
du OPA
=
d
( A2uOPA ) = A2 SROPA .
dt MAX dt MAX

Therefore the SR of the composite amplifier in Fig. 9 increases in accordance with

increasing value of |A2|.

And ωT also increases. A qualitative description of gains in the composite amplifier


is in Fig. 10. If we find the gain Am on frequency fm as shown in Fig. 10, we can easily
determine that (Am – no dB here)
f T ≈ Am f m .

Figure. 10: Qualitative description of gains in the composite amplifier in Fig.9

In [29] we find (Rf = 1 MΩ, Rg = 10 kΩ, RLOAD = 1,8 kΩ, R = 4,7 kΩ, C = 1,5 nF,

C1 = 15 pF) that for fm = 100 kHz (greater than f3 = 80 kHz) is Am = 55,6 (34,9 dB),

32
thus fT = 55,6·105 = 5,56 MHz.

8. Conclusions

The Wien oscillator has been analyzed by means of nodal voltage analysis.

There have been derived simple formulas (19) and (20), these are useful for amplifiers

with VFAs and CFAs.

It is evident that in practice is ω3/ ω0 = Ω3 > 15 very good solution. A choice of

smaller ratio results in frequency errors over 10 %, see Table 4. The needed K0 (for Ω3 >

15) is nearly 3 and for VFA we can use directly the simple formulas (19) and (20),

substituting K0 = 3:

1
Ω0r ≈ = ≅ 1 − 4,5 ΩT
1 + 9 ΩT 9 ΩT 〈〈 1

thus

 f 
f 0 r ≅ f 0 1 − 4,5 ⋅ 0  .
 fT 

Otherwise, we have to use more complex formulas given in the text.

If the amplifier K is realised by means of CCII+, we have no Ω3. We must use


more complex eqs. (42) and (43) for Fig. 6a and eqs. (48) and (49) for Fig. 6b.
However, if ΩV3 > 5, then follows for Fig. 6a from eq. (43)
1
Ω 02 r =
3
1+
m
ΩV 3
m +1

An expression
mΩV 3
Ω 3E =
m +1

„plays a role” of an ekvivalent Ω3E.

33
For Fig. 6b we get from eq. (48)
1
Ω 02 r =
3
1+
m
ΩV 3
m +1 3

thus
mΩV 3
Ω 3E = .
m +1 3

The main advantages of amplifiers „with current on demand” (CFA, CCII, VFA

based on CFA) are large slew rate and considerably large ω3 respect to their classic VFA

counterparts.

The slew rate affects f0r slightly, it mainly acts on the distortion level. To

suppress nonlinear distortion “bound to SR“ (only), we can choose “HiFi criterion”:

 SR 
 
2π ⋅ U 0
f0 ≤  .
(4 ÷ 8)
If we want to vary the oscillator frequency, we have to use an amplifier that

meets all of the requirements above at the maximum required frequency f0MAX.

Procedures when it is necessary to adjust the amplifier parameters for each frequency

are somewhat inconvenient in practice.

If ω3/ω0 > 150 (!) the accuracy of the Wien oscillator is mainly limited by only

the precision of the used passive components. The manufacturer's stated tolerance may

produce unacceptable error in the design unless extremely expensive components are

used there.

Acknowledgements

This paper has been elaborated in the framework of the project Opportunity for

34
young researchers, reg. no. CZ.1.07/2.3.00/30.0016, supported by Operational

Programme Education for Competitiveness and co-financed by the European Social

Fund and the state budget of the Czech Republic.

References

[1] MOHYLOVÁ, Jitka and Josef PUNČOCHÁŘ. Theory of electronic circuits:

students book. 1. vyd. Ostrava: Vysoká škola báňská - Technická univerzita Ostrava,

2013, 340 pp. ISBN 978-80-248-3112-1.

[2] SMITH, Adel S. Sedra; Kenneth C. Microelectronic circuits: students book. 5.

ed., international student ed., [Nachdr.]. New York: Oxford University Press, 2007, 340

pp. ISBN 01-951-4252-7.

[3] J. Punčochář, Josef. Analysis of Sellen and Key low – pass filters with real

operational amplifiers, 1. vyd. Ostrava: Vysoká škola báňská, 1999, vol. V, No 1, pp.

177 – 186, ISBN 80-7078-740-6

[4] PUNČOCHÁŘ, Josef. Admittance models of modern linear amplifying

structures, 1. vyd. Ostrava: Vysoká škola báňská, 2003, 200 pp. ISBN 80-248-0223-6.

[5] MIKULEC, Milan and Václav HAVLÍČEK. Basic circuit theory. Vyd. 1. Praha:

ČVUT, 2000, 331 pp. ISBN 80-010-2127-0.

[6] GONZALEZ, Guillermo. Foundations of oscillator circuit design. Boston:

Artech House, 2007, xi, 422 pp. ISBN 15-969-3162-0

[7] PUNČOCHÁŘ Josef. Stabilizace amplitudy kmitů oscilátoru diodami v obvodu

kladné zpětné vazby (Stabilization of the oscillator amplitude by means of diodes in the

positive feedback), Sdělovací technika, No. 11, pp. 403–404 (1978)

[8] HILEMAN D. Common silicon diodes stabilize oscillator. Electronics. Vol. 48.1

35
pp. 111, ISSN 0883-4989 PENTON PUBL INC 1100, 1975

[9] PUNČOCHÁŘ Josef. Wien-bridge oscillator with real op amp. 11th

international Czech-Slovak scientific conference: conference proceedings: May 10-11,

2001, Brno, Czech Republic. Brno: University of Technology, Faculty of Electrical

Engineering and Computer Science, 2001, 442 pp. ISBN 80-214-1861-3.

[10] HE, Fan, Raymond RIBAS, Cyril LAHUEC and Michel JÉZÉQUEL.

Discussion on the general oscillation startup condition and the Barkhausen criterion.

Analog Integrated Circuits and Signal Processing. 2009, vol. 59, issue 2, pp. 215-221.

[11] CELMA, S., P.A. MARTINEZ and A. CARLOSENA. Current feedback

amplifier based sinusoidal oscillators. IEEE Transactions on Circuits and Systems I:

Fundamental Theory and Applications. vol. 41, issue 12, pp. 906-908.

[12] BORYS, Andrzej, P.A. MARTINEZ and A. CARLOSENA. Sinusoidal

oscillators with slew-rate type nonlinearity. European Transactions on

Telecommunications. 1991, vol. 2, issue 4, s. 447-451. DOI: 10.1002/ett.4460020411.

[13] PUNČOCHÁŘ, Josef. Operační zesilovače v elektronice. Praha: BEN, 1996,

479 pp. ISBN 80-901-9843-0.

[14] PUNČOCHÁŘ Josef. Souvislost mezi rychlostí přeběhu a činitelem nelineárního

zkreslení u operačních zesilovačů (The relationship between slew rate and distortion of

op amp), Sdělovací technika, No. 11, pp. 417–420 (1980)

[15] HUERTAS, J.L., A. RODRIGUEZ-VAZQUEZ and B. PEREZ-VERDU. High-

frequency design of sinusoidal oscillators realised with operational amplifiers. IEE

Proceedings G (Electronic Circuits and Systems). 1984, vol. 131, issue 4.

[16] PUNČOCHÁŘ, Josef. Current conveyors in linear circuit theory. In: AL, Instytut

Elektrotechniki Teoretycznej i Przemysłowej Politechniki Śląskiej [et a Eds.Magdalena

36
Umińska-Bortliczek] EDS.: ZYGMUNT PIĄTEK. IC-SPETO '99: XXII

Międzynarodowa Konferencja z Podstaw Elektrotechniki i Teorii Obwodów, Gliwice-

Ustroń, 19-22.05.1999. Gliwice: Politechnika Śląska, 1999. ISBN 8385940219

[17] ABUELMA'ATTI, Muhammad Taher, Sa'ad Muhammad AL-SHAHRANI and

Ahmed M. SOLIMAN. A Minimum Component Grounded-Capacitor CFOA-Based RC

Oscillator. Active and Passive Electronic Components. 1997, vol. 19, issue 4, pp. 85-

105. DOI: http://dx.doi.org/10.1007/978-1-4614-1383-7-4.

[18] MARTÍNEZ, Pedro A., Santiago CELMA and Inmaculada GUTIÉRREZ. Wien-

type oscillators using CCII. In: Analog Integrated Circuits and Signal Processing. 1995,

pp. 139-147. ISSN 0925-1030. DOI: 10.1007/BF01239168.

[19] Analog Devices AD844 data sheet.

[20] Drechler, W., Two – stage current - feedback amplifiers, Analog Dialogue 29: 1-2,

1995. www.analog.com/library

[21] SENANI, Raj, D BHASKAR, A SINGH a V SINGH. Current feedback

operational amplifiers and their applications. New York: Springer, 2013, xvii, 249 p.

Analog circuits and signal processing series. ISBN 978-1-4614-5188-4.

[22] PUNČOCHÁŘ, Josef. Operační zesilovače: historie a současnost. 1. vyd. Praha:

BEN - technická literatura, 2002, 66 s. ISBN 80-730-0047-4.

[23] BRANDSTETTER, Pavel and Lukáš KLEIN. Novel Wien bridge oscillator design

using functional block structure with current conveyors. Advances in Electrical and

Electronic Engineering. 2012, Vol. 10, issue 1, pp. 13-16. ISSN: 13361376

[24] BAJER, Josef, Abhirup LAHIRI a Dalibor BIOLEK. Current-Mode CCII+ Based

Oscillator Circuits using a Conventional and a Modified Wien-Bridge with All

37
Capacitors Grounded. Radioengineering. Praha: Fakulta elektrotechnická ČVUT -

Středisko vědeckotechnických informací, 1992, Vol. 20, issue 1, pp. 245-251.

[25] SOLIMAN, Ahmed M. Generation of CCII and ICCII based Wien oscillators using

nodal admittance matrix expansion. AEU - International Journal of Electronics and

Communications. 2010, vol. 64, issue 10, pp. 971-977.

[26] BUDAK, A. and K. NAY. Operational amplifier circuits for the Wien-bridge

oscillator. IEEE Transactions on Circuits and Systems. 1981, vol. 28, issue 9, pp. 930-

934. DOI: 10.1109/TCS.1981.1085060

[27] CARLOSENA, A., P. MARTINEZ and S. PORTA. An improved Wien bridge

oscillator. IEEE Transactions on Circuits and Systems. vol. 37, issue 4, pp. 543-546.

DOI: 10.1109/31.52758

[28] Op amp applications handbook. Editor Walt Jung. Amsterdam: Elsevier 2005, 878

pages. ISBN 07-506-7844-5

[29] PUNČOCHÁŘ Josef. Zesilovací struktura s operačními zesilovači a zlepšenými

dynamickými vlastnostmi, Sdělovací technika, No. 12, pp. 463-465 (1978)

[30] GELB, Arthur and Wallace E VANDER VELDE. Multiple – input describing

functions and nonlinear system design. pp. 655, McGraw-Hill, 1968.

38

View publication stats

You might also like