Sunday 12 August 2012

Total Quality Management in Supply Chain


International Business Research
Total Quality Management in Supply Chain  
Guangshu Chang  Zhengzhou Institute of Aeronautical Industry Management 

Abstract
Since 1980’s, the competition between enterprises has become the one between supply chains. Therefore, the  implementation of total quality management (TQM) in supply chain system but not only in enterprise has become an  exquisite premise for the survival of enterprise. This paper discussed the application of the eight modern TQM  principles of ISO9000 in supply chain quality management, namely customer focus, leadership, involvement of people,  process management, system management, continual improvement, factual approach to decision-making, and mutually  beneficial supplier relationships.

Keywords: Supply Chain, Total Quality Management, ISO9000 

Introduction  
In nowadays, the core ideas of TQM set forth by W. Edwards Deming, Joseph Juran, and Kaoru Ishikawa gained  significant acceptance and has become something of a social movement. The series standards of ISO9000 are  implementing in many industries, such as manufacturing, service, health care, nonprofit organizations, educational  institutions, even public bureaucracies. In the introduction of Quality Management System of ISO9000:2000, eight  principles of TQM are proposed, namely customer focus, leadership, involvement of people, process management,  system management, continual improvement, factual approach to decision making, and mutually beneficial supplier  relationship. The eight principles generalize the success experience of the advanced enterprises in the developed  countries.  In the current buyer’s market with global hard competition, enterprises cannot respond rapidly to the customers’  demand through traditional operation mechanism. Thereupon, a kind of new operation mechanism, i.e. supply chain  management, emerges as the times require [2]. In supply chain circumstance, the majority of enterprises, especially  some excellent enterprises, rely on their suppliers more and more heavily. The product quality and manufacturing  process of suppliers has great effect on the quality of final product of core enterprise. It means that the emphasis of  research and practice of TQM has transferred from enterprise focus to supply chain focus. Not only the high quality of  product and service but also the high level of quality control of the whole supply chain system ensures the competition  advance.Up to now, researchers has studied some related problems of quality management in supply chain. For example, Noori  investigated the implementation of continuous collaborative improvement activities in the supply chains of Canadian  industries, including the automotive, electronics and aerospace sectors.

2. Supply chain quality management based on the TQM principles

2.1 Customer focus
Customer focus is the core principle and idea of TQM because quality effort comes of customer’s needs and ends with  customer’s acceptance. In supply chain circumstance, customer includes not only the end user but also many in-between  users, such as suppliers, manufacturers, sellers, etc. However, more than half of the quality problems in supply chain are  resulted by specifications because of the inadequate communications between the members of supply chain. In many  cases, the procurement specifications released by buyers are equivocal while suppliers dare not to argue against buyers  on the specifications in the bidding process [3]. Therefore, the core enterprise must pay attention to the needs and  expectation of end users, and all the members of supply chain must pay attention to the needs and expectation of their  backward users. The needs and expectation of end users should be deployed layer upon layer in the whole supply chain  system. The end users will satisfy if all the member of supply chain can satisfy the needs of their backward users.  Moreover, the operation efficiency of supply chain system can be improved through the satisfaction level of the end  users. In supply chain quality management, some traditional tools of TQM are also effective. For example, we can use  Quality Function Deployment (QFD) to identify the distinct and potential needs and preferences of users, use Fishbone  Chart to investigate the factors affecting the satisfaction level of users and then use Pareto Chart to find out the key  factors.

2.2 Leadership
The effective of quality management depend on the effective of leadership because quality effort can get actual effect  only with the recognition and support of the leadership. In supply chain circumstance, the core enterprise play as the  leadership since it establishes the development strategy and operation targets of supply chain affect the actual efficiency  and effectiveness of the quality effort of all the other members. Therefore, the core enterprise must act as leadership to  consider adequately the needs and expectation of the other members, establish a clear, realizable and coincident holistic  target, and then lead and inspire the other members to strive jointly for the target. At the same time, the core enterprise  should foster more leaders of TQM in each layer of supply chain and make them take their responsibility zealously.

2.3 Involvement of people
The exertion of enthusiasm and creativity of all the employees is the precondition of the actual effect of quality  management. In supply chain circumstance, an up-and-coming excelsior work atmosphere should be established to  inspire the enthusiasm and creativity of the employees of all the members. Each employee should understand his/her  role and responsibility in the supply chain system, solve the problems forwardly as mastership, and learn the principles,  skills and technologies of TQM and ISO9000. Here, we can foster the ethos of self-motion and self-knowledge in  supply chain through 5S, i.e. seiri, seiton, seiso, seiketsu, and shitshke. Furthermore, we can make all the employees  participate into supply chain quality management and strive for the satisfaction of users jointly through the  establishment of QC teams that cross function or even enterprise.

2.4 Process management
 The focus of modern quality view is the process quality management but not the product itself of traditional quality  view. It is the requirement of the quality management system of ISO9004:2000 and the essential difference of modern  and traditional quality view. In each step of supply chain, there are many correlative processes, such as procurement,  logistics, production, inventory, selling, service, etc. These processes have their own independent objectives and  programs. There are usually conflicts among the objectives and programs. Therefore, the processes and their mutual  effects should be identified and managed to ensure the harmonious operation of supply chain. Then, all the processes,  especially the key processes, can realize high quality, i.e. small variation, small waste, and more increment, through the  continuous improvement and total quality control in all the nodes of supply chain system, as shown in Figure 1.  Insert Figure 1 Here.

2.5 System management  
The application of system approach in quality management is to view the quality management system as a big and  holistic system, identify and manage the sub-systems respectively. Then, the coordinated effect and mutual promotion  among the sub-systems will make the whole effect greater than the sum of the improvement of each sub-system and  improve the validity and efficiency of the realization of final targets [8]. In supply chain circumstance, enterprise should  confirm the mutual dependence relationship among the processes in supply chain system, break the boundary among  supply chain members, construct and integrate the processes in supply chain system. Then, many well operation  sub-systems can be constructed to collocate the resources rationally among the sub-systems. Finally, the whole supply  Vol. 2, No. 2 International Business Research  84  chain system, including supply, transport, production, distribution, inventory, etc., can realize the target and policy of  quality through the optimal operation mode.

2.6 Continual improvement 
Continual improvement is one of the focuses of modern quality research and practice. Enterprise must improve the  quality of product and service continually and reduce the cost to make customer satisfactory. In supply chain  circumstance, the pressure of continual improvement is more and more pressing because the market competition is more  and more hard. Not only the core enterprise but also the other members, such as suppliers, sellers, and logistics  providers, must improve their product and service respectively so as to construct the continual improvement of products  and services all over the supply chain process. Then, the continual, stable and harmonious ability of quality assurance  can be established. Furthermore, the core enterprise and other members must find the ways and practices improving  performance in or out of supply chain through benchmarking to make the continual improvement speed fast than the  one of rivals. However, it is ironical that one of the most important reason in the predicament of Xerox, which initiated  benchmarking practices, was exactly its slow reaction with the fast changing environment.

2.7 Factual approach to decision making
The sufficient and adequate data and information is the foundation of making right and effective decisions. Up to now,  many enterprises have began to collect and deal with all kinds of data and information by utilizing many advanced  information technology, e.g., EDI, MRP??, ERP, POS, Intranet/Extranet/Internet, so as to provide foundation for  making effective decision. In supply chain circumstance, enterprise should collect data and information of not only  itself but also the other members of supply chain to record and analyze the current operation situation of each member.  Therefore, the potential problems in any step of supply chain can be found duly according to the results of data analysis.  Then, the corresponding correct and timely decision can be made to avoid or rectify the problem.

2.8 Mutually beneficial supplier relationships  
What impact can suppliers have in achieving quality?
TQM authorities recommend that organizations work directly  with raw material suppliers to ensure that their materials are of the highest quality possible. Currently, at  least 50 percent of TQM organizations collaborate with their suppliers in some way to increase the quality of  component parts. Often these organizations send out “quality action teams” to consult with their major suppliers.  The objective is to help suppliers use TQM to analyze and improve their work processes . Suppliers can contribute  to quality in a number of other ways.  Therefore, the organization and its supplier are mutually dependent. Maintaining the mutually beneficial relationships  between them can improve the ability of creating value both of them. In supply chain circumstance, the product quality  is performed and ensured by all the members of supply chain because the production, sales and service process must be  performed by all the members.

3. Concluding remarks  The series standards of ISO9000 are made for the standardization of quality management and quality assurance.  Therefore, in supply chain circumstance, the implementation of ISO9000 is the basic assurance for an enterprise to  provide acceptable product or service and improve the quality level in a certain supply chain. The application of the  eight modern TQM principles of ISO9000 in supply chain quality management will promote the improvement of  operation efficiency and competition ability of the whole supply chain system. 

Friday 10 August 2012

Design Of Chip Ceramic Capacitor


Design of Chip Ceramic Capacitor

Ceramic Capacitor Basics
  • A capacitor is an electrical device that stores energy in the   electric field between a pair of closely spaced plates
  • Capacitors are used as energy-storage devices, and can also be used to differentiate between high-frequency and low-frequency signals. This makes them useful in electronic filters
  • Capacitance Value: Measure of how much charge a capacitor can store at a certain voltage
  • MLCC: Multilayer Ceramic Chip Capacitor L-ayers of ceramic and metal are alternated to make a multilayer chip
Process of Making Capacitor:

he process of making ceramic capacitors involves many steps.

Mixing: Ceramic powder is mixed with binder and solvents to create the slurry, this makes it easy to process the material.

Tape Casting: The slurry is poured onto conveyor belt inside a drying oven, resulting in the dry ceramic tape. This is then cut into square pieces called sheets. The thickness of the sheet determines the voltage rating of the capacitor.

Screen Printing and Stacking: The electrode ink is made from a metal powder that is mixed with solvents and ceramic material to make the electrode ink. The electrodes are now printed onto the ceramic sheets using a screen printing process. This is similar to a t-shirt printing process. After that the sheets are stacked to create a multilayer structure.

Lamination: Pressure is applied to the stack to fuse all the separate layers, this created a monolithic structure. This is called a bar.

Cutting: The bar is cut into all the separate capacitors. The parts are now in what is called a ‘green’ state. The smaller the size, the more parts there are in a bar.

The process of making ceramic capacitors involves many steps.

Mixing: Ceramic powder is mixed with binder and solvents to create the slurry, this makes it easy to process the material.

Tape Casting: The slurry is poured onto conveyor belt inside a drying oven, resulting in the dry ceramic tape. This is then cut into square pieces called sheets. The thickness of the sheet determines the voltage rating of the capacitor.

Screen Printing and Stacking: The electrode ink is made from a metal powder that is mixed with solvents and ceramic material to make the electrode ink. The electrodes are now printed onto the ceramic sheets using a screen printing process. This is similar to a t-shirt printing process. After that the sheets are stacked to create a multilayer structure.

Lamination: Pressure is applied to the stack to fuse all the separate layers, this created a monolithic structure. This is called a bar.

Cutting: The bar is cut into all the separate capacitors. The parts are now in what is called a ‘green’ state. The smaller the size, the more parts there are in a bar.


Firing: The parts are fired in kilns with slow moving conveyor belts. The temperature profile is very important to the characteristics of the capacitors.

Termination: The termination provides the first layer of electrical and mechanical connection to the capacitor. Metal powder is mixed with solvents and glass frit to create the termination ink. Each terminal of the capacitor is then dipped in the ink and the parts are fired in kilns.

Plating: Using an electroplating process, the termination is plated with a layer of nickel and then a layer of tin. The nickel is a barrier layer between the termination and the tin plating. The tin is used to prevent the nickel from oxidizing.

Testing: The parts are tested and sorted to their correct capacitance tolerances.
At this point the capacitor manufacturing is complete. The parts could be packaged on tape and reel after this process or shipped as bulk.

Types of Material Systems Used to make capacitors
There are two material systems used today to make ceramic capacitors: Precious Metal Electrode and Base Metal Electrode. The precious metal system is the older technology and uses palladium silver electrodes, silver termination, then nickel and tin plating. Today this material system is mostly used on high voltage parts with a rating of 500V and higher. The base metal system is a newer technology and uses nickel electrodes, nickel or copper termination, and nickel and tin plating. This material system is typically used for parts with voltage ratings lower than 500VDC.

Precious Metal Vs Base Metal System



MLCC Basics
The capacitance value of a capacitor is determined by four factors. The number of layers in the part, the dielectric constant and the active area are all directly related to the capacitance value. The dielectric constant is determined by the ceramic material (NP0, X7R, X5R, or Y5V). The active area is just the overlap between two opposing electrodes.

The dielectric thickness is inversely related to the capacitance value, so the thicker the dielectric, the lower the capacitance value. This also determines the voltage rating of the part, with the thicker dielectric having a higher voltage rating that the thinner one. This is why the basic trade off in MLCCs is between voltage and capacitance.





Critical Specifications

Material
Dielectric Constant
% Capacitance Change
DF
NP0
15-100
<0.4% (-55 to 125C)
0.1%
X7R
2000-4000
+/-15% (-55 to 125C)
3.5%
Y5V
>16000
Up to 82% (-30 to 85C)
9.0%

  • Dissipation factor: % of energy wasted as heat in the capacitor
  • Dielectric Withstanding Voltage: Voltage above rating a capacitor can withstand for short periods of time
  • Insulation resistance: Relates to leakage current of the part (aka DC resistance)


The critical specifications of a capacitor are the dielectric constant, dissipation factor, dielectric withstanding voltage, and insulation resistance. Dielectric constant: this depends on the ceramic material used. The table shows different dielectrics and some of their specifications. As you can see NP0 has the lowest dielectric constant, followed by X7R which has a significantly higher constant, and Y5V which is higher still. This is why the capacitance values for X7R capacitors are much higher than NP0 capacitors, and Y5V has higher capacitance than X7R. The capacitance change vs temperature is very small for NP0 parts from -55C to 125C, and gets larger for X7R, then even larger for Y5V. So, the more capacitance a material provides, the lower the stability of capacitance over temperature. Dissipation Factor: this is the percentage of energy wasted as heat in the capacitor. As you can see, NP0 material is very efficient, followed by X7R, then Y5V which is the least efficient of the three materials. Dielectric withstanding voltage: this refers to the momentary over voltage the capacitor is capable of withstanding with no damage. Insulation resistance: this is the DC resistance of the capacitor; it is closely related to the leakage current.

Characteristics of Ceramic Capacitors




Low impedance, equivalent series resistance (ESR) and equivalent Series Inductance (ESL). As frequencies increase, ceramic has bigger advantage over electrolytic

The final part of this presentation will cover the characteristics of ceramic capacitors. MLCCs have low impedance when compared with tantalum and other electrolytic capacitors. This includes lower inductance and equivalent series resistance (ESR). This allows ceramic capacitors to be used at much higher frequencies than electrolytic capacitors.

Temperature Coefficient: Describes change of capacitance vs. temperature. Ceramic materials are defined by their temperature coefficient



Voltage Coefficient: Describes change of capacitance vs voltage applied. Capacitance loss can be as much as 80% at rated voltage. This is a property of ceramic materials and applies to all manufacturers



Voltage Coefficient of Capacitance: describes change of capacitance vs DC voltage applied. This is a property of ceramic materials and applies to all manufacturers. The graph shows typical voltage coefficient curves for 500VDC rated X7R and NP0 capacitors. Note that the capacitance of the NP0 remains stable with applied voltage, while the X7R material can have a capacitance loss of 80% at rated voltage.


Aging: X7R, X5R, and Y5V experience a decrease in capacitance over time caused by the relaxation or realignment of the electrical dipoles within the capacitor.



  • For X7R and X5R the loss is 2.5% per decade hour and for Y5V it is 7% per decade hour, NP0 dielectric does not exhibit this phenomenon
  • De-Aging: aging is reversible by heating the capacitors over the “Curie Point” (approx 125°C), the crystalline structure of the capacitor is returned to its original state and the capacitance value observed after manufacturing.


Aging: X7R, X5R, and Y5V experience a decrease in capacitance over time caused by the relaxation or realignment of the electrical dipoles within the capacitor. For X7R and X5R the loss is 2.5% per decade hour and for Y5V it is 7% per decade hour, NP0 dielectric does not exhibit any aging. Aging is reversible by heating the capacitors over the “Curie Point” (approx 125°C), the crystalline structure of the capacitor is returned to its original state and the capacitance value observed after manufacturing.


Johanson Part Number Breakdown


  
This slide is for reference and shows the Johanson Dielectrics part number breakdown.

Summary
  • Manufacturing process and basic structure of ceramic capacitors
  •  Material systems and basic specifications of ceramic capacitors
o   Precious Metal vs Base Metal
o   Critical Specifications of MLCCs
  •        Characteristics of ceramic chip capacitors
o   Low impedance, temperature coefficient, voltage coefficient, aging

Tuesday 7 August 2012

Executive Summary of Manufacturing of Ceramic Capacitor


Design of Ceramic–Capacitor VRM’s with Estimated Load Current Feed forward 
Angel V. Peterchev Seth R. Sanders  
Department of Electrical Engrineering and Computer Science  
University of California, Berkeley, USA  

Abstract—
For voltage regulator module (VRM) designs with  ceramic output capacitors, the capacitor size has to be chosen  sufficiently large to allow for the use of relatively large inductor  values. This enables operation at conventional switching frequencies,  while meeting load transient response specifications. Due to the small effective series resistance (ESR) time constant of  ceramic capacitors, this may result in designs with output capacitor ESR substantially lower than the desired output impedance.  This is in contrast to conventional VRM implementations with  electrolytic capacitors, where the desired output impedance is  closely related to the output capacitor ESR. The feedback bandwidth is limited by stability constraints linked  to the switching frequency. The use of load current feedforward  can extend the useful bandwidth beyond the limits imposed  by feedback stability constraints. Load current feedforward  is used to handle the bulk of the regulation action, while  feedback is used only to compensate for imperfections of the  feedforward and to ensure tight DC regulation. An experimental  converter demonstrates tighter output regulation with estimated  load current feedforward, than with pure feedback control.

INTRODUCTION 
The specifications for modern voltage regulator modules  (VRM’s) require that the microprocessor supply voltage Vo  follows a load line,  Vo = Vref - Rref Io, 
where Vref is the reference voltage, Rref is the desired  load line slope (or regulator output impedance), and Io is  the current drawn by the microprocessor load.
A method for load-line regulation , where the closed-loop output impedance is set  equal to the output capacitor effective series resistance (ESR),  was introduced. With this approach, the  nominal system closed-loop bandwidth is tightly related to the  output capacitor ESR time constant. With conventional  electrolytic capacitors having such a time constant on the  order of 10 µs, it is straightforward for this approach to work  with conventional switching frequencies in the range of 200–  500 kHz. For modern VRM applications, ceramic capacitors  present an attractive alternative to electrolytics due to their  low ESR and low effective series inductance (ESL), better  reliability, and low profile. However, ceramic capacitors have  ESR time constants of about 0.2 µs, yielding the conventional  load-line design framework unworkable, since it would require  switching frequency on the order of 10 MHz.
Due to the  small ESR time constant of ceramic capacitors, this may  result in designs with output capacitor ESR substantially lower  than the desired output impedance. This is in contrast to  conventional VRM designs with electrolytic capacitors, where  the desired output impedance is closely related to the output  capacitor ESR.  In ceramic capacitor designs with conventional feedback  control, the required loop bandwidth is inversely proportional  to the output capacitor size. Extending the bandwidth can  result in cost and board area savings, since it can reduce  the required number of capacitors. However, bandwidth in a  feedback-controlled converter is limited by stability constraints  linked to the switching frequency.
In this approach, the load current feedforward is  used to handle the bulk of the regulation action, while the  feedback is used only to compensate for imperfections of the  feed forward and to ensure tight DC regulation. unloading current step.  Previous derivations of the critical inductance, use assumptions which yield the critical inductance value  directly proportional to the output capacitor ESR time constant  
                     tC = rCC, where C is the output capacitance and rC is  the ESR.
 As a result, these analyses suggest the need for  very low critical inductance values for VRM’s using low-  ESR ceramic output capacitors, implying, in turn, the need for  high switching frequencies. Further, these derivations assume  infinite load current slew rates (tI = 0), and no unloading  overshoot (?Vp = 0), while VRM specification list finite  values (Table I). Here, we present an extended analysis, which  reveals a more detailed relation between the critical inductance  value and other key power train parameters.  Fig. 2 shows a model of the VRM response for a large  unloading transient. The unloading current step can be modeled  by a magnitude ?Io and a time constant tI which  characterizes the slew rate,  Io(t) = Io(0) - ?Io(1 - e  -t/tI ), (2)  for t = 0. The maximum control effort the VRM controller  can exert after a large unloading transient, is to saturate the  duty ratio to zero after some delay td inherent in a physical  implementation. The most important observation is that the  critical inductance depends strongly on the output capacitance,  rather than on the ESR time constant, when the output  impedance is allowed to assume values different from the  capacitor ESR. This implies that in ceramic capacitor VRM  designs, the output capacitance has to be chosen sufficiently  large to allow for a reasonably high inductance value, and  thus enable operation at conventional switching frequencies  (e.g., < 1 MHz). Due to the small ceramic capacitor ESR time  constant, this results in designs where the ESR is smaller than  the specified output impedance Rref .
For a design example, consider the critical inductance  for the VRM specifications. Assume a 4-phase  converter with Vref = 1.3 V, C = 800 µF, tC = 0.2 µs, and  td = 100 ns. Then we obtain Li,crit ˜ 318 nH  per phase. These power train parameters allow efficient operation  with a 1 MHz switching frequency.Further,  the critical inductance for the loading transient is approximately  1.58 µH—much larger than Li,crit. This confirms the  observation that the unloading transient presents the critical  conditions for the transient design.
While the previous section addressed power train design  considerations for all-ceramic capacitor VRM’s, here we discuss  the appropriate control strategies for regulating the output  impedance to meet the specifications. A. Output Impedance Regulation  Conventional load-line VRM control sets the desired closedloop  impedance equal to the output capacitor ESR . However,  the discussion above indicates that in ceramic  capacitor VRM designs, the output capacitor ESR may be substantially  smaller than Rref , to enable operation at moderate  switching frequencies.  With this approach the output impedance  is specified dynamically, as a generalization of the resistive  output impedance specified in conventional load-line control.  This behavior is illustrated in Fig. 3. Importantly, the controller  has to be designed so that the output impedance is regulated  to Zref and not to Rref , since the latter approach will  result in undesirable behavior, when during a load transient,  the controller initially acts to change the inductor current  in direction opposite to the load step, eventually producing  additional output voltage overshoot.

Output Current Feed forward  It has been shown that for conventional VRM designs  with current-mode control, where Rref = rC, the voltageloop  bandwidth should be equal to 1/tC rad/s, to ensure  appropriate output impedance control. In VRM designs with voltage-mode load-line control the loop bandwidth has to be larger than 1/tC, to provide  tight load-line regulation. Since in both cases the feedback  controller bandwidth is constrained by the switching frequency  for stability reasons, these approaches require very high  switching frequencies, on the order of 10 MHz, for ceramic capacitor designs. On the other hand, in the generalized load-line regulation  approach discussed here, for Rref = rC, the feedback loop  bandwidth should exceed 1 refC, which corresponds to the  dominant time constant characterizing the output impedance. Thus, if conventional feedback control is used, the output  capacitor should be selected sufficiently large, so that  1 refC is less than the practical feedback loop bandwidth.  Clearly, in this case there is a trade-off between the number  of output capacitors required and the switching frequency  used. To eliminate this constraint, and thus enable operation  at moderate switching frequencies (fsw < 1 MHz), with a  small number of ceramic capacitors (C < 1 mF), we introduce  load current feedforward. Load current feedforward can  decrease the converter response time, without an increase of  the switching frequency, since the gain and bandwidth of  the feed forward are not limited by stability considerations,  in contrast to pure feedback regulation. The use of load  current feed forward to speed up the load transient response  in current-mode converters with stiff voltage regulation, has  been demonstrated
The modulation of the feedback signal,  however, can possibly cause subharmonic instability resulting  from the closed-loop bandwidth approaching relatively near  the switching frequency due to the small power train energy  storage element values [8]. To prevent this from happening,  a sample-and-hold (S/H) operating at the effective switching  frequency fsw,eff = fsw  of an N-phase converter, could  be introduced in the feedback path, thus eliminating switching  ripple and reducing the bandwidth of the control signal. The  sample-and-hold could be preceded by a resettable integrator  (    ) averaging the feedback signal over each effective  switching period (1/fsw,eff ), and thus providing good DC  accuracy of the feedback control. The sample-and-hold and the  resettable integrator would introduce some additional delay in  the feedback path, however this is not critical to the overall  speed of response since fast load changes are handled by the  feed forward path. All  of these have turn-off latency equal to or less than the steadystate  on-pulse-width, which is about a tenth of the switching  period in 12-V VRM’s. Hysteretic modulation also offers very  fast response, however its switching frequency is not fixed,  and it is difficult to generalize it to multi-phase power trains. 

EXPERIMENTAL RESULTS
To test the concepts discussed in this paper, a 1 MHz,  120 A, 4-phase synchronous buck converter board (International  Rectifier IRDCiP2002-C) was modified to incorporate  estimated load current feed forward and load-line regulation  with the controller structure. The on-board voltage  mode PWM modulator with phase current balancing (Intersil  ISL6558) was used A unloading transient with corresponding estimated load current  and output voltage with and without load current feedforward.  and unloading transients between 60 A and 112 A. Due to  hardware constraints on the pulsed load circuit, the loading  current step is relatively slow with a time constant of about  500 ns. The unloading current step, which tests the critical  performance of the converter, is much faster.
Estimated load  current follows very well the measured current with a delay of  about 100 ns. The 4 MHz switching noise present in the load  current estimate results from parasitic coupling to the sense  wires which were soldered on top of the VRM board. The  switching noise does not affect the DC regulation precision  because it is attenuated by the PID controller. Further, in a  dedicated implementation, the sensing can be done through  buried PCB traces, thus reducing both electrostatic and magnetic  pickup.  From the loading transient in Fig. 7 it can be seen that  with combined feedback and feedforward control, the output  voltage follows very well the desired load line. With only  feedback, however, there is an extra sag of about 40 mV  reflecting the inability of the feedback controller to tightly  regulate the output impedance. Note that this overshoot is due  to bandwidth limitations of the feedback controller, since the  feedback loop crossover frequency is not significantly larger  than 1 refC, as required in Section III-B. The observed  ^ Io  1 us/div  Io  10 mV/div  Vo  Vo  8 A/div  8 A/div  feedback only  Fig. 9. A 8 A unloading transient with corresponding estimated load current  and output voltage with and without load current feedforward.  overshoot is not a result of duty ratio saturation, since the  phase inductor value is substantially smaller than the critical  inductance of 1.58 µH for loading transients, calculated in  Section II-A.  In the unloading transient in Fig. 8, an extra overshoot of  about ?Vp ˜ 85 mV can be observed, which is expected  given the prototype parameters listed above. This response  corresponds to the duty ratio being saturated to zero about  300 ns after the beginning of the load step. Here too, the  combined feedback and feedforward control produces a better  voltage response than the feedback alone, implying a faster  transition to duty ratio saturation. In implementations using  a faster modulator, the advantage of the feed forward scheme  is expected to be even greater for large unloading transients,  since the duty ratio can be driven to saturation even faster.  Finally, Fig. 9 shows a smaller 68-to-60 A unloading transient  which does not drive the duty ratio to saturation. Analogously  to the loading transient example in Fig. 7, it is clear that  the combination of feedback and feedforward provides tighter  output impedance regulation than feedback alone. 
CONCLUSION   
For representative VRM designs with ceramic output capacitors,  the capacitor size has to be chosen sufficiently large to  allow for the use of inductor values in the range of hundreds  of nH, thus enabling efficient operation at conventional submegahertz  switching frequencies.For  designs with ceramic capacitors, the loop bandwidth required  with conventional feedback control, is inversely proportional to  the output capacitor size. Extending the bandwidth can result  in cost and board area savings, since it can reduce the required  number of capacitors. However, bandwidth in a feedback controlled  converter is limited by stability constraints linked  to the switching frequency. In both current-mode and voltage mode  control, load current feedforward can extend the useful  bandwidth beyond that achievable with pure feedback, since  feedforward is not limited by stability constraints. The load current can be estimated from the inductor  and capacitor voltages with simple RC networks. More  sophisticated and robust estimation schemes, using the input  current, for example, can be developed in the future. Different  types of modulators can be used with the load feedforward  scheme, as long as they have low latency with respect to  unloading transients. The ability of estimated load current  feedforward to provide tighter output impedance regulation  than that with pure feedback control, was demonstrated with  an experimental 12-to-1.3 V, all-ceramic capacitor