142 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO.

1, JANUARY 2019

On the Misunderstanding of the

Ziegler-Nichols’s Formulae Usage
Alexandro Garro Brito

Abstract—In the Ziegler-Nichols’s method of reaction curve, case is not discussed herein. The second experiment involves
the proportional gain should be calculated as an inverse relation the plant response to a step input. Based on measurements
of the plant steady-state gain. One of the reasons behind this from an S-shaped response – also known as “reaction curve”
is to avoid an excessively high loop gain, which can jeopardize
many required characteristics of the closed loop. However, many – the controller gains are calculated.
reports, scientific papers and books have been neglecting such A key detail of this latter method is that the proportional
gain compensation in the tuning formulae. gain should be normalized by the plant steady-state gain.
This brief presents a comprehensive discussion about such This procedure avoids an excessively high loop gain, which
uncompensated tuning rules. The main paper finding is that
can affect many important characteristics of the closed loop.
either the stability margin or the disturbance rejection is reduced
in this case. A theoretical analysis is performed to obtain the main However, we had noted that many references consider a
result. Moreover, a consistent simulation study is also performed tuning formulation which do not include such plant gain
to show the impact of the lack of compensation on performance. compensation [2]−[15].
This brief proposes to study these uncompensated tuning
Index Terms—Controller tuning, process control, reaction rules from a technical point of view. The aim is to comprehend
curve. their effects on process control. It is demonstrated that such
strategy can conduct to inefficient or even unstable closed
I. I NTRODUCTION loops.
Particularly, it is shown that the stability margin and the
T HE Ziegler-Nichols’s methods [1] can be seen as a mile-
stone in process control. Due to their wide acceptance by
technical personnel and academy, these methods have attracted
disturbance rejection are strongly affected by the lack of
gain compensation. The conclusions are achieved by both
theoretical and simulation analysis. This study is useful to alert
increasing interest even nowadays. The main contribution of
technical personnel, students and scientists on the drawbacks
those authors was to provide a fast method for PID gain
related to such tuning approach.
tuning through some simple measurements from the plant.
The paper is structured as follows. Section II presents a
They aimed to answer some questions like “how can the
general description of the original ZN reaction curve method
proper controller adjustments be quickly determined on any
and the uncompensated formulation. Section III discusses
control application?”. Then, the objective was to discuss the
several examples from literature. Section IV analyzes the
controller tuning from a strictly practical point of view. To
effects of the lack of compensation from the theoretical point
perform such task, those authors analyzed the time response
of view. It also performs a deep simulation study, based on an
provided by several common plants. They proposed some
already well-established test batch. Finally, Section V draws
simple PID tuning rules, so that a proper closed loop behavior
the main paper conclusions.
is achieved. The approach was based on maintaining a stable
closed loop with a amplitude ratio around 25 % (between two
successive peaks of the oscillatory response). It also looks for II. T HE Z IEGLER -N ICHOLS ’ S R EACTION C URVE M ETHOD
a damped oscillation period around 1/8 from the respective
undamped closed loop. One of the experimental methods proposed by Ziegler and
Two experiments were proposed to obtain the PID gains Nichols for PID tuning involves the plant open loop step
[1]. The first uses a closed loop where the plant is fed back response. It assumes an S-shaped response – also known as
with a proportional controller, which is adjusted for sustained “reaction curve” – similar to Fig. 1. Firstly, one obtains the
oscillations. According to the period of the oscillations and maximum response rate R (reaction rate), which occurs at the
the gain to achieve them, the PID gains can be obtained. This curve inflection point. By taking a tangent line to this point,
two values can be graphically picked: 1) the intersection with
Manuscript received August 2, 2018; accepted September 25, 2018. Rec-
ommended by Associate Editor Qinmin Yang. the time axis (L), and 2) the intersection with the amplitude
Citation: A. G. Brito, “On the misunderstanding of the Ziegler-Nichols’s axis (−RL). The authors argued that L can represent all the
formulae usage,” IEEE/CAA J. Autom. Sinica, vol. 6, no. 1, pp. 142−147, possible time delays (lags) contained in the system.
Jan. 2019.
A. G. Brito is with the Federal University of Santa Catarina, Rua Another version considers non-integrative systems, which
Dona Francisca 8300, Joinville/SC, 89.218-035, Brazil (e-mail: alexan- the step response converges to the plant steady-state gain K
dro.brito@ufsc.br). (Fig. 1). Here, a line is traced at the gain level. The intersection
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. between this line and the tangent line occurs at a time L + T .
Digital Object Identifier 10.1109/JAS.2019.1911336 The interval T is assumed to be the approximate time constant
BRITO : ON THE MISUNDERSTANDING OF THE ZIEGLER-NICHOLS’S FORMULAE USAGE 143

TABLE II
of the system. Then, a first-order plus time delay (FOPTD)
T UNING RULES FOR C ONTROLLER D ESIGN W ITHOUT P LANT
model can be obtained as:
G AIN C OMPENSATION

Ke−Ls Kp Ti Td
Gp (s) = . (1) P controller T
∞ 0
Ts + 1 L
PI controller 0.9 T
L
L
0.3
0
PID controller 1.2 T
L
2L 0.5L

III. E XAMPLES F ROM L ITERATURE

This section presents some references which the tuning
rules in Table II were applied. Our objective is not to discuss
the validity of the results therein. We only aim to show
the influence of those tuning rules on the control system
performance. When it was possible, some simulation analysis
is performed.
A PI controller was tuned for a Luo-converter [4]. The
authors found the following values for the plant: K = 2,
T = 0.0222 s and L = 0.0167 s. Based on these values, they
tuned a PI controller as Kp = 2.89 and Ti = 53.59 s. However,
such FOPTD model should conduct to a PI controller with
gains Kp = 1.2 and Ti = 0.056 s by Table II. Then, the paper
adopted a proportional gain that is four times bigger than the
Fig. 1. Reaction curve for dynamic systems. The Ziegler-Nichols’s tuning
ZN original proposition. The integral time is also high. Even
rules are based on the parameters K, L, T and R.
if Table II was strictly applied, the proportional gain would
be the double from the original ZN method. By analyzing the
The original ZN tuning rules considered R and L values. paper results, one can observe a very high closed-loop percent
However, one can also use K, L and T values since R = K/T . overshoot.
This is a preferred format nowadays because it presents the In [3] the performance of a particle-swarm optimized con-
complete information about the FOPTD model. Based on their troller was studied for a DC motor. A PID controller – tuned
analysis, Ziegler and Nichols proposed controller tuning rules through Table II – was used for the sake of comparison. The
for the proportional action, the integral action (there named DC motor model presents a steady-state gain around 41. As the
automatic-reset) and the derivative action (there named pre- plant gain was not compensated, this resulted in a considerably
act). The PID controller then assumes the format: high control loop gain. The consequence was a closed-loop
response with percent overshoot of 69 % and a settling time
· ¸ of 2.3 s. These values are much bigger than those obtained by
1
C(s) = Kp 1 + + Td s . (2) the particle-swarm optimization. A better comparison would
Ti s
probably be assessed if the original ZN tuning rules was used
instead.
The ZN tuning rules are summarized in Table I. A PI controller was also applied to a HVAC system [4].
TABLE I
For a typical plant, the FOPTD parameters were obtained as
O RIGINAL Z IEGLER -N ICHOLS ’ S T UNING RULES FOR P, PI AND
K = 14.3, L = 60 s and T = 150 s. The tuning rule in Table
PID C ONTROLLERS
II was applied. This should result in PI gains Kp = 2.25,
Ti = 200 s. However, we have noted that different controller
Kp Ti Td gains were used in the paper. If Table II was strictly used,
T
P controller KL
∞ 0 the closed loop would be unstable. In contrast to this, a stable
T L
PI controller 0.9 KL 0.3
0 closed (27 % of overshoot, settling time of 700 s) is achieved
T
PID controller 1.2 KL 2L 0.5L with the original ZN formulation.
A PID controller was calculated for an experimental batch
The proportional gain is normalized by the plant steady-state distillation column [5]. The FOPDT parameters were K =
gain K in Table I. The reason behind this is to avoid an undue 0.055, L = 1380 s and T = 2200 s. The authors used the
high loop gain, which can give poor closed loop performance. following PID gains: Kp = 1.93, Ti = 2760 s and Td = 690 s.
Some recent reports, papers and textbooks have been apply- The question here is the very low steady-state gain of the
ing the Ziegler-Nichols’s method by neglecting the correction plant. Such plant gain could be compensated to provide a faster
term on Kp . These uncompensated tuning rules are shown in closed loop response. In fact, our simulation showed that the
Table II. The closed loop performance for such uncompensated system would be 11 times faster if the original ZN tuning was
tuning rules is under study in this paper. The lack of gain applied.
compensation can provide a very poor behavior, as we show A PI controller was proposed for a negative hydrogen ion
herein. source for a particle accelerator [6]. By applying the reaction
144 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019

curve method, the authors obtained L = 1.54 s, T = 5.27 s examples:

and K = 0.2. The controller parameters were calculated as 1
G(s) = . (3)
Kp = 3.08 and Ti = 5.13 s by Table II. For the original ZN (s + 1)(s + 4)
method, Kp = 15.4. We used the respective FOPTD model The FOPTD modeling resulted in K = 0.25, L = 0.125s
to compare the closed-loop behavior of these tuning methods. and T = 1.59 s. These values were used to obtain two sets of
The achieved responses appeared considerably slower when tuned PID gains: by Table I (Kp = 61.107, Ti = 0.2494 s and
compared to the results presented by the authors. Table II Td = 0.0623 s) and by Table II (Kp = 15.28). Fig. 3 shows
provides a non-oscillatory slow closed-loop response (settling the results for both closed loops.
time of 27 s). The original ZN formulae achieved a faster
convergence (settling time of 14 s) with a considerably higher
overshoot (40 %). This reinforces the clear discrepancy be-
tween the tuning rules in Tables I and II.
In another recent paper, a PID controller was tuned for
a precision planter [7]. The open loop model has a steady-
state gain of 35 rev/min. A limiter was introduced to maintain
the command duty-cycle between 0 and 100 %. The FOPTD
modeling provided L = 0.021 s and T = 0.118 s. By
using Table II, the following PID controller was achieved:
Kp = 6.743, Ki = 160.548 and Kd = 0.071. However, the
controller gains would be Kp = 0.1927, Ki = 4.587 and
Kd = 0.02 by the original ZN method (Table I). We performed
closed loop simulations with the same conditions and block
diagram proposed in the paper (Fig. 2).

Fig. 3. PID control of the system in Eq. (3). The original ZN tuning rules
provided a faster and less oscillatory response than Table II.

We have observed that the controller based on Table I

is considerably superior in terms of performance. We have
also noted that the original ZN method is as fast as the
optimized NMS method proposed in the paper (although a
higher overshoot occurred for the ZN method).
Some other papers used the uncompensated formulation in
Table II for either control design or comparison purposes. We
can cite as examples: the control of a giant magnetostrictive
actuator [9]; the performance comparison with a flow fuzzy
controller [10]; the control of an electric power-flow [11]; and
the load frequency control in power systems [12]. An analysis
was not possible for these cases. However, we are confident
Fig. 2. PID closed-loop response for a precision planter. The original ZN that similar conclusions can be also addressed to them.
(OZN – Table I ) and uncompensated gain (NCG – Table II) tuning rules are
compared. Notice that command saturation occurred for Table II. IV. A D EEPER A NALYSIS ON THE C LOSED L OOP
P ERFORMANCE
An important feature from a controller tuning method is
We observed that the uncompensated methodology provided
to provide a specified performance for processes with similar
a slightly faster response than ZN. But this occurred due
dynamics. The controller gains are adjusted according to the
to PWM duty-cycle saturation during a part of the system
plant-model parameters so that such performance is achieved.
operation. In other words, the uncompensated closed loop is
For a PID controller, a common tuning practice is to
not working at the linear operation mode. On the other hand,
normalize the proportional gain as an inverse relation of the
Fig. 2 shows that the maximum duty-cycle is around 30 %
plant gain. This is true for almost every PID tuning method
when the original ZN approach is used. This is advantageous
based on reaction process [18]. The reason behind this is
because it permits that a higher planter speed can be achieved
to avoid an excessively high loop gain. By considering the
– before occurring a switching operation.
Ziegler-Nichols’s case, the 25 % amplitude ratio criterion is
A PID tuning method based on neutrosophic similarity mea- not possible if such compensation does not occur.
sure (NSM) was proposed [8]. The closed loop performance Important impacts on performance are noted when Table II
provided by NMS method was there compared to Table II. is applied. The main objective of this section is to present a
We discuss here the results concerning one of the simulation profound analysis about this issue.
BRITO : ON THE MISUNDERSTANDING OF THE ZIEGLER-NICHOLS’S FORMULAE USAGE 145

A. Theoretical Analysis

Some relations were derived for phase and gain margins by

considering a FOPTD plant, a PI controller, and well-known
tuning formulae [16]. The following equations apply to the
Ziegler-Nichols’s tuning method:
µ ¶
π ωg L
φm = + tan−1 − tan−1 (ωg T ) − ωg L (4)
2 0.3
s
ωp L2 ωp2 T 2 + 1
Am = (5)
0.9αT ωp2 L2 + 0.32
q ¡ ¢
π + π 2 − 4πL 0.3 L − T
1
ωp ≈ (6)
v 4L
u ³ p ´
u 0.3 β + (−β)2 + 3.24²2 Fig. 4. Phase margin for different values of plant gain (K) and a PI controller
t
ωg = (7) based on Table II. The increasing of K results in a reduction of the stability
2LT 2 margin. The closed loop becomes unstable for any L/T ratio if K ≥ 3.
à µ ¶2 !
L Tα
β= 0.81 −1 (8)
0.3 L
αT 2
²= (9)
L

where φm and Am are phase margin and gain margin respec-

Fig. 5. System closed loop with disturbance input.
tively. ωg and ωp are the frequencies which such margins oc-
cur. The parameter α depends on the plant gain compensation
Define C(s) = Kp C̃(s) and Gp (s) = K G̃p (s). This proce-
by the controller, according to:
dure is done to isolate the gain in both the controller and the
½ plant. Then, the sensitivity function can be defined as
1, for Table I
α= (10) 1
K, for Table II. S(s) = (12)
1 + αγ C̃(s)G̃(s)
A known result is that the stability margins vary consider- where α is given by (10), γ = δT /L, and δ = 1, 0.9, 1.2 for
ably as function of L/T for the original ZN method. In fact, a P, PI and PID design respectively. The sensitivity function
1.7 ≤ Am ≤ 2.3 and 23◦ ≤ φm ≤ 93◦ when 0.1 ≤ L/T ≤ 1 represents the transfer function between the disturbance signal
for the PI case [16]. Similar fact occurs for the PID case [17]. D(s) and the output Y (s). By taking the modulus of S(s) and
Let us analyze how the stability margins are affected by the the triangle inequality, one obtains
lack of plant gain compensation. Define Am−C and Am−N C ³ ¯ ¯´−1
¯ ¯
the gain margin for the tuning rules in Tables I and II |S(s)| ≥ 1 + αγ ¯C̃(s)G̃(s)¯ . (13)
respectively. By applying (5), it is straightforward to note that:
Assume SC (s) and SN C (s) the sensitivity functions when
Am−C Tables I and II are respectively applied. For a plant with
Am−N C = . (11) steady-state gain K < 1, (13) gives |SN C (s)| > |SC (s)|. In
K
other words, the disturbance rejection of the uncompensated
Then, the gain margin of the uncompensated rules (Table II) tuning rules are reduced in comparison to the original ZN
is a fraction of the compensated rules (Table I ) when K > 1. method.
In other words, the gain margin can be considerably reduced The conclusion is that either stability or disturbance rejec-
if the uncompensated tuning rules are used. tion can be jeopardized if Table II is used. The lack of gain
compensation affects dramatically the system performance.
The phase margin can be obtained by the computation of
The closed loop can even become unstable for moderate values
(4), (7) and (10) as function of L/T . Fig. 4 presents the results
of K. Because of this, such tuning rules should be avoided.
when Table II is applied and 0.7 ≤ K ≤ 3. Notice that the
phase margin is considerably affected by K. In addition, the
closed loop becomes unstable when K ≥ 3 for any value L/T . B. Simulation Results
The lack of plant gain compensation can also affect the A comparative analysis on the closed loop performance for
disturbance rejection. This is an important fact, since ZN both tuning strategies (Tables I and II) is now performed. This
tuning method also aims at a proper disturbance rejection [1]. study was inspired by the test batch method presented in [19].
Consider a feedback control system (Fig. 5), with plant and Three generic plants, which are very representative of common
controller given by (1) and (2) respectively. industrial processes, were used:
146 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 6, NO. 1, JANUARY 2019

Fig. 6. Simulation results for the batch test in Eqs. 14 to 16. The circle-marked curves represent closed loop with PID controllers designed through the
original Ziegler-Nichols’s tuning rules (Table I). The cross-marked curves show the results for the uncompensated tuning rules (Table II). Each curve represents
the closed loop for a specific plant gain. Notice that Table II provides closed-loop responses that are strongly influenced by the plant gain.

Ke−s P1 closed loop is unstable if K ≥ 2 and τ ≥ 0.77. P6 (s)

P1 (s) = (14)
T1 s + 1 conducts to an unstable system if K ≥ 2 and τ ≤ 0.3. This
K is a consequence of the uncompensated gain, which produces
P3 (s) = (15)
(s + 1)(T3 s + 1)2 an undesired high loop gain. Such fact was already concluded
Ke−L6 s by the theoretical analysis in Section IV-A.
P6 (s) = (16) Fig. 6 shows that the overshoot, settling time and ITAE are
s(T6 s + 1)
insensible to the plant gain variation when the original ZN
K = 0.01, 0.05, 0.1, 0.2, 0.4, 0.8, 1, 1.2, 2, 4, 8, 15, 30. method is used. This is a quite obvious result since the gain
The model values are the same from [19]. The gain K compensation in Table I. However, Table II provided closed
was introduced to test the performance dependence with the loops that are very sensible with the plant gain variation.
plant steady-state gain. The real values of time delay (L) Finally, we noted a very depreciated performance of closed
and time constant (T ) were obtained numerically for each loop when the plant gain is not compensated. This tuning
plant, according to Section II. The analysis was performed by strategy can provide a lower overshoot and settling time for
simulating the closed loop step response with two individual some values of K (specially for K < 1). However, very poor
controllers: the original ZN PID (Table I) and the uncom- behavior is obtained for many plant configurations (mainly for
pensated PID (Table II). A second order Padé approximation K À 1). This can also be seen from the ITAE curves. Notice
was used to represent the time delay for P1 (s) and P6 (s). the ITAE index provided by Table II is considerably higher
The closed loop behavior was analyzed through the percent than those provided by the original ZN formulae.
overshoot, the settling time and ITAE index (integral of time
multiplied by absolute error). The results are summarized in V. C ONCLUSIONS
Fig. 6, where the criteria above are shown as function of The Ziegler-Nichols’s method is a paradigm for PID con-
τ = L/(L + T ), 0 < τ ≤ 1. troller tuning still nowadays. This influential aspect is due to its
The first conclusion is that all the plants in (14) to (16) wide application in process control along the decades. Because
produced stable closed loop when the original ZN tuning of this, several reports and scientific papers keep studying
method is applied. This is an expected result since the strategy improvements on the ZN technique. Other references apply the
proposed by Ziegler and Nichols. On the other hand, several ZN method as a performance metric for new developments.
closed loops become unstable when Table II is used. In fact, However, several references applied a PID tuning that does
BRITO : ON THE MISUNDERSTANDING OF THE ZIEGLER-NICHOLS’S FORMULAE USAGE 147

not consider the plant gain compensation (Table II). This fact algorithm”, Computers and Electronics in Agriculture, vol. 136, no. 15,
directly affects the closed loop performance. pp. 184−192, Apr. 2017.
This paper presented a comprehensive analysis on the ef- [8] M. Can and O. Ozguven, “PID Tuning with Neutrosophic Similarity
Measure”, Int. J. Fuzzy Systems, vol. 19, no. 2, pp. 489−503, Feb. 2017.
fects of such lack of compensation in control loops. The study
[9] H. Lui, Z. Jia, F. Wang and F. Zong, “Research on the constant output
comprises some examples taken from the literature. A con- force control system for giant magnetostrictive actuator disturbed by
sistent theoretical and simulation analysis is also performed. external force”, Mechatronics, vol. 22, no. 7, pp. 911−922, Oct. 2012.
Then, useful information was obtained, intending to answer [10] P. Solatian, S. Abbasi and F. Shabaninia, “Simulation study of flow
some questions related to such strategy. control based on PID ANFIS controller for non-linear process plants”,
The main paper finding is that either the stability margin American Journal of Intelligent Systems, vol. 2, no. 5, pp. 104−110,
Oct. 2012
or disturbance rejection is reduced if Table II is applied.
[11] P. Kasinathan, R. Vairamani and S. Sundramoorthy, “Dynamic perfor-
Moreover, the response overshoot, settling time and other mance investigation of d model with PID controller-based unified power-
performance criteria are also influenced by the gain lack of flow controller”, IET Power Electronics, vol. 6, no. 5, pp. 843−850, Jul.
compensation. This contrasts to the originally proposed ZN 2013.
method, where the performance is basically insensible to the [12] N. Bahgaat, M. El-Sayed, M. Hassan and F. Bendary, “Load frequency
control in power system via improving PID controller based on par-
plant gain. ticle swarm optimization and ANFIS techniques.” Intern. Journ. Syst.
Some further studies are possible based on the conclusions Dynamics App., vol. 3, no. 3, pp. 1−24, Jul. 2014.
from this paper. An example of future work is to assess the [13] K. Ogata. Modern Control Engineering (4th edition). Upper Saddle
effects of lack of compensation on other tuning methods. River, NJ, USA: Prentice Hall, 2001.
With this paper, we hope to provide useful information for [14] M. Srivastava, M.C. Srivastava and S. Bhatnagar. Control Systems. New
technical personnel, control students and scientists. Our results Delhi, India: Tata McGraW-Hill, 2009.
show that the tuning rules in Table II should be avoided. The [15] S. Konami and T. Nishiumi. Hydraulic Control Systems: Theory and
Practice. Toh Tuck Link, Singapore: World Scientific Publ., 2017.
reason is the low control performance that it can provide.
[16] W. Ho, C. Hang, J. Zhou, “Performance and gain and phase margins of
well-known PI tuning formulas”, IEEE Transactions on Control Systems
R EFERENCES Technology, vol. 3, no. 2, pp. 245−248, Jun. 1995.
[17] W. Ho, O. Gan, E. Tay, E. Ang, “Performance and gain and phase
[1] J. Ziegler and N. Nichols, “Optimum settings for automatic controllers”, margins of well-known PID tuning formulas”, IEEE Transactions on
Transaction of ASME, vol. 64, no. 11, pp. 759−765, Nov. 1942. Control Systems Technology, vol. 4, no. 4, pp. 473−477, Jul. 1996.
[2] T. Narmadha, J. Velu and T.Sudhakar, “Comparison of Performance [18] A. O’Dwyer. Handbook of PI and PID Controller Tuning Rules (3rd
Measures for PV based Super-Lift Luo-Converter using Hybrid Con- edition), London, UK: Imperial College Press, 2009.
troller with Conventional Controller”, Indian Journal of Science and
Technology, vol. 9, no. 29, Aug. 2016. [19] K. Aström and T. Hagglund, “Revisiting the Ziegler-Nichols step re-
sponse method for PID control”, Journal of Process Control, vol. 14,
[3] A. Myrtellari, P. Marango and M. Gjonaj, “Analysis and Performance no. 6, pp. 635−650, Jan. 2004.
of Linear Quadratic Regulator and PSO algorithm in optimal control of
DC motor”, Intern. Journ. Latest Research in Engin. and Tech., vol. 2,
no. 4, Apr. 2016.
[4] T. Salsbury, and B. Chien, “A new sequencer controller for multistage
systems of known relative capacities”, HVAC&R Research, vol. 8, no.
4, pp. 403−428, Oct. 2002. Alexandro Garro Brito received the B.S. and M.S.
[5] N. Habobi and S. Yaseen, “Dynamics, Simulation, and Control of a degrees from the Federal University of Minas Gerais,
Batch Distillation Column using Labview”, Intern. Jour. Current Eng. Brazil, in 2003 and 2004 respectively. He also
and Tech., vol. 6, no. 1, Feb. 2016. received the Ph.D. degree from the Technological
Institute of Aeronautics, Brazil, in 2013. He is cur-
[6] I. Hong, Y. Song and Y. Cho, “A control system of a negative hydrogen
rently an Adjunct Professor at the Federal University
ion sources by using an EPICS-based proportional-integral-derivative”,
of Santa Catarina, Brazil. He is a member of the
Review of Scientific Instruments, vol. 79(02C706), pp. 1−3, Feb. 2008.
Brazilian Society of Automatic Control. His research
[7] X. Hea, T. Cuia, D .Zhanga, J. Weib, M. Wanga, Y. Yua, Q. Liua, interests include robust control, networked control
B. Yana, D. Zhaoa and L. Yanga. “Development of an electric-driven systems, and vehicular engineering.
control system for a precision planter based on a closed-loop PID