Proceedings of the 13th WSEAS International Conference on CIRCUITS On Stability of Electronic Circuits HASSAN FATHABADI Electrical Engineering Department Azad University (South Tehran Branch) Tehran, IRAN h4477@hotmail.com NIKOS E. MASTORAKIS Industrial Engineering Department Technical University of Sofia, Sofia 1000, BULGARIA mastor@wseas.org Abstract: -In this paper, a necessary and sufficient condition for robust stability of electronic circuits at high frequency which contain differential pair (emitter-coupled pair) is proposed. In fact, when emitter-coupled pairs are used as one input signal – one output signal, they have uncertainty in their transfer functions at high frequency. Even as will be shown, this uncertainty can be caused instability in closed loop electronic circuits at high frequency. The uncertainty will be modeled as multiplicative perturbation in the transfer function of the differential pair at high frequency. Based on this uncertainty model, a necessary and sufficient condition for robust stability of above electronic circuits at high frequency will be presented. This condition guarantees internal stability of electronic circuits at high frequency with respect to above uncertainty. Key- Words: - Stability, robust, uncertainty, perturbation, electronic, circuit. 1 Introduction When Vo = VO 2 , Vi = V1 and V2 = 0 , differential A group of electronic circuits that contain emittercoupled pairs are very important. They have many applications in amplifiers, communication circuits and etc [1]. The differential pair or emitter-coupled pair is an essential building block in modern integrated circuits (IC) amplifiers [2]. This circuit is shown in Fig. 1. This paper considers this group of electronic circuits at high frequency. As we know, emittercoupled pair has two main properties: A) It rejects most of common noises that arrive from two bases. B) Its transfer characteristic has linear region larger than a stage which consists of only one BJT. As we know, the property (A) is decreased by increasing frequency [1]. About property (B), in fact, for small difference voltage ( Vd pair has one input signal( Vi ) and one output signal ( V0 ) as shown in Fig.2. Thus, we can consider differential pair as SISO (Single Input Single Output) block which shown in Fig.3. 4VT 〉 V1 − V2 = Vd in Fig.1), the differential pair behaved as a linear amplifier [2]. differential pair which shown in Fig.1. Consider Corresponding author’s address: Hassan Fathabadi, No. 27-Ghazanfar Khavand Alley -30 Metry Jey BloverTehran-Iran .P.O.Box: 13517-53593. E-mail: h4477@hotmail.com ISSN: 1790-5117 Fig. 1. Emitter-coupled pair. 43 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS Uncertainty in systems is one of the most difficult problems [3], [4], [5]. Almost, all systems, such as linear, nonlinear, discrete, neural and etc have uncertainty in their transfer function [6], [7], [8]. In recent years, there are many researches in analysis of circuits and robust stability [9], [10], [11]. In this paper, we show, the block which shown in Fig. 3 has uncertainty at high frequency. This uncertainty is in the place of its poles. We consider only domain pole and its uncertainty will be modeled as multiplicative perturbation at high frequency in the transfer function of the above block. Then base on this uncertainty model, we will study the robust stability of the closed loop electronic circuits which contain above block as a section of their plant at high frequency. Finally a necessary and sufficient condition for robust stability and stabilization of above electronic circuits at high frequency will be proposed. 2 Multiplicative Perturbation Suppose that the nominal plant transfer function is P (s ) and consider perturbed plant transfer function of the form ~ P ( s ) = [1 + Δ ( s )W ( s )]P ( s ) . (1) Here W (s ) is a fixed stable transfer function, the weigh, and Δ (s ) is a variable stable transfer function satisfying Δ ( s ) ∞ ≤ 1 further more, it is assumed that no unstable poles of P ( s ) are ~ canceled on forming P (s ) [8]. Thus, P (s ) and ~ P ( s ) have the same unstable poles. Such a perturbation Δ (s ) is said to be allowable. The idea behind this uncertainty model is that Δ ( s )W ( s ) is normalized plant perturbation away from (1): ~ P ( s) − 1 = Δ( s)W ( s) P( s ) hence if Δ ( s ) ∞ (2) ≤ 1 , then ~ P ( s) − 1 ≤ W ( s) , P( s ) ∀ω . (3) So W (s) provides the uncertainty profiles. The inequality (3) describes a disk in the complex ~ P (s) plane. In fact, at each frequency the point P( s ) lies in the disk with center 1 and radius W ( s) . Typically, W (s) is a (roughly) increasing function of ω . In other word, uncertainty increases with increasing frequency. The main purpose of Δ (s ) is to account for phase uncertainty and to act as a scaling factor on the magnitude of the perturbation. Thus, this uncertainty model is characterized by a nominal plant P (s ) together with a weighting function W (s ) [12]. Fig. 2. Differential pair with one input and one output. 3 Perturbed Transfer Function of Differential Pair Consider differential pair which shown in Δ Fig.1. We define: Vid ( s ) = V1 ( s ) − V2 ( s ) , Δ Vod ( s ) = VO1 ( s ) − VO 2 ( s ) , Fig. 3. SISO equivalence block. ISSN: 1790-5117 44 ISBN: 978-960-474-096-3 (4) Proceedings of the 13th WSEAS International Conference on CIRCUITS [V1 ( s ) + V2 ( s)] , 2 Δ [V ( s ) + V O 2 ( s )] Voc ( s) = O1 2 Δ ADM ( s ) ≈ Vic ( s) = (5) Δ Δ V ( s) Vod ( s) , ACM ( s ) = oc Vid ( s ) Vic ( s ) (12) As we know, domain pole p1 is computed from following equation [1]: and ADM ( s ) = − g m RC . s (1 + ) p1 (6) p1 = where Vid (s ) is differential-mode input Vod (s ) is differential-mode output Vic (s) is common-mode input Voc (s ) is common-mode output ACM (s ) is common-mode gain 1 R {Cπ + C μ [(1 + g m RC ) + RC R ] + C C cs } R R (13) where R = ( RS + rb ) rπ . In above Equations RS is total resistance of differential –mode input signal source. rπ , rb , ADM (s) is differential-mode gain As we saw, when Vo = VO 2 , Vi = V1 and V2 = 0 , the Cπ , C μ and C cs are shown in Fig.4 which is equivalence circuit of BJT. differential pair which shown in Fig.2 is obtained. Thus, we have 1 Vo ( s ) = − Vod ( s ) + Voc ( s ) 2 so Vo ( s ) = − 1 ADM ( s )Vid ( s ) + ACM ( s )Vic ( s ) 2 (7) 1 1 ADM ( s ) + ACM ( s )]Vi ( s ) 2 2 (8) and Vo ( s ) = [− From (8), we can write the transfer function of the differential pair which shown in Fig.2 as V ( s) A ( s) 1 ~ = [1 − DM ] ACM ( s) . P (s) = o Vi ( s) ACM ( s) 2 Fig. 4. Equivalence circuit of BJT. (9) It is reminding that, p1 also is upper (high) half-power frequency of differential pair. Just as we see, p1 is a multivariable function of rπ , rb , Also we know that [1] RC (10) (1 + sRE C E ) 2 RE where C E and RE are output capacity and resistance ACM ( s ) ≈ − Cπ and C μ . In a BJT, rπ is input dynamic resistance, rb is base pin resistance, Cπ is internal capacity between internal base and internal emitter and C μ is internal capacity of current source, respectively. The differential-mode gain is obtained from following equation [1]: ADM ( s ) = − g m RC s s (1 + )(1 + ) p1 p2 between internal base and internal collector. On the other hand, we know rπ , rb , Cπ and C μ are (11) related to internal properties of BJT, such as WB where is transfer conductance of BJT . Also p1 and ( width of base ), N D (concentration of donor p2 are poles of above transfer function. In fact, p2 and p2 >> p1 . Thus, we can ignore of atoms), N A (concentration of acceptor atoms), Dn or D p (constant distribution of minority carriers), β 0 and etc [1], [2]. consider only domain pole p1 . So we have ISSN: 1790-5117 45 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS ADM ( jω 〈〈 ACM ( jω From (13) and above notes, we conclude that, domain pole p1 is changed when a BJT in emittercoupled pair is replaced by another BJT (for example, when we repair electronic circuits which contain this differential pair). Thus, domain pole p1 has uncertainty. By replacing (10) and (12) in equation (9), it is not difficult to verify that and from (8), it follows that − RC 1 ~ P ( s ) ≈ ACM ( s ) = (1 + sRE C E ) , 2 4 RE this is the same result which has been obtained in (17). Even, reader can note that, although at low ~ frequency, the phase of P ( jω ) is about 0° , at high frequency, its phase is about 180 ° .This causes to change negative feedback to positive feedback and thus, instability in closed loop electronic circuits at high frequency. As we see, in (15), Δ (s ) is variable (because 2gm RE R −1 ~ P(s) = [1+ ( )( )](− C )(1+ sRECE ) s 4RE (1+ ) 1+ sRECE p1 (14) So equation (14) is the transfer function of the differential pair which shown in Fig.2. In (14), we define of uncertainty of p1 ) stable transfer function Δ( s) = is fixed stable transfer function. Also in (17), P (s ) is the nominal transfer function with no −1 , s (1 + ) p1 2 g m RE . W (s) = 1 + sRE C E with Δ( s ) (15) ~ ~ replacing (15), (16), (17) in (14), P ( s ) is become (16) ~ as the form P ( s ) = [1 + Δ ( s )W ( s )]P ( s ) ,which is the same equation (1) and as we saw, is multiplicative perturbed transfer function. Thus, the differential pair which shown in Fig. 2 or its equivalence SISO block diagram which shown in Fig.3 at high frequency, has multiplicative perturbed transfer function ~ P ( s ) which has been presented in (14) with Δ (s ) , W (s ) and P ( s ) which have been obtained in (15), (16) and (17), respectively . − RC ~ lim P ( jω ) = lim (1 + jωRE C E ) . ω →∞ 4 R E ω →∞ So at high frequency ( ω 〉〉 p1 ), we have RC (1 + sRE C E ) , 4 RE (17) ~ and P (s ) is the nominal transfer function of P ( s ) . Consider equation (8), in fact, at low frequency (ω ≤ 1 ), we have RE C E 4 Internal Stability Consider a basic feedback loop as shown in Fig. 5. The signals shown in Fig.5, have the following interpretations: r reference or command input y output and measured signal n sensor noise d1 , d 2 external disturbances ADM ( jω 〉〉 ACM ( jω , in the other word CMRR( jω ) = ADM ( jω 〉〉1 , ACM ( jω where CMRR ( jω ) is "Common Mode Reject x1 , x2 , x3 , x4 state variables P1 , P , P2 forward plants F feedback transfer function or the transfer 1 , the RE C E and CMRR ( jω ) is decreased with slope -20 dB Dec for ω 〉 p1 , the CMRR ( jω ) is decreased with slope Ratio". When ω is increased, at ω 〉 -40 dB Dec function of sensor. Note that, all of above signals and transfer functions are in Laplace domain. For example P1 = P1 ( s) , x1 = x1 ( s) and etc. In Fig.5, we have [1]. So for ω 〉〉 p1 , we have ISSN: 1790-5117 ≤ 1 and in (16), W (s ) (the weigh) canceled unstable pole in forming P ( s ) . By As we see in (14), P( s) = − ∞ 46 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS ⎧ x1 = r − Fx4 ⎪x = d + P x ⎪ 2 1 1 1 . ⎨ x d Px = + 3 2 2 ⎪ ⎪⎩ x4 = n + P2 x3 ( P1 PP2 F in Fig. 5). Then, at least one internal signal of this electronic circuit is unbounded. This means that, at least one signal is clipped. Consequently, this electronic circuit can not operate as linear circuit or amplifier (about the operating point). (18) 5 Robust Stability Suppose that, P (s ) in Fig. 5 is replaced by ~ P ( s ) which has uncertainty and belongs to a set such as M . This is shown in Fig. 6. Definition 2: The system shown in Fig. 6 has robust stability if it provides internal stability for ~ every P ( s ) ∈ M [12]. ~ Assume that, in Fig. 6 P ( s ) has uncertainty as multiplicative perturbation form. In other ~ word, P ( s ) is described with equation (1). Theorem 1: The closed loop system which shown in Fig. 6 has robust stability iff Fig. 5. Basic feedback loop. In matrix form these are ⎡ 1 ⎢− P ⎢ 1 ⎢ 0 ⎢ ⎣ 0 0 1 0 0 1 −P 0 P2 F ⎤⎛ x1 ⎞ ⎛r⎞ ⎜ ⎟ ⎜ ⎟ ⎥ 0 ⎥ ⎜ x2 ⎟ ⎜d ⎟ = ⎜ 1 ⎟. ⎟ ⎜ 0 ⎥ x3 d ⎜ 2⎟ ⎥⎜⎜ ⎟⎟ ⎟ ⎜ 1 ⎦ ⎝ x4 ⎠ ⎝n⎠ P1 PP2 ( s) 〈 1. 1 + P1 PP2 F ( s) ∞ Proof: Assume (⇐) W ( s) F ( s) (19) W ( s) F ( s) ⎡ 1 ⎢ P ⎢ 1 ⎢ P1 P ⎢ ⎣ P1 PP2 − PP2 F 1 P PP2 − P2 F − P1 P2 F 1 P2 − F ⎤⎛ r ⎞ ⎜ ⎟ − P1 F ⎥⎥⎜ d1 ⎟ − P1 PF ⎥⎜ d 2 ⎟ ⎥⎜ ⎟ 1 ⎦⎜⎝ n ⎟⎠ P1 PP2 ( s) 〈 1 . Construct the 1 + P1 PP2 F ( s) ∞ P2 ( s) in Re( s ) ≥ 0 . Fix an allowable Δ (s ) . Construct the Nyquist plot of ~ P1 P P2 F ( s ) = [1 + Δ ( s )W ( s )]L( s ) . P1 ( s) and (20) Definition 1: If the sixteen transfer functions in (20) are stable, then the feedback system which shown in Fig. 5 is said to be internally stable [12]. As a consequence, if the exogenous inputs are bounded in magnitude, so too are x1 , x2 , and x3 , and hence u , y and v . So, for an electronic circuit, internal stability is very important, because it guarantees bounded internal signal for all points of electronic circuit. Suppose that, a electronic circuit has BIBO stability but don’t have internal stability. In other word, there is at least one pole–zero cancellation in Re( s ) ≥ 0 when the closed loop transfer function of the electronic circuit is formed ISSN: 1790-5117 that Nyquist plot of L( s) = P1 PP2 F ( s) , indenting D to the left around poles on the imaginary axis. Since the nominal feedback system is internally stable, we know this form the Nyquist criterion: The Nyquist plot of L does not pass through -1 and its number of counterclockwise encirclements equals the number of poles of P (s ) in Re( s ) ≥ 0 plus the number of poles of Thus, the system is well-posed if and only if the above 4× 4 matrix is nonsingular, that is, the determinant 1+ P1 PP2 F is not identically equal to zero [12]. From (19), it follows that ⎛ x1 ⎞ ⎜ ⎟ 1 ⎜ x2 ⎟ ⎜ x ⎟ = 1 + P PP F 1 2 ⎜ 3⎟ ⎜x ⎟ ⎝ 4⎠ (21) No additional indentations are required since Δ ( s )W ( s ) introduces no additional imaginary axis poles. We have to show that the Nyquist plot of [1 + Δ ( s )W ( s )]L ( s ) does not pass through -1 and its number of counterclockwise encirclements equals the number poles of [1 + Δ ( s )W ( s )]P ( s ) in Re( s ) ≥ 0 plus the number of poles of P1 P2 F ( s) in Re( s ) ≥ 0 ; equivalently, the Nyquist plot of [1 + Δ ( s )W ( s )]L( s ) does not pass through -1 and encircles it exactly as many times as does the 47 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS Nyquist plot of L (s ) . We must show, in other words, that the perturbation does not change the number of encirclements. The key equation is 6 Robust Stability of the Circuit Consider an electronic circuit that its block diagram at high frequency shown in Fig. 7. As we see, this circuit contains a differential pair, which has one input signal and one output signal. As we obtained in (14), the transfer function of this differential pair at high frequency is ~ P ( s ) which has uncertainty as multiplicative perturbation with Δ (s ) and W ( s ) that were derived in (15) and (16). Also the nominal ~ transfer function of P ( s ) , is P ( s ) which was obtained in (17). 1+ [1+ Δ(s)W (s)]L(s) = [1+ L(s)][1+ Δ(s)W (s) P PP (s) ] F(s) 1 2 1+ P1 PP2 F(s) (22) Since Δ(s)W(s)F(s) P1PP2 (s) PPP(s) ≤ W(s)F(s) 1 2 〈1 1+ P1PP2F(s) ∞ 1+ P1PP2F(s) ∞ (23) P1 PP2 ( s ) always The point 1 + Δ ( s )W ( s ) F ( s ) 1 + P1 PP2 F ( s ) lies in some closed disk with center 1, radius 〈 1 , for all points s on D . Thus from (22), as s goes once around D , the net change in the angle of 1 + [1 + Δ ( s )W ( s )]L( s ) equals the net change in the angle of 1 + L ( s ) . This gives the desired result. P1 PP2 ( s) ≥ 1. 1 + P1 PP2 F ( s) ∞ We will construct an allowable Δ (s ) that destabilizes P1 PP2 ( s ) is the feedback system. Since F ( s ) 1 + P1 PP2 F ( s ) (⇒ ) Suppose, W ( s ) F ( s) Fig. 6. Basic feedback loop with uncertainty. In Fig. 7, P1 ( s) and P2 ( s) are the total transfer functions of all stages which are located before and after differential pair, respectively. Also F (s ) is the transfer function of negative feedback in circuit. Theorem 2: The electronic circuit which shown in Fig. 7, has robust stability with respect ~ to uncertainty of P ( s ) ( the transfer function of this differential pair) at high frequency, iff strictly proper, at some frequency ω , we have W ( jω ) F ( jω ) P1 PP2 ( jω ) = 1. 1 + P1 PP2 F ( jω ) Suppose that ω = 0 , then W ( 0) F ( 0 ) (24) P1 PP2 (0) is a real number, either +1 1 + P1 PP2 F (0) or -1. If Δ (0) = −W (0) F (0) P1 PP2 (0) , then Δ (0) 1 + P1 PP2 F (0) P1 ( s ) P2 ( s ) F ( s) 1 g m RC 〈1 R 2 1 − C (1 + sRE C E ) P1 ( s) P2 ( s ) F ( s) 4 RE ∞ is allowable and 1 + Δ (0)W (0) F (0) P1 PP2 (0) =0. 1 + P1 PP2 F (0) (25) From (22) with respect to (25), we conclude that, the Nyquist plot of [1 + Δ ( s )W ( s )]L ( s ) passes through the critical point, so the perturbed feedback system is not internally stable. ISSN: 1790-5117 (26) Proof: (⇔ ) As we see, the block diagram of electronic circuit which shown in Fig. 7 and the block diagram of the closed loop system which shown in Fig. 6, both are the same. 48 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS g m = 0.01 Ω −1 , rπ 2 = rπ 3 = R = 3.76 β gm = 10 K and 10 K = 2.73 K . By replacing above values in (13), we obtain that, p1 ≈ 6.06 M rad sec . For BJT ( T1 ), we have I C = 1 mA , g mT = Fig. 7. Block diagram of electronic circuit at high frequency. rπ = β g mT 1 Ω −1 , 25 = 2 .5 K , R LT = 4.7 K 22 K 128 K = 3.76 K , In fact, the block with multiplicative perturbed ~ transfer function P ( s ) in Fig. 6 has been replaced by differential pair which has multiplicative perturbed ~ transfer function P ( s ) , as was obtained in (14) with Δ (s ) , W (s ) and P (s ) that were derived in (15) , (16) and (17). So, from theorem 1, by replacing (16) and (17) in (21), we have Rπ 0 = [(0.05K 68K 68 K ) + rb ] rπ = 0.31K and AV = − g mT .RLT = −1.27 . 1 + g mT (6.8 K 5.2 K ) By replacing above values in following equation [1]: − RC (1+ sRECE )P1 (s)P2 (s)F (s) 2gm RE 4RE 〈1 . (1+ sRECE ) 1− RC (1+ sR C )P (s)P (s)F (s) E E 1 2 4RE ∞ 1 p ≈ Rπ 0 {Cπ + Cμ [(1 + g mT RLT ) + (27) the domain pole of p ≈ 73 M rad It is not difficult to verify (26). This completes the proof. sec RLT RLT ]+ Ccs } Rπ 0 Rπ 0 (28) T1 , is computed as . Comparing simulated circuit in Fig. 8 and the block diagram in Fig. 7, it follows that Example 1: Consider the simulated circuit which shown in Fig. 8 with following parameters. C μ = Cπ = C cs = 0.25 pF , For BJT ( T1 ): β = 100 , rb ≈ 0 Ω and VT ≈ 25 mV . For BJTs ( T2 and T3 ): Cμ = Cπ = Ccs = 1 pF , β = 100 , rb ≈ 0 Ω and VT ≈ 25 mV . We have ( is the symbol of the parallel resistances) R S = 4 .7 K RC = 10 K 22 K 128 K = 3.76 K , 12 K = 5.45 K . For biasing of the BJTs ( T2 and T3 ), we have I CT = I CT = 0.25 mA . 2 3 So we have ISSN: 1790-5117 Fig.8. Simulated electronic circuit. 49 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS − 1.27 6 .8 K , F ( s) = ≈ 0.57 s 6 .8 K + 5 .2 K 1+ 73 × 10 6 and P2 ( s ) = 1 . P1 ( s ) = 1 P1 ( s ) P2 ( s ) F ( s ) 〈1 g m RC RC 2 1− (1 + sRE C E ) P1 ( s ) P2 ( s ) F ( s ) 4 RE ∞ and thus, simulated circuit has robust stability and also it is unstable for ω < 488.82M rad sec . The input of the simulated circuit is small signal sinusoidal form. As we see, the results of simulation which shown in Fig. 9, validate the presented proposal. Taking these and other values in (26), we have − 1.27 × 0.57 s ) 6 1 73 ×10 × 0.01× 5.45 × 103 5.45 − 1.27 2 [1 − (1 + 33 × 27 × 10−9 s) × 0.57] s 132 (1 + ) 73 × 106 (1 + sup 19.7263 1.0299 + (0.0403 × 10 −6 ω ) 2 2 So, for ω > 488.82 M rad sec = ∞ . >> p1 7 Conclusion In this paper, a group of electronic circuits which contain an emitter-coupled pair with one input and one output signal, has been considered. We showed that, the transfer function of this differential pair at high frequency has uncertainty. This uncertainty has been modeled as multiplicative perturbation. Base on this multiplicative uncertainty model, the necessary and sufficient condition for robust stability at high frequency has been presented in (26). This condition guarantees internal stability of electronic circuit with respect to the uncertainty of the transfer function of differential pair at high frequency. A direct benefit of the result in this paper is a condition which must be satisfied, when we design integrated circuits at high frequency. It is an interesting research topic to obtain similar conditions for other differential pairs which contain MOSFET and JFET. Acknowledgment: Implementation of the project has been done in Azad University (South Tehran Branch). References: [1] P. E. Gray, P. J. Hurst, S. H. Lewis and R. G. Meyer, Analysis and Design of Analog Integrated Circuits, New York: Wiley, 2001. [2] J. Millman and A. Grabel, Microelectronics, 2nd ed. New York: Mc Graw-Hill, 1988. [3] C. Yuhua, M. J. Deen and C. Chih-Hung, MOSFET modeling for RF IC design, IEEE Trans. Electron Devices, vol. 52, 2005, pp. 1286 - 1303. [4] M.N.A. Parlakci, Robust stability of uncertain time-varying state-delayed systems, Fig.9. The results of simulated circuit. a) ω = 666.7 M rad sec b) ω = 490.0 M rad sec c) ω = 266.7 M rad sec . ISSN: 1790-5117 50 ISBN: 978-960-474-096-3 Proceedings of the 13th WSEAS International Conference on CIRCUITS IEE Proc. Control Theory and Applications, vol. 153, 2006, pp. 469-477. [5] Sanqing Hu and Jun Wang, Global robust stability of a class of discrete-time interval neural networks, IEEE Trans. Circuit and Systems I, vol. 53, 2006, pp. 129-138. [6] A. Schmid and Y. Leblebici, Robust circuit and system design methodologies for nanometer-scale devices and single-electron transistors, IEEE Trans. Very Large Scale Integration (VLSI) Systems, vol. 12, 2004, pp. 1156 – 1166. [7] N. Ozcan and S. Arik, Global robust stability analysis of neural networks with multiple time delays, IEEE Trans. Circuit and Systems I, vol. 53, 2006, pp. 166 - 176. [8] L. Liu and J. Huang, Adaptive robust stabilization of output feedback systems with application to Chua's circuit, IEEE Trans. Circuit and Systems II, vol. 53, 2006, pp. 926-930. ISSN: 1790-5117 [9] A. Buonomo and A. Lo Schiavo, Analysis of emitter (source)-coupled multivibrators, IEEE Trans. Circuit and Systems I, vol. 53, 2006, pp. 1193-1202. [10] A. Ghulchak and A. Rantzer, Robust control under parametric uncertainty via primal-dual convex analysis, IEEE Trans. Automatic Cont., vol. 47, 2002, pp. 632-636. [11] S. Dasgupta, P. J. Parker, B. D. O. Anderson, F. J. Kraus and M. Mansour, Frequency domain conditions for the robust stability of linear and nonlinear dynamical systems, IEEE Trans. Circuit and Systems, vol. 38, 1991, pp. 389 - 397. [12] J. C. Doyle, B. A. Francis and A. R. Tannenbaum, Feedback Control Theory, New York:Macmillan,1992. 51 ISBN: 978-960-474-096-3